THE SELF-CORRECTING ENTERPRISE

POZNAŃ STUDIES
IN THE PHILOSOPHY OF THE SCIENCES AND THE HUMANITIES

VOLUME 92

EDITORS

Jerzy Brzeziński Izabella Nowakowa

Andrzej Klawiter Katarzyna Paprzycka (managing editor)
Piotr Kwieciński (assistant editor) Marcin Paprzycki
Krzysztof Łastowski Piotr Przybysz (assistant editor)
Leszek Nowak (editor-in-chief) Michael J. Shaffer

ADVISORY COMMITTEE

Joseph Agassi (Tel-Aviv) Leon Koj (Lublin)

Étienne Balibar (Paris) Władysław Krajewski (Warszawa)
Wolfgang Balzer (München) Theo A.F. Kuipers (Groningen)
Mario Bunge (Montreal) Witold Marciszewski (Warszawa)
Nancy Cartwright (London) Ilkka Niiniluoto (Helsinki)
Robert S. Cohen (Boston) Günter Patzig (Göttingen)
Francesco Coniglione (Catania) Jerzy Perzanowski (Toruń)
Andrzej Falkiewicz (Wrocław) Marian Przełęcki (Warszawa)
Dagfinn Føllesdal (Oslo) Jan Such (Poznań)
Bert Hamminga (Tilburg) Max Urchs (Konstanz)
Jaakko Hintikka (Boston) Jan Woleński (Kraków)
Jacek J. Jadacki (Warszawa) Ryszard Wójcicki (Warszawa)
Jerzy Kmita (Poznań)

Poznań Studies in the Philosophy of the Sciences and the Humanities

is partly sponsored by SWPS and Adam Mickiewicz University

Address: dr Katarzyna Paprzycka . Instytut Filozofii . SWPS . ul. Chodakowska 19/31

03-815 Warszawa . Poland . fax: ++48 22 517-9625
E-mail: [email protected] . Website: http://PoznanStudies.swps.edu.pl
 NEW TRENDS IN PHILOSOPHY

Katarzyna Paprzycka (editor-in-chief)

Instytut Filozofii . SWPS
ul. Chodakowska 19/31 . 03-815 Warszawa . Poland
[email protected]

New Trends in Philosophy is a new subseries of the Poznań Studies in the

Philosophy of the Sciences and the Humanities book series. It publishes
collections of papers that deal with new or underrepresented topics in
philosophy.

Forthcoming:
S. Tennenbaum (ed.), Moral Psychology
This page intentionally left blank
POZNAŃ STUDIES IN THE PHILOSOPHY OF THE SCIENCES AND THE HUMANITIES, VOLUME 92
NEW TRENDS IN PHILOSOPHY

THE SELF-CORRECTING
ENTERPRISE
ESSAYS ON WILFRID SELLARS

Edited by

Michael P. Wolf and Mark Norris Lance

Amsterdam - New York, NY 2006

The paper on which this book is printed meets the requirements of "ISO
9706:1994, Information and documentation - Paper for documents -
Requirements for permanence".

ISSN 0303-8157
ISBN-10: 90-420-2144-6
ISBN-13: 978-90-420-2144-0
©Editions Rodopi B.V., Amsterdam - New York, NY 2006
Printed in The Netherlands
 CONTENTS

Mark N. Lance and Michael P. Wolf, Preface. . . . . . . . . . . . . . . . . . 9

Michael P. Wolf, Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Terry Pinkard, Sellars the Post-Kantian? . . . . . . . . . . . . . . . . . . . . 21
Willem deVries, Folk Psychology, Theories, and the Sellarsian Roots . . 53
Timm Triplett and Willem deVries, Is Sellars’s Rylean Hypothesis
Plausible? A Dialogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
David Forman, Learning and the Necessity of Non-Conceptual Content
in Sellars’s “Empiricism and the Philosophy of Mind” . . . . . . . . . 115
Jeremy Randel Koons, Sellars, Givenness, and Epistemic Priority . . . . . 147
Susanna Schellenberg, Sellarsian Perspectives on Perception and
Non-Conceptual Content . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Matthew Burstein, Prodigal Epistemology: Coherence, Holism, and the
Sellarsian Tradition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Mark Owen Webb, Meeting Others in the Space of Reasons: Fallibilism
for Sellarsians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Michael P. Wolf, Sellars on the Revision of Theoretical Commitments . . 233
Jaroslav Peregrin, Developing Sellars’s Semantic Legacy: Meaning as a
Role . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
This page intentionally left blank
 PREFACE

This volume was the brain-child of Katarzyna Paprzycka who, nearly four
years ago, approached Mark Lance with the idea of a volume of new work on
the philosophical significance of Wilfrid Sellars. This proposal was
enthusiastically accepted, and solicitations were sent out to a wide range of
scholars working in the Sellarsian tradition. In soliciting new works, an effort
was made both to focus on young scholars who were likely to bring fresh
insights to Sellars’s work, though without ignoring many of the better known
interpreters, and also to bring in scholars who would connect Sellars’s work to
traditions outside the confines of analytic philosophy narrowly conceived.
After this initial flurry of activity more than the usual delays began, some
due to a tendency to over-commit academically. Others arose from the
decision of the US administration to invade Iraq, a decision Mark felt morally
obligated to oppose in ways that often amounted to a second occupation. Thus,
as often occurs, political commitment clashed with academic passion.
Thankfully Michael P. Wolf stepped up to help, taking over the bulk of the
major tasks of editing, correspondence with authors, formatting, etc. At long
last, we find ourselves with a finished volume, one which we believe sheds
new light on the work of an extraordinarily deep philosopher who is still too
little known in the philosophical world.
Our sincere thanks go out to the many wonderful authors represented here,
many of whom deserve to have seen their articles in print long ago. Thanks as
well to Katarzyna Paprzycka who has shown unending patience with the
project.
We dedicate this book to Joseph Camp, Jr., teacher, mentor, and friend,
who taught so many of us so much about Sellars.

Mark Norris Lance

Michael P. Wolf
3 March 2006
This page intentionally left blank
Michael P. Wolf

INTRODUCTION

Wilfrid Sellars came and shook things up in philosophy. Around the mid-point
of the 20th Century, causal confidence had set in on a number of fronts of the
Anglo-American philosophical world, particularly regarding central issues in
epistemology, the philosophy of science and the philosophy of mind. Sellars’s
challenges to those orthodoxies ran so deep that two generations later,
philosophers continue to mine his writings for insights into problems and
programs that have arisen in his wake. It might be argued that the mark of truly
great work in philosophy is not solely the arguments given within a book’s
pages, but also in the continuing returns for those who revisit its themes.
Devoted readers of Sellars’s work will agree that he offered us especially
fertile lines in this way.
In many ways, Sellars was a philosopher out of step with his times. For all
its virtues, the early stages of analytical philosophy were resolutely ahistorical,
even anti-historical at times, following the Logical Positivists’ lead in
suggesting that the debates of the past were based in large part on vacuous
distinctions and best overcome by severing our ties to them, rather than further
searches for resolution. Into this milieu stepped Sellars, an analytical
philosopher in some sense and certainly one well-versed in Carnap’s work, but
one who also considered pursuing a purely historical approach at various
points in his early career. As he himself said,
I soon came to see that a dialectical use of historical positions is the most
reliable way of anchoring arguments and making them intersubjectively
available. In the limiting case, this use of history is illustrated by
correspondence and controversial exchanges with contemporaries. Even on
paper, philosophy becomes explicitly what it has always really been, a
continuing dialogue. (1975, pp. 292-293)
More than any major philosopher of his time, Sellars put this method at the
heart of his work. Alongside volumes of his work on language, mind and
metaphysics, one will find essays on prominent figures from every major
period in Western philosophy. This approach was not mere bookkeeping, a
pursuit of minutiae for minutiae’s sake. Sellars saw these “continuing
dialogue[s]” as continuing right up to the present and incorporating the
developments of the natural sciences, just as many of those earlier figures had
12 Michael P. Wolf

aspired to do. This sense of history’s most important strains prepared him to
offer more trenchant critiques of the present and the way forward.
Analytical philosophers also began to shape their work in the mold of
scientific research programs, with tremendous logical rigor and greater
concentration on focusing each piece of work on the smallest, best-defined
problems. The succinct, precise article replaced the grand treatise as the
medium of serious philosophical inquiry. While breaks from the past are
sometimes necessary and Hegelians’ reputations for excess are not wholly
undeserved, this sort of specialization often bred more myopic and less
systematic approaches to philosophical problems. Sellars stands out from his
contemporaries in this respect. With all the conceptual rigor of the best
analytical philosophy, Sellars pursued a broad and systematic approach to the
key issues in the philosophy of language, philosophy of mind and
epistemology, which in turn informed a set of views in the philosophy of
science, ontology and ethics. Sellars offered his most concise and most famous
presentation of his central themes in “Empiricism and the Philosophy of
Mind,” initially published in 1956. Amid the attack on analyticity in Quine’s
“Two Dogmas of Empiricism” and the attack on prevailing analytical
approaches to meaning in Wittgenstein’s Philosophical Investigations, the
most resonant and distinctive theme may have been Sellars’s challenges to
epistemological foundationalism. However, part of what distinguished his
work there and throughout his corpus was that these themes were interwoven
into a critique of current approaches and a systematic vision for the path to
take forward.
Sellars was not the first anti-foundationalist, nor the first anti-Cartesian,
nor the first to offer a holistic theory of meaning. What was novel about his
approach was his insight that all three of these themes would have to be
developed conjointly in order to allow any one of them to prevail. Leave the
Cartesianism intact, and a collapse into some kind of foundationalism is
inevitable; permit some part of the language to have a meaning unrelated to the
rest of the language and one would inevitably be compelled to explain such
content in terms of an immediately grasped state, in keeping with
foundationalism and Cartesianism. Entering an account whose parts are so
densely interrelated can prove difficult, but the best entry point might be
Sellars’s account of the relation between thought and language. Offering the
theme that best resonated with analytical philosophy of the time, Sellars
suggested that having a thought and making use of a piece of the language
were one and the same; to have a thought was simply to be disposed in the
same way that one would be, except less candidly or overtly. Sellars classified
that sameness of disposition in broadly functionalist terms. For an expression
to have meaning and a thought to have content were for such states to be so
 Introduction 13

organized with other word- and sentence-types (in the case of meaning) and
states (in the case of mental content) in distinctive ways in a larger linguistic
economy. Like all functionalist approaches, the central idea was that a
particular state or utterance would be the product of particular sorts of causal
inputs, and would result in a characteristic set of outputs in the form of
utterances, inscriptions or other sorts of behavior. But unlike most
functionalists, Sellars did not see this as a merely causal story. The economy in
question took shape as a medium for practices of giving and asking for
justifications for one’s utterances and actions, creating what Sellars
metaphorically called “the space of giving and asking for reasons,” rather than
as a purely regular natural process. The rationality and normativity implicit in
these suggestions was an irreducible feature of thought and language. To speak
the language or to have thought was not merely to be causally induced into a
particular state (though one would be, obviously), but also to recognize and
take those states and utterances to have the force of reasons for other such
utterances and actions.
This approach, both functional and normative, informed Sellars’s critique
of foundationalism in all its forms. Traditionally, attacks on foundationalism
offered very localized critiques of various species of beliefs (e.g. reports of
sense impressions, analytic truths, etc.) and rejected their suitability as
“regress-stoppers” that could stand without the need for further justification.
Success with this strategy never really addressed the fundamental assumptions
behind foundationalism and the need for some sort of basic belief, and so the
beast always slinked away and returned to fight another day. “Perhaps sense
impressions/analytic truths/etc. are not the source of basic beliefs,” it could
always be said, “but without some foundation, there will be no knowledge at
all.” Sellars went directly after the assumptions behind this view in attacking
what he would call the “Myth of the Given.” Empirical knowledge, he said,
suffers from an inconsistent triad of assumptions:
A. X senses red sense content s entails x non-inferentially knows that s is red.
B. The ability to sense sense contents is unacquired.
C. The ability to know facts of the form x is Φ is acquired. (1997, §6/p. 21)
The tensions between these lead us to posit a source of direct and unmediated
knowledge, the given. Sellars’s resolution, in the briefest of terms, was to
reject the very idea of a given as incoherent and replace it with the sort of
rational, inferential economy described above, in which each piece of
knowledge would require the knower to have mastered a larger battery of
concepts into which that piece of knowledge could fit as a reason for some
things and other things might serve as a reason for it. This did not preclude
some forms of non-inferential knowledge, as when I open my eyes under
14 Michael P. Wolf

normal conditions and observe that something before me is blue without

deliberation or overt discussion of the finer points of that concept’s usage. This
sense of immediacy, of something just popping into our heads, remains intact.
What has been dispelled is the notion that some piece of potential knowledge
like that could be held by itself, prior to and independent of the rest of our
conceptual capacities and epistemic statuses, and thus serve as a support for
the inferential knowledge that would “rest upon it.”
There is clearly some point to the picture of human knowledge as resting on a
level of propositions – observation reports – which do not rest on other
propositions in the same way as other propositions rest on them. On the other
hand, I do wish to insist that the metaphor of “foundation” is misleading in that
it keeps us from seeing that if there is a logical dimension in which other
empirical propositions rest on observation reports, there is another logical
dimension in which the latter rest on the former. (1997, §38/pp. 78-79)
The mention of “observation reports” here is, of course, not incidental. In
the three centuries since Descartes reshaped the epistemological terrain for
Western philosophy, confidence had waned in the prospects of deploying
ordinary perceptual beliefs as a foundation, and phenomenalism had
reemerged as the most promising strategy for many devoted empiricists. Send
evil demons after me, or suck my brains out and put them in vats as long as
you like, I have direct and unmediated access to the contents of my mind such
as sense-impressions, the foundationalists might say. Without a gap, as we
supposedly find between appearances and the world itself, there was no room
for error or deception. So long as we were willing to talk about sense contents
themselves (to use Sellars’s term) in a slightly tortuous fashion, a foundation
would be available. Here, the link between Cartesianism in the philosophy of
mind and foundationalism in epistemology becomes evident, but more
importantly, we confront perhaps the most immediately compelling aspect of
both views. Come what may, nothing seems so clear and well-known to me as
the contents of my own mind, and some of them seem to need none of the
complex conceptual entangelements of empirical knowledge or scientific
theories. To preserve the insights offered thus far, Sellars would have to make
the case that just like empirical knowledge, knowledge of our own mental
states involved the deployment of acquired conceptual capacities and not the
exercise of some innate capacity for introspection.
 Introduction 15

To this end, Sellars offered his own myth, the “Myth of Jones,” as he called
it.1 Sellars asks us to imagine a community of speakers, perhaps in our
Paleolithic past, who engage the world in various ways but do not conceive of
themselves as possessors of mental states as we understand them. In some
sense, our pets and fellow creatures do this much, but Sellars wants his
subjects to be a bit sharper than our dogs and cats. The Ryleans, as he calls his
mythical community, should be in possession of a reasonably sophisticated
language, including the basic first-order logical operators and subjunctive
conditionals, and a reasonably large battery of concepts for ordinary medium-
sized drygoods around them. To some extent, these people should know their
way around their world. To that end, they should also engage in some
reflective consideration of the behavioral dispositions of those around them
without appealing to any sort of internal states (in the distinctly mental sense)
as causal explananda. Thus, a Rylean would be ready to say things like, “Smith
always steals my food if I do not watch him, and he would do it again given
the chance, so I will watch him until the food is eaten.” Given these resources,
we should expect that the Ryleans would be able to navigate their world
reasonably well, perhaps even developing more sophisticated theories about
various observable phenomena, though we should also expect them to have
unexpected disappointments in those areas where their theories of the world
are most crude or as yet unformed. Clumsy as they might be, Sellars insists
that there is nothing logically or conceptually problematic about imagining
such a group of speakers.
Things pick up a bit for the Ryleans when a genius emerges in their midst.
We shall call this genius “Jones.” Jones suggests that the Ryleans might help
themselves to explain the behavior with which they have had so little success
by postulating a set of states that belong to each of them individually and that
both arise in response to their environment and lead them to act in various
ways. These would be like things that the Ryleans already know in some
crucial ways and different in others. As models, Jones suggests that the
Ryleans might imagine a set of internal states that were like acts of speaking,
but which did not involve overt noises. That is, as they found themselves

1Myths need not be false, of course, nor need they be empirically verifiable to serve their purpose.
Hence, one should not think that our concern for Jones and the Ryleans would be answered by
discovering extensive evidence of ancient tribes with mental-term-free languages, nor does the
absence of such evidence undermine this account. The scenario is a philosophical device, not an
empirical retrodiction. Surely, there is some room for empirical consideration here, if, say,
cognitive psychologists made startling discoveries about how the brain develops, but the Myth is
ultimately here to tell us something about what it is to conceptualize our own mental states. See
deVries’ contribution to this volume for more on this point.
16 Michael P. Wolf

disposed to act in various ways and that would involve speech acts, they could
suppose some internal state that arose in just the same circumstances and led to
just the same sorts of things. The affinity would thus be limited, but it would
augment their capacities to explain, predict and understand their own behavior.
Similar things could be done taking physical objects as models, giving the
Ryleans what we think of as sensations. As subsequent generations adopt
Jones’s theories and they enhance the Ryleans’ capacities in the appropriate
ways, the Ryleans might become so familiar with these strategies that they
begin to notice the sorts of features that would prompt such a theoretical claim
in a much more spontaneous way. Rather than theorizing, they begin using
these as observation concepts and have learned to track and understand their
own mental lives. It is crucial to note that Sellars is not suggesting that
thoughts and impressions are unreal or useful fictions. He takes them quite
seriously and believes them to be real. The point of the Myth of Jones is to
deliver a scenario that illustrates how we might develop an understanding of
our own mental states without appeal to some primitive faculty of self-access.
If this succeeds, then the last vestige of the Cartesian program has been
removed. We do not possess thought and self-consciousness of our mental
states by beginning with direct access to them and later acquiring public
language to express that understanding (as both classical Rationalism and
Empiricism presumed and early analytical philosophy adopted as its own).
Rather, we first learn to deal with the world around us and interact with others
who share our social practices and then use those conceptual resources to
understand and articulate the features of the mental states we experience. The
Ryleans always had thoughts and sensations; what they did was figure out a
way of grasping them.
It would be fair to say that these conclusions have been contentious, but
they have also served as a springboard for many philosophers and helped usher
in various strands of contemporary philosophy of language, cognitive science
and epistemology. Sellars also did all of this while steadfastly maintaining a
realist stance in the philosophy of science and adding tremendous
sophistication to accounts of explanation and the role and nature of models in
theories. Many see his criticism of the Positivists “layer cake” approach to
theories, with separate levels of discourse for observation statements,
empirical generalizations and laws as a turning point in the move away from
the strictures of Positivism towards a richer, more robust account of scientific
inquiry. His themes in ontology and ethics also merit serious consideration, but
we will end our introduction of Sellarsian themes here, since these will serve
as the backbone for so much of the discussion that follows. Where appropriate,
the various authors will address any additional material.
 Introduction 17

In “Sellars the Post-Kantian?” Terry Pinkard investigates Sellars’s relation

to post-Kantian German philosophy. Sellars made no bones about his
admiration for both Kant and Hegel, both decidedly unfashionable choices in
mid-20th century analytical philosophy. Sellars’s conception of the manifest
and scientific images of man, discussed at greater length by Pinkard, resembles
the concerns with human freedom in the face of scientific revolution that
preoccupied post-Kantians like Schelling and Fichte. This suggests a need for
Hegel as an element in any interpretation of Sellars, a suggestion that Sellars
himself would surely have endorsed.
In “Folk Psychology, Theories, and the Sellarsian Roots,” Willem deVries
addresses Sellars’s relation to other contemporary philosophers and cognitive
scientists who embrace the suggestion from the Myth of Jones that
psychological concepts function like theoretical concepts. Making our
psychological concepts acquired has led some to conclude that like all other
theoretical concepts, they are potentially revisable and even replaceable.
Sellars’s view that the scientific image was gradually supplanting the manifest
image has led eliminiativists like Paul Churchland to suggest that Jones’s
development of folk psychology demonstrates that it could be replaced by a
fully developed neuroscience. Cognitive scientists, especially those in
developmental psychology, have taken this suggestion as an empirical
hypothesis about the underlying cognitive mechanisms that express themselves
in our acquisition of various sorts of competence. DeVries examines these
views for both their affinity to Sellars’s view and their plausibility in light of
Sellars’s criticisms, staking out a more moderate position for Sellars on
eliminativism and throwing some cold water on the more radical claims of the
developmental psychologists.
In “Is Sellars’s Rylean Hypothesis Plausible?: A Dialogue,” Willem
deVries and Timm Triplett offer us a lively extended dialogue on Sellarsian
themes in the philosophy of mind. As Triplett notes, it includes “the first time
in the history of philosophy that Jean-Paul Sartre has been invoked in the
defense of Wilfrid Sellars!” Never let it be said that the authors failed to
deliver cutting-edge philosophy. Sellars’s critics are well represented here.
In “Learning and the Necessity of Non-Conceptual Content in Sellars’s
‘Empiricism and the Philosophy of Mind’,” David Forman addresses an
objection to Sellars from John McDowell regarding non-conceptual content.
McDowell contends that Sellars’s rejection of the “Myth of the Given” (which
McDowell endorses) leads to a “frictionless coherentism” when coupled with
his inferentialism in a theory of meaning. Forman defends Sellars against this
charge, appealing to existence in the perceiver of some sort of non-conceptual
sensory states whose qualities systematically correspond to the ones one is
learning to identify. There must then be a way of conceptualizing these
18 Michael P. Wolf

non-conceptual states so that they may serve a role in the collection of

inductive reasons necessary for recognizing the authority of a perceptual report
type. Forman argues that this allows Sellars to keep his inferentialism without
falling into the “Myth of the Given.”
In “Sellars, Givenness, and Epistemic Priority,” Jeremy Koons addresses
recent criticisms of Sellars’s views on the Given. While critics rightly go after
the purported Sellarsian view that sensations can have no role in justifying
observational knowledge, Koons contends that Sellarsians can accommodate
the more plausible elements of these criticisms. Doing so requires some
elaboration of Sellars’s view that observation reports rest “in some sense” on
other beliefs and his commitments to internalism. The end result is an attack
on the Given made all the stronger by greater clarity.
In “Sellarsian Perspectives on Perception and Non-Conceptual Content,”
Susanna Schellenberg addresses the role of non-conceptual content in Sellars’s
account of perception. She argues that there is a legitimate sense in which we
may both say that all perception is concept-dependent and that something is
given to us in perception. Her account addresses both the insufficiency of rules
of application in directing thought towards particular objects, and the need for
a “full-blown language” in doing so. A clearer account of Sellars’s positions
on sensations and intuitions emerges from this discussion.
In “Prodigal Epistemology: Coherence, Holism, and the Sellarsian
Tradition,” Matthew Burstein argues that Sellars’s attack on foundationalism
in “Empiricism and the Philosophy of Mind” does not entail a commitment to
coherentism, as many readers both friendly and critical have assumed.
Coherentism demands various forms of argumentation where they are not
needed and wrongly limits the domain of possible forms of justification. While
the Sellarsian tradition does commit us to certain forms of holism, we may still
“stanch the slide” from antifoundationalism into coherentism.
In “Meeting Others in the Space of Reasons: Fallibilism for Sellarsians,”
Mark Webb offers an examination of fallibilism along Sellarsian lines. Giving
up on certainty is a viable possibility with a little help from Sellars’s
conception of empirical knowledge as a “self-correcting enterprise.” He
examines a number of works from the contemporary literature and offers his
own formulation of an appropriate principle of fallbilism.
In “Sellars on the Revision of Theoretical Commitments,” I consider both
the epistemological and semantic features of the revision of commitments. If
we assert that languages are made up of sentences and subsentential
expressions with distinctive inferential roles, then we commit ourselves to
some specific role for each of them. Yet we also find ourselves compelled to
revise such commitments, particularly with respect to empirical claims and
theoretical claims in the sciences. If we commit ourselves to specific roles but
 Introduction 19

must adopt other commitments as inquiry proceeds, incommensurability

looms. I suggest that we can resolve this with a pair of Sellarsian themes. On
the one hand, we understand conceptual change as an ongoing evaluative
procedure of endorsing a continuity of sufficiently affiliated versions of related
comments, thereby abandoning univocality but replacing it with a diachronic
process of establishing affinities between speakers. This process is in turn
guided by the constraints of a pair of “rules of criticism” guiding the revision
of theoretical commitments that jointly serve to both drive empirical inquiry
and cull the options for replacements.
In “Developing Sellars’s Semantic Legacy: Meaning As a Role,” Jaroslav
Peregrin investigates the status of logical models and formal semantics with
respect to Sellarsian inferentialist semantics. He argues that Sellars’s account
might be augmented by the use of what he calls “Euclidian” models, following
the usage introduced by Imre Lakatos. Inferentialism has often been seen as
fundamentally opposed to the rigid structures of formal semantics developed in
the tradition of Carnap, but Peregrin contends that they there is no reason for
this division and that these may be tools of great value for the inferentialist.

Michael P. Wolf
Department of Philosophy
California State University, Fresno
Fresno, CA 93740
United States of America
e-mail: [email protected]

REFERENCES

Sellars, W. (1975). Autobiographical Reflections. In: H.N. Castañeda (ed.), Action, Knowledge,
and Reality: Critical Studies in Honor of Wilfrid Sellars, pp. 277-293. Indianapolis: The
Bobbs-Merrill Company, Inc.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
This page intentionally left blank
Terry Pinkard

SELLARS THE POST-KANTIAN?

ABSTRACT. In Kant’s “fact of reason,” there is an apparent paradox of our being

subject to laws of which we must regard ourselves as the author, while at the same time
being normatively bound by the same laws that we cannot see ourselves as authoring.
Working out the implications of this apparent paradox generated much of the response
to Kant in post-Kantian idealism. Wilfrid Sellars notes the same paradox when he
speaks of the “paradox of man’s encounter with himself” in “Philosophy and the
Scientific Image of Man.” Like some of the idealists, Sellars thus opted for “two track”
system of philosophy that combines the two tracks in a metaphor of “stereoscopic
vision.” This paper argues that understanding Sellars’s own thought in terms of the
issues that formed the dynamic of post-Kantian philosophy in Germany puts us in a
better position to understand why Sellars’s own conception of experience and of the
unreality of the manifest image is not completely consistent with the argumentative
direction of his thought. Understanding Sellars in this way helps us understand the
limits of post-Kantian idealism just as understanding the dynamic of post-Kantian
idealism gives us a more nuanced version of Sellars’s conception of experience.

A rule is an embodied generalization which to speak loosely but suggestively, tends to

make itself true.
(Sellars 1953, p. 299).

Mindedness (Geist) existing in and for itself is not the mere result of nature but instead is
in truth its own result; it produces itself out of the presuppositions that it makes for itself,
from out of the logical idea and external nature, and is the truth of the former as much as
of the latter.
(Hegel 2004, §381)

Wilfrid Sellars gave the subtitle, “Variations on Kantian Themes,” to his one
full-length book treatment of his ideas (Science and Metaphysics, an expanded
version of his Locke lectures of 1966), and he notoriously, although a bit
tongue in cheek, characterized his other major statement of his views,
“Empiricism and the Philosophy of Mind,” as a series of Meditations
Hegeliennes. But if anything, the “Kantian” and “Hegelian” elements might
seem to be only very general: The “Kantian” element is quite general, in that

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 21-52.
Amsterdam/NewYork: Rodopi, 2006.
22 Terry Pinkard

Sellars holds that there is some version of the “synthetic a priori” and some
sense in which we can be said to bring certain conceptual elements to
experience rather than deriving them from experience; and the Hegelian
element might be perceived to be even more superficial, consisting only a very
general kind of “holism” that both he and Hegel share.
Indeed, Sellars’s work is characterized by the kind of concerns that
animated the early analytic and, more specifically, positivist reactions to the
revolutionary developments in late nineteenth and early twentieth century
movements in science and philosophy. What has been called the “second
scientific revolution” in physics (and likewise in chemistry and biology)
seemed to call for a more serious approach to science than anything that the
older versions of the idealist Naturphilosophie had to offer; the corresponding
revolutionary developments in mathematics since Hegel’s day also led many to
believe that the so-called “dialectical logic” of the Hegelians was, most
politely put, “outmoded”; and, finally, the revolutionary development in the
approach to philosophical problems concerning the relation between thought
and reality that focused on the representational character of language
(particularly on issues of reference and meaning) gave philosophers a better
and more concrete grasp on the issues than the all too general notion of
“thought’s relation to being,” and the rigor of Frege’s and Russell’s
mathematical logic offered a far better apparatus for understanding these issues
than did anything like “dialectical logic.”
Sellars’s invocation of Kant was, of course, part of a long-building
movement in contemporary philosophy. Many recognized that issues brought
to the fore by the new science, the new mathematics, the new logic, and the
new analytical approach called for a similarly new philosophical approach, not
just for the rejection of Hegelian idealism, and those same concerns, as
Michael Friedman has shown, led to a revival of Kantianism on the continental
side of the philosophical divide, where Kantianism came to be seen as a kind
of approach to those issues that did not take the “analytical” turn towards
issues of meaning and reference to be the preferred method for tackling the
new problems that arose in the second scientific revolution. (See Friedman
2000). German Neo-Kantianism, of course, was not so much philosophically
defeated as it was simply cut short by the poison of the Nazi regime in
Germany and the war that followed. In the Anglo-American world, the Kant
revival had to wait until the 1960’s, when in 1959, P.F. Strawson’s book,
Individuals, appeared with the provocative subtitle, “An Essay in Descriptive
Metaphysics.” (One nowadays has to remind oneself just how shocking it was
in 1959 to see the word “metaphysics” in the subtitle of a book by a prominent
British analytical philosopher.) Sellars’s own Locke lectures in 1966, along
with Strawson’s 1966 book on Kant (The Bounds of Sense, published at the
 Sellars the Post-Kantian? 23

same time as Jonathan Bennett’s, Kant’s Analytic) helped to stoke the interest
in Kant’s theoretical philosophy, which was followed by a tsunami of interest
in Kant’s ethics fueled by John Rawls’ A Theory of Justice in 1971 put Kant
firmly back on the map in Anglo-American philosophy. Hegel, however,
remained an outsider, since he seemed that he had little to offer the dominating
interests in philosophy of language, Fregean logic, or philosophy of science.
The other object of attack from the new analytical movement was
American pragmatism, with Dewey’s “logic” suffering the same kind of
withering dismissal from Russell that Hegel’s “logic” had earlier received.
What divided, in particular, Dewey from the older, Hegelian tradition – it is
worth being reminded that Dewey began his philosophical career as a kind of
Hegelian – was Dewey’s belief that Darwinian evolutionary biology had
changed the entire playing field for philosophy and that the challenge to
modern thought came from that direction, not from the new emphasis on
Fregean logic and the representational structure of language.1
Sellars mixed some of these streams together, not only being an analytical
philosopher (even with “Kantian variations” as his theme), but also a
pragmatist of a sort.2 What I want to suggest here is that Sellars own systematic
views, seen most generally as a mixture of pragmatism and analytical
philosophy, can be put into an even broader context of what is nowadays
called “post-Kantian” philosophy (with the term referring to that broad
movement in Germany in the late eighteenth up to the mid-nineteenth century
that took Kant as its starting point and then tried to use Kant to get out of
Kant).3 The term, “post-Kantian” has been used to characterize the movement,
since the older term, “German Idealism,” neither characterized all the
participants in the movement, nor did adequately capture the differences even
among the various so-called “idealists” that it supposedly described. Indeed, it
was a set of very similar issues that motivated some pragmatists (most
especially, Dewey) and the issues that spurred both the development of
analytical philosophy and the Neo-Kantian movement in Germany which

1 See Burke (1994) for a reconstruction and defense of Dewey in terms of the renewed interest in
evolutionary approaches to epistemology.
2 Robert Brandom in his “Study Guide” published as an appendix to the reprint of Empiricism and

the Philosophy of Mind notes that “Sellars’s pragmatism dictates that issues of conceptual priority
be translated into questions of the relative autonomy of different strata of language – that is, into
questions concerning what language games can be played independently of and antecedently to
which others” (Sellars 1997, p. 170).
3 This characterization of “using Kant to get out of Kant” was the point stressed by Rolf-Peter

Horstmann in his classic, “Die Grenzen der Vernunft,” (1991) a study of the themes and issues
motivating the “post-Kantian” philosophers.
24 Terry Pinkard

provided the impetus for much of the much earlier post-Kantian development.
To make this point requires the telling of a rather generalized narrative about
that period of philosophy – or, to paraphrase Sellars, of how the impulses of
that movement, in the broadest senses of the terms, hung together in the
broadest sense of the term. My concern here, however, is not just to place
Sellars a different context from where he is usually found; I hope that doing so
will also offer us a different way of understanding his thought.

1. Some Post-Kantian Issues: Reinhold, Fichte, and Schelling

The story of the move from Kant to his immediate successors is complex and
multifaceted, and I have tried to tell that story in more detail elsewhere (see
Pinkard 2002). To get a handle on it, though, it is best to begin with a well-
known Sellarsian theme: The Myth of the Given. The problem seemed
particularly acute for the post-Kantians, starting with Karl Leonhard Reinhold,
who recognized that the Kantian system, spread across three Critiques and a
variety of other works, needed some kind of interpretive reordering. In
particular, there was the usual concern about Kant’s notion of unknowable
things in themselves and a view that by the time of the Critique of Judgment,
Kant had effectively shown that the more dualistic understanding that the first
Critique seemed to present (with its sharp division among the faculties of
understanding and sensibility) was not to be maintained and that both faculties
were somehow to be united. Very striking for the post-Kantians emerging in
the last decade of the eighteenth century was the third Critique’s assertion that
in aesthetic judgment we have a normative, universally communicable
judgment (although not, strictly speaking, a conceptual judgment because we
cannot state a rule for it), and that in making this kind of judgment we orient
ourselves by the “indeterminate concept (namely, of the supersensible
substrate of appearances)” (1987, §57) which is “neither nature nor freedom
and yet is linked with the basis of freedom, the supersensible” (1987, §59).
One line of post-Kantian thought took that an invitation to unite Spinoza’s
monistic conception of substance with the Kantian metaphysics of freedom;
but another line of post-Kantian thought (which put great stress on Kant’s own
emphasis on what we would call the “intersubjectivity” of judgments) took
that instead as Kant’s own admission that the strict separation of concepts and
 Sellars the Post-Kantian? 25

intuitions was untenable.4 In particular, in light of Kant’s later texts, passages

from the first Critique stood out, such as: “The same function which gives
unity to the various representations in a judgment also gives unity to the mere
synthesis of various representations in an intuition; and this unity, in its most
general expression, we entitle the pure concept of the understanding.”5 This
seemed to indicate, most forcefully, that even for Kant himself intuition could
not be conceived as a separate and independent “faculty” that provided some
kind of foundation for knowledge that could be apprehended without the
exercise of any conceptual faculties being needed or brought into play.
To Reinhold, this suggested that the key concept was instead that of
representation in general, since both intuitions and concepts were species of
representation, and if we could understand the representative function of
“mindedness” itself, we could put the Kantian system into proper order. For
Reinhold, a sensation, for example, was merely a subjective mental occurrence
open to empirical (although only introspective) investigation. It is another
thing altogether, Reinhold argued, to take that sensation as a representation of
something different from the sensation, and its taking on that representative
function is a result of our bestowing what Reinhold called “representational
form” on the sensation, that is, giving what we could anachronistically call
“semantic content” to the sensation. This was all based on what Reinhold
called the “fact of consciousness,” an awareness that we were conscious of
objects in the world and of ourselves as the subject to which those conscious
states were ascribed; this “fact” was supposed to serve as the foundation for all
other claims about representations.
Reinhold’s own version of this kind of reworking of Kant ran into some
severe difficulties, which we need not go into here.6 Fichte in particular picked

4 The relevant text in the Critique of Judgment was the following: “Instead, we must [here] take
sensus communis to mean the idea of a sense shared [by all of us], i.e., a power to judge that in
reflecting takes account (a priori), in our thought, of everyone else’s way of representing, in order
as it were to compare our own judgment with human reason in general and thus escape the illusion
that arises from the ease of mistaking subjective and private conditions for objective ones . . . we
compare our judgment not so much with the actual as rather with the merely possible judgments of
others, and put ourselves in the position of everyone else, merely by abstracting from the
limitations that happen to attach to our own judging” (1987, §40).
5 Kant (1964, A79/B104). This judgment was repeated in even stronger form in the well known

footnote to §16 of the B-Transcendental Deduction, where Kant said, “The synthetic unity of
apperception is therefore that highest point, to which we must ascribe all employment of the
understanding, even the whole of logic, and conformably therewith, transcendental philosophy.
Indeed this faculty of apperception is the understanding itself.”
6 I cover the objections to Reinhold in chapter five of Pinkard (2002).
26 Terry Pinkard

up on the difficulties in Reinhold’s position as stemming from Reinhold’s

foundationalist approach to Kantianism. If the so-called “fact of
consciousness” is to serve as a foundation, it must itself be conceptually
apprehended, which means in turn that it would beg all the questions to use it
as a foundational “fact” since the report of such a fact itself already
presupposes that other conceptual capacities are in play (or that we are in
possession of them), and the issue has to do with how we can claim entitlement
to those concepts. As Fichte rephrased the issue, if there simply can be no base
level “fact” of any sort that we could know without knowing anything else or
which could be apprehended without bringing any conceptual capacities into
play, then the issue in rethinking Kantianism was much more severe than
Reinhold himself had seen.
Fichte saw that if this were the case, then a particular problem that had
surfaced in Kant’s practical philosophy between the publication of Kant’s
Groundwork in 1785 and the Critique of Practical Reason in 1788 covered all
of Kant’s philosophy. In the Groundwork Kant had declared that
[T]he will is therefore not merely subject to the law, but is so subject that it
must be considered as also giving the law to itself and precisely on this account
as first of all subject to the law (of which it can regard itself as the author).
(1998, p. 98/AA431)
However, as Kant himself clearly saw, taken literally, this would be
paradoxical; a lawless will cannot obligate us, and a choice of practical
principle made on the basis of no law would be simply arbitrary, so if the will
were to give itself a law, it would need another law to choose that law; but if
only a self-legislated law was obligatory, that prior law could not itself be
obligatory. By the time of the Critique of Practical Reason, Kant had decided
to call this the “fact of reason” (using the term, “Faktum,” with its overtones of
“making” instead of the more ordinary “Tatsache,” a fact in the non-normative
sense, indeed, in the sense that Reinhold put into play). Kant clearly
recognized that the very notion of there being a kind of moment of existential
choice in which we somehow freely elect to impose a binding law on ourselves
without any prior commitment is itself a pure fantasy (although something like
that seems at least to have been Jean-Paul Sartre’s position at one point); we
always find ourselves in the space of reasons, bound by the commitments of
reason itself, and there is no getting outside of that space, no non-normative
foundation on which the normative space of reasons could metaphorically sit.
 Sellars the Post-Kantian? 27

Yet Kant’s use of the “fact of reason” also showed that he did not think that
he had disposed of the paradox, since he acknowledged that we nonetheless
must see this law as having been self-authored at least in the sense that
“reason” is said to “originate” (to be legislative, “gesetzgebend” for) the law.7
If indeed, however, as Kant claimed, “the same function which gives unity to
the various representations in a judgment also gives unity to the mere synthesis
of various representations in an intuition,” then there can be no “given” of any
sort, the apprehension of which could possibly serve as the basis for other
epistemic claims. If “pure intuitions” put no normative constraints on the
“synthesis of various representations in an intuition,” then, so Fichte saw, the
paradox in practical reason extends itself to cover all problems of normative
authority. For Fichte, giving up the “given” means that the “Kantian paradox”
takes center stage.8
Fichte thus put forward the view that the problem of the
“self-authorization” of reason was simply the philosophical problem that had
to be solved, and to that end, Fichte devised an ingenious answer. (Here we
shall have to content ourselves only with the barest of sketches of Fichte’s
immensely subtle and complex position.) First, Fichte argued that even
assertions of apparent certainty, such as “A = A,” in fact get their necessity
from something like a prior inference license to the effect of “if A, then A,”9
which, as he saw, raised the question of how or from what such inferences are
licensed; and, since he had ruled out any apprehension of some kind of object
(even a pure meaning), it followed that its authority must come from us, that
is, from what he called the “I,” indicating by that term the subject that
supposedly both institutes the license and also simultaneously authorizes itself
to institute such licenses. This “I” must thereby come to grasp itself not as a
natural “thing,” and “object” out there in the world, but as something more like
a normative status, which it quasi-paradoxically attributes to itself. To cut a
long story (Fichte’s story) short, the “I,” so Fichte claimed, could attribute

7 “The consciousness of this fundamental law may be called a fact of reason, since one cannot
ferret it out from antecedent data of reason, such as the consciousness of freedom (for this is not
antecedently given) and since it forces itself upon us as a synthetic a priori proposition based on
no pure or empirical intuition . . . one must note that it is not an empirical fact but the sole fact of
pure reason, which by it proclaims itself as originating law.” (Kant 1956, p. 31/AA 31)).
8 This is the name I give to this general problem of the authority of reason in Pinkard (2002).
9 Fichte formulates this in the terms of a logic that is not at all familiar to us nowadays; he says in

one place, for example, that the proposition, “A = A,” “has content only under a certain condition.
If A is posited, it is naturally posited as A, as having the predicate A. But this proposition still tells
nothing as to whether it is actually is posited, and hence whether it is posited with any particular
predicate” (Fichte 1982, p. 96).
28 Terry Pinkard

such a status to itself only by virtue of positing that there was something (the
“not-I”) that authorized it to do so. The tension (if not contradiction) had to do
with the “not-I’s” getting its authority from the “I,” even though the “I” was
authorizing it to be the non-normative basis of the “I’s” assertions. That “not-
I,” so Fichte went on to claim, is to be conceived quasi-causally, quasi-
normatively as an Anstoß, a “stimulus,” that is, a “check” on the “I’s”
otherwise unconstrained power to posit things. To put it in less Fichtean terms:
The world is experienced as a constraint (both causally and normatively) on
what would be our otherwise unchecked inferential powers, but the normative
force of that constraint is something we bestow on it, not something that it
wears on its sleeve as it presents itself to us. The “Not-I” is, on Fichte’s view, a
necessary posit since without it, we would be left with a view of our
mindedness as simply frictionlessly spinning in the void (to appropriate a
phrase from John McDowell). However, he just left the mind-world
connection at the level of a causal event on which we bestowed normative
significance, with its remaining rather mysterious how we were to apprehend
that quasi-causal event, the Anstoß, without its already having some kind of
conceptual status. (Fichte later amended this in his practical philosophy to
make the “Not-I” into another “I,” that is, another agent, since he came to
believe that only another agent and not a natural fact simply impacting on me
can give me a reason for believing something; but that is another, longer
story.)
There was a group of young people attending Fichte’s lectures who more
or less came to the conclusion that Fichte’s solution did not so much resolve
Fichte’s problem as it just renamed it. (They came to be known as the “early
Romantics.”) As Fichte saw, we require a doctrine of receptivity (of the
“Not-I”) to make any sense of our own activity, but that receptivity is itself a
“posit” made by the spontaneous subject. The very idea of the “stimulus” (the
Anstoß, in Fichte’s terms) served to point out the difficulty of rejecting the
given and avoiding the kind of frictionless spinning in the void (that many of
Fichte’s readers took him to be doing anyway). Some of them (particularly
Hölderlin, Schelling, and Novalis) took the problem to be Fichte’s way of
thinking the distinction between the “I” and the “Not-I” was something
established by the “I” itself; or, to put it differently, Fichte’s idea that since the
distinction between the subjective and the objective was itself not a natural
distinction to be found in the world, it was therefore itself a normatively
established and therefore a subjective distinction. That way of putting things,
so the early Romantic thinking went, would take you nowhere except perhaps
to the point where subjectivity is conceived along the lines of (only
self-constrained) inferentialism, and the relation to the world would be
 Sellars the Post-Kantian? 29

conceived as purely causal, and, ultimately, Kant’s doctrine of unknowable

things in themselves would have to be brought back into play.
The early Romantics instead tried to show how we could intelligibly get a
grip on both items: On our own spontaneity and even creativity in fashioning
accounts of the world, and on our nonetheless being responsive to a world that
is not in any terms the product of our creation and which does indeed put
normative constraints on our beliefs.10 To do that, some of them implicitly
abandoned the kind of inferentialism that Fichte endorsed, holding instead that
there are variety of entitlements that must be in play in our epistemic activities
that do not normally themselves count as knowledge or as evidence for a claim
but which must be in play (or can be at least partly made conscious) for us to
be able to make even the ordinary perceptual judgments we make.11
Schelling (along with Hölderlin and Novalis) took that a step farther and
proposed that we see the distinction between the subjective and the objective
as being itself neither subjective nor objective but as something itself
established by something more basic, which Schelling took to calling the
“absolute.” Starting the argument with the distinction between subjects and
objects already in place, so these early Romantics argued, is starting with a
kind of separation between elements that should itself not be taken as final. In
Schelling’s account, the idea that there is a metaphysical line dividing the
subjective from the objective is misleading; in our own experience, we are
always on both sides of the line, alternately seeing ourselves as subjects with a
point of view on the world and as objects in the natural world (or, as subjects
who are one among many other subjects, and as objects who are one among
many other objects).
On Schelling’s account, naturalists tend to want to make everything an
object, whereas subjective idealists (like Fichte) tend to want to make
everything a posit from the “I,” and therefore, so Schelling had concluded by
the late 1790s, philosophy had to pursue a two-track method: On the one hand,
it had to understand that we were organic beings in a natural world and that
there was no need to explain the status of agency by appeal to some other kind
of non-empirical “stuff” (such as mental substance) that would supposedly
form our constitution, which led him to construct a Naturphilosophie in which
he developed a conception of nature as a series of “powers” (Potenzen, in
Schelling’s terminology) that played off each other, resulting in new and more

10 This is a theme that has been taken up by a number of commentators and defenders of early
Romanticism. See Eldridge (1989 and 1997); Larmore (1996); Seyhan (1992); Frank (1997 and
1989).
11 See Manfred Frank’s (1997) masterly account of this movement.
30 Terry Pinkard

complex forms of organization in nature that culminated in human communal

self-consciousness; and since we also had to take the reality of human
subjectivity equally seriously, he constructed an independent Kantian-Fichtean
transcendental philosophy that culminated in an account of human communal
self-consciousness that was the same as that offered by the Naturphilosophie.
For Schelling, the mutual irreducibility of the subjective to the objective and
the necessity of keeping faith with both points of view required us therefore to
understand that neither could be simply be added to the other, nor could one be
reduced to the other, nor could we simply be satisfied in holding onto with two
competing points of view (the subjective and the objective); the unity of the
two points of view (or the two tracks that philosophy had to follow) was
therefore not itself something that was up for argument or inferential
articulation. Instead, it was somehow itself the product of a way of seeing the
world that necessarily included both viewpoints within itself, indeed, was the
viewpoint from which both were to be derivative but which could not be stated
in the terms appropriate to any one of those viewpoints. In keeping with the
terminology of the time (some of which he had introduced), Schelling called
this an “intellectual intuition” of the absolute, something that could be shown
(or metaphorically seen) but not adequately said.
However, to bring this off, Schelling realized that he had to do more than
simply offer this as a piece of therapy for perplexed readers or just to leave it
at the level of advice on how to avoid both the Fichtean dead-end of
“subjective idealism” and the Kantian dualism between deterministic
appearance and the practical necessity of presupposing a kind of noumenal,
transcendental freedom. Instead, he had to construct an appropriate theory of
nature as it includes human agency within itself, and he had to construct an
appropriate theory of subjectivity that required that same conception of nature
for its own self-image to be rational. To carry that project out without reducing
subjectivity to any form of naturalism, however, Schelling came to the
conclusion that the account of disenchanted nature that Kant had constructed
(a view of nature as a machine championed by the Newtonians) instead had to
be replaced with a more or less re-enchanted account of nature in terms of the
metaphor of an organism, and he also thought such a view of nature was also
fully compatible with all the findings of the natural sciences; indeed, it was
supposed to be (in a sense that Schelling never fully worked out) an
implication of the findings of the natural sciences.
In Schelling’s initial conception of his project, part of it was to supposed to
be therapeutic in that it diagnosed a set of philosophical problems as stemming
from an inadequate picture of how things hung together in the broadest sense
of the term, substituting instead another picture in which those problems were
not so much solved as they were dissolved. As Schelling put it about the so-
 Sellars the Post-Kantian? 31

called problem of skepticism about the existence of objects apart from our
mental representations of them: “Hence this question cannot be dissolved
(aufgelöst) except in the way in which Alexander dissolved the Gordian knot,
that is, by canceling and superseding (aufzuheben) the question. Hence it is
quite simply unanswerable, because it can be answered only in such a way that
it can never again be raised” (1927, p. 175). On the other hand, Schelling was
not content to leave it that, since he also thought that we needed some kind of
general systematic account of how things hang together (in the broadest sense
of the term). That unity of the two viewpoints – grasped through the
intellectual intuition of the absolute – joined the picture of human agents as
free beings with a picture of nature that was both dynamic and (so Schelling
went to great lengths to show) consistent with modern science.
There were of course some developments Schelling could not have
foreseen, such as the development of Darwinian evolutionary theory, even
though recent reinterpretations of Schelling have put him as an essential figure
for the development of such views (see Richards 2002). What split Schelling
off, however, from later developments was his thoroughgoing antipathy to any
kind of “disenchanted” nature. This seems to have been a bad bet on
Schelling’s part; certainly the sciences since Schelling’s day have developed
more and more in the direction of vindicating the “disenchanted” view of the
world.

2. Philosophy and the Scientific Image of Man

For those who know Sellars’s own work, the very general comparison being
suggested here is perhaps obvious. In some ways, Sellars too opts for a “two
track” system of philosophy, and in his early essays, he shows himself to be
struggling with the same kinds of issues about “saying” and “showing” that
worried Schelling and his followers namely, how one is to express the unity of
the two tracks (but now transposed by Sellars into the more contemporary
notion of the unity of different “frameworks”). In the later works, such as
“Philosophy and the Scientific Image of Man,” Sellars argued that what is at
stake are not simply modern, science-inspired visions of “disenchanted” nature
in contrast to a view of agency that is indebted to Kantian conceptions of
spontaneity but something even deeper, to which he gave the names of the
“scientific” and the “manifest images.” The manifest image is that view of
ourselves that we have developed over the eons in our development of what
Husserlian phenomenologists call the “life-world.” However, the scientific
image is more than just an outgrowth from that world (or an idealization of the
life-world, as Husserl himself thought); it is instead another framework
32 Terry Pinkard

altogether, and which, although it grew out of that framework of the life-world,
detached itself from it and became a competitor to that framework. (On
Sellars’s view, science postulates imperceptible entities to explain the
inductively established correlations drawn from the life-world of the manifest
image, explaining even why those inductive correlations sometimes do not
hold.) The problem with the two images is that, as competitors, there is no
obvious way to unite them, since each claims different sorts of things to be the
basic objects of its framework – each has a different view of what really exists
in the world – which is to say that the scientific image makes claims about
what is the true account of the world, all things considered, that is
fundamentally at odds with that of the manifest image, the life-world.12 The
scientific image is not merely a different framework that has historically
emerged from out of the manifest image; it is a competitor to the manifest
image.
Sellars’s metaphor for how to deal with that competition is, as is well
known, that of stereoscopic vision (suggesting again the difficulty both he and
Schelling faced as they each drew on metaphors of vision to express how it is
that we could grasp the seemingly incompatible frameworks as one unity). As
the metaphor of the stereoscope suggests, the two frameworks had to be seen
as filling out the full image in “depth,” as opposed to the heavily detailed but
ultimately “flat” images provided by the manifest and scientific frameworks.
The problematic aspect in providing any “depth” to those two images is the
(manifest image’s) conception of a person, an agent, which as Sellars admits,
cannot be reconstructed within the framework of the scientific image; such a
reconstruction is, as he puts it, “in principle impossible, the impossibility in
question being a strictly logical one” (1963, p. 38) – namely, because the
conception of an agent is a normative conception – that is, that agency itself is
a kind of norm – and as normative, it can find no place in a rigorously
scientific account of the world.
If, however, neither Cartesian mental/physical, nor Kantian
phenomenal/noumenal dualisms are live options to reconcile the two, then as
Sellars puts it, “the conceptual framework of persons is not something that
needs to be reconciled with the scientific image, but rather something to be
joined to it,” and, Sellars admits, “we can only realize this direct incorporation

12 See Sellars (1963, p. 9), and also Sellars’s claim there: “It is not the denial of a belief within a
framework, but a challenge to the framework. It is the claim that although the framework of
perceptible objects, the manifest image of everyday life, is adequate for the everyday purposes of
life, it is ultimately inadequate and should not be accepted as an account of what there is all things
considered” (1963, p. 27).
 Sellars the Post-Kantian? 33

of the scientific image into our way of life only in imagination. And to do so is,
if only in imagination, to transcend the dualism of the manifest and scientific
images of man-in-the-world.”13 But leaving it at that sounds suspiciously close
to Fichte’s own idea of joining the “I” and the “Not-I” (the source of all
normative authority and the idea that normative authority comes from
something not itself normative, to which we are simply responding) as
constituting an “infinite task,” a kind of regulative idea of the unity of nature
and freedom – indeed, begins to look suspiciously close to Kant’s own
conception of how the practical necessity of our freedom is irremediably
coupled with its theoretical unintelligibility.
Indeed, as Sellars notes, something virtually identical to the “Kantian
paradox” is at work in this conception of agency. To be a person (an agent) is
both to hold oneself to norms and to be held to norms by other such agents; or,
in Sellars’s words, to think of something “as a person is to construe its
behavior in terms of actual or potential membership in an embracing group
each member of which thinks of itself as a member of the group.”14 Being a
person therefore involves taking a self-conscious stance toward being a person,
since it involves not merely being a member of a group but actually thinking of
oneself as such a member. This in turn leads to the Kantian paradox, since it
means that our self-conceptions as agents make us into the agents we are, yet
(barring a “given” that Sellars argues against), we could not become these
agents unless we already had some kind of self-conception already in hand.15
(In developmental terms, the Kantian paradox points out that we cannot
conceive of a group of beings who are not concept-users instituting the norms
necessary to make themselves concept-users, which is the version that Sellars
seems to invoke.)
However, the paradox within the “Kantian paradox” is itself not even easy
to see, as Sellars himself obliquely notes. Since the ability to guide one’s

13 Sellars (1963, p. 40). (In this citation, Sellars’s text speaks of “man-of-the-world,” which I

presume is a typographical error in the printed edition.)

14 Sellars (1963, p. 39) and: “Yet the essentially social character of conceptual thinking comes

clearly to mind when we recognize that there is no thinking apart from common standards of
correctness and relevance, which relate what I do think to what anyone ought to think. The
contrast between ‘I’ and ‘anyone’ is essential to rational thought” (1963, pp. 16-17).
15 See Sellars (1963, p. 6): “For it is no merely incidental feature of man that he has a conception

of himself as man-in-the-world, just as it is obvious, on reflection, that ‘if man had a radically
different conception of himself he would be a radically different kind of man’. I have given this
quasi-historical dimension of our construct pride of place, because I want to highlight from the
very beginning what might be called the paradox of man’s encounter with himself, the paradox
consisting of the fact that man could not be man until he encountered himself.”
34 Terry Pinkard

thoughts and deliberations is constituted by the ability “to measure one’s

thoughts by standards of correctness, of relevance, of evidence,” (1963, p. 6)
we seem to be driven to the rather Kantian conclusion that because we must
always presuppose the “fact of reason,” reason itself is therefore not only prior
to any act of self-legislation, it is indeed itself even constitutive of it, and thus
there is no paradox in acting or thinking in accordance with the standards that
constitute acting and thinking itself. However, in invoking the “fact of reason,”
even Kant realized that such an invocation involves a kind of circularity in that
such a “constitutive standards” view – that there are standards which are
constitutive of something like thinking such that anything that does not meet
those standards simply does not count as thinking – cannot presuppose that we
can acquire or come to be acquainted with those standards in any way that
does not already presuppose those standards itself. That is, presupposing the
“fact of reason” does not get one out of the paradox unless one also
presupposes that “reason” has some other kind of metaphysical status as
existing prior to us, as the timeless essence that establishes the standards of
rational thought prior even to the emergence of self-conscious life itself, and
that conformity to that metaphysical essence (or to the “meaning” of the terms)
is all that is required of reason.16 (That conception of our conforming our
maxims to the standards of reason is, interestingly enough, almost identical to
the Wolffian view of reason that Kant so vehemently rejected (see Pippin
forthcoming). Or, that this is just the meaning of the word, “thinking” or
“thought,” and that such meanings can be fixed in advance of any linguistic or
behavioral expression of them.
Given that difficulty of confronting the Kantian paradox, the temptation
might seem so great as to be inescapable either, first, to construe normativity
in terms of non-normative facts (especially in light of the obvious explanatory
potential of the approach of the scientific image), or, second, to posit some
normative facts and just leave it that (leaving oneself content to point out the
inconsistencies of those who try to explain away the normative in terms of the
non-normative).17 Sellars focuses his energy on trying to avoid the first

16That conception itself was argued against by Sellars as ultimately requiring a kind of direct

apprehension of meanings and rules that is itself unintelligible (or at least is impossible to square
with a scientific view of the world). See Sellars’s (1953, p. 337): “There is nothing to a conceptual
apparatus that is not determined by its rule, and there is no such thing as choosing these rules to
conform with antecedently apprehended universals and connexions, for the ‘apprehension of
universals and connexions’ is already the use of a conceptual frame, and as such presupposes the
rules in question. The role of the given is rather to be compared to the role of the environment in
the evolution of species.”
17 For the latter approach, see Thomas Nagel’s short but subtle The Last Word (2001).
 Sellars the Post-Kantian? 35

temptation while providing overall a picture that avoids having to think of the
normative as simply a “fact of reason” requiring no further explanation as to
how it fits into the scientific image, especially into that part having to do with
evolutionary biology. To that end, he focuses most of his attention (or at least
at greatest length) on classical empiricist doctrines of the “given” (leaving his
attack on rationalist doctrines of things like “intellectual intuition” for other
papers).18 Sellars’s attack on the empiricist version of the “given” is well
known, and the commentaries on it are legion. The rudiments of it involve
arguments against the view that we could have non-inferential knowledge of
anything like atomistically construed data of experience without also being
able to exercise conceptual capacities in the apprehension of those data (and
hence against the view that that apprehension cannot be used to explain our
apprehension of the concepts themselves), and arguments to the effect that the
ease of maintaining the illusion that we could have such non-inferential
knowledge is to be explained by the fact that the concepts (or conceptual
capacities) at work in the apprehension of such data (typically expressed in
“looks” expressions, such as, “This looks green to me”) are not typically there
on the surface of the apprehending activity itself; or to put it another way, one
must already know a good bit more than one is stating when one says
something like, “This looks green to me” in order to be able to make that
statement in any cognitively significant way at all. Otherwise, one would have
to start with an apprehension of something like “redness,” which itself would
have to be itself something we do not acquire by any learning.
Part of Sellars’s argument reprises an argument from Kant to the effect that
some rules cannot be acquired by virtue of other rules; otherwise, there would
be a infinite regress of rule learning. The idea that one learns the meaning of
“red” by learning something like the rule, “Say ‘red’ when confronted by red
objects” already presupposes that one has learned the rule by which one
discriminates red objects from not-red objects. There must therefore be a way
of acquiring rules (and norms) which is not itself a matter of acquiring other
rules. If the behaviorist model of rule-acquisition and learning is to be rejected,

18 Sellars does not, of course, ignore those parts in “Empiricism and the Philosophy of Mind,” but

they are not the focus of his arguments there. (In the early part of “Empiricism and the Philosophy
of Mind,” Sellars does note that “Many things have been said to be “given”: sense contents,
material objects, universals, propositions, real connections, first principles, even givenness itself.”
(1997, §1/p. 14). The early papers on “Inference and Meaning” and especially “Language, Rules,
and Behavior” contains perhaps his clearest statements of the impossibility of basing one’s grasp
of conceptual norms (which he there calls “rules”) in terms of some kind of mind’s eye
apprehension of ideal entities (such as universals).
36 Terry Pinkard

then there must be some other way of acquiring rules that is not incompatible
with their normative character. To that end, in “Empiricism and the Philosophy
of Mind,” Sellars focuses his attention on “looks” talk and what it purports to
report. If such talk (as in “It looks green to me”) does not report atomistically
construed items such as “this green here now” (which themselves supposedly
also do not require any conceptual capacities for their apprehension), on what
does it report? Sellars’s answer is a version of the realist notion that it reports
(when it does accurately report) on the presence of green objects in the
perceiver’s environment. For that account to work, the realist must be able to
give a satisfactory account of the status of inaccurate reports, where “It looks
green to me” turns out to have been made in the presence of something not
green but which seemed to be green. Sellars’s own account turns on his
argument that “looks” talk is parasitic on “is” talk, so that learning to say
things like “It looks green to me” is parasitic on learning to say things like “It
is really is green.” (Sellars thus reverses the implicit doctrine of classical
empiricism, which holds the opposite: We begin with things seeming to be
such and such, like “green,” and on the basis of that we then go to learn how to
infer the presence of things really being green.) Even Sellars seems to think,
however, that this point, difficult as it is to make, is not the hard part of his
account. The “decisive hurdle” (his term) to be taken is coming to terms with
the fact
[T]hat to be the expression of knowledge, a report must not only have authority,
this authority must in some sense be recognized by the person whose report it
is . . . not only must it be a symptom or sign of the presence of a green object in
standard conditions, but the perceiver must know that tokens of ‘This is green’
are symptoms of the presence of green objects in conditions which are standard
for visual perception. (1997, §35/pp. 74-75)
Sellars actually seemed to have two different responses to his own
self-proclaimed “hurdle.” On the one hand, Sellars claims that “not only must
the conditions be of a sort that is appropriate for determining the color of an
object by looking, the subject must know that conditions of this sort are
appropriate” (1997, §19/p. 44). That is, there must be a self-conscious stance
taken toward to the situation in which the agent finds himself; the agent has to
know, in some sense, what he is doing in making those claims. At one point in
“Empiricism and the Philosophy of Mind,” Sellars takes that to imply that for
Jones to know that the object is green “requires that Jones could now cite prior
particular facts as evidence for the idea that these utterances are reliable
indicators” of the presence of green things (1997, §37/p. 77). But this makes
the criteria for self-conscious possession of these concepts too stringent, since
 Sellars the Post-Kantian? 37

it requires the agent to be able to adduce evidence for everything that she
claims to know.
To shift the terms of exposition a bit: this way of putting Sellars’s own
views is too Fichtean. For Fichte, something could function as a piece of data
for such reports only to the extent that it was self-consciously taken up by the
perceiver as data, that is, taken up as something having a cognitive potential,
which Fichte neatly summed up as: “No activity of the self, no check” (1982,
p. 191). As Fichte also put it, “I can be conscious of any object only on the
condition that I am also conscious of myself, that is, of the conscious subject.
This proposition is incontrovertible,” by which he meant not that we are
directly aware of any internal mental states in addition to being aware of the
object but that we had to have an “intellectual intuition” of the various “posits”
carried out by the “I” – or, in other words, that we had to be able, in principle,
to cite evidence for all our claims if they were to count as claims and not just
random noises or merely well formed sentences that expressed neither
assertions nor judgments.19 If one were to take Sellars’s “Kantian” variation in
that Fichtean direction, one would end up with something very much like the
Fichtean picture: one would have the “I” (now construed in more
contemporary terms as a “inferential network” of inferences and inference
licenses) which responds to the “Not-I” by conferring some kind of authority
on it, but in which the “Not-I” itself must be construed as the non-normative,
that is, the causal order of things. In seeing Sellars that way, one would then
take him to be putting forward something like this: The person is a biological
entity who learns through socialization to make inferences (which themselves
can in principle be mapped out into something like a flow chart); this being
then learns to respond non-inferentially to the causal inputs of the world (as
effects on the sensory organs, responding to what Fichte calls the “check,” the
Anstoß) by acquiring reliably responsive dispositions to respond to, for
example, the presence of green things by making utterances such as “This is
green” (or “This seems green to me”), which in turn he or she can then
integrate into the network of inferential connections and thus transform it from
being merely a natural response into something that counts as knowledge (or,
in Fichte’s terms, the “check” becomes knowledge only when the “I” has made
that “Not-I” into a posit – i.e., something integrated into an inferential network
– carried out by the “I”).20

19 Fichte (1994, p. 112; 1971, pp. 526-527). I discuss Fichte’s views as relating to his conception

of the “I” as a normative status in chapter five of my (2002).

20 To turn the tables and put Fichte into Sellarsian terms: Fichte wanted to deny the “given” by

transforming it into something that is cognitively meaningless until taken up as a posit by the “I,”
38 Terry Pinkard

However, even in “Empiricism and the Philosophy of Mind,” and in other

essays, Sellars qualifies that claim in a manner reminiscent of the early
Romantics’ critique of Fichte’s exuberant idealism (or, what we could call
without too much anachronism, Fichte’s exuberant inferentialism). Sellars
notes that “while this does not imply that one must have concepts before one
has them, it does imply that one can have the concept of green only by having
a whole battery of concepts of which it is one element” (1997, §19/p. 44). In
some other texts, Sellars qualifies what this “battery of concepts” might be in
terms of his notion of a “lived rule” – as he puts it, “a rule, properly speaking,
is not a rule unless it lives in behavior.”21 The notion of the “battery of
concepts” and the “lived rule” is, of course, metaphorical, but it suggests that
Sellars is open to something like a more Wittgensteinian (or even
Merleau-Ponty inspired) idea of the “battery of concepts” as a kind of implicit
background practice that functions not as evidence for a claim but as a forming
a kind of pre-reflective background of various conceptual capacities that we
must already have and which therefore are in some sense already at work even
though they are not in any way explicitly present for consciousness for us to be
able to make such claims. Indeed, such pre-reflective knowledge would be of
things that we might find difficult, perhaps even impossible, to articulate.22

which he did by calling his “Not-I” the Anstoß, thus positioning it ambiguously between
something that has a cognitive standing and something that is a mere causal input, or something
offering a “resistance” to the subject’s otherwise unimpeded cognitive activity.
21 Here are two citations from Sellars on the idea of “lived rule”: “On the other hand, we must not

say that a rule is something completely other than a generalization. The mode of existence of a
rule is as a generalization written in flesh and blood, or nerve and sinew, rather than in pen and
ink.” And: “We saw that a rule, properly speaking, is not a rule unless it lives in behavior . . .
Linguistically we always operate within a framework of living rules. To talk about rules is to
move outside the talked-about rules into another framework of living rules . . . In attempting to
grasp rules as rules from without, we are trying to have our cake and eat it. To describe rules is to
describe the skeletons of rules. A rule is lived, not described” (1971, p. 299; p. 315).
22 This reading of Sellars texts forms one of the many points of disagreement between the

influential readings offered by John McDowell and Robert Brandom. Brandom’s reading is
“Fichtean” in that Brandom takes Sellars to be arguing for our direct connection to the world to
consist in “reliable dispositions,” whose reliability is not determined by the agent whose
dispositions they are but by some other agent – as Brandom works that out in his overall account
of “scorekeeping,” in which the contents of one’s commitments are “administered” by someone
other than the agent whose commitments they are; the reliability of the response is thus not a
matter of “knowledge” on the part of the agent making the response but is only a reliable
disposition to utter phrases like, “This is green” when in the presence of green things; on
Brandom’s account, the agent may thus be entitled to make the claim, “This is green,” even
though she does not know that she is entitled; the entitlement is bestowed by somebody else (the
“scorekeeper”). (In this way, Brandom, in following out a version of the dialectic that was
 Sellars the Post-Kantian? 39

Articulating Sellars’s position along those (early Romantic) lines puts him in
the position of being able to claim what he quite clearly intends to claim: We
can give up the “Myth of the Given” without giving up the idea that we are in
possession of non-inferential knowledge because such non-inferential
knowledge nonetheless always presupposes a background of non-articulated
conceptual capacities that have to be at work even if not, perhaps ever, being
made explicit or even being able to be made explicit (and one thus avoids the
regress of making the knowledge too self-conscious so that one would need a
rule to apply the rule to apply the rule, etc., without having to reduce that
background to a set of mere non-cognitive dispositions).
However, by giving the life-world (or the “manifest image”) no priority –
in effect, by cutting off an entire movement in phenomenology that relies on
something like the priority of “originary human experience” to undercut the
scientistic (not the scientific) misunderstanding of human agency – and by
insisting on the truth of the scientific image in its competition with the
manifest image, Sellars might seem to be putting himself firmly on the side of
the “scientistic” school of thought and denying any significance to human

originally at work in post-Kantian thought, in the way he makes a move very similar to that made
by Fichte, who, as he worked out his practical philosophy, came to see that the real “check” had to
come from another agent, not from non-agent objects in the world. For Brandom, that goes on, as
it did for Fichte, among individuals, with Brandom using the much more determinate device of
“scorekeeping” to work the idea out, whereas Fichte simply kept it at a fairly general level of
“checking” each other; but in that way, Brandom also has to come to terms with Hegel’s criticism
of Fichte: “First of all, by virtue of that the ‘I’ keeps the significance of being an individual actual
self-consciousness over and against the universal, the absolute, that is, against spirit, in which it
itself is only a moment; for individual self-consciousness is just this, that it remains fixed and
standing on one side vis-à-vis an other. When the “I” consequently was called the absolute
essence, there was an enormous scandal, because “I” in fact only makes its appearance in the
sense of being an individual self-consciousness, that is, the subject, opposed to the universal.
Second, Fichte [for that reason] never arrives at the idea of reason” (1971, vol. 20, p. 408).
McDowell makes the Wittgensteinian point against Brandom’s suggestion that we thereby just
eschew what Sellars’s called his “decisive” hurdle: “Sellars is driven to his implausible picture
because he assumes that for a reporter to count as recognizing her own authority (which he only
says is required “in some sense”), she must be able to adduce evidence of her reliability –
premises for an argument in which the claim that she is reliable figures as conclusion. But that is
not the only possible sense for the idea of recognizing one’s own observational authority. Along
with most adults, I know that, at least in a good light, I can tell a green thing when I see one. That
is, I am cognizant of my authority as a reporter of greenness. But I would be at a loss if someone
pressed me for evidence of it. It stands firm for me in a quite different way, about which I would
suggest we can find the beginning of wisdom in Wittgenstein’s On Certainty. There is no need to
decline Sellars’s second hurdle. We simply need to adopt a different posture when we jump it”
(McDowell forthcoming-b, pp. 13-14).
40 Terry Pinkard

subjectivity (in effect, dissolving it, seeing it as a kind of illusion, a distorted

version of the truly objective). That is certainly not Sellars’s intention. It is
another story altogether, but part of the project of “Empiricism and the
Philosophy of Mind” is to introduce a notion of “inner episodes” (that is,
thoughts) which, although modeled on “outer behavior” (in other words, on
speaking and communicating in language), can nonetheless become the objects
of a kind of non-inferential knowledge. That is, we can non-inferentially know
that we are thinking such and such without having to make an inference based
on any outward behavior and we acquire that capacity to immediately know
our thoughts by learning to use the language of inner episodes; in that way,
Sellars opens the door to preserving a relatively rich sense of subjectivity when
it is “joined” to the scientific image without having also to buy into any kind
of dualistic or reductionist picture of subjectivity. Moreover, in holding that
“humanity” is (in a Kantian and post-sense) an “Idea” that humanity gives to
itself, Sellars is arguing against a view that would see the subjective and the
objective as divided by any kind of bright metaphysical line (as both the
Cartesian and the classical empiricist pictures would have it). The difference
itself is a normative difference, something constituted within the space of
reasons, not a bedrock metaphysical fact about human agents and the world.
But, as we noted, that is another story altogether, important as it is to Sellars’s
overall scheme of thought.

3. Experience

Nonetheless, by developing this conception of the status of “inner episodes,”

Sellars puts himself in a position to endorse the (B-Edition) Kantian
conception of the unity of concepts and intuitions by interpreting it to mean
that such a unity of experience is the same kind of unity as that expressed in
overt linguistic judgments and assertions, and this puts Sellars in the position
of having a recognizably richer, post-Kantian conception of experience itself.
In “Empiricism and the Philosophy of Mind” and in Science and Metaphysics
(with its metaphorically musical subtitle: Variations on Kantian Themes),
Sellars takes himself to have adequately constructed a conception of
experience that forsakes the classical empiricist, atomistic conception in favor
of a richer conception of experience as containing claims within it(mostly
epistemic, but leaving the account open for other types of claims). That is,
experience is not first of independently cognized “data” about which we then
proceed to make claims; experience itself, for it to be human experience,
already contains claims within it, even though those claims are not usually
self-consciously present or self-consciously entertained. Instead, the
 Sellars the Post-Kantian? 41

experience of, say, seeing a bird outside the window, is possible as a seeing-a-
bird, etc., only by virtue of a pre-reflective background of conceptual
capacities that have to have already been acquired and be at work in order for
that type of (non-inferential) experience to be possible in the first place.
As Sellars notes, that only ups the ante for justifying the authority of that
background “battery” of concepts. Sellars’s own commitment, however, to the
idea of a “framework” as the proper way to think of concepts and their relation
to each other (and of human behavior that evinces or employs concepts) leads
him to some conclusions that run against the grain of the rest of his thought.
For Sellars, there are two such “frameworks,” as we have seen, and the truth of
one (the “scientific image”) leads us inexorably to the conclusion that the
experiences we have under the older and original framework (that of the
“manifest image”) are in some deep sense false, even illusory. However, the
superiority of one framework over another is something Sellars finds it
difficult to justify. In his more mature writings, one finds him committed to
something like the explanatory superiority of the scientific image, a superiority
that is supposed to be evident even to those operating completely within the
manifest image. It does not take the intellectual elegance of a Foucault or a
Rorty, though, to show that that way of justifying the superiority of one
“framework” over another inevitably begs all the questions it should be
answering if all our conceptual capacities really are relative to a framework.23
It is therefore not surprising that even Sellars finds himself more or less
resorting to something like the claim that there are simply some “constitutive
standards” for thought (or perhaps, if one wants to stay Kantian, to claim that
these constitutive standards are the “transcendental presuppositions” of
thought and agency); that brings two commitments with it: First, anything that
does not meet those standards is simply ruled out of court as counting as
thought, and second, these constitutive standards must be themselves rich
enough to give us some very concrete guidance in choosing, for example,
among the manifest and scientific images, or among competing modes of
conduct.24 The “constitutive standards” view does rule the Kantian paradox as

23 Neo-Sellarsians like Richard Rorty draw the valid conclusion from those premises that we can

only allow ourselves to say that we find one framework more attractive since though we cannot
give any ultimate reasons to those operating within another framework to opt for our framework
which in their own terms would count as reasons for them (even if those reasons are, from within
our framework, decisive reasons).
24 This appeal to “constitutive standards” is also one of the ways of avoiding the Kantian paradox,

since if it is just “constitutive” of thought to obey certain determinate norms (rich enough to guide
us in selecting or constructing other determinate norms), there is no paradox to be dealt with. This
line is also taken by McDowell himself, who claims to reconcile the spontaneity and autonomy of
42 Terry Pinkard

being out of bounds (in the manner of something like a judicial ruling on the
basis of some kind of reading of what is implied by the cosmic constitution),
but it does not fit into Sellars’s own view of what a “joining” of the manifest
and scientific images is supposed to be. One suspects, in fact, that Sellars
himself kept the doubts he expressed in his 1949 essay, “Language, Rules and
Behavior,” when he said:
How is the adoption of a set of rules to be justified? . . . The more I brood on
rules, the more I think that Wittgenstein was right in finding an ineffable in the
linguistic situation, something which can be shared but not communicated . . .
To learn new rules is to change one’s mind. Is there a rational way of losing
one’s reason? . . . The convert can describe what he was. Can he understand
what he was? But here we are on Wittgenstein’s ladder, and it is time to throw it
away. (1949, pp. 314-315)
In even toying with any notion of a “constitutive standard,” Sellars is not
only pushing against the direction of his own self-professed pragmatism and
his conception of the historical synthetic a priori, he is also avoiding coming to
terms with the basic post-Kantian direction he gave to his very own thought.
There are two ways in which that comes to the fore. First, there is Sellars’s
own invocation of the social nature of thought; and second, there is Sellars’s
own (at least implicit) working out of a specifically Kantian insight of the “B”
edition of the first Critique about the unity of concepts and intuitions (where,
as we noted earlier, Kant says that “the same function which gives unity to the
various representations in a judgment also gives unity to the mere synthesis of
various representations in an intuition; and this unity, in its most general
expression, we entitle the pure concept of the understanding” (1964,
A79/B104)). Such a unity of concepts and intuitions itself undermines the
notion that anything like a “framework” can work in the way that Sellars
seems (or at least sometimes seems) to think it does, particularly when it is set
against his conception of experience as “containing claims” within itself (in
the sense that human experience involves the possession of conceptual
capacities that are at work there, although not necessarily consciously in play,

norm-setting thought with the more common sense notion (itself surely correct) that if we do not
submit ourselves to norms not of our own individual choosing, we are not submitting to norms at
all: “The sense in which the source of the norms is in us is just that the norms are constitutive of
the practice of thinking, and the practice of thinking is not optional for us” (McDowell
forthcoming-a, p. 16). The term ‘constitutive standard’ is also put to similar use by Korsgaard in
her “Locke Lectures,” made available in a preliminary format at her website:
http://www.people.fas.harvard.edu/~korsgaar/#Locke%20Lectures
 Sellars the Post-Kantian? 43

and which function as a kind of pre-reflective background for our

deliberations).
“Experience” in the Anglo-American philosophical literature has in fact
been so overlaid with atomistic, empiricist interpretation that it is perhaps
difficult to see how close Sellars’s own conception of “experience as
containing claims” actually is to the later post-Kantian conceptions of
experience that one finds, particularly in Hegel. If indeed we cannot, on
Sellarsian grounds, separate experience into “given” data which are then
organized according to conceptual form (or in terms of “frameworks”), then
we must approach experience more holistically; and if there is no real appeal
to “constitutive standards,” then the authority of those conceptual capacities
must be dealt with in some way that acknowledges that they can change but
which preserves the idea that there is such a things as critical, rational revision
of those norms and which does not lapse into the fantasy that we can, out of
nowhere, freely will to construct those standards.
Sellars’s own hunch about this – “I should like to be able to say that one
justifies the adoption of rules pragmatically, and, indeed, this would be at least
a first approximation to the truth” (1949, p. 314) – certainly points to the idea
that what is at stake is something more along the lines of Wittgenstein’s
conception of a “form of life.” What is at stake, that is, is not a series of
different ways of conceptually organizing the same “given” data but of “unities
of concepts and intuitions,” which, to make the last post-Kantian leap, would
be to conceive of those unities as embedded in practices, that is, in whole ways
of living. The conceptual capacities that we acquire are indeed social in a deep
sense, not merely in the sense that we acquire them by being initiated into
forms of life (which just about everybody acknowledges) but also in the sense
that their normative authority has to do with their being socially instituted and
sustained over time. The “Kantian paradox” is more paradoxical (even
unintelligible) when it is construed as a problem for an individual agent who
would be making himself into a thinking being (governed by norms) by an act
of thinking already governed by norms. The more intelligible picture is the one
that is also the more obvious and commonsensical: We never begin by
legislating norms for ourselves; we instead are born into a form of life, we are
initiated into various practices (among which, obviously, is intelligible
language use), and we thus always, already find ourselves bound to norms not
of our own choosing.
The experience we have of the world is thus deeply social in nature, and
that experience can, and does, change in the course of human history. To say
that our experience of the world changes (and not to make the mistake of
concluding, as subjective idealists must, that the world thus changes, at least
not in anything other than a metaphorical sense) is to say something more
44 Terry Pinkard

along the lines that very basic concepts that function as authoritative for us
themselves change, and that therefore some very basic orientations for our
lives change. Indeed, part of the alleged superiority of what Sellars calls the
“scientific image” has to do with how it has built on the failures of past
attempts (from Aristotelian physics to medieval alchemy), that is, how it is the
result of (to use the common-sense expression) of a kind of communal
“learning from experience” about what has worked and what has not. And this
is the sense of “experience” that figures in more widespread, although
common phrases, such as a “literary experience” and a “historical experience.”
Such a view captures, so we might put it, the “true motion” of Sellars’s
thought (in distinction from the “apparent motion,” if one only sees it in the
terms of post-war analytical philosophy), even if it might not be the motion in
which Sellars thought he was moving; moreover, that “motion” also expresses
the direction in which the internal dialectic of post-Kantian philosophy found
itself, which then found its expression in the Meditations Hegeliennes carried
out by Hegel himself. After all, Hegel characterized his 1807 Phenomenology
as the “science of the experience of consciousness,” where (as he put it a few
years later in his Logic), one dispenses with the “abstract form of apriority
over and against aposteriority,” looking instead at the unities in question
(namely, those of intuition and concept) “in terms of their own particular
content,” and a “phenomenology” (in physics) distinguishes true from merely
apparent motion (Hegel 1971, vol. 5, p. 62).
There are three points to make about this conception of experience. First,
although one might want to call it an “enriched” conception, the classical
empiricist conception is so impoverished that it should really count more as the
“common-sense” conception. Second, it is both tied to the kinds of beings we
are as organisms that have evolved on the planet (although Hegel himself
denied the validity of evolution, a point to which I shall return, although very
briefly) and to the kinds of social beings that we are by virtue of both our
sociality and our history. Third, it puts us in a position to understand the
Kantian paradox of our being subject only to those norms of which we can
regard ourselves as the authors as being better rephrased in terms of Hegel’s
dialectical conception of thought. To take up the third point first: When Hegel
says things like, “the concept . . . is that alone which has actuality
(Wirklichkeit), specifically in the way that it gives itself actuality,” he is giving
expression to the way he understands the “Kantian paradox” to be transformed
 Sellars the Post-Kantian? 45

into his “dialectic.”25 “Actuality” is for Hegel one of his technical terms that he
takes over from Aristotle, and he gives it the Aristotelian sense of indicating
what is at work in reality; in particular, Hegel is concerned is what concepts
are at work in the way in which humanity conceives of itself and what
concepts are at work in the kinds of experience that we have. (This conception
of being “at work” is bound up with the German, wirklich.26) Just as Kant held
that we find ourselves always, already committed to the normative priority of
reason in undertaking any deliberations at all, Hegel too holds that what
“actuality” reason has for us is to be found in the way that the practices of
giving and asking for reason explain our actions, such that in holding ourselves
and others to norms, we make them real, or, socially viewed, those norms tend
to make themselves true in that the practices tend to reinforce the way in which
people on the whole hold to these norms and to shape certain types of practical
conclusions into being drawn. What Hegel in his more colorful (and in many
ways more accurate) language calls “a shape of spirit” is structured by the
kinds of authoritative norms that hold it together – by a sense of what counts
absolutely for it, what makes certain things worth sacrificing, enduring,
promoting, and holding onto (sometimes at all cost). These norms cannot
become “actual,” cannot come to be at work, until they become
generalizations, matters of disposition, to be matters therefore of conformity
and regularity – to be, as Hegel puts it, “second nature” – which therefore sets
up the unavoidable tension between facticity and normativity that characterizes
our mindedness and like-mindedness.
In Hegel’s words, “the concept’s” determinations and laws are those it
“gives itself, not those it already has within it and which it finds,” which is to
say that such concepts only have their full reality, their actuality – that is, can
only be genuinely at work – in the practices of giving and asking for reasons,
and that the model for how concepts come to be at work in experience
(including those that are passively drawn into our intuitive apprehension of the
world in what Husserl, straining for the right terminology, called “passive
synthesis”) is best captured in a developmental, historical model and not in any

25 See Hegel (1971, vol. 7, §1). (“ . . . daß der Begriff . . . allein es ist, was Wirklichkeit hat und

zwar so, daß er sich diese selbst gibt.”) (Italics in the body of the text above added by me.)
26 For example, in his Logic, Hegel says that “what is actual is what is efficacious” (“was wirklich

ist, kann wirken”) (1971, vol. 6, p. 208). A.V. Miller translates the relevant phrase as “what is
actual can act” (1989, p. 546).
46 Terry Pinkard

kind of appeal to a “constitutive standard” (not what, as Hegel says, it merely

“finds” within itself).27
That sets the stage for the second point: We must be able to make
intelligible to ourselves how we, as the natural organic beings we are, could
develop our capacities for conceptual reasoning, and, as Schelling saw, that
itself requires a rethinking of nature. It would obviously require much more
room than we could give it here, but the Hegelian outline of such a conception
would run, very roughly, along the following lines. Contrary to what Schelling
thought, nature in the modern, scientific image of it, must be conceived as
fundamentally meaningless, de-divinized, devoid of spirit (entgöttert and
geistlos, as Hegel calls it28 – in Max Weber’s well known term, nature in the
modern world must be seen as thoroughly disenchanted; and Hegel
provocatively insists that this modern disenchantment is precisely the problem
with nature, since as properly explained by the means of modern natural
science, nature displays itself to modern reflection as simply incapable on its
own of organizing itself into anything like “better exemplifications” of itself,
(something which Hegel ironically calls the “impotence (Ohnmacht) of
nature”29) and hence cannot provide any normative standard for practice. It is
only when organic life appears in nature that it first makes sense to speak of
any particular things in nature as “better” and “worse” since only organisms
display the kind of self-directing, internally functional teleological structure
that can make the application of such terms meaningful. As Hegel stresses, this
becomes apparent in our use of the term, “disease”; it makes sense to speak of
an organism as diseased when it is in some kind of state where it prevented
from achieving goals proper to the internal teleology of that organism; but it
makes no sense to speak of a diseased planetary system, a diseased structure of
crystals, a diseased mountain range, and so forth. Moreover, disease is not
something we project onto the world as kind of subjective addition to nature
devoid of spirit; a diseased animal or plant really is diseased, really is hindered
in its ability to be the organic creature it is and to achieve what is proper to its
well-functioning; it is not just an animal or plant that we see as diseased.30 (In

27 The discussion of “passive synthesis” occurs throughout Husserl’s later writings, but it is to be

found in its clearest form in Husserl (1985).

28 The reference to “de-divinized” is from Hegel (1971, Vol. 13, p. 27). The reference to

“geistlos” is from Hegel (1971, Vol. 14, p. 137).

29 There are a variety of references to the Ohnmacht der Natur: See (1971, §250; §24, Zusatz 2)
30 In my (2002), I did not state this point with nearly enough accuracy. The language used there

suggests that the attribution of disease is solely based on Geist’s interests and thus also suggests
that disease and defects in organisms are only something we “read into” the world. This is
misleading: That the concept of disease is relative to Geist’s interests is not the same as saying
 Sellars the Post-Kantian? 47

other words, although only humans have the concept of disease, there are
diseased things in the world even though they do not themselves have the
concept of disease.)
This is not to say that organisms (or the concept of disease) introduce
anything non-natural into nature; the same type of causality remains at work
there as elsewhere. It is only to note how we can get a handle, in a way that
Deweyian pragmatists have always appreciated, as to how human reason
develops out of organic nature. Because organisms, especially the so-called
higher organisms, have this kind of internal teleology to them, they can be said
to have things that are good or bad for them. Thus, organisms develop means
of registering those items in their environment that are good and bad for them
as tools for survival and flourishing; that there are these goods for various
organisms, moreover, opens the space for the further claim that within the
terms of the modern, fully scientific image of nature, since organisms have
goods, they may also have reasons to act even if those organisms themselves
cannot be aware of those reasons as reasons (or, to put it differently, they may
register those reasons but not register them in any way as reasons). To put it in
Hegelian terms: There can be goods, and therefore reasons, in nature even
though reason itself cannot be said to be at work in nature (since neither nature
as a whole, nor, for the most part, nature in any of its parts is guided by any
kind of concept of itself); or, to put it another way, organisms can have
reasons, although it is only we as reasoning beings who can note that about
them. (Indeed, we might say that it is only a Davidsonian prejudice that links
reasons with beliefs and thus prevents us from seeing this distinction.)
It is therefore only human beings (although this will surely be disputed by
some) who have developed ways not only of registering what counts as a
reason but have developed the capacity to reason about those reasons (who
have developed what Sellars calls “rules of criticism”). Even if some higher
animals (perhaps dolphins, some apes, and maybe even some elephants) can be
said to make inferences, they do not, as far as we know, make inferences about
their inferences, since that requires the presence not just of a sophisticated
system of communication but also of a fully fledged language with its capacity
to speak about itself on various meta-levels, which itself requires (besides the
obvious neurological basis) a social group of a particular sort.

that when we form that kind of concept, we are merely “reading” it into neutral matter, which on
its own is not diseased. The diseased lion is a diseased lion independently of our descriptions; but
nature, as geistlos, cannot form this concept of itself and cannot orient itself in terms of better or
worse, cannot have a concept of itself.
48 Terry Pinkard

Second, although humans as organisms obviously have reasons, as do all

organisms (since there are things that are naturally good in this sense for
them), what is distinctive about humans is that, as reasoning beings, they are,
to use Charles Taylor’s term, “self-interpreting animals” such that even the
natural goods of organisms become transformed and reordered in light of other
goods that are non-natural (although not for all that in any sense supernatural).
Their lives are not structured completely around the good functioning of the
organism but in terms of, as we might put it, their concept of themselves, or, in
Sellars’s words, “the fact that man couldn’t be man until he encountered
himself,” (1963, p. 6) and the proper account of how it is that we come to
grasp ourselves and revise our self-conceptions requires something like the
very complex Hegelian phenomenology of our own mindedness, that is, a
philosophical history of our collective self-conceptions structured in terms of
how the very determinate failures of earlier self-conceptions not only led to
their successors but explain why their successors (taken as “shapes of spirit” or
forms of life) took on the determinate shape they did (in a process Hegel
colorfully calls “determinate negation”) and in turn how our modern
self-conception (including our conception of our own place within the
scientific image) is only to be understood not as our having finally drawn the
proper conclusions from the premises of the “constitutive standards” of
thought but in the way they have been required by those past failures to take
on the shape they have.
The proper metaphor, therefore, as Sellars saw, had to do not with how to
replace that account of the historical and social nature of our self-conceptions
but how to join that view of agency to the scientific image. That image of
“joining,” however, itself pulls the carpet out from under Sellars’s own refusal
to see how his own reading of Kant’s B edition version of the unity of
concepts and intuitions itself undermined his view that the objects of
perception in the manifest image were unreal, the illusory effects of
imperceptible things in themselves which were only adequately grasped in the
scientific image; Sellars seemed to think (as did Kant in an analogous way)
that only such a distinction preserved the idea of independently existing
objects exercising any kind of normative authority for us. The latter view,
though, is possible only from the standpoint of what Hegel called “reflection,”
or what (following John McDowell’s nice metaphor) we might call the
“sideways-on” view. The “sideways-on” view steps outside of the practices of
giving and asking for reasons and looks at the complex of agent-and-world as
if it consisted of a subject on one side, an object on the other, and some kind of
metaphysical line separating them – exactly the picture, one would have
thought, that Sellars’s richer notion of experience and his critique of the Myth
of the Given would have prevented him from taking. The distinction of
 Sellars the Post-Kantian? 49

“subject” and “object” is already a distinction made from within the space of
reasons – from within pretty much what Hegel called the “Idea” (the Idee, to
distinguish it from a Vorstellung, an “idea”-as-a-representation, something
which naturally suggests the “sideways-on” view). What Hegel called the
“opposition of consciousness” (the “sideways-on” view, which he thought
Kant stubbornly adhered to while undermining it at the same time) is to be
superseded and integrated into a larger story in which the existential
independence of objects of nature is not undermined, but their normative
authority comes to be situated in a much larger developmental, historical, and
social story of how the space of reasons came to have the determinate shape it
now does.
Much, much more needs to be said about this, but in general we can get the
outlines of how a post-Kantian conception would try to make out how there
can be reasons in nature without having to deny the reality of the scientific
account of nature; that conception (again, in outline) gives us a way to
understand and bring down to earth our own non-natural self-conceptions and
to join the images without having to opt for the idea that only one of them tells
us what there really is. The details of that story would require not just a few
more pages but several other books, but in Hegel’s telling of that story (in his
“phenomenology of consciousness”) such self-conceptions develop in light of
a self-instituted liberation from nature on the part of humanity not in any
metaphysical but in a totally normative sense such that those conceptions came
less and less to depend on natural facts (such as procreative facts about men
and women) for their normative force. Indeed, what it means to be a human
being in general, and in particular what it means to be a father, daughter,
friend, business associate, citizen, artist, professor, natural scientist and so on
involve telling a story of how certain options vanished for us because of our
own collective self-legislation in human time, that is, human history, in which
“man encountered himself,” and, in so encountering himself, began to initiate
that self-instituted normative liberation from nature.

4. Post-Kantianism?

This might make it sound as if all Sellars needs is some Hegelian correction so
that his “Kantian variations” would once and for all merge with his
Meditations Hegeliennes. That, of course, would be to deny the obvious.
Hegel’s own Naturphilosophie was, after all, written in the first half of the
nineteenth century, and, as the cliché rightly has it, there has been a lot of
water under the bridge since that time. Even worse: In his Naturphilosophie,
Hegel notoriously denied that explanation of the structure of organisms in
50 Terry Pinkard

terms of evolution was a satisfactory explanation at all. Hegel’s reasons for

this had to do with his acceptance (as did so many Germans of his generation)
of a certain model of comparative anatomy favored by Georges Cuvier, the
basic tenets of which held that, first, each species of organism was so perfectly
adapted to its environment and was therefore a self-contained whole, and,
second, because of that, any slight change in its makeup would make it
impossible to for that species to continue living, and thus, third, the kind of
gradual change that evolution requires has to be ruled out.31 (Hegel also
thought that the leading model of the competing evolutionary theory given by
Lamarck was anchored in too much external teleology to make it believable.)
Hegel extended that model into his philosophy of history (and thus also into
his phenomenology of the development our mindedness), in that he came to
see each “form of life” as a self-contained “shape of spirit” whose appearance
on the stage of history vanishes when its reason for being (its internal self-
conception of what it is and what counts absolutely for it) vanishes (when its
own internal constitution breaks down under the weight of its self-contained
contradictions and existential tensions). Needless to say, Sellars’s conception
of nature is not burdened by that kind of conception, and his ethics need not be
burdened by it. Nonetheless, Sellars only had a very general, generally
progressivist, “social achievement” story to tell (with an overlay of a general
Kantian approach to ethics), with no determinate account of how one might
rationally comprehend the move from one ethical community to the another
except in terms of changes in empirical belief.
One of the things that prevented a revival of Hegel at the same time as the
Neo-Kantian revival was getting into full swing was the failure of the
Hegelians to come up with a philosophy of science (whereas the Kantians had
little trouble with that). If we were, however, to rephrase Hegel’s own views in
a self-consciously anachronistic way, he would be taken to have given his
Naturphilosophie pride of place over any kind of more narrowly conceived
modern “philosophy of science”; Hegel would not comprehend the authority of
natural science as resting in a special methodology reserved to it or resting in a
structure of theories that is independent of what it is studying; instead, he
would see the authority of the scientific approach – including the virtues
involved in being a scientific researcher – in the very disenchantment of
nature that is itself a condition of the modern normative conception of
subjectivity, that is, of the modern “encounter of man with himself”). The pure
“externality” of nature (in Hegel’s jargon) makes it a fit object for the

31 I have tried to tell that story in some more detail in Pinkard (2002).
 Sellars the Post-Kantian? 51

mathematical techniques of natural science, just as the relative “externality”

(still his jargon) of modern economic life makes it also the fit object of
mathematical modeling.
Amidst the conflicts circulating within the Hegelian school after Hegel’s
death, David Friedrich Strauss made the joke that, just as in the French
Revolution (which Hegel championed for his whole life), the Hegelian school
had dissolved into the quarrel between the “left” and the “right” Hegelians.
There is a similar joke about Sellars’s school (usually attributed to Richard
Rorty but the origins of which are apparently disputed) splitting into quarrels
between the “left” and “right” Sellarsians. More recent historians of the
Hegelian school have pointed out that the “left/right” distinction covers over a
more fine-grained set of distinctions: There were in fact old-left, old-right,
new-left, and new-right Hegelian schools. If we wanted to keep the jest going,
we might likewise hold that there is an old-right Sellarsianism (with its
emphasis on the scientific realism and naturalism in Sellars’s views), and there
is a Sellarsian old-left, sharing a version of its own of the naturalism it has in
common with the old-right. But perhaps there is also a new-left Sellarsianism
that does not eschew Sellarsian naturalism but tames it. (Perhaps John
McDowell is the forefather, maybe even the Eduard Bernstein, of that new
left.) Perhaps the new-left Sellarsians can make the case that Sellars himself
was slyly really a new-left Sellarsian all along; and maybe there is an even
newer-left formation on the Sellarsian front that would run his Haydnesqe
variations on Kantian themes all the way up to Hegelian symphonies.
However that turns out, to make sense of Sellars, we need to bring in Hegel
as a interpretive background; and, of course, to make sense of Hegel, we need
to understand more deeply Sellars’s own contribution to the post-Kantian
debate that began in the late eighteenth century, an earthquake whose
aftershocks we are still experiencing.

Terry Pinkard
Department of Philosophy
Georgetown University
Washington, DC 20057
United States of America
e-mail: [email protected]

REFERENCES

Burke, T. (1994). Dewey’s New Logic: A Reply to Russell. Chicago: University of Chicago Press.
52 Terry Pinkard

Eldridge, R. (1989). On Moral Personhood: Philosophy, Literature, Criticism, and

Self-Understanding. Chicago: University of Chicago Press.
Eldridge, R. (1997). Leading a Human Life: Wittgenstein, Intentionality, and Romanticism.
Chicago: University of Chicago Press.
Fichte, J.G. (1971). Werke. Edited by I.H. Fichte. Berlin: Walter de Gruyter and Co.
Fichte, J.G. (1982). The Science of Knowledge. Edited and translated by P. Heath and J. Lachs.
Cambridge: Cambridge University Press.
Fichte, J.G. (1994). An Attempt at a New Presentation of the Wissenschaftslehre. In: J.G. Fichte
(ed.), Introductions to the Wissenschaftslehre and Other Writings. Indianapolis: Hackett
Publishing Company.
Frank, M. (1997). Unendliche Annäherung. Frankfurt a. M.: Suhrkamp.
Frank, M. (1989). Einführung in die frühromantische Ästhetik. Frankfurt a.M.: Suhrkamp.
Friedman, M. (2000). A Parting of the Ways: Carnap, Cassirer, and Heidegger. Chicago: Open
Court.
Hegel, G. (1971). Werke in zwanzig Bänden. Edited by E. Moldenhauer and K. Markus Michel.
Frankfurt a.M.: Suhrkamp Verlag, 1971. (Abbreviated as Werke and volume number).
Hegel, G. (2004) Encyclopedia of the Philosophical Sciences. Oxford: Oxford University Press.
Hegel, G. (1989). Science of Logic. Translated by A.V. Miller. New York: Prometheus Books.
Horstmann, R. P. (1991). Die Grenzen der Vernunft. Frankfurt a.M.: Anton Hain Verlag.
Husserl, E. (1985). Erfahrung und Urteil. Hamburg: Felix Meiner.
Kant, I. (1956). Critique of Practical Reason. Translated by L. White Beck. Indianapolis: The
Bobbs-Merrill Company.
Kant, I. (1964). Critique of Pure Reason. Translated by N.K. Smith. London: Macmillan and Co.
Kant, I. (1987). Critique of Judgment. Translated by W. Pluhar. Indianapolis: Hackett Publishing
Company.
Kant, I. (1998). Groundwork of the Metaphysics of Morals. Cambridge: Cambridge University
Press.
Larmore, C. (1996). The Romantic Legacy. New York: Columbia University Press.
McDowell, J. (forthcoming-a). Autonomous Subjectivity and External Constraint. Manuscript.
McDowell, J. (forthcoming-b). Transcendental Empiricism. Manuscript.
Nagel, T. (2001). The Last Word. Oxford: Oxford University Press.
Pinkard, T. (2002). German Philosophy 1760-1860: The Legacy of Idealism. Cambridge:
Cambridge University Press.
Pippin, R. (forthcoming). On Giving Oneself the Law. Manuscript.
Richards, R.J. (2002). The Romantic Conception of Life: Science and Philosophy in the Age of
Goethe. Chicago: University of Chicago Press.
Schelling, F. (1927). Philosophische Briefe über Dogmatismus und Kritizismus. In: M Schröter
(ed.), Schellings Werke, pp. 47-112. Munich: C. H. Beck und Oldenburg.
Seyhan, A. (1992). Representation and Its Discontents: The Critical Legacy of German
Romanticism. Berkeley, CA: University of California Press.
Sellars, W. (1949). Language, Rules, and Behavior. In: S. Hook (ed.), John Dewey: Philosopher
of Science and Freedom, pp. 289-315. New York: The Dial Press.
Sellars, W. (1953). Inference and Meaning. Mind 62, 313-338.
Sellars, W. (1963). Philosophy and the Scientific Image of Man. In: Science Perception and
Reality, pp.1-40. Atascasdero, CA: Ridgeview Publishing Company.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Willem deVries

FOLK PSYCHOLOGY, THEORIES, AND THE SELLARSIAN ROOTS

ABSTRACT. Almost fifty years ago, Wilfrid Sellars first proposed that psychological
concepts are like theoretical concepts. Since then, several different research programs
have been based on this conjecture. This essay examines what his original claim really
amounted to and what it was supposed to accomplish, and then uses that understanding
of the original project to investigate the extent to which the later research projects
expand on it or depart from it.

1. Introduction: Folk Psychology as Theory

We use mentalistic or psychological terms constantly in describing,

explaining, understanding, and generally coping with the behavior of others
and ourselves. For several hundred years in the modern era (until, say, 50 years
ago), the bad old explanation of how we do that and what exactly is going on
when we do seemed, well, not unproblematic, but at least uncontroversial in its
fundamental outlines: We enjoy direct and privileged access to our own mental
states. This enables us, first, to abstract a set of psychological concepts from
actual and self-intimating instances of psychological states. Furthermore, since
the psychological concepts we have are abstracted directly from self-
intimating instances, first-person application of those concepts is
unproblematic. It is incorrigible and provides each of us with certainty about
his or her own psychological states. This knowledge is primordial. On its basis
we build up knowledge of everything external to our minds. From the
materials we have direct access to we (somehow) construct and apply concepts
of physical objects and events. More problematically, once we have the core
set of concepts and some knowledge of our own psychological states, we can
begin to apply psychological concepts to describe, explain, understand,
interpret, and generally cope with others as well. This requires the assumption
that the observable behavior of others is tied to their internal psychological
states in the same ways and in the same kinds of patterns that we find connect
our own psychological states and behavior. The argument for that assumption
is a pretty weak induction by analogy, and this was a stumbling block for the

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 53-84.
Amsterdam/NewYork: Rodopi, 2006.
54 Willem deVries

picture as a whole, but the rest of the picture seemed so solid, so obvious, that
most preferred trying to find some way to solve the associated “problem of
other minds” rather than rethinking the whole picture.
About fifty years ago some brave souls started rethinking the whole
picture.1 Principal among the attempts to revise the whole picture explicitly
was Wilfrid Sellars’s trail-blazing effort in “Empiricism and the Philosophy of
Mind.” Sellars proposes that our psychological concepts and terms are like
theoretical concepts and terms. In Sellars’s picture our most primordial
knowledge is not our first-person acquaintance with our own mental states, but
a knowledge of “medium-sized dry goods,” the public, physical objects and
events that form the arena within which we conduct our lives. Sellars argues
that we should reverse the hitherto standard picture that our knowledge grows
“from the inside out”: if we begin with a knowledge of the public objects and
events that shape our lives and at least a rudimentary set of metalinguistic,
particularly semantic, concepts, we would have all the material necessary to
developing the psychological concepts that Cartesians think are simply given
elements of our mental lives.
Sellars’s proposal, radical at the time, struck a chord and has since become
enough of a commonplace that many no longer know its provenance. Sellars’s
original idea has served as inspiration for subsequent developments that have
picked up the notion that psychological concepts are like theoretical concepts.
If psychological concepts are like theoretical concepts, then our use of those
concepts, even in our everyday attributions of psychological states to ourselves
and others, is like using or applying a theory. Commonsense psychology is
itself, then, really like a theory.
While Sellars’s proposal has influenced almost everything in subsequent
philosophy of mind, there are two different projects that have drawn particular
inspiration from this aspect of Sellars’s proposal that I want to focus on.2 The
first, whose principal spokesmen have been Richard Rorty (1965) and Paul
Churchland (1979), both of whom worked with Sellars, is eliminative
materialism. They point out that a salient property of theories is susceptibility
to replacement by a better theory. If commonsense psychology is a theory,
then it, too, is susceptible to replacement by a more adequate theory. Given its

1 Hegel had already rethought it from the ground up, but left little mark on English-speaking
philosophy, the domain I am most concerned with here. And Wittgenstein started rethinking it in
the 1930s, but nothing appeared in print until the 1950s, though there were many reports of his
thought circulating.
2 Other projects, such as the functionalist analysis of intentional states, have drawn inspiration

from other aspects of Sellars’s proposal.

 Folk Psychology, Theories, and the Sellarsian Roots 55

hoary past and extended run, they argue that commonsense psychology is not
only susceptible to replacement, we should expect it to be replaced. It is not
only a theory, it is a bad theory. It took several thousand years for folk physics
to be superceded, even after Aristotle codified it well enough to enable some
more or less clear-cut tests. Folk psychology has endured through the first few
centuries of serious efforts at scientific psychology, but as neuroscience
advances, it, too will eventually succumb. We will find ourselves in the
seemingly paradoxical position of denying that the psychological concepts in
terms of which we (currently) understand our own behavior really designate
any robust features of the world. In our current lingo: we’ll end up believing
that there are no beliefs.
The second major project that takes off from Sellars’s proposal is not
concerned with the large-scale picture of the conceptual evolution of
psychological concepts in human society, but the small-scale picture of the
individual’s development and application of psychological concepts. If
psychological concepts are like theoretical concepts, then one should
investigate whether the kinds of processes by which psychological concepts
are acquired and applied by the individual are like the kinds of processes by
which theoretical concepts are acquired and applied. Indeed, in (probably
unbeknownst) tribute to Sellars’s suggestion, the investigation of the human
ability to attribute psychological states to oneself and others and utilize such
attributions to understand, explain, and sometimes predict human behavior is
often called inquiry into (or the theory of) Theory of Mind.3 A number of
psychologists are vigorously pursuing this project, e.g. Janet Astington, Alison
Gopnik, Josef Perner, Henry Wellman, Alan Leslie, and Paul Harris. Since
‘theory of mind’ is now a general name for a complex field of investigation,
however, it covers a wide range of specific proposals that take the phrase and
attendant analogy more or less seriously. Gopnick and Wellman (1992) take
the phrase ‘theory of mind’ most literally; she argues that the “theory of mind”
that children acquire and the process by which children acquire and learn to
use it is very much like a scientific theory and the corresponding process of

3 Consider, for example, the titles of representative contributions to the field:

“What Is Theoretical About the Child’s Theory of Mind? A Vygotskyan View of Its
Development” (Astington 1996); “The Theory of Mind Deficit in Autism: How Specific Is It?”
(Baron-Cohen 1991); “Does the Autistic Child Have a Theory of Mind?” (Baron-Cohen, et al,
1985).; Theories of Theories of Mind. (Carruthers and Smith 1996); Introduction to Theory of
Mind: Children, Autism, and Apes. (Mitchell 1997); “Theory of Mind Is Contagious: You Catch It
from Your Sibs.” (Perner, et al, 1994); “Insights into Theory of Mind from Deafness and Autism”
(Peterson and Siegal 2000); “Does the Chimpanzee Have a Theory of Mind?” (Premack and
Woodruff 1978); The Child’s Theory of Mind (Wellman 1990).
56 Willem deVries

theory acquisition and revision in the sciences. Other contributors to the field
discount the analogy to varying degrees, claiming, e.g. that our “theory of
mind” is a highly implicit set of representational capacities. Some claim that
our “theory of mind” is contained in a set of informationally encapsulated,
innate modules that need only appropriate triggers at the proper point in one’s
developmental history to blossom. By this point, ‘theory of mind’ functions
arguably only as a label and has lost any descriptive functionality. In this paper
we will focus on those who take the “theory of mind” metaphor quite
seriously.
Both of these Sellars-inspired projects have drawn substantial critical
response. Eliminative materialism has been vilified as well as criticized, and
the literature examining the claims of eliminative materialism is very
extensive. Sellars has himself been branded an eliminativist, though I will
argue that this is unjustified guilt by association. Against those who take
‘theory of mind’ to be a description and not a mere label, a group of
“simulationists,” e.g. Robert Gordon, Alvin Goldman, and Jane Heal, have
argued that we cannot appropriately understand our own use of psychological
concepts as anything like an instance of theoretical practice. Though there may
be some abstract analogy between theoretical concepts and psychological
concepts, the mechanisms by which we apply psychological concepts are not
at all like the kinds of mechanisms theorists employ in their work.
I am now finally in a position to state my intent in this paper. I want to go
back to the original source in Sellars of the idea that psychological concepts
are like theoretical concepts, see what his original claim really amounted to
and what it was supposed to accomplish, and then use that understanding of
the original project to investigate the extent to which the later research projects
expand on it or depart from it. There is some purely historical interest in my
effort here, but I think we will also find that it helps us get a better
understanding of the terrain in which all these programs operate. Eliminative
materialism is a broad, general challenge to a pervasive and fundamental
aspect of our conceptual framework; the debate about the theoreticity of our
“theory of mind” is a fairly narrow and ultimately empirical debate in
psychology. Understanding the connections and disconnections among them is
important to understanding the current state of the philosophy of mind.

2. Sellars’s Original Proposal

Sellars’s stated goal in “Empiricism and the Philosophy of Mind” (1997) is to

attack the “Myth of the Given” (§1/p. 14). Sellars believes this myth is not a
purely epistemological fiction, however. It needs and finds support in a set of
 Folk Psychology, Theories, and the Sellarsian Roots 57

metaphysical assumptions that must also be overthrown if the attack is to

succeed. In one sense, Sellars’s attack on the Myth of the Given is complete
halfway through the paper, in Part VIII, where he pulls together the threads of
his critique of traditional notions of the given and sketches his own
nonfoundationalist theory of observational knowledge as an alternative. Why
does he not stop there?
Through the halfway mark in “Empiricism and the Philosophy of Mind,”
Sellars has worked hard to displace first-personal knowledge of one’s mental
states from the foundation (or the core, if you prefer that metaphor) of our
knowledge. Having achieved that, however, he must then develop an
alternative, non-Cartesian explanation of the privacy and authority of
first-person knowledge as well as of the nature and structure of our knowledge
of others’ minds. For Sellars does not want to deny that our self-ascriptions of
mental states usually have a special, first-person authority, nor that we have in
some respectable sense privileged access to them. And it is crucial to his
overall view that our intersubjective knowledge of others’ mental states can be
as firm as our purely subjective knowledge of our own states.
So the situation at this point in the essay is that Sellars believes he has
given an adequate, non-Cartesian reconstruction of the human capacity for
observational knowledge of public, physical objects and events, a
reconstruction that notably does not itself require that humans have prior
concepts or knowledge of their own (or others’) mental states as inner, private,
or privileged. He then needs to show that he can account for concepts and
knowledge of one’s own (and others’) mental states in ways that explain
privacy, first-person authority, and intersubjectivity without falling back into
Cartesian conceptions of givenness.
Sellars’s strategy is fairly direct, given the situation. He tells a little myth
himself about some people who exemplify the dialectical situation reached at
this point in the essay: the “Ryleans” have rudimentary observational
knowledge. Indeed, they exhibit fairly sophisticated abilities to describe and
explain the behavior of public physical objects, utilizing the subjunctive
conditional and other sophisticated syntactic forms. But they have no
conception (and therefore no knowledge) of their own mental states as inner.
Could these Ryleans suitably acquire concepts and knowledge of mental states
(their own and others’) as inner, private states to which they have privileged
access?
Sellars’s answer is yes, if they also have (1) semantical concepts they can
and do apply to their language, and (2) the ability to theorize, i.e., to formulate
the conception of a new domain, not otherwise perceptible, modeled on one’s
conception of an already familiar and more or less well comprehended domain.
Sellars’s mythical character Jones provides the theory, formulating a new
58 Willem deVries

conception of an inner domain consisting of two kinds of internal states. One

kind of state (thought) is attributed properties that are modeled on the
semantical properties of speech acts. Such states have meanings, entailments,
truth, etc. The other kind of state (sense impression) is attributed properties
that are modeled on the perceptible properties of physical objects. Such states
can be blue or red, sharp or flat, bitter or sweet.
Jones develops his theory (or is it theories?) in order to explain certain
kinds of behavior that appear anomalous in the Rylean framework.4 Once he
has developed the theory and taught others how to use it, though, “It is but a
short step to the use of this language in self-description . . . [I]t now turns out –
need it have? – that Dick can be trained to give reasonably reliable
self-descriptions, using the language of the theory, without having to observe
his overt behavior” (1997, §59/p. 106). Thus are born the notions of privileged
access and privacy.
If one accepts the possibility (not the actuality) of Sellars’s myth, then the
Cartesian story is not the only one in town. We have available an
understanding of our concepts of the mental that is both methodologically and
ontologically compatible with empirically and naturalistically respectable
science and that does not require any given. Sellars’s attack on the given has
been brought to completion: he has not only shown us the problems with the
given, he has shown us how to do without it.

3. Eliminativism

This is the basis for attributing to Sellars the view that folk psychology is a
theory, and it is a strong basis for that attribution. But I want now to argue that
things are not as straightforward as they might seem, for we also have to be
careful to ensure that “folk psychology is a theory” is construed the same way
by all the parties to this conversation.

3.1. Theories and Models

There is one thing to be clear on from the outset, something that some
commentators on Sellars have not recognized: whereas Jones’s theory is about
our behavior and mental states, Sellars’s theory is about the nature and status

4For more detailed discussion of what Jones’s theories allow him to explain, see deVries and
Triplett (2000), pp. 142-144 and 159-164.
 Folk Psychology, Theories, and the Sellarsian Roots 59

of our concepts of mental states; Jones was the first psychologist, Sellars is a
philosopher.
Sellars does offer us a theory, and it is part of his overall view that theories
are often constructed using models. A model is a domain of which one
presumes to have an understanding, and which one proposes to be in some way
analogous to the domain one is currently trying to understand. Interestingly, in
his metaconceptual inquiry, Sellars uses as his model for the domain of our
mentalistic concepts (our psychological language), which we are trying to
understand, theoretical concepts (theoretical language), which he presumes 20
or 30 years of philosophy of science has helped us get a good grasp of.
Now, it is also part of Sellars’s understanding of the use of models in
theory formation that (1) the analogy is always limited, and (2) the terms of a
theory can acquire “surplus value” from their use in their new context, which
entails that they become independent of the model and can proceed to develop
in their own way in confrontation with experience. To say that the analogy is
limited is to say that the analogy is always accompanied, more or less
explicitly, by a commentary that specifies which aspects of the model domain
are denied to the domain theorized and which aspects of the model domain are
attributed to the domain theorized. These conditions entail that it is a mistake
to identify the domain of the theory with the domain of the model unless there
is independent reason to do so. The presumption should be against such an
identification.
This already affords us a quick and easy argument – though not a deeply
convincing one – that it is a mistake to say that Sellars thinks of folk
psychology as a theory.5 He employs theoretical concepts and theoretical
language as his model in order to understand psychological concepts and
language, but we should assume that, as in all theoretical models, the analogy
is limited and that psychological concepts have some “surplus value” above
and beyond the analogous theoretical concepts that disrupts the analogy still
further.

5 The reason this is not a deeply convincing argument is that the theory Sellars is proposing here is
not a scientific theory that postulates a new domain of theoretical entities – the kind of theory to
which his speculations on the nature of theories and models directly applies. Philosophical or
interpretive theories have to work in a different way, and the role of a model in a philosophical
theory will correspondingly be different. But I see no reason why the fundamental principle
defended in the argument – that it is a mistake to simply identify the domain of the explanandum
with the model in a theory – would itself cease to apply to philosophical theories and the models
used in them.
60 Willem deVries

This argument invites us then to investigate more deeply the analogies and
disanalogies between theoretical concepts and theoretical language sensu
stricto and folk psychological concepts and language. If there are systematic
and strong disanalogies, then we should indeed abandon the claim that Sellars
believed that folk psychology is itself a theory.

3.2. Theory and Observation

The first item in the “commentary” on the theory-model that Sellars is using to
help understand folk psychology I want to investigate is the theory-observation
contrast. In Sellars’s treatment this is a very important theme. It is in the
context of a discussion of the theory-observation contrast that Sellars most
definitively rejects identifying mentalistic concepts as theoretical concepts:
. . . I am going to argue that the distinction between theoretical and
observational discourse is involved in the logic of concepts pertaining to inner
episodes. I say ‘involved in’ for it would be paradoxical and, indeed, incorrect,
to say that these concepts are theoretical concepts. (§51/p. 97)
This is a complex topic, but I will try to be brief. Sellars does not hold that all
theories postulate unobservables, but he does hold that this is true of
(scientific) theories in their “most developed or sophisticated form.”
(§51/p. 94). There is a continuum of theoretical activity from very low-level
empirical generalizations entirely in observation vocabulary to highly
mathematized theories that postulate complex domains of unobservable
entities. It is important for Sellars that “the ways in which the scientist seeks to
explain empirical phenomena are refinements of the ways in which plain men,
however crudely and schematically, have attempted to understand their
environment and their fellow men since the dawn of intelligence” (§51/p. 97).
For instance, commonsense often postulates an unobserved, though
observable, entity or event to help explain some occurrence. How big a step is
it to postulate a kind of systematically unobserved event to help explain other
occurrences? How large is the gap between the unobserved and the
unobservable? Is there not a difference in kind between the unobservable and
the observable, whether in fact observed or not?
It is important to see that for Sellars, any such difference is not ontological.
The line between the observable and the unobservable is not fixed by
ontology, but by our current methods, and it is correspondingly movable as our
 Folk Psychology, Theories, and the Sellarsian Roots 61

methods change (and hopefully improve).6 For Sellars, “theoretical entity” and
“observable entity” denote methodological statuses not ontological genera.
Given the theory of observational knowledge that Sellars sketched in Part VIII
of “Empiricism and the Philosophy of Mind,” anything that one can make a
noninferential report of, knowing that that report is a reliable indicator of the
matter reported, is thereby observable. Vocabulary that gets its start in a theory
(e.g. “There is an alpha-particle track”) can come to be used in noninferential
reports. If such usage becomes widespread, that vocabulary thereby enters the
observation language.
We can now see why, in the context of a discussion of the
theory-observation distinction, Sellars says “it would be paradoxical and,
indeed, incorrect, to say that these concepts are theoretical concepts”
(§51/p. 97). To call a term or concept theoretical is to say something
methodological about it; it characterizes the conditions of its use. Among other
things, it characterizes it in contrast to observation terms as not having a use in
noninferential reports.7 But when a concept that perhaps originated in some
theory becomes regularly used in direct and immediate noninferential reports,
it has lost its status as theoretical. And this is precisely the situation in which
we find mentalistic vocabulary in Sellars’s myth. Jones’s disciples learn to
apply mentalistic language to themselves directly, immediately, and
noninferentially. “What began as a language with a purely theoretical use has
gained a reporting role” (§59/p. 107). Sellars’s phrasing seems to hold out the
possibility that some terms can retain a theoretical use while adding to it a
reporting role. But in gaining a reporting role, mentalistic language also loses a
principal characteristic of the theoretical.

6 Many have been tempted to believe that, though the line between the observable and the
unobservable is mutable, there are determinate and significant limits in principle to how far it can
be moved. I am not sure anyone has clearly maintained that there are entities or features of the
world that are necessarily unobservable (and can therefore be known only theoretically).
(Classical early modern theorists would not have thought so, because there is nothing in the world
that can be known only inferentially, since God’s knowledge is not inferential.) But the idea that
there are certain features of the world that are necessarily observable – in particular, the proper
sensibles – has been a leitmotif in modern epistemology. It is, however, tied to the Myth of the
Given, the idea that there are certain features of the world that are such that when they present
themselves to us (or when we encounter them) we thereby know them.
7 Notice that having a use in sentences of the theory is not sufficient to make a term a theoretical

term. Presumably, there are some points at which the theory makes contact with observation, and
therefore there will be sentences in which both theoretical and observation terms occur. This does
not make the observation terms theoretical. We have to conclude that calling something a
theoretical term involves denying to it an observation or reporting role.
62 Willem deVries

Thus, when Sellars says that “the distinction between theoretical and
observational discourse is involved in the logic of concepts pertaining to inner
episodes,” we cannot take this to mean that concepts pertaining to inner
episodes are to be straightforwardly construed as theoretical and therefore as
opposed to observational. At most, he is pointing again to the idea that the
theory-observation contrast offers us a useful model for illuminating the inner
episode-outer behavior relationship.

3.3. Theory and Self-Reports

As I noted above, Sellars seems committed to the idea that anything we can
immediately and knowingly report is observable, but we do not normally
consider our self-reports as expressions of observations. (Imagine the fit Ryle
would throw, and properly so, at the idea that I observe my mental states!) In
normal usage ‘observation’ seems tied to the use of our five senses, and our
self-reports about our mental states do not employ the five senses as
intermediaries (though sensory states may be the object of such reports).8
Undoubtedly, this is one reason Sellars employs the more neutral vocabulary
of “reports.” Sellars’s story does help us understand the constant temptation
(which some philosophers have succumbed to) to assimilate our mentalistic
self-reports to sensory observation reports: the justificational structures are
very similar, even though the causal mechanisms involved differ in important
respects.
Sellars does not bring out clearly enough that there are two different kinds
of cases in which the originally theoretical vocabulary proposed by Jones
could gain a reporting role. It could be the case that Jones’s disciples learn to
apply his new mentalistic vocabulary to others directly and noninferentially. In
this case, they would learn to report (in Sellars’s special sense) the mental
states of others. To the Cartesian, of course, this is a blasphemous idea: we
have and can have no immediate access to the mental states of others. And it is
certainly true that very often we must figure out, laboriously, what other
people are thinking or feeling. But it has been pointed out that, at least
phenomenologically, we sometimes see another’s anger or happiness. There
does not seem to be a Sellarsian reason against saying that, in the proper
conditions, I can be both a reliable and a knowing meter of a[n] (admittedly

8 Proprioception is a difficult case here. It is not counted among the basic 5 senses, yet does not
seem reducible to their output. Self-reports of some mental states, e.g. (some?) sensations such as
having a pain in one’s foot, seem like proprioceptive observations. In any case, self-reports of
intentional states are not proprioceptive nor are they comfortably treated as observations.
 Folk Psychology, Theories, and the Sellarsian Roots 63

limited) range of the mental states of others.9 Perhaps the fact that there are a
great many caveats on my ability to report immediately and noninferentially
on the mental states of others (e.g. I may be able to do so only for a very
limited range of people – my family and good friends or maybe the members
of my cultural community – and/or for only a very limited range of mental
states – principally stronger emotions – and under fairly stringent conditions)
discouraged Sellars from mentioning this possibility. But it is worth taking
seriously, for it is an indication of how far Sellars’s position moves us from the
traditional Problem of Other Minds, and it emphasizes the important point that
in Sellars’s theory, psychological vocabulary is univocal across first- and
third-person application.
Sellars does focus on the other case in which the Jonesian vocabulary
becomes a reporting or observation vocabulary: “And it now turns out – need
it have? – that Dick can be trained to give reasonably reliable self-descriptions,
using the vocabulary of the theory, without having to observe his overt
behavior” (§59/p. 106). The parenthetical ‘need it have?’ is important, for it
raises the question of the necessity of privileged access.10 In one important
sense, that question is already settled by Sellars’s myth. Some within the
Cartesian tradition think that privileged accessibility is a defining
characteristic of the mental. There are, as has been noted, varieties of
privileged access, however. The most hard-core Cartesians might then believe
that it is a defining characteristic of a mental state that it is self-intimating:
mental states are such that if one is in mental state F, then, necessarily, one
knows one is in F. If Sellars’s Rylean myth is at all coherent, however, then it
is not a conceptual truth that mental states are self-intimating. But weaker
versions of privileged access are also threatened by Sellars’s myth. Since the

9 As Jay Garfield points out, (personal communication) there is good reason to believe that our
expression of and abilities to perceive emotions and other mental states evolved as mechanisms to
facilitate social interaction. It is very natural to treat emotion detection mechanisms as perceptual,
because they are fast, mandatory, domain specific etc.
10 But Sellars’s ‘need it have?’ is also frustrating, for it raises important questions without even

hinting at an answer. Could there be a linguistic community in which there is a robust

psychological vocabulary in use, but in which first-person reports using that vocabulary have no
special status? If I read him correctly, Davidson argues that this is not really a coherent possibility:
objective knowledge is possible only when all three legs of the triangle self-others-world are, in
some sense, immediately available to the knower. (See the essays in Donald Davidson 2001.)
Sellars seems to be contemplating the possibility of a linguistic community with a psychological
vocabulary but without first-person authority. But he certainly does not commit himself to such a
possibility, nor does he explore the extent to which citizens of such a community would differ
from us.
64 Willem deVries

Ryleans cannot be thought of as operating with exactly our current concepts of

the mental, one lesson of the myth is that there could be respectable concepts
of mentality that do not include privileged access as an indispensable feature.
This is important in deciding what to make of animals, for instance.

3.4. The Malleability of Observation

Churchland claims that the malleability of the theory/observation distinction
plays in his favor, because it shows that we could learn to report (more
adequately) in the vocabulary of his projected ideal neuroscience the presence
of those states we now normally report on (less adequately) in psychological
vocabulary. There are two problems with this view, however. First, it proves
too much for Churchland, and second, it does not fully appreciate the
complexity of the relationship between the observational and the theoretical.
It proves too much for Churchland because it ultimately allows us no good
reason to stop at the level of neuroscience in our replacement of psychological
language. If we can be trained to report our inner states and episodes in
neuroscientific terms and thereby do a better job of describing and explaining,
why not dispense with the neuroscience and report those states and episodes
directly in the vocabulary of physics and do the best possible job of describing
and explaining? Churchland is unimpressed by the pragmatic difficulties of
using the complex language of neuroscience in our self-reports, so he could
not use similar pragmatic difficulties to block the call for discarding
neuroscience in favor of flat-out fundamental physics.
More importantly, this view oversimplifies the relationship between
observation language and theoretical language. Churchland seems to think of
our ordinary observation language as consisting of a variety of modules, each
dealing with some subject area, and each liable to replacement by some
scientific theory that would improve it. Slowly, bit by bit, then, we will
upgrade our language until we find ourselves at home in the world as it really
is, as described and explained by science. But this is not the way Sellars sees
things, even though it does retain traces of Sellars’s influence. The problem is
that Churchland’s view as I have sketched it does not make provision for any
overarching structure that would organize the various subject-matter modules
in the language, what Sellars would call “framework principles.” The issue
runs deep for Sellars, for it concerns the nature of the clash between the
“manifest” and the “scientific” images of the world – another piece of Sellars’s
philosophical apparatus that has broken off and acquired a life of its own in
philosophy.
 Folk Psychology, Theories, and the Sellarsian Roots 65

3.5. Categorial Structure and the Manifest and Scientific Images

In Sellars’s view, though science may begin as a sophistication of
commonsense methods of explanation, and may thus at first propose revisions
and replacements of pieces of our manifest image of the world, it ultimately
comes to pose a challenge to that manifest image as a whole. This means that
the fundamental structures by which things (in the broadest possible sense of
the term) are identified and individuated in the manifest image are challenged
in the developing scientific image, and the entire ontology of the manifest
image is revealed as ill grounded (ultimately) compared to the ontology
developed in the sciences. The sciences perform an ontological coup, but not
simply by replacing talk of pains with, e.g. talk of brain states. The sciences
will reorganize our conceptions of objects, events, properties, processes,
particles, space, time, etc., the very categories that organize the world for us.
The entire world of the manifest image will have been revealed as phenomenal
in a more-or-less Kantian sense: a representation of reality that is well-founded
in relation to the native abilities we (as evolved and socially developed
creatures) bring to the world, but that cannot claim to be an adequate
representation of the world as it is in itself. Sellars believes that the framework
under development in the sciences does have a reasonable claim to be an
adequate representation of the world as it is in itself. (In this, Sellars departs
radically from Kant, however many other Kantian inspirations he retains.) This
is the source of Sellars’s Scientia Mensura: “in the dimension of describing
and explaining the world, science is the measure of all things, of what is that it
is, and of what is not that it is not” (§41/p. 83).
Sellars’s scientific realism, summed up in the Scientia Mensura, seems to
support the eliminativists. The sciences, presumably including neuroscience,
will tell us what exists. The categories of the manifest image will be replaced.
Does that not mean that mental states as we now understand them will be
revealed as “merely phenomenal,” and thus not real? But, of course, things are
more complex than that.
First, in order to infer the truth of eliminativism from Sellars’s scientific
realism, one needs the premise that folk psychology is not only false, but
radically false. And it turns out this does not follow automatically from the
claim that the manifest image is merely phenomenal. Eliminativism would not
follow from Sellars’s position if the following is possible: Though the
manifest-image concepts of mental states, both intentional states and sensory
states, are false in the sense that they are part of an overall framework that
cannot withstand the rigors of repeated, exhaustive, empirical test, there is
nonetheless in the successor framework that replaces the manifest image a
distinctive set of concepts that retains enough of the general structure of folk
66 Willem deVries

psychology that we cannot say that folk psychology has been so falsified that it
has been abandoned. Rather, we have to say that folk psychology is the
predecessor of scientific psychology, a scientific discipline that retains some
independence from the neuroscientific investigation of the embodying
substratum of psychology. Churchland needs to claim that folk psychology is
so empirically inadequate that there will be no successor science that will
retain even its most general structural features, and Churchland’s arguments
that that will prove to be the case have been widely criticized.
Here’s a revealing parallel: The concept of a brain is clearly present in the
manifest image framework. If the phenomenality of the manifest image
established the falsehood or inapplicability of its component concepts, then we
would have to conclude that there could be no brain science in the projected
scientific millennium, since brains do not really exist. Churchland’s beloved
baby, neuroscience, would be thrown out with the bath water of folk
psychology. The inference from the supercession of the manifest image to the
elimination of the psychological proves too much, for it would do away with
brain science (and physics) as well. The further premise that there is no
scientific-image successor to psychology is also necessary, and that could be
established only by the failure to develop any such successor, which
Churchland is not in a position to ascertain.
Second, Sellars does not think that we are destined to end up with the
scientific image, pure and simple. He argues instead that, though science is
king “in the dimension of describing and explaining the world,” (§41/p. 83)
describing and explaining the world are not the only activities of interest to us.
We are not only cognizers of the world, we are agents in the world, and Sellars
does not believe that the concepts essential to our being agents in the world are
going to be reconstructed within the scientific image. Therefore, the scientific
image will not (and cannot) totally displace the manifest image; we must
marry the agent-constituting concepts of the manifest image to the scientific
image in order to obtain a “synoptic vision” of humanity’s place in the world.
The concepts essential to us as agents include precisely the concepts of folk
psychology: intentions, desires, beliefs, etc. Psychological concepts are clearly
ineliminable, according to Sellars.
This enables us also to see another dimension in which it would be wrong
by Sellars’s lights to treat folk psychology as a theory in Churchland’s sense.
Psychological concepts do not, in Sellars’s view, constitute a self-contained
module within the manifest image that could be unplugged and replaced with
something else. They rather infect the whole manifest image, influencing its
structure and “framework principles.” To the extent that this is the case, to the
extent that psychology is present not only in our vocabulary, but in the
syntactic structures available to us, in the very structure of the language
 Folk Psychology, Theories, and the Sellarsian Roots 67

through which our experience of the world is filtered, we cannot treat

psychology as merely one theory among others, to be adopted or discarded as
dictated by the success or failure of a certain research program. Psychological
concepts are provided for by the very structure of the language/framework by
which the world is constituted for us.11
With these considerations we have emerged beyond the theory/observation
distinction, in the following sense. Is ‘physical object’ an observation concept?
In one sense, of course it is. We can directly, immediately, and
noninferentially apply the concept “physical object” to items of experience.
But that does not entail that every physical object is observable, nor that the
concept “unobservable physical object” is paradoxical or contradictory.
Thinking of “physical object” as one concept alongside many others, in fact,
gets things wrong. “Physical object” is better understood as a category, a
fundamental classification that designates a most basic type in the framework
or language. Calling something a physical object determines what kinds of
things can (and must) be said about it, what kinds of properties it can be said to
have, what forms of explanation can be applied to it, etc. It determines the
most generic forms of language that apply to it. The concept of a mental event
or state is no less categorial for Sellars, determining the fundamental
linguistic/conceptual forms that apply to such things. Some mental states may
well be observable or reportable, some may well not be. The notion of a
mental state is so deeply woven into our language that it is not obvious how it
might be related to the notion of a physical object. Sellars’s Rylean myth gave
us a deep and fruitful insight into that relationship. But it does not have the
consequence that we should expect the progress of science to eliminate the
concept of the psychological from our repertoire.
There is, thus, no sense of ‘theory’ in which Sellars would have conceded
that folk psychology is an eliminable theory, not even as a consequence of his
claim that the manifest image is ultimately to be superceded by the scientific in
matters ontological. The eliminativists, as far as Sellars is concerned, took his
original insight down a garden path.

11 The biggest impediment to taking Sellars’s Rylean myth seriously, my experience indicates, is

the immense difficulty of trying to imagine a language or point of view on the world that is really
bereft of the concepts and consequent structures of the psychological.
68 Willem deVries

4. “Theory of Mind,” Psychological Concepts, Theoretical Concepts, and

Empirical Hypotheses

We can now turn to consider whether those who take Sellars’s suggestion at
face value and make it the inspiration for a research program in developmental
psychology are just unpacking elements already prefigured in Sellars, or
creatively extending Sellars’s original suggestion, or have somehow
misconstrued the meaning of Sellars’s insight. As noted, ‘theory of mind’
covers a broad range of proposals explaining the development and structure of
our abilities as folk psychologists. I will focus here on those who take most
seriously the idea that our “theory of mind” is really theory-like, such people
as Alison Gopnick. I will call this position the “theory-theory.”

4.1. The Prima Facie Argument against the “Theory-Theory”

We have already seen that it is a mistake, from Sellars’s point of view, to
identify without further ado the domain about which Sellars is theorizing
(psychological concepts) and the domain of the model Sellars uses in his
theoretical activity (theoretical concepts). Just as it did with the eliminativists,
this affords a prima facie argument against a Gopnick-style “theory-theory”
approach to the theory of mind.
But the situation is, in fact, different in the theory-theory case from the
eliminativism case. The eliminativists assume the theory-likeness of
psychology and infer the presence in folk psychology of a particular property
of theories strictly so-called, namely their eliminability in favor of some better
theory. It is an argument from analogy and suffers the weaknesses of such
arguments. That is not the overall form of the theory-theorists’ project. Rather,
they take the original Sellarsian suggestion, not as an assumption, but as a
hypothesis, and investigate the extent to which it can inform an empirically
adequate theory, not of our psychological concepts, but of our actual
psychological competence, our ability to describe, explain, understand, and
generally cope with the behavior of others and ourselves. The theory-theorists
are not assuming that the analogy is good and arguing from there, they are
empirically investigating how far the first-order form of the analogy can be
pushed.
One way to see the theory-theorists, then, is as neo-Jonesians, who, now
puzzled by our psychological competence in particular, follow their Master
(now fading into mythological history) by explaining those behavioral abilities
by postulating inner episodes that are like a specific kind of linguistic activity,
namely the activity of employing a theory to inform or control one’s behavior.
This is a further specification and extension of Jones’s strategy, it seems, and
 Folk Psychology, Theories, and the Sellarsian Roots 69

thus in keeping with the overall thrust of Sellars’s story, especially if it makes
sense out the empirical data. The theory-theorists want to see whether Sellars’s
analogy affords not just a philosophical (meta-)theory about the nature of
psychological concepts, but also a psychological theory about a certain set of
psychological abilities.
Thus, the theory-theory seems at least consistent with Sellars’s original
suggestion. But we can also ask whether it might be what Sellars originally
had in mind. In order to do this, we need a more detailed examination both of
Sellars’s text and of the characteristics of theories. Did Sellars believe that the
analogy between psychological and theoretical concepts would provide a good
empirical explanation of our folk psychological skills? I will start this next
phase of the investigation by stating what seem to be the most important
relevant characteristics of theories, then examine Sellars’s text for clues to
what he thought might be the empirical import of his analogy.

4.2. Theories and the Attractions of the Theory-Theory

What are the characteristics of scientific theories that “theories of mind” might
share? The details differ slightly depending on which theory-theorist one looks
at, but the core relevant properties of theories are the following. Theories are
defeasible structures of (generally) propositional knowledge used to provide
explanations and (falsifiable) predictions (even in new or counterfactual cases)
by postulating a set of unobservable states or entities that are related in certain
specifiable, lawlike ways, which relations implicitly define the unobservables
in question.12 To the extent that a conception of the human “theory of mind”
deviates from this picture of the “theory” involved, it is not trying to exploit
Sellars’s analogy explicitly, and is not a “theory-theory” as I am using the
term.
Botterill (1996) points out four different kinds of reasons for finding the
theory-theory attractive:
There are epistemological attractions, associated with the special epistemic
status of theories – whether one goes on to add “It is a theory, so it may be
radically wrong” á la Churchland, or “It is a theory that works very well, so
probably it is broadly correct” á la Fodor. There are semantic attractions,
associated with the idea that theories provide implicit definitions: on a
functionalist view this promises to solve the problem of how we understand the
meaning of mental state terms. There are developmental attractions, associated

12 This characterization picks up themes found in, for instance, Gopnik and Wellman (1992) and

Botterill (1996). Whether this is an adequate characterization of theories is not a topic for today.
70 Willem deVries

with the way in which theories are discovered and elaborated, which may be
supported either by identification of specific proto-theoretical stages, or merely
by the claim that a theoretical structure would facilitate an otherwise formidable
learning task. Finally, there are cognitive processing attractions, associated
with the application of theory, which hold out the hope of answering the “How
do we do it?” question by employing a frequently successful strategy – i.e.,
positing a body of tacit theoretical knowledge. (Botterill 1996, p. 106)
How does Sellars’s treatment of the psychological stand with respect to these;
which attractions are the ones that drew Sellars? First, notice that these
attractions fall into two groups. The first and second, the epistemological and
semantic attractions, are philosophical in the sense that construing the
psychological as theory-like offers interesting answers to (non-empirical)
questions usually posed by philosophers, questions about the nature of
knowledge and understanding or meaning. The other two, the developmental
and the cognitive processing attractions are empirical. In these cases the
psychological as the theory-like promises interesting answers to essentially
empirical questions about how we acquire and apply our psychological
competence. We expect some kind of causal theory to respond to both
developmental and cognitive processing questions, but there is no presumption
that the answers to epistemological or semantic questions would be contained
in a causal theory.
It is clear that the first, epistemological, attraction, was very significant for
Sellars. Remember the context of Sellars’s Rylean myth: a critique of the
given. The significance of this for Sellars, its importance in overcoming a
faulty Cartesian picture of our relation to the world, has already been discussed
above. I want to reserve discussion of the second, semantic attraction, for later,
for it turns out also to be very significant for Sellars. So let us turn now to the
empirical attractions of taking Sellars’s analogy seriously. And let us start by
returning to Sellars’s text to see whether it contains clues about whether
Sellars thought his approach would bear empirical as well as philosophical
fruit.
Before we move on, however, a remark on the relations among these four
attractions. Though they divide into philosophical and empirical pairs, it would
be a major advantage to have a unified treatment of the psychological that
responds effectively in all four of these dimensions. Notably, a strictly
Cartesian answer in the first two, epistemological and semantic dimensions, an
answer which would treat the mind as a separable realm from the bodily,
makes it difficult to understand how to even begin answering questions about
development or cognitive processing. We should expect that Sellars’s
anti-Cartesian epistemology and semantics of the mental would put some
 Folk Psychology, Theories, and the Sellarsian Roots 71

constraints on reasonable answers to the developmental and cognitive

processing questions.

4.3. Learning Psychology: The Reception of Jones’s Theory

Looking for further detail on how other humans come to appropriate and apply
Jones’s theory, perhaps unsurprisingly, we find the text frustratingly brief:
[O]nce our fictitious ancestor, Jones, has developed the theory that overt verbal
behavior is the expression of thoughts, and taught his compatriots to make use
of the theory in interpreting each other’s behavior, it is but a short step to the
use of this language in self-description. Thus, when Tom, watching Dick, has
behavioral evidence which warrants the use of the sentence (in the language of
the theory) “Dick is thinking ‘p’” (or “Dick is thinking that p”), Dick, using the
same behavioral evidence, can say, in the language of the theory, “I am thinking
‘p’” (or ‘I am thinking that p’.) And it now turns out – need it have? – that Dick
can be trained to give reasonably reliable self-descriptions, using the language
of the theory, without having to observe his overt behavior. Jones brings this
about, roughly by applauding utterances by Dick of “I am thinking that p” when
the behavioral evidence strongly supports the theoretical statement “Dick is
thinking that p”; and by frowning on utterances of “I am thinking that p,” when
the evidence does not support this theoretical statement. Our ancestors begin to
speak of the privileged access each of us has to his own thoughts. What began
as a language with a purely theoretical use has gained a reporting role (1997,
§59/pp. 106-107).
And again, about sensations:
Let us suppose that as his final service to mankind before he vanishes without a
trace, Jones teaches his theory of perception to his fellows. As before in the
case of thoughts, they begin by using the language of impressions to draw
theoretical conclusions from appropriate premises. . . . Finally he succeeds in
training them to make a reporting use of this language. He trains them, that is,
to say “I have the impression of a red triangle” when, and only when, according
to the theory, they are indeed having the impression of a red triangle (1997,
§62/p. 115).
In both cases, Jones’s protégées first apply psychological concepts to
others in a strictly theoretical manner. That is, they note the presence in the
other of certain kinds of nonpsychologically characterized observable
behavior. They then infer, on the basis of Jones’s theory, the presence of
certain psychological states, and, if called upon to make a further inference to
other psychological states or a prediction of behavior, also do so via the
mediation of Jones’s theory. They first apply Jones’s theory to themselves on
exactly the same basis. Later, Sellars tells us, they come to be able to apply
72 Willem deVries

Jones’s new vocabulary to themselves directly and make reports of their own
states. But notice that Sellars does not tell us anything further about the
second- or third-person cases. Does Sellars think that Jones’s protégées can
also be trained to report immediately on the psychological states of others? I
argued in the previous section of this paper that being able to report on a fact
or entity, in Sellars’s view, removes it from the realm of the theoretical sensu
stricto. And I also argued that at least phenomenologically speaking, we can
and do often report immediately and directly on the mental states of others.
(These points are most directly relevant to the theory-theory as an account of
our use of our mature psychological competences. We might still acquire that
competence in theory-like ways, even if we eventually learn to apply it in non-
theory-like ways.) If these claims are correct, then in neither the first-, second-,
nor third-person cases need we be literally applying a theory when we exercise
our psychological abilities.

4.4. Explicit and Implicit Theories

So here is one point on which the theory analogy cannot be pushed. But this
point need not faze a theory-theorist, for there is a common move available
(and noted by Botterill) that allows the theory-theorist to hold on to the
analogy in a fairly strong form while denying that in the exercise of our
psychological abilities we are always applying a theory in an explicit fashion.
That is, the theory-theorist can attribute to us a tacit rather than an explicit
theory. Although phenomenologically we often respond directly and
immediately to our own and other’s psychological states, those responses are,
the theory-theorist claims, mediated by unconscious internal processes that are
like those involved in the application of a theory. So, a theory-theorist might
say, when Jones’s protégées come to be able to report on psychological states,
they have become so fluent in the application and manipulation of the
theoretical vocabulary that the cognitive operations involved have become
automatic and habitual, so automatic and habitual that they recede behind the
line of consciousness. Just as the musician practices until he no longer has to
think about what fingers to move where, until it is all so automatic that he need
only think about the music itself, we become expert at on-the-fly psychological
theory. For the musician, it can be the death of music if she begins to think
about her movements, and she could not possibly construct an explicit account
of those movements; similarly, we may be incapable of any explicit account of
our psychologizing and may get significantly worse at it when we think about
it.
Possessing and acting on a tacit theory differs in some important respects
from possessing and acting on an explicit theory. Phenomenologically the two
 Folk Psychology, Theories, and the Sellarsian Roots 73

are very different. But that does not entail that it is wrong to think of them as
two different ways of possessing and acting on a theory. The notion of tacit
knowledge has been utilized by philosophers since Plato, but it has been at the
very heart of the cognitive revolution in psychology: explain complex
behavioral capacities by attributing to the organism some internal
representational structure(s) that mediate(s) its responses to its environment.
Very often, the representational structures that enable a particular capacity or
competence are thought to be (like) theories of the relevant domain. That is,
the representational structures are taken to consist of items with propositional
form that are interrelated by law-like generalizations that are also elements in
the system, knit together by some inference engine. Such tacit theories share
many characteristics with explicit theories; differing with regard to our
explicit, conscious command and manipulation of the theory is not itself
sufficient to destroy the otherwise fruitful analogy.
Is there any indication that Sellars thought that our actual, empirical
competence at psychology would be best accounted for by literally moving
Jones’s theory inside our heads and retaining it as implicit? No, there really is
not. As I noted earlier, Sellars is (and knows himself to be) a philosopher and
not a psychologist. As such, he is not interested in either the developmental
story behind our psychological competence (note, there is no mention in the
myth of Jones of how children can come to acquire psychological expertise by
four or five, well before they would be capable of mastering an explicit
Jonesian theory), nor is he interested in the cognitive processing story beyond
the commitments contained in Jones’s theory that something, somehow like
overt language is going on in us when we think and that something, somehow
like a replica of a physical object is present within us in perception. In this
light, we have to conclude that the latter-day theory-theory is a creative
extension of Sellars’s proposal that, though consistent and coherent with his
original proposal, is not already contained or intended therein. As such, the
theory-theory is subject to empirical confirmation and refutation, not
philosophical support or refutation. Should the theory-theory prove not to be
the best available empirical explanation, Sellars’s original suggestion remains
untouched, for the theory-theory goes beyond anything Sellars himself
claimed.
Indeed, the hard-core theory-theory, in the light of current research, faces
an uphill battle as an empirical theory:
It requires an implausible degree of inductive and hypothesis-testing
competence in three- and four-year-olds, an implausible universal, uniform and
rather selective fixation on the mind as a domain of study, and fails to explain
why [Theory of Mind] would be selectively impaired in those like
high-functioning autistics who are nonetheless theoretically competent in other
74 Willem deVries

domains (that is, it does not account for the false picture/false belief
performance dissociation). (Garfield et. al. 2001, p. 526)

5. Beyond the Theory-Theory: Sellarsian Constraints on the Theory of

Mind

5.1. Semantics and the Mental

The astute reader will undoubtedly have noticed that I have so far ignored the
second attraction of the theory-theory mentioned by Botterill, the semantic
attraction. That is because the story here gets both more complex and more
interesting. As Botterill notes, since theories constitute implicit definitions of
the primitive terms of the theory, the theory-theory offers us an approach to
what is involved in our comprehension of psychological terms. But notice that,
for Sellars, this is not a particular advantage of the theory-theory approach, for
Sellars possesses a general theory of meaning that is thoroughly
functionalistic. For him, the fact that theoretical terms possess their meaning in
virtue of functional relationships to other terms, to observations, and to action
does not distinguish them from any other kinds of terms. Theoretical terms are
distinguished by the fact that they have no reporting use in observation
statements, but their meanings are otherwise determined in much the same way
any other terms’ meanings are determined. The story of Jones offers a
distinctive and illuminating account of the semantics of psychological terms
insofar as the Cartesian-empiricist tradition presumes that psychological terms,
if any, get their meanings from direct connection to or acquaintance with the
very phenomena they mean. That is, the myth of Jones can help us leave
behind the old Cartesian-empiricist theory that there must be some terms that
get their meaning by direct connection to the object, event, state, or property
meant (i.e., by ostensive definition), and that the meanings of all other terms
are constructions out of or derived from this foundational level of meaning.
But once we have left the old foundationalist conception of meaning behind,
we realize that there is nothing special in this regard about psychological
terms.
But, I believe, there is a deeper level at which the semantic attraction is a
powerful draw for Sellars. The story of Jones gives us not just a way to
understand how psychological terms can be meaningful, but also a way to
understand what meanings they have. Combined with Sellars’s theory of
meaning, it gives us a (theory of) psychological nomenclature, a first but very
important step towards a theory of the development of and the cognitive
processes underlying our psyches. But, as we will see, it does not give us
 Folk Psychology, Theories, and the Sellarsian Roots 75

anything like a set of distinctive, lawlike psychological generalizations, and

thus falls short itself of being the kind of psychological theory empirical
psychologists are looking for.
It is part of Jones’s theory that intentional states are inner states that are
tied (on the action side) both to non-verbal behavior and (most importantly in
this connection) to verbal episodes in which they are expressed. It is in virtue
of this last fact about intentional states that the semantic categories that apply
in the first instance to the linguistic episodes in which intentional states are
expressed also apply by analogical extension to those intentional states
themselves. Thus, Smith’s inner state that would typically express itself in an
assertion that it is cold outside is a belief with that same content, that it is cold
outside.
Attributing meaning or content to an intentional state is analogous,
according to Sellars/Jones, to giving the meaning of an utterance. So what is it
to give the meaning of an utterance, according to Sellars? His analysis of
meaning sentences is a very important theme in Sellars’s work, but has not
been given very much attention. In his view, a meaning sentence such as,
‘Gelb’ (in German) means yellow
should not be understood as stating a kind of relation between a German word
type and yellow, (which is what: an abstract object (the property yellowness),
the set of yellow objects (actual and/or possible), the scattered object that is the
sum of all yellow things?). Rather, according to Sellars, ‘means’ functions in
such assertions as a specialized form of the copula. That means that he
construes meaning statements as statements in which we classify a symbol
type. How does this form of classification work? Well, Sellars argues that in
the above meaning statement the symbol ‘yellow’ occurs in a very special
way. It certainly is not straightforwardly used – nothing is being described as
yellow. Neither is it straightforwardly mentioned – the sentence is not about
the English symbol ‘yellow’. There is a third way in which a symbol can occur
in a sentence like a means sentence that is neither use nor mention, though it
has some affinities with each. Sellars called such terms “illustrating sortals”
and developed his somewhat notorious dot-quote device to capture the
intention. But Sellars’s extensive technical discussions of distributive singular
terms, illustrating sortals, and dot-quotes did not, apparently, get his point
across very well, for as I have said, his analysis of meaning has not received
much attention. So let me try a different approach here.
Jane Heal (1997, 2001, unpublished) has lately been developing a related
conception that many might find more intuitive than Sellars’s somewhat
baroque terminology. Heal points out that, though indexical reference has been
given a great deal of attention, there is no reason to believe that indexicality is
76 Willem deVries

confined to the phenomenon of reference alone. She therefore develops in a

most interesting fashion a notion of indexical predication. And (as is often the
case with a nice piece of philosophy), we can immediately recognize a familiar
phenomenon. Suppose one is redecorating. Having chosen the central furniture
group, one then seeks coordinating drapes. Some people might be able to
describe in language alone the exact color and pattern to be matched, but most
of us would need to bring along a sample of the fabric or a color swatch we
have matched to the fabric directly. Then we can search for drapes that are that
color, where we exhibit an example of the relevant property in conjunction
with our assertion or thought. Similarly, one might describe what someone did
by saying “He did this,” accompanied by an exhibition of the relevant action or
motion.
Armed with this notion, we can say that Sellars thinks that meaning
statements are indexical predications the subject of which is the symbol on the
left-hand side (‘Gelb’ in the example above) and the predicate of which
involves an indexical, illustrative, or exemplary use of something with the
relevant property on the right-hand side, in this case the English word
‘yellow’. According to Sellars, the above meaning statement conveys the
information that in the German language the symbol type ‘Gelb’ has a (set of)
function(s) relevantly similar to those possessed by the symbol ‘yellow’ in the
language of the speaker. It conveys this information, but it does not literally
say it. It does not mean in the German language the symbol type ‘Gelb’ has a
(set of) function(s) relevantly similar to those possessed by the symbol ‘yellow’
in the language of the speaker, any more than ‘He did this’ combined with a
picking of one’s nose literally says that he picked his nose.13 When we say
what an expression means, we do so by exhibiting another expression in our
language with the same meaning (or as close as we can get). Although Sellars
thinks that having a meaning is playing a functional role in a complex
“linguistic economy” of language-entry moves, intralinguistic transitions, and
language-exit moves, and that which meaning an expression has is a matter of
which functional role it plays, we do not specify the meaning of an expression
by describing which role it plays. We give the meaning by exhibiting another
expression that plays a relevantly similar role. Could we say which functional

13 A quick test: translate both “‘Gelb’ (in German) means yellow” and “In the German language

the symbol type ‘Gelb’ has a (set of) function(s) relevantly similar to those possessed by the
symbol ‘yellow’ in the language of the speaker” into French. They are clearly not the same, since
in the former ‘yellow’ is translated, while the ‘yellow’ that occurs in the latter would not be.
 Folk Psychology, Theories, and the Sellarsian Roots 77

role a particular expression plays, and thus avoid the indexical predication?14
Perhaps in some cases, but if so, only a few.15

5.2. Classificatory vs. Causal Theories

In Sellars’s view, then, when we attribute intentional states to people, we are
effectively classifying their states. Consider a standard propositional attitude
attribution, e.g.
Ralph believes that it is raining.
In Sellars’s view this functions much like a piece of indirect discourse such as
Ralph asserts that it is raining.
In indirect discourse (roughly), as in meaning statements, we are classifying
some item (in this case Ralph’s assertion) using a kind of indexical predication
in which we exhibit an item in our background language that possesses the
relevantly similar property. Since Sellars thinks that having a specific meaning
or content is possessing a specific functional role, such a classificatory
statement will be true iff Ralph’s internal state is involved in a complex
functional system in which it plays a role analogous to the role assertions of ‘it
is raining’ plays in English. Such attributions presume a complex system of
causal relations, but they do not say what those causal relations are. The hard
work is done (or perhaps it is avoided) by counting on our linguistic
competence. Because I am an accomplished speaker and interpreter of English,
I have an inarticulable sense or command of the role ‘It is raining’ plays. (I
may be able to say many things about the proper use of the phrase, but could
not finally articulate all the “rules” at work in my know-how concerning even
this simple phrase.) I can employ my linguistic know-how to generate
psychological insight, for my knowledge of the appropriate situations for
asserting that, denying that, apologizing for, or praying that it is raining can be
utilized to recognize appropriate situations for believing that, disbelieving that,
regretting that, hoping that it is raining, etc., and the consequences thereof.
It is Jones’s theory of intentional states that he and his comrades possess
internal states with semantic properties (such as meaning) that are highly

14 Heal names kinds our principal cognitive relations to which are narrowly indexical in this way

“Lagadonian,” a reference to Swift and Gulliver’s Travels. Meanings, in Sellars’s view, are
Lagadonian kinds.
15 Truth tables give the function the connectives have. But, notoriously, the closest English

equivalents are not simply and straightforwardly the truth-functional connectives. A complete
specification of the functional role of ‘and’ in English would be very complex.
78 Willem deVries

analogous to the semantic properties of overt utterances. So he can

commandeer (or perhaps better, transmute) metalinguistic forms of speech to
talk about the properties of these internal states. With such a move a great deal
comes along for free. We not only get a method of distinguishing in a very
fine-grained way among the contents attributed to these internal states, but the
notion that we can take different attitudes towards a given content. We can,
that is, not only differentiate between being in internal states with content p
and content q, we can distinguish being in the kind of state that is normally
expressed in an assertion that p and one that is expressed in a demand that p or
a request that p or a question whether p. In what is close to a fell swoop, Jones
acquires a rich nomenclature of internal states, which we can perfectly well
call a classificatory theory of mental states.16
Does he acquire thereby a rich causal theory of internal states, that is, a set
of connected causal generalizations that would have empirical consequences
for the developmental and cognitive processing questions? I think the answer
here is clearly no.
Jones and his protégées, while still in the awkward early stages of
mastering his theory, can reason as follows:
Ralph is in a state that he would candidly express by asserting that p, so
Ralph believes that p.
Ralph is also in a state that he would candidly express by asserting that
if p, then q, so Ralph believes that if p, then q.
So, ceteris paribus, it is likely that Ralph believes that q (though I have
no direct behavioral evidence of that).
Jones’s reasoning here is perfectly good and is familiar to all of us. But it is
not explicitly a form of causal reasoning. Jones and company get the
connection between these states of mind for free via their semantic
characterization. They (and we) can assume that there are causal connections
among our internal states that enable those internal states to relate to each
other in ways that conform to their semantic characterization – but how those
causal connections work, what kinds of mechanisms instantiate them, how
they might develop and how they might break down, is left totally untouched.
The “internal structures” recognized in Jonesian psychology are almost
completely parasitic on the structure of the language Jones and company
speak.

16 It seems worth calling it a “theory” because there are a wealth of implications about relations

among the states we attribute to people with this nomenclature. We can therefore use it to generate
new knowledge not otherwise available.
 Folk Psychology, Theories, and the Sellarsian Roots 79

5.3. Constraints on “Theory of Mind”

In developing his theory, Jones has very cleverly exploited a model (language)
from which he can import a vast array of distinctions, distinctions which
presume causal underpinnings but make no direct commitment to them. Jones
thereby jumps into a rich and complex classificatory theory of mental states
that is fairly impoverished as a causal theory. However the causal story
underlying our mental states gets filled in, however, it will be expected to
preserve in fair degree the classificatory scheme and generic causal ties that
provided the initial fruit of the theory. Thus Jones’s theory (and Sellars’s) does
not have a developmental component built into it, nor a theory of cognitive
processing; it is compatible with many such theories. For instance, the
theory-theory would be a consistent extension and development of Jones’s
theory, but it is not, as it were, an automatic extension of it, nor even a
particularly natural extension of it.
One of the reasons Jones’s theory can be impoverished when considered as
a causal theory is that we do not need to do a great deal of causal calculation in
order to be able to use the theory effectively and efficiently. We can exploit
the classificatory scheme the theory makes available without employing
complex causal laws to connect the various states and state transitions because
we can connect them semantically. Because of Jones’s cleverness, our mastery
of our language does most of the work in our psychological competence. To
flip Haugeland’s dictum around, by taking care of the semantics, we can let the
syntax (the causal) take care of itself. Thus, we do not need or have (at least in
the immediate post-Jonesian community) a separable psychological theory as
an extensive set of lawlike causal generalizations governing interrelations
among our psychological states. We learn rather to exploit our linguistic
competence “in a new key.”17
If our linguistic competence is itself to be explained by the possession of a
theory, then, of course, the theory theorist could still be happy. But it is not
clear that Sellars believes that our mastery of language is to be explained as
possession of a theory. A thoroughgoing investigation of Sellars’s
understanding of linguistic competence goes beyond what we can hope to do
here, but Sellars certainly rejects the Fodorian line that we have, essentially, an
innate language that we command because we possess (innately) a theory of its

17 Indeed, the view we have discovered in Sellars explains why most of the time it certainly does
not “feel like” we are applying a theory when we explain and predict the behavior of others: much
of the work is done by whatever mechanisms are responsible for our linguistic processing, and
language comprehension and performance does not “feel like” theory application either.
80 Willem deVries

grammar and semantics. In the first place, any such position flies in the face of
Sellars’s critique of the given, for it presumes
. . . that the process of teaching a child to use a language is that of teaching it to
discriminate elements within a logical space of particulars, universals, facts,
etc., of which it is already undiscriminatingly aware, and to associate these
discriminated elements with verbal symbols (1997, §30/p. 241).
Despite his trenchant critique of empiricism, Sellars has no intention running
into the arms of an innatist rationalism, even with regard to the linguistic
abilities he believes are the central core of our cognitive lives.
This point can be generalized a bit: Sellars would also have to reject the
“modularist” position as it is most often conceived. Many modularists (e.g.
Baron-Cohen, Swettenham) maintain that our psychological competence is to
be explained by our possession of a theory, but this theory is not acquired by a
complex process of theory revision. Rather, it is encapsulated in an innately
specified module that develops in response to environmental stimuli. Sellars, I
believe, would find modularism of this type a bit of a muddle. Such a
psychology module does not develop by any standard form of learning, it
grows; furthermore, it is a special purpose device, to some important degree
independent of our other cognitive abilities. Yet insofar as it contains or
applies a theory, it is supposed to operate by applying concepts, performing
inferences and other logico-conceptual, that is, rational operations. But
theories are subject-specific data-structures or knowledge-structures,
possession of which allows us to bring our most powerful general tool,
Reason, to bear upon that particular subject-matter. In Sellars’s view there may
be innate, genetically determined, special-purpose, relatively encapsulated
capacities employed in recognizing and understanding psychological states,
but it would muddle important issues to characterize these capacities as
constituting a theory. If the “theory” that accounts for our psychological
competence is not only implicit, but cut off from our general rationality both in
its synchronic use and its diachronic development, what is the point of still
calling it a “theory”? Of course, these remarks are consistent with the
empirical truth of modularism, namely that our psychological abilities are best
explained by the presence of an encapsulated, innately specified, mandatory
set of processes. The point is only that explaining our psychological
competence by such an innate module really abandons the idea that
psychological concepts are – at the empirical as well as at the semantic or
 Folk Psychology, Theories, and the Sellarsian Roots 81

philosophical level – significantly like theoretical concepts.18 A proponent of

innate psychology modules talks of the “theory of mind” only by courtesy or
convention.
Besides the oxymoronic character of talk of innate, modular “theories,”
there is a deeper question: what is the relationship between innate, modular
cognitive capacities – and there are without much doubt some such capacities,
though which ones is still in question – and an epistemological given? But that
is a topic that demands a paper on its own.

6. Bridging Another Gap: Theory and Simulation

From the 1980s the “theory-theory” has been contrasted to the so-called
“simulation” theory, an approach that argues on philosophical grounds that we
do not apply psychological concepts in any way similar to the application of
theoretical concepts, because there is a distinctive kind of use of one’s own
psychology to model the psychological structures attributed to others.
Proponents of the simulation approach have at times attacked Sellars, hoping
to discredit the father of the theory theory and similar approaches and thereby
all his progeny as well (see Gordon 2000).
But if the arguments I have assembled here are right, Sellars’s own position
is hardly a full-fledged theory-theory position. Rather, Sellars is committed
only to the limited idea that, epistemologically, our psychological concepts are
analogous to theoretical concepts; he remains steadfastly uncommitted on the
causal mechanisms that underlie our use of those psychological concepts.
Those causal mechanisms might be like those used in scientific theories, but
they need not be. The simulationists, however, are concerned primarily about
the mechanisms by which we apply our psychological concepts.
The simulationists maintain that in attributing psychological states to
others we utilize our own psychological mechanisms to model the psychology
of the other. As others have pointed out, this seems to assume prior and
independent access to our own psychological states.19 And that smacks of a
return to the inside-out Cartesian view that we have primitive, privileged
access to our own psychological states as such, from which we can construct

18 For other, empirically based arguments against the notion that our psychological competence is

to be explained by reference to an innate theory-constituting module, see Garfield et. al. (2001).
They argue there that there is a perfectly acceptable sense in which our psychological competence
may be modularized, but that it should not be thought of as an innate module.
19 Gordon works hard to avoid this commitment, but it is clearly evident in Goldman’s work.
82 Willem deVries

other knowledge. Sellars would clearly not be happy with any such position,
because he wants to be able to explain and not presuppose our access to our
own psychologies, but we can nonetheless identify simulation-like elements in
his view.
A Sellarsian story would not presume our ability to identify our own
psychological state independently of the psychological states of others. Our
ability to identify, understand, and attribute psychological states is
intersubjective from the root up in a Sellarsian story. But, because many
important relationships among our psychological states are mirrored in
corresponding relationships between the speech acts in which their contents
are expressed, language learning creates a competence that can also be
recruited to perform psychological work, especially when combined with an
increasingly sophisticated understanding of and competence in social
interaction.20
In Sellars’s view, then, our attributions of psychological states rests on a
kind of pre-theoretic21 know-how, but it is not knowing how to invoke our own
psychology (which would have to be known to us in a different way) in order
to understand another’s, it is knowing how language and social situations work
“from the inside.” This knowledge can then be recruited to make sense out of
both our own and other’s behavior without having to take a detour through
externalization in an explicit theory of psychology.22

7. Conclusion

“Empiricism and the Philosophy of Mind” marked a major turn in the

philosophy of mind. Some turned too far, for lack of a deeper understanding of
the place psychological concepts play in our conceptual framework. It is
unfortunate that Sellars’s writing career was pretty much finished before most
of the empirical research into “theory of mind” occurred and before the
theory-theory vs. simulation debate really took hold. It is unlikely Sellars

20 Heal’s version of simulationism emphasizes its relevance to our understanding of psychological

content. Given the similarities between Sellars’s and Heal’s treatments of content, it is not
surprising to find them echoing each other in this regard.
21 Sellars can call these competencies “pre-theoretic” despite the Myth of Jones, because children

acquire these competencies well before they acquire any conception of theory or observation.
Remember, to call something “theoretical” is a methodological distinction, and so, therefore, is the
notion of the pre-theoretical.
22 For a more detailed picture of how this view fits into the empirical data currently available, see

the article cited above by Garfield et. al. (2001).

 Folk Psychology, Theories, and the Sellarsian Roots 83

would have found any of the standard alternatives fully satisfying, and most
likely he would have cast interesting new light on the debate. Thinking his
position through forces us to be more careful about the alleged analogy
between the theoretical and the psychological. Without such care, confusion
threatens on many fronts. It is important to remember that Sellars’s original
philosophical point about the epistemological status of psychological concepts
is separable from empirical debates about the mechanisms and the
developmental process by which we come to employ those concepts.

Willem deVries
Department of Philosophy
University of New Hampshire
Durham, NH 03824
United States of America
e-mail: [email protected]

Visiting Fellow, Philosophy Programme

University of London
London WC1E 7HU
United Kingdom

REFERENCES

Astington, J. (1996). What Is Theoretical About the Child’s Theory of Mind? A Vygotskyan View
of Its Development. In: P. Carruthers and P. Smith, (eds.), Theories of Theories of Mind,
pp. 184-199. Cambridge: Cambridge University Press.
Baron-Cohen, S. (1991). The Theory of Mind Deficit in Autism: How Specific Is It? British
Journal of Developmental Psychology 9, 301-314.
Baron-Cohen, S., A. Leslie, and U. Frith (1985). Does the Autistic Child Have a Theory of Mind?
Cognition 21, 37-46.
Botterill, G. (1996). Folk Psychology and Theoretical Status. In: P. Carruthers and P. Smith,
(eds.), Theories of Theories of Mind, pp. 184-199. Cambridge: Cambridge University Press.
Churchland, P. (1979). Scientific Realism and the Plasticity of Mind. Cambridge: Cambridge
University Press.
Davidson, D. (2001). Subjective, Intersubjective, Objective. Oxford: Oxford University Press.
deVries, W.A. and T. Triplett (2000). Knowledge, Mind, and the Given: Reading Wilfrid Sellars’s
‘Empiricism and the Philosophy of Mind’. Indianapolis, IN: Hackett Publishing.
Garfield, J.L., C.C. Peterson, and T. Perry (2001). Social Cognition, Language Acquisition and
The Development of the Theory of Mind. Mind & Language 16, 494-541.
Gopnik, A. and H.M. Wellman (1992). Why the Child’s Theory of Mind Really Is a Theory. Mind
and Language 7, 145-171.
84 Willem deVries

Gordon, R.M. (2000). Sellars’s Ryleans Revisited. Protosociology 14, 102-114.

Heal, J. (1997). Indexical Predicates and their Uses. Mind 106, 619-640.
Heal, J. (2001). On Speaking Thus: The Semantics of Indirect Discourse. The Philosophical
Quarterly 51, 433-454.
Heal, J. (unpublished). Lagadonian Kinds. Manuscript.
Mitchell, P. (1997). Introduction to Theory of Mind: Children, Autism, and Apes. London: Arnold
Publishers.
Perner, J., T. Ruffman, and S. Leekam (1994). Theory of Mind Is Contagious: You Catch It From
Your Sibs. Child Development 65, 1228-1238.
Peterson, C. and M. Siegal (2000). Insights into Theory of Mind from Deafness and Autism. Mind
and Language 15, 123-145.
Premack, D. and G. Woodruff (1978). Does the Chimpanzee Have a Theory of Mind?
Behavioural and Brain Sciences 1, 515-526.
Rorty, R. (1965). Mind-Body Identity, Privacy, and Categories. The Review of Metaphysics 19,
24-54.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Sellars, W. (1963). Science, Perception and Reality. London: Routledge and Kegan Paul.
Wellman, H. M. (1990). The Child’s Theory of Mind. Cambridge, MA: The MIT Press.
Timm Triplett and Willem deVries

IS SELLARS’S RYLEAN HYPOTHESIS PLAUSIBLE?

A DIALOGUE

ABSTRACT. In order to provide an alternative to the Cartesian myth that knowledge

of our thoughts and sensations is “given,” Sellars posits a community of “Rylean
ancestors” – humans at an early stage of conceptual development who possess a
language containing sophisticated concepts about the physical world and about their
own language and behavior, but who lack any concepts of thoughts or sensations.
Sellars’s presentation of this thought experiment leaves many important details sketchy.
In the following dialogue, we offer our differing assessments of how well those details
could be filled out. TT questions how the Rylean hypothesis could provide a plausible
account of human thought and sensory experience at any stage of human conceptual
development. WdeV responds to TT’s challenge by filling out the picture of what the
Ryleans’ conceptual world would look like. TT and WdeV debate the merits and the
plausibility of the resulting picture.

Wilfrid Sellars devotes the final sections of “Empiricism and the Philosophy of
Mind” (1997, hereafter: EPM) to outlining and discussing a thought
experiment about “our Rylean ancestors.” After his famous critical attack on
the traditional empiricist notion of the “given,” Sellars turns to the Rylean
scenario to provide the key positive element in his reconstruction of a
post-empiricist account of mind and our epistemic access to its contents.1
In order to move from the Cartesian myth that knowledge of our thoughts
and sensations is “given” – known by us first and best, immediately accessible
to us epistemically by the mere fact of their occurrence, and the foundation or

1 To be sure, the critique of the “given” itself includes a statement and defense of a positive
account of knowledge in Part VIII of EPM. We discuss it in our commentary on EPM –
“Knowledge, Mind, and the Given” (deVries and Triplett 2000, see especially pp. 77-107). But
Sellars devoted many more pages of EPM (Parts XII through XVI) to discussing the Rylean
scenario than to elaborating his Part VIII positive epistemic principle. And it is the former that has
had the most subsequent influence. It had and continues to have a major influence on the course of
developments in the philosophy of mind.

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 85-114.
Amsterdam/NewYork: Rodopi, 2006.
86 Timm Triplett and Willem deVries

base out of which we construct our knowledge of physical objects and other
persons – Sellars has us consider a community of humans in an early stage of
conceptual development who possess a language containing an array of
concepts about their physical world and about their own language and
behavior, but who lack any concept of thoughts or sensations. They do not
conceive of themselves as possessing inner mental states; indeed, they do not
conceive that there are such things as mental states. And yet their language is
otherwise quite sophisticated:
Its total expressive power is very great. For it makes subtle use not only of the
elementary logical operations of conjunction, disjunction, negation, and
quantification, but especially of the subjunctive conditional. Furthermore, I
shall suppose it to be characterized by the presence of the looser logical
relations typical of ordinary discourse which are referred to by philosophers
under the headings ‘vagueness’ and ‘open texture’. (1997, §48/p. 258)
Because this language allows for elaborate description, analysis, and
prediction of human behavior (without, of course, referring to motivations,
intentions, decisions or other mental events – unless behavioristically
construed), Sellars dubs the members of his hypothetical community
“Ryleans,” in honor of Gilbert Ryle and his attempt in The Concept of Mind to
analyze mental states and events in terms of behaviors and dispositions to
behave (see Ryle 1949).
The Rylean scenario provides an approach to our conception of the mental
that is diametrically opposed to that provided by the Cartesian “givenist.” It
suggests that human mental life need not be understood as something we are
intimately and immediately acquainted with. Rather, from concepts pertaining
to what is publicly observable – physical objects and events, bodily behavior,
and linguistic utterances – the Ryleans develop concepts of the mind as
something inner in a theoretical or quasi-theoretical way.2
In order to explain how this development of mental concepts can proceed,
Sellars posits the genius Jones, a member of the Rylean community who
conceives of and introduces mental concepts to the rest of the community. On
the basis of his native Rylean concepts of publicly observable objects and
events, Jones develops concepts pertaining to thoughts and sensations (called
“impressions” by Sellars in the relevant later sections of EPM) by utilizing a

2 Sellars says that mental concepts are analogous to the theoretical concepts introduced in, e.g.

scientific discourse, but are not purely theoretical. See our commentary (deVries and Triplett
2000, pp. 154-55) for elaboration.
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 87

purportedly analogous model. The details of the developments of concepts of

thoughts and concepts of impressions differ in significant ways.
In developing his concept of thoughts, Jones uses words and language as
his model. We summarize Jones’s specific theory-construction in our
commentary on EPM as follows:
One day . . . Jones worries about how it can be that his fellow Ryleans exhibit
intelligent behavior when they are not exhibiting overt verbal behavior. Since
the Ryleans do have the capacity to conceptualize unobservable entities in an
explanatory role, Jones does just this with respect to the problem he wants to
solve. As his model for the new domain of entities that he is postulating, Jones
adopts overt linguistic episodes and says that episodes like these occur within
us, the normal effect of which is intelligent behavior, including the production
of the very linguistic episodes that the inner episodes are modeled on. (This is
not a vicious circle, any more than explaining the behavior of billiard balls in
terms of tiny corpuscles modeled on such physical objects entails a circle.)
(deVries and Triplett 2000, p. 142)
Jones then teaches this theory of inner speech to other Ryleans. It is
important to understand that the way in which they first learn to recognize
thinking within themselves is not by simply attending to some feature of the
world – their thinking – that they had failed to notice but are able to readily
and directly notice as soon as Jones draws their attention to it. Rather, an
individual Rylean first recognizes thinking by observing others’ behavior and
by reasoning, like any good scientist, from that behavioral data to an
unobserved cause. He notices that others (sometimes) act in complexly
intelligent ways without engaging in outward verbal behavior and infers that
they must be thinking – that events that are not publicly observable must be
occurring within them that are in relevant ways like speech.3 The individual
Rylean can also apply this reasoning to observations of his own behavior and
conclude that he himself is now thinking. Eventually, he can learn to recognize
instances of his own thinking without going through this inferential process,
much as a Renaissance astronomer comes to recognize – and perhaps even
claims to “just see” – the moons of Jupiter without going through the
inferential process taught him by Galileo. Here is Sellars’s description of this
final – non-inferential – stage:

3See our commentary on EPM for discussion of Sellars’s cautions about what does and does not
get carried over from the features of public language to the constructed theory of thought (deVries
and Triplett 2000, pp. 144-145).
88 Timm Triplett and Willem deVries

And it now turns out – need it have? – that Dick [one of Jones’s pupils who has
learned to ascribe the activity of thinking to himself based on his observations
of his outward behavior] can be trained to give reasonably reliable
self-descriptions, using the language of the theory, without having to observe
his overt behavior. Jones brings this about, roughly, by applauding utterances
by Dick of “I am thinking that p” when the behavioral evidence strongly
supports the theoretical statement “Dick is thinking that p”; and by frowning on
utterances of “I am thinking that p,” when the evidence does not support this
theoretical statement. Our ancestors begin to speak of the privileged access each
of us has to his own thoughts. What began as a language with a purely
theoretical use has gained a reporting role. (1997, §59/pp. 106-107)
It is important to note that, for Sellars, Jones’s account of thinking is, while
the unsophisticated “germ of a theory” (1997, §58/p. 267), nonetheless correct
in its essentials (taking account of its sketchiness). Sellars has been
misinterpreted as an eliminativist who treats thoughts and other mental states
as the product of a bad theory that will be discarded come the scientific
millennium. But for Sellars, thoughts are real enough – they are indeed the
causes of our intelligent behavior. What is wrong with the traditional account
of thoughts is the belief that our epistemic and conceptual access to our
thinking is somehow directly and unproblematically presented to us – that the
contents of our minds are transparent to us and the source of epistemically
certain or incorrigible judgments that can found the rest of our knowledge.4
But Jones’s theory of thoughts is just half the story. As noted, Sellars
(rightly) understands thoughts and sensory impressions to be quite distinct
types of mental events. Sellars’s account of the Ryleans’ acquisition of the
concept of a sensory impression is similar to his account of their acquisition of
the concept of a thought in that the concept of an impression does not present
itself to the mind directly and immediately upon the occurrence of an
impression. Jones must exercise just as much theoretical ingenuity here as in
the previous case. And, as in the case of thoughts, Sellars does not deny that
the Ryleans have impressions. Ryleans are human beings with the sensory
organs of human beings – they thus can see and hear, feel pain, taste their food
and smell odors. They get hungry, experience the satisfaction attendant to
relieving that hunger or quenching a thirst, and enjoy sexual pleasure. One can

4 This is not all that is wrong with the traditional conception of thoughts – equally problematic is
the traditional substance dualist ontology and its taking thoughts to be nonphysical. Sellars opts
for a functionalist account of thoughts that can allow an ultimately neurophysiological account in
the case of human thoughts. But regarding the Rylean scenario that is our concern here, it is the
conception of thoughts rather than their ontology that is of primary relevance.
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 89

also presume that they can imbibe perception-altering substances that would
cause them to believe in the presence of objects near them – pink elephants,
say – that are not actually there. Also, they will be subject to the same
perceptual illusions that we are (sticks half in the water, apparent lakes in
deserts, etc.), though they will not describe or understand them as perceptual
illusions, nor of course will they understand their visual perceptions, pains, etc.
as having to do with inner, nonbehavioral states of themselves.
Jones’s theory of impressions, though, differs from his theory of thoughts
in that the former are modeled not on anything linguistic but on physical
objects.
The theory of impressions is interestingly more complex than the theory of
thoughts, and it would take us further afield than we need to go here to include
all the significant details (see our commentary on EPM 2000 for that). One
detail only to be stated here rather than explained is that, as part of his theory
of thoughts (not impressions), Jones develops an account of (veridical) seeing
(and hearing, touching, etc.) that expands Rylean language to include such
perceptual locutions as “Fred sees a boulder,” “Tom hears the thunder.” But
these locutions are still explicable, in Jones’s theory, as referring to ways of
thinking about things, i.e., as attributing thoughts to the perceiver. As long as
the perception is veridical (there really is a boulder, the thunder really is
occurring) there is not yet a need to add a second type of mental entity. It is
only in order to explain the inevitable nonveridical perceptions – based on
what we understand as perceptual illusions, hallucinations, etc. – that Jones
needs to posit impressions as entities in addition to and distinct from thoughts.
Our commentary again presents Jones’s specific reasoning:
Fred says that he sees pink elephants dancing in front of him. But Charlie, who,
like Fred, meets all of Jones’s conditions for seeing, reports that he sees no pink
elephants in that location. Yet Fred seems sincere (Jones can have seems
sincere locutions or something close to them, given his already developed
theory of thoughts), and he acts truly frightened. Since other Ryleans enter the
scene and claim to see no pink elephants, it is a straightforward matter to
conclude that no dancing pink elephants are in fact before them.
Jones concludes that sometimes a person can be in just that kind of internal
state that he would normally be in if he saw pink elephants dancing in front of
him, even though there are in fact no such pink elephants. It is for Fred just as if
there were pink elephants. But how could this situation be? Clearly, it is not a
case of Fred’s seeing that there are dancing pink elephants. Jones realizes that
there is a need to adopt still further innovations in how we describe situations.
Because perceptual locutions take perceptual objects, Jones models the
grammar of his new way of talking on this aspect of perception talk, and invents
the term ‘impression’ to serve as the object when there is in fact no physical
90 Timm Triplett and Willem deVries

object that is perceived. Thus, Fred has the impression of a dancing pink
elephant.
The model for this idea of an impression is a physical object. An impression of
a dancing pink elephant is in some ways like a dancing pink elephant, except, of
course, that it is not an actual dancing pink elephant, and indeed it is not an
external thing at all but rather, Jones hypothesizes, something internal to the
perceiver that nevertheless shares certain structural similarities with physical
objects. (deVries and Triplett 2000, p. 164)
Once Jones has introduced and taught his impressions theory to his fellow
Ryleans, the concept impressions again progresses, as did the concept thought,
from a theoretical to a reporting role:
As before in the case of thoughts, [the Ryleans] begin by using the language of
impressions to draw theoretical conclusions from appropriate premises. . . .
Finally [Jones] succeeds in training them to make a reporting use of this
language. He trains them, that is, to say ‘I have the impression of a red triangle’
when, and only when, according to the theory, they are indeed having the
impression of a red triangle.
Once again the myth [of the Ryleans] helps us to understand that concepts
pertaining to certain inner episodes – in this case impressions – can be primarily
and essentially inter-subjective, without being resolvable into overt behavioral
symptoms, and that the reporting role of these concepts, their role in
introspection, the fact that each of us has a privileged access to his impressions,
constitutes a dimension of these concepts which is built on and presupposes
their role in inter-subjective discourse. (1997, §62/p. 115)
Jones leaves his theory of impressions in a relatively sketchy form.
Although he initially presents his hypothesis of a Rylean stage of humanity
as a myth, Sellars’s final point in EPM is to suggest that we take it seriously,
not as a myth but as a description of a stage in the development of human
awareness and self-awareness:
I have used a myth to kill a myth – the Myth of the Given. But is my myth
really a myth? Or does the reader not recognize Jones as Man himself in the
middle of his journey from the grunts and groans of the cave to the subtle and
polydimensional discourse of the drawing room, the laboratory, and the study,
the language of Henry and William James, of Einstein and of the philosophers
who, in their efforts to break out of discourse to an arché beyond all discourse,
have provided the most curious dimension of all. (1997, §63/p. 117)
Sellars’s Rylean hypothesis, with its story of the nature and development of
mental concepts, is unquestionably innovative and important. It has had a
continuing influence in the philosophy of mind. Sellars’s functionalist account
of the nature of thoughts has long been understood to have anticipated the
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 91

heyday of functionalism that began in the 1960s.5 But what has not been
recognized is that his insistence on treating impressions differently from
thoughts also anticipated the later questions that arose regarding popular
attempts to provide a wholly functionalist account of the mental. Sellars would
have been surprised neither by the rise of functionalist accounts nor by the
subsequent recognition by many philosophers that nonfunctionalizable
“qualia” (the currently fashionable term for sensory impressions or sensory
experience) create serious problems for functionalism.6
Sellars’s Rylean hypothesis has not been without its critics.7 We ourselves
note in our commentary, regarding the move from the inferential to the
noninferential stage of the Ryleans’ ability to recognize thoughts, that “the
means by which a Rylean introduced to Jones’s theory comes directly to
observe his thoughts, and the explanation for how he is able to do so remains
rather troublingly mysterious in Sellars’s telling of the story” (deVries and
Triplett 2000, pp. 153-154). But this is only to note that there is an explanatory
gap in that story. Many criticisms are much stronger in claiming that the
hypothesis itself is incoherent: necessarily, there could not be human beings
with just the conceptual abilities Sellars attributes to them – the Ryleans are
logically or conceptually impossible. We have in progress a separate dialogue
in which we consider one of the most sophisticated and fully developed of
these incoherence arguments – that of Ausonio Marras (1973a, 1973b, 1977).
Not unexpectedly, we disagree with each other regarding the cogency of
Marras’s critique.
TT: But here, as a foundationalist and a critic of Sellars, I would like to
take a different approach from Marras in criticizing Sellars’s Rylean
hypothesis. Even if the hypothesis was conceptually coherent, its
plausibility would seem questionable to me. One can weave a fairy
tale or science fiction story that is logically possible and internally
coherent, but entirely unlikely as a story about actual human beings in
the real world. Whatever the ultimate success of incoherence
arguments like Marras’s, I suspect that the particular combination

5 Less positively, eliminativists such as Paul Churchland took the cue for their view of mental

entities as posits of a false folk theory of psychology from Sellars’s account of mental concepts as
theoretical, though their treating Sellars himself as an eliminativist is in our judgment a serious
misreading of the Rylean hypothesis.
6 See Jackson (1982) for one of the most influential early works critical of functional treatments of

qualia.
7 The first and best known criticism is that of Roderick Chisholm in his correspondence with

Sellars shortly after EPM was published (Chisholm and Sellars 1957). See also Marras (1973ab,
1977), Young (1973), Gordon (2000), Lehrer and Stern (2000), and Rosenberg (2000).
92 Timm Triplett and Willem deVries

Sellars requires of the concepts that the Ryleans have and those that
they lack is too implausible to count as a realistic description of any
human life. It is important to take account of implausibility arguments
because, if cogent, not only would they provide an independent line
of criticism in addition to the incoherence arguments, but they would
also tell us something important about the nature of sensations and
thoughts. The conclusion would not necessarily be that the traditional
Cartesian account has to be true after all. But we could conclude that
our concepts of ourselves as thinking and sensing beings are at least
on the same evidential level as, if not more fundamental evidentially
than, our concepts of ourselves as linguistic beings in a physical
world.
Certainly, Sellars’s Rylean hypothesis as it stands cries out for
some further elaboration. For the Ryleans’ conceptual world is
manifestly different from ours, and it is reasonable to ask: can the
details of that conceptual world be plausibly worked out? The
question is a natural and forceful one because Sellars grants that the
Ryleans are just like us in that they have sensations and thoughts.
They just do not recognize those states as such. How can they be so
like us constitutively and so unlike us conceptually? Sellars needs to
say more about this than he does, and, as a Sellarsian and a student of
Sellars, you are in as good a position as anyone to help him out.
Specifically, given that a Rylean has sensations8 – to focus on these
rather than thoughts for the moment – and that these are inner
episodes, what conceptual alternatives are open to the Sellarsian?
Clearly, one alternative is that the Rylean has no concept at all in any
way analogous to our concept of sensation. The very idea just never
enters her consciousness. I suppose this is conceptually possible, but
it seems to me like the proverbial Thurber characters who somehow
manage to get through their days without ever acknowledging the
elephant that one day just appears in their bedroom and takes up
residence there.
A second alternative is to conceive of sensations as external
objects like tables, publicly accessible to all, and presumably existing,
like the table, when everyone has left the room. But no one could get
things this wrong. Surely the Ryleans would be aware that the

8 Here and in what follows we use the more common term ‘sensation’ rather than Sellars’s term
‘impression’.
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 93

phenomena we identify as sensations are more closely connected to

individuals than this.
In fact, it seems to me that the Ryleans would have to understand
sensations as aspects of an individual’s behavior. On this third
alternative, the pain sensation is the sudden bodily movement or
vocalization (“Ouch!”). Now, while this is more plausible than the
other alternatives, it still does, by Sellars’s own account, get things
wrong. For Sellars, these inner episodes exist, it is just that the
Ryleans never grasp that these episodes are the cause of these
behaviors. Sellars acknowledges that, prior to Jones’s conceptual
innovations, the Ryleans have a bad theory of themselves, just as in
human history there have been bad theories of mental illness, such as
the demon hypothesis. But it does baffle me how the Ryleans could
get this sort of thing wrong. Is it really just that, lacking any concept
of pain, they simply notice that they and others “react differentially”
(to use some behaviorist jargon) to a hand too close to the fire by a
very quick move of that hand away from the fire? The hand moves
away because of an intense sensation. It seems clear to me that
anyone who knows what pain is like from the inside understands why
there is a differential reaction! Do you at least see why some of us
find the Sellarsian alternative implausible?
WdeV: Things are getting jumbled here. First, it is not clear that the hand
moves away because of the pain: there is evidence to indicate that the
hand begins to move away before the pain is registered. The reflex
uses a shorter pathway than the pain signal.
TT: Touché. There are indeed cases like this where the pain does not
cause the movement. But in many or most cases, surely sensations do
play a causal role. Reflex actions probably appear early from an
evolutionary perspective, but I do not think that organisms could have
evolved very far if they had only reflex actions to cause them to get
out of harm’s way or to undertake activities necessary to their
survival. I would think that the experience of pain immeasurably
improves (over the more primitive reflex action mechanism) the
taking of appropriate action.
In addition, not all pains are of the instantaneous-onset type that
best fit the sorts of cases you have in mind. The onset of the sensation
of hunger, for example, arises and increases gradually.
WdeV: I do not know what you mean by “improves the taking of appropriate
action” in your above scenario. The evolution of pain certainly makes
possible the insertion of other forms of behavioral control that is more
flexible and sophisticated than a reflex. But we should not engage in
94 Timm Triplett and Willem deVries

armchair biology here. Nor should we assume that in order to perform

its job, a pain must be conceptualized as a pain, as something inner.
Let us get back to the alternatives you have offered for how the
Ryleans think of the sensations they have.
The alternatives you propose really break into two categories:
(1) The Ryleans have no concepts that apply to inner states,
either pains or other sensations;
(2) They have some concepts that apply to inner states and
treat them as belonging to a category already familiar to
them – either:
(a) the category of external public objects (tables,
trees, etc.), or
(b) the category of externally observable behaviors
and dispositions to such behaviors.
Your essential point is that even in case (2) the Ryleans make at best
an ongoing category mistake: inner states are necessarily neither
external public objects nor external behaviors nor mere dispositions to
such behavior.
Now, there is no problem with the Ryleans having concepts of
behavioral complexes or syndromes. In Sellars’s story, Jones
develops an explanatory theory for such things by developing a new
category of concepts modeled on (but not identified with or belonging
to the category of) external public objects (for sensations) and public
verbal behavior (for thoughts). So your alternatives have the right
elements. You think the chronology could not plausibly orchestrate
the way Sellars says it might. As I now understand your point, you do
not think that the Ryleans could ever plausibly have a self-
understanding (a theory of themselves) that is so faulty that (from the
perspective of its successor theory), it is absolutely pervaded with
category mistakes.
TT: What I called the third alternative, which you reformatted as your
(2b), can, considered abstractly, perhaps seem reasonable. But if we
look more closely, its implausibility seems to me manifest. Let us
look specifically at the story Sellars has to tell about something like
the sensation of hunger:
How Humans Discovered Hunger
Humans, of course, have always been hungry, in that they have
had the sensations we now identify as hunger from time
immemorial. But, although they thought about many other items,
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 95

features, and relationships in their world and communicated

about these matters via an elaborate language, they never knew
they were hungry or had any conception or idea of hunger as an
inner state. They did notice correlations between types of
behavior: for example, that if an individual went without food for
a time, this was typically followed by unusual persistence and
vigor in the stalking and killing of game or other food-acquiring
behavior. But they never had the thought that they experienced
hunger or any other sort of inner episode in such contexts. This
situation continued for generations. Then one day Jones came
along. After living, and thinking, into his adulthood much as his
compatriots lived and thought, he one day grasped a brilliant new
idea, reasoning as follows: “I notice that sometimes going
without food does not result in the usual correlated behavior, for
example when I eat just a bit of the leaf of this plant (suppose it
contains an appetite-suppressing drug) I do not engage in the
usual stalking and killing behavior even when I have been
without food for quite awhile, and when I have a bite of the root
of this other plant (an appetite-enhancer) I engage in such
behavior even when I have recently had a nice meal. This has
happened regularly enough that the mere fact of going without
food cannot be the full explanation for the presence or absence of
the usually associated food-acquiring behavior. I speculate that
there is some episode inside me that is more directly responsible
for this behavior.” And this is how humans first got the idea that
what we now know and describe as hunger existed.
You would agree that this is how Sellars would characterize the
Ryleans’ situation in this case?
WdeV: Agreed.
TT: You do not find this story implausible on the face of it?
WdeV: Not at all.
TT: Hmmm. Let me try to get at my sense of its implausibility by asking
you some questions about your own experience, OK?
WdeV: Fire away.
TT: You feel pains, and you get hungry, right?
WdeV: Sure.
TT: And you, as a conceptually sophisticated adult in this culture, think of
a sharp pain or a gnawing hunger as an inner episode of some kind?
96 Timm Triplett and Willem deVries

WdeV: Of course, although there will typically be correlations, sometimes

suppressible, between these inner sensations and observable
behaviors.
TT: Fine. In fact, sometimes these sensations are so intense that it is hard
to attend to anything else, right? And it might be nearly or even
totally impossible to suppress the behavior because of that intensity.
WdeV: All right.
TT: And sometimes thoughts can be similarly intrusive, especially the
emotionally powerful ones?
WdeV: OK, I grant all this. So are you thinking that this means that anything
so intrusive has to be conceptualized in one particular way? That does
not follow.
TT: Well, I’m trying to set up a comparison between external physical
objects and events, on the one hand, and internal sensations and
thoughts, on the other. Sometimes you attend to external items and
features, like the waves on the pond outside your house, or the words
on the page. And you have a choice about what to attend to, and
whether to attend to anything at all. But sometimes the external events
command your attention, like the truck suddenly rounding the curve
and bearing down on you as you try to cross the street.
WdeV: OK.
TT: And one can say the same about sensations and thoughts. Sometimes
one can choose to conjure up an image, or to revel in a pleasant
sensation. Other times, as with strong pains or gnawing hungers or
emotionally laden thoughts, they make it difficult to attend to
anything else or they even positively preclude attention to anything
else.
WdeV: I still do not see what this has to do with how we conceptualize them.
You cannot assume that how we now conceptualize these things is the
only plausible way it could be done.
TT: My point is that some sensations and thoughts are just as dominating
and intrusive as some physical objects or external events. So why
should we form concepts of physical objects and not of sensations and
thoughts?
WdeV: Large electrical discharges between the sky and the ground are
sometimes dominating features of our experience. But that does not
mean that it is implausible that minimally sophisticated thinkers
might lack concepts of electricity. The only thing you can conclude is
implausible is that there is no concept at all in the repertoire of the
relevant group which they can utilize in attending to this feature of
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 97

their experience. And we can assume that the Ryleans have a concept
of food deprivation they can use to notice their state.
The difference between our conceptual awareness of external
physical objects, events, and features, on the one hand, and such
awareness of thoughts and sensations, on the other, is that our entire
orientation, evolutionarily speaking, has been toward perception of
objects and features of the physical world. Phylogenetically, there is
every reason to think that humans came to form concepts of the
physical world first. Perception of physical objects, events, and
conditions, whether outside of the body or within it, is what is
essential to our survival, and what evolution has selected for.
Reflection on and awareness of the mental processes by which we
come to have such awareness of the physical world is an evolutionary
afterthought, something it is reasonable to expect humans came to late
phylogenetically. As Nick Humphries has suggested, such meta-
awareness probably became important in adapting to changing and
complex social structures.
TT: We react not to an electrical discharge per se but to a brilliant flash of
light and a loud noise – if there were the discharge without the
sensory manifestations, there would be no dominating feature of our
experience. So it is not surprising that we can have the concept of a
brilliant light or a loud noise without having the concept of an
electrical discharge. Similar points could be made about the
underlying neurophysiology of pain. It is not hard to understand how
we could lack the concept of a neuron – what is hard to understand is
how we, as adults who possess other concepts, could lack the concept
of pain.
WdeV: You’re treading close to the notion of a given here. The concepts of
light flashes or loud noises are not concepts of sensory states, so I do
not think they help you very much. The Ryleans do have concepts
with which to attend to and deal with those states we think of as
hunger or pains. You just want to insist that the concepts they have
must be the concepts we have. But the only reason you’ve given is
that you cannot imagine it otherwise.
TT: I’d like to think that by this stage I have offered something a bit more
substantive than a subjective appeal to my imaginative powers. I am
posing the question of what specifically the Ryleans do conceptualize
– with respect to the phenomenon of hunger sensations, the existence
of which we both acknowledge – given that they do not conceptualize
the sensation of hunger itself. And your most recent answer – that
they have a concept of food deprivation – seems quite unsatisfying.
98 Timm Triplett and Willem deVries

For the fact of food deprivation, while clearly associated with the
phenomenon of hunger, does not track it all that well. Did you ever
notice how, if you go without food long enough, the hunger
moderates and eventually disappears? Early human cultures must
have noticed this frequently as they dealt with the inevitable famines
and scarcities. Conversely, there can be hunger in the midst of plenty;
even hunger immediately after a meal (think of the appetite-enhancing
drug mentioned in my story). The upshot of the significant non-
tracking of the two phenomena is that it is reasonable to think that the
Ryleans would notice the hunger as something distinct from the fact
of food deprivation.
As for flashes of light and noises, while it is obvious that electrical
discharges and neuron firings can occur in a normal adult human’s
immediate environment throughout that person’s lifetime without him
ever having the concept of an electrical discharge or a neuron, this is
not so clear in the case of flashes of light or noises. My point does not
depend on the claim that concepts of flashes and noises are sensory
concepts. I’m just saying there are cases and there are cases. We need
to see whether the hunger case is closer to the electrical discharge and
neuron model or to the flash of light and noise model. And one can
see some problems here for Sellars’s hypothesis. Humans have
probably all experienced flashes of light which obviously occur
without external causes, for example when one rubs one’s closed
eyes. It hardly seems to require an innovative conceptual leap to
understand that such visual images occur within oneself, not in the
world outside.
WdeV: That’s a measure of the distance between us. I think that combining a
grasp of the difference between the internal and the external with a
fine enough sense of the typical forms of external causation so that
one can conclude that some particular events must be internal and not
external requires a tremendous amount of conceptual sophistication.
I think you’re misconstruing what’s really in question here: you
say above that you’re “posing the question of what specifically the
Ryleans do conceptualize.” But that’s not really the question we’re
after; we both think they’re conceptualizing hunger. The question for
us here is how Ryleans conceptualize it. Focusing on what they
conceptualize, the “object” they are dealing with, makes it sound as if
there could not be any difference between them and us, since
everyone is concerned with the same thing, hunger. But the manner in
which they conceptualize hunger could easily be very different from
the way we find so obvious.
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 99

The Ryleans certainly do have concepts available to track hunger,

for besides the concept of food deprivation, they would have
subjunctive constructs thereof. They could say that someone acts as if
food-deprived, even though he’s just eaten, or that someone else,
despite not having eaten for a long time in the famine, no longer is or
acts food-deprived. They could say of someone who is in the midst of
a fast that if he were to hold out a bit longer, the obsession with food
would pass. Behaviorists that they are, there is no reason the Ryleans
cannot realize that food-deprivation behavior (that is, behavior
typically caused by food deprivation) is not always correlated with
food deprivation. They can do a great deal without a concept of
hunger as an inner state.
And do not forget: the Ryleans are unsophisticated behaviorists:
their “theory of mind” is a bad one and leaves a great deal
unexplainable. That’s why Jones’s contribution is so valuable.
TT: Do not forget, either, that the Ryleans are not stupid. You’re right that
Ryleans can refer to food deprivation in counterfactual situations. But
what’s more likely: that they will still keep thinking only in terms of
food deprivation – even when there are often puzzling disparities
between perception and reality forcing them to use complex
subjunctive conditionals – or that they will know exactly what’s going
on: they feel hungry!
WdeV: You keep insisting that there is just something so obvious about
sensations as inner experiences that they must manifest themselves to
human beings as such.
TT: I’m not simply insisting that something is obvious; I’m saying that
you owe us an account that shows how the Ryleans can intelligently
explain their world and themselves without having concepts of inner
episodes. Since we have these concepts, Sellars needs to fill out his
story of the Ryleans who do not. Just what are the Ryleans thinking in
a specific case such as the hunger episode? Just how can a sensory
feeling intrude on one’s consciousness, as you acknowledge, and yet
be conceived by the Ryleans not as anything inner but as a set or
pattern of behaviors ?
WdeV: I have been giving you an account of how the Ryleans can
intelligently explain their world without having concepts of inner
episodes. They cannot do so as well as we can; they miss some things
that seem obvious to us, but that’s only to be expected at their early
stage of development.
How feelings can affect them yet not be conceived of as inner is
easily explained. Given that humans do conceptualize features in the
100 Timm Triplett and Willem deVries

physical world and that they must form concepts of such features if
their conceptualizations are to have any survival value, a sensation
can be transparent to conceptual attention. For example, if a Rylean is
attacked – struck with a weapon, say – he may indeed feel great pain,
but not focus at all on the pain as an inner experience. All his
attention would presumably be focused on the external object that
threatens him – his enemy – or on the physically inner state – the
damage done to his body. The evolutionary point of such sensations,
again, is to allow the organism to respond to features of the physical
environment (including, again, spatially inner states of one’s body)
that are significant to that organism’s survival.
In fact, we have to be trained to focus on our sensations
themselves, as opposed to the physical objects or conditions that our
sensations are meant to direct our attention to. Being able to
appreciate and compare the particular sensory features of fine wines,
for example, is a skill that takes considerable time and effort to
acquire.
TT: The idea of the transparency of sensations is an interesting one, and I
think that the phenomenon does occur. But I think that you’ve
selected your specific examples with some care. Yes, in the case of a
physical object in the form of a weapon-wielding enemy, one focuses
directly on the object that is a threat to one’s life. And in the case of
mild and very subtle sensory differences, such as the wine-tasting
example, we would all agree that sensory discrimination can be a
matter of skill and practice.
But remember, I’m not saying that every instance of sensory
experience is conceptually manifest to a person as an inner sensation.
So let’s stick with the examples I have been working with, though I
would note that even in the enemy-with-the-weapon example, the
Rylean who experienced the attack, assuming he survives it, could
very well focus after the fact on the pain caused, and get angry and/or
frightened all over again.
Hunger is a common enough experience, and unlike the cases you
noted, it is both a strong sensation and one that can occur over an
extended period of time and that often does not require or allow
immediate physical action.
Such a case is I believe difficult to accommodate to your
transparency scenario. Temporally, there is often a long gap in the
case of hunger between behavioral cause and behavioral effect. One is
hungry, but cannot go hunting because of weather or famine or threats
from rivals. So one sits at home and feels hungry. It is very hard to
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 101

see how the feeling could be transparent and how the Rylean would
always think only in terms of some behavior that is not even
occurrently manifest.
Consider the differences between the behavior and its sensory
cause: one is external, the other internal; one publicly observable, the
other not; one is a sequence of macro movements of the body
occurring during one temporal sequence, the other is caused by
neurological episodes within the body having, most frequently, quite
a different temporal sequence (think of the time spent hungry at
home, not engaged in any hunger-resolving behavior). So it cannot be
that the behaviors and the sensation are so similar that the
transparency of the sensation could be explained in terms of a natural
failure to notice the one because of its similarity to the other.
On those grounds alone, the idea that the sensation will always be
entirely transparent – that one just sees through the sensation, as it
were, and does not recognize its presence – seems quite incredible.
WdeV: The Rylean is not confusing the sensation of hunger with a pattern of
behavior. Certainly not because they’re similar, for they are not. She
does not have the concept of a sensation of any kind, so she is
certainly not committing a massive category mistake. The Rylean
may well be obsessively thinking “I have to get some food, I have to
get some food” (not that she realizes that what she is doing is
thinking), she may be made greatly uneasy by her hunger, tempted to
take ever more drastic action in order to obtain food.
TT: I grant that there is no confusion. I am saying that the sensation of
hunger is categorially distinct from the behavior, and yet it is, I would
argue, just as striking a feature of experience as features of the
physical world. So, like those features, the sensation itself will be an
object of conceptual attention.
The inner episode occurring to the Rylean is just as intense, as a
sensation, as any intense hunger we experience, and we all know that
that this can be a very strong sensory feature of our experience. I
guess I’d like to hear from you, while you’re thinking about the
strongest hunger pangs you’ve felt, that you think it is entirely
plausible to hold that, for generations, humans could have had such
pangs and yet thought only of correlated patterns of behavior and
never directed any conceptual attention to those hunger pangs at all.
WdeV: Just as we see via our sensations, but it is wrong to say we see our
sensations, I think that we are at first and primarily aware of the
world via our sensations. And it requires a very significant shift for us
to change our focus to the sensations themselves. Consider, for
102 Timm Triplett and Willem deVries

example, what it takes to be a good representationalist artist: one has

to work incredibly hard to unlearn one’s assumptions about what one
sees. One has to learn to see the world anew. I was terrible at it, try as
I might. So, no, I have no trouble thinking that for generations
humans connected transparently to their world through their
sensations and only later began to develop the ability to focus on
those sensations themselves. I take it to be virtually self-evident that a
transition like this has to occur somewhere in animal development,
for few, if any, other animals are aware of their sensations as such.
TT: Obviously, sensations can occur without conceptions of them,
because animals low on the evolutionary ladder and human infants
have sensations yet presumably possess no concepts. But I am not
arguing here that concepts of sensations must have been the first
concepts to have occurred in the evolution of minds (human or
animal) or in the development of the individual; only that, for any
human with the conceptual capacity to understand subjunctive
conditionals, for example, it is implausible to maintain that she would
never understand that she has inner experiences like hunger. Your
learning-to-paint example again draws attention to subtle aspects of
sensory discrimination, which, I have already noted, is a diversion
from the point at hand. The row you have to hoe requires otherwise
conceptually sophisticated humans to never conceptually notice any
of their sensations: they must always remain transparent.
WdeV: The learning-to-paint example is not a highfalutin’∗ case that draws
attention only to the “subtle aspects of sensory discrimination.” I
think it is striking evidence of the extent to which understanding and
appreciating the structure of the sensory manifold as such is a learned
ability. And since one cannot grasp the nature of the sensory manifold
unless one also has some grasp of its structure, grasping its nature, I
conclude, is also a learned ability.
Have you ever read Julian Jaynes’s book The Origin of
Consciousness in the Breakdown of the Bicameral Mind? (1976) A lot
of it is pretty implausible, even for me, but the first part, where Jaynes
discusses at great length how competent in the world a being without
a consciousness of its inner states can be, is a very valuable antidote
to Cartesian prejudices.

∗Editors’ note: Meaning, in colloquial American English, something like “overly pretentious and
academic.”
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 103

TT: It may be a good exercise in rethinking prejudices, but Jaynes’s

speculations are evidentially very thin.
WdeV: Jaynes’s evidence for the specific thesis that humans gained
self-consciousness around the time of Homer is thin, but the evidence
behind the claim that we are capable of a great deal of cognitive
competence without self-consciousness is very strong, with plenty of
laboratory documentation. And that is the aspect of his project of
interest here.
I take it that the principal way in which we can and do become
aware of our sensations as inner is by becoming aware that the
(objective) structure of the world does not coincide with the
(subjective) structure of our representations (to use the Kantian lingo).
And we usually gain that kind of awareness in cases of
representational failure. Sellars’s proposal is plausible, it seems to me,
because it asks us to think about the Ryleans as not yet being able to
account for such representational failures. Being able to account for
representational failure is an order of magnitude more sophisticated
than having representational success, so it seems clear to me that there
might have been a stage in our development where we had achieved
significant representational power (like the Ryleans) but were not yet
able to account for our representational failures.
TT: Given the differences between kinds of sensations, your claim that
“the principal way in which we can and do become aware of our
sensations as inner is by becoming aware that the (objective) structure
of the world does not coincide with the (subjective) structure of our
representations” is not properly supported by your painting example.
Why could not it be, instead, that sensations like pain give us the clue
to understanding the inner nature of all sensations, including visual
ones?
So, keeping the focus on the hunger example, I want to return to
your answer to my question: In spite of the intrusiveness of the
sensation of hunger, which you acknowledge, in spite of its temporal
and categorial distinctness from patterns of observable behavior, or
from facts such as food deprivation, you insist on the plausibility of a
scenario in which the hunger is always perfectly transparent and never
becomes the object of the Ryleans’ conceptual attention. It is an
answer which I cannot fathom.
WdeV: My assertion that our awareness of our sensations depends on a
mismatch between the objective structure of the world and the
subjective structure of our sensations is not touched by what I guess
you take to be a counterexample. It could well be that pains in fact
104 Timm Triplett and Willem deVries

“give us the clue” to developing a conception of sensations, but how

could they do that? – Because of some form of representational
failure: pains without corresponding bodily damage or bodily damage
without corresponding pain. Pains would lose their “transparency” in
a breakdown in the normal connection between bodily state and pain.
You say that pains could give us the clue to the nature of
sensation. That is, of course, perfectly compatible with Sellars’s story.
Since pains can also be involved in a kind of representational failure,
they will raise questions that a theoretical move like Jones’s could
respond to. But if we come to understand and notice our sensations by
such a theoretical move, then Sellars’s Rylean myth is vindicated, not
impugned. And the only significant alternative I see on the horizon is
the Cartesian insistence that pains are necessarily self-intimating,
which you disavow.
TT: Any full assessment of this issue is going have to closely compare
alternatives. There may be alternatives to the Rylean story Sellars tells
that do not take us all the way back to traditional Cartesian
assumptions about the self-intimating, incorrigible nature of our
thoughts and sensations. That’s one of many interesting questions we
cannot explore here, since our focus is on filling out the all-too-
sketchy story in EPM of the Ryleans’ conceptual world.
So far, though, our discussion of the plausibility of a fuller
explanation of the Ryleans’ conceptual scheme has focused on
sensations rather than thoughts. But because Sellars, rightly, treats
sensations and thoughts as distinct types of mental events, it is
important to discuss the plausibility of the Ryleans’ conception of
thoughts as well. Our discussion of sensations suggests how my own
challenge would go regarding thoughts: Some thoughts, particularly
those connected with powerful emotions, will, like sensations, not be
amenable to moves like your transparency thesis. Consider a Rylean
man who is head-over-heels in love with an unattainable woman. She
rejects his advances utterly – her reasons do not really matter here.
Once rejected, our lustful Rylean never expresses or acts on his love –
the shame of rejection is too great to advertise his fate. It seems
reasonable to think that such a Rylean would both be aware of his
beliefs and desires and of their inner and private nature. He would
not, I would claim, be likely to identify them with linguistic
utterances which he has never even articulated, and he would be
aware that sometimes at least he could prevent others from knowing
his real thoughts and desires.
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 105

WdeV: I think there is a perfectly coherent way to tell such a story consistent
with the Rylean hypothesis. It might read a bit Sartrean, for the
unrequited lover would experience the woman as attractive, as
drawing him to her almost magnetically, while other forces – e.g. the
shame of rejection were he to act on that attraction – pull in the
opposite direction. He feels himself torn: impelled towards the
woman, yet impeded by the certainty of rejection. Clearly, we would
say that he is in emotional turmoil, riven by the conflict between his
desire for the woman and his shame at rejection. But he need not
experience or understand it that way. He may well see himself as torn,
but as torn, not by anything internal, but by the purely external facts
of the situation: the attractiveness of the woman and the shame of
rejection or the scorn of his fellows.
That our emotions are originally transparent to us, that we
experience them as ways in which the world is organized for us, and
that we need to acquire the ability to objectify them and become
aware of them as emotions, has been argued by several philosophers,
not just Sartre. So one cannot assume that conflicting emotions either
must or naturally are experienced as internal states.
TT: This may be the first time in the history of philosophy that Jean-Paul
Sartre has been invoked in the defense of Wilfrid Sellars! Certainly,
such cross-pollination between the traditions is often insightful –
perhaps it can be so here as well.
But still, let’s look at the details. The Sartrean psychology is
suggestive, but it is only that at this point. It is not clear how it can
explain significant elements of the unrequited Rylean’s conceptual
situation.
You say the lustful Rylean sees himself torn only by the external
facts. But it is hard to see how he can make this kind of mistake, since
typically nothing in the external world will correspond at all closely
with his internal situation. Notice that the facts that you cite in
explaining what the lustful Rylean sees himself being torn by are all
unchanging background facts: the attractiveness of the woman, and
societal norms. (By background facts I just mean facts that are
unchanging relative to some changing facts in the foreground.) But
these facts in themselves represent no conflict. If he weren’t in love
with the woman, there would be no threat of rejection, and, of course,
her attractiveness is something that can be noted without inducing
love. Suppose that he accepted all these background facts without
qualm until one day he saw or thought that he saw her glance at him
in a suggestive way. Suddenly he cannot get her out of his mind.
106 Timm Triplett and Willem deVries

Something has decisively and overwhelmingly changed in his life,

and he cannot reasonably attribute this entirely to facts, which have
not changed. His thoughts (and not just his emotions, which your last
remarks reduce the discussion to without warrant) will be constantly
changing in relation to the background facts. One minute he’s
hatching a plot, the next he realizes it would never work, and so on.
He’s about as likely to attribute these features of his life to the
background facts as he is to take a hummingbird for a tree sloth.
WdeV: I think this is just wrong: the facts in terms of which he understands
and explains his behavior are not all unchanging background facts.
That he is enduringly torn he sees as a function of enduring facts, e.g.
the attractiveness of the woman and the danger of rejection; the
shorter-term states and changes in his plans and activities he sees as
functions of shorter-term states and changes in the world, e.g.
changing possibilities of access, new opportunities to curry favor,
possible changes in the strength of her disposition to reject his
advances. The world is so rich that externally-based explanations of
one’s behavior could be quite satisfying to someone who altogether
lacks our notion of intentional explanation.9
TT: That’s just what I am questioning. Your move here is to point out
(correctly) that there are lots of quickly changing external facts
relevant to our Rylean’s plight that are part of his understanding of
the world and to suggest (vaguely) that he will understand his
situation solely in terms of those sorts of facts and never in terms of
any internal facts.
But I do not see how you can tell a plausible story to that effect.
Exactly which (set of) these constantly changing or just-changed
external facts could constitute a reasonable explanation of his
situation? Could it be the woman’s suggestive glance mentioned
above? Unlikely. Whether it was real or imagined, it will not track the
lovelorn Rylean’s obsessive thoughts in the typical case. The glance
occurs in an instant; his obsession lasts for weeks or months. And he
cannot use his speech acts as the external facts by which he could
explain to himself what we understand to be his obsessive thoughts, if
he is unable to confide in anyone. He does a lot of his stewing,
fretting, and plotting in silence.

9 Robert Gordon is currently writing some interesting things emphasizing how seldom we actually
(that is, outside of philosophical discussions) cite our beliefs and desires in explaining our actions.
See also his (2001).
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 107

Could he associate his changing states with the changing

likelihood of rejection or with the changing degrees of loveliness of
the woman? Again unlikely in typical cases. His shame might make
him avoid the woman at all costs. Taken individually or collectively,
these external facts – or our Rylean’s knowledge of them – will very
often radically fail to correspond to the temporal changes in his
thinking. His obsession and its swirl of thoughts is the major feature
of his life, almost constantly present, and there is no external fact to
hook it on to. He will surely understand that it is an internal feature of
himself.
WdeV: You demand explanation of our hypothetical lover’s “situation” and
claim I cannot give one, but that’s very vague. I claim that for any
particular legal explanandum you want, I can produce a plausible
explanation. What I mean by a “legal” explanandum is one that is not
characterized mentalistically. Why does he alternate between boldly
making ready to claim his love for his own and dejectedly hunkering
at home? Because different aspects of his environment evoke in him
different responses. For instance, the (possibly misinterpreted) glance
he caught from his beloved clearly changed his dispositions towards
her. It does not need to “track” anything. It caught his attention and
interest, it began a cascade that changed him profoundly, but it is all
changes in his dispositions.
I also think you’re selling the Sartrean approach to emotions (and
strong beliefs) short. Love and fear are experienced, on the Sartrean
story, as particular ways in which the world is organized. In love, one
finds, at every turn, things that emphasize the attraction of the loved
one, one’s need of the beloved. A particular smell, a glimpse of a
color that reminds one of her hair, virtually anything can be the
occasion of a tear or a love-sick sigh, some sign of one’s striving to
attain the beloved. But any number of things can also evoke a fear-of-
rejection response.
Thus, the Ryleans will have a different way of thinking about such
turmoil. Even without trying to discover constant changes in the
external situation, we can imagine the young swain saying “I’m in
turmoil. At one and the same time I’m ready to run to her and
proclaim my love, and yet I could not bear the burden of another
rejection. I am not sleeping, for in the night it is often as if she is
before me, and then I am drawn to her, heart aflutter, while
simultaneously unable to move, lest she scorn me yet again. I cannot
do my work, I cannot sit still, I cannot bring myself to peace. Oy
Vey!” What you see as conflicting thoughts, the Rylean understands
108 Timm Triplett and Willem deVries

as conflicting dispositions. What you see as rapidly changing plans,

attempts to think of a way to satisfy his desires, the Rylean
understands as changing dispositions to act supplanting each other,
none of them holding forth the promise of success. There is simply
nothing implausible about this.
TT: Perhaps not. But look at the position you have arrived at here. You
began by saying that the Rylean will see himself as torn “not by
anything internal, but by the purely external facts of the situation” like
the woman’s attractiveness or the proscription against adultery. But
you end by maintaining that the Rylean understands his conflicted
situation as changing dispositions to act. You have significantly
switched your ground here, for these dispositions are not external
facts – not part of the furniture of the world external to the Rylean.
They are characteristics of the Rylean himself. And, unless he is
insane or delusional, it is reasonable to think that the Rylean will
understand this. You yourself have him noting that it is “as if she is
before me.” He is not so deluded as to think she is really there.
Similarly for his disposition to make certain utterances: “It is as if I
were proclaiming my love to her,” he says (to himself). But he knows
all too well that this as-if-proclaiming does not occur at his beloved’s
feet, nor even as a spoken utterance in his own home, lest the woman
of the house boot him out.
And if he understands these imagined but not acted-upon
scenarios as silent, private features of himself, then he is already in
effect a Jonesian. Jones says that something is going on within him
that is like speech and yet is clearly not publicly observable behavior.
Our Rylean has come to the same conclusion. But if you have to grant
that the ordinary Rylean will have this conception of his situation, you
are depriving the Jones story – and thus Sellars’s Rylean hypothesis –
of its force. The point of the hypothesis was that Ryleans could live in
ignorance of inner thoughts and sensations for generations until Jones
arrived with his genius-level insight. In fact, it looks hard to maintain
that the typical Rylean living the typical life would not already
understand what Jones understands.
WdeV: I do not believe I have switched my ground at all. We’ve always
agreed that the Ryleans had a conception of dispositions. Sometimes,
especially when the underlying disposition is stable, the explanation
of an event is best accomplished by reference to the new conditions
that evoke the manifestation of the disposition. Sometimes, though, an
explanation is accomplished by reference to the disposition,
especially when the dispositions of the objects involved change. The
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 109

fact that a machine part does not break under a certain force may be
explained by saying that the steel is tempered. When the tempering
process was first developed, no one knew exactly why it worked,
what it did to the microstructure of the metal; they just knew that
tempered steel had different dispositional properties (tensile strength,
etc.) from untempered steel. There is no reason to deny the Ryleans
the full use of their acknowledged conceptual resources.
From this point of view, Jones does not just exploit his knowledge
of human dispositions to explain the behavior of his fellows, he offers
a theory about the structure underlying our dispositions, what’s going
on inside us that accounts for why our dispositions work as they do.
TT: So you say you can produce a plausible explanation, but you haven’t
offered a specific explanation for the situation I described, where the
Rylean understands all too well that his dispositions do not manifest
themselves in terms of his actual behavior. He knows that his loving
conversations, the visions of his beloved, are not real. Given that his
thoughts and imaginings are very intense, constantly changing and,
we are supposing, the most important thing in his life, it seems
practically inevitable that his attention will turn to them, and he will
understand that there are active, constantly changing nondispositional
features of his inner life. Indeed, he would understand that these are
the causes of his conflicting dispositions, now to approach, now to
retreat, and so on.
WdeV: I think you are projecting too much of who we are and how we
conceive matters onto the Ryleans. You seem to think that the Rylean
will, on any plausible account, have to understand his imagined
conversations as internal states of himself. But even actual human
cultures have failed to understand what we know to be internal mental
states in that way. Dreams, for example, have been taken in some
cultures to be travels to other worlds. There is no reason why the
Rylean could not suppose that his conversations with the object of his
affection actually do take place in a different and, for him, much
happier world.
TT: You are supposing, in other words, that he is deluded.
WdeV: In the sense of insane, no; in the sense of systematically mistaken,
sure. But Sellars builds the possibility of such systematic error into
his account, and there is nothing unrealistic in general about
supposing whole human cultures can get things systematically wrong,
even in their broad conceptions of the nature of the world they live in.
Indeed, there are plenty of examples in the actual history of human
cultures.
110 Timm Triplett and Willem deVries

TT: You make a good point. I do think that the case I have been imagining
is different from the dream case. In a dream, one is presented with a
seemingly self-contained world and oneself in it. One can see how a
culture might take that to be a separate but equally real world. It is
much harder to imagine how, in waking life, invented conversations,
deliberate plotting, conjured visual images of the object of affection
could be taken either to be a separate world or an externalized aspect
of this world. There is no presentation of a self-contained world, and
there are too many close-to-hand features of the immediate
environment that tell against the reality of the imagined objects of
thought (or the very acts of thinking). So it does not seem a
reasonable possibility to me. And yet I understand that this could
simply be a failure of imagination on my part. While I think
arguments from implausibility have a place, they also have their
limits, and one of them is the danger of just such failures of
imagination. At this level of speculation, it is difficult to argue either
side of the case regarding what is a plausible conceptual world for
human beings. I think we’ve arrived at a draw here.
I do think, though, that this applies only to Sellars’s account of the
Ryleans’ conceptions with respect to thoughts, and not to sensations.
The reason that I cannot press my case for the implausibility of the
Ryleans failing to have the concept of thinking is because, as you
point out, it is possible to give a dispositional analysis of thinking
when such thinking does not manifest itself in actual behavior. But
Ryle himself was dissatisfied with his attempt to give a dispositional
analysis of sensations, and for a good reason. While I have to concede
that it is hard to argue the plausibility or implausibility of the Rylean
projecting his imagined conversation onto the actual world, there is
no analogue in the case of the sensation of hunger. Since thoughts are
about situations and objects, it is always possible to imagine the
Rylean projecting these thoughts as actual features of the external
world. But hunger is not about anything, and so your move is blocked
in that case.
WdeV: I’d noticed that your arguments tended to rely on the phenomenology
of thought to reveal thoughts to us, and was going to remark on that.
Classically, thought itself does not have a phenomenology, and that’s
a major reason why a functional-dispositional analysis of thought has
been convincing. The Ryleans have the concepts crucial to
dispositional analyses, so they seem to be able to do a more thorough
job in explaining the kinds of things we explain by reference to
thoughts than of those things we explain by reference to sensory or
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 111

qualitative states. I wonder whether, if your argument here worked, it

wouldn’t really get you the categories of imagination and verbal
imagery, rather than that of thought proper.
TT: It is true that I have been assuming that the Ryleans have a
phenomenology associated with their thinking. (I take it by
‘phenomenology’ here you mean, for example, the stream of silent
words that courses through many people’s minds as they go through a
typical day – see Joyce’s Ulysses – or the visual images a person
refers to in saying, for example, “I can vividly recall the expression
on her face even though it was years ago.”) But I’m not “relying on
the phenomenology of thought to reveal thoughts to us” if you mean
by that making the argument that the very existence of the
phenomenology entails the awareness of it as an inner experience.
Sensations are inner experiences but I haven’t presupposed that they
are, simply by virtue of that fact, conceptualized as such. I gave
Sellars the benefit of the doubt that it was possible for sensations to
exist without being conceptualized as what they are, and I’m making
the same assumption here – there is some phenomenology, but by
hypothesis it is not conceptualized as such. I then have tried to show
that reflection on the details of what such a conceptual world would
look like lands us in some implausible scenarios.
If you are suggesting that the Ryleans have no phenomenological
accompaniment to the type of thinking under discussion here, I think
that you’ll find that a very hard row to hoe. If, as has been known to
happen with human beings, one person’s sexual attraction to another
is primarily physical, then it is hard to imagine how the one person
could think about the other in the other’s absence without visual
imagery of that person’s body and features. Or suppose our hapless
Rylean plots how he might contrive to place himself where he will
cross paths with the woman. He would rely on visual memory of the
layout of his village and its buildings and of the pattern of her
comings and goings among them, as he considers how he might
engineer an encounter. It is reasonable to assume that specific visual
imagery based on this memory would accompany his plotting.
If my argument only shows that the Ryleans have concepts of
visual and verbal imagery rather that concepts of “thought proper,”
that does not weaken my case at all. I myself took verbal imagery to
be an example of thought proper and did not encounter arguments
against this until I began studying Sellars. I imagine that most people
who are not philosophers and cognitive scientists take their verbal and
visual imagery to be instances of thinking. If my arguments show that
112 Timm Triplett and Willem deVries

Ryleans will most likely conceive of themselves as having inner,

private verbal and pictorial images, I have made my case: they have a
concept of thinking that is pretty much like that shared by most actual
persons.
WdeV: That’s a conception of thought like the conception that everything
yellow and sparkly is gold. These arguments return to the theme that
the qualitative aspects of mind – what you’d tend to call the “What it
is like” aspects – are somehow too obvious for someone to have them
and not conceptualize them. I do think that most people recognize that
there need be nothing that it is like to have a thought, even though
they often, and sloppily, equate thinking with internal imagery. So I
do not think your arguments show that there is anything implausible
about Sellars’s story concerning thoughts.
Let me try another tack that might reveal something important
about our disagreement. You’re willing to grant that mental states are
not self-intimating, that it takes some (though minimal) inferential
work or theoretical creativity to develop a conception of subjectively
inner states, and that the spur to such work would be the need to
explain and understand our behavior. But Sellars’s own story seems
implausible to you because it makes this move too difficult: you think
that once someone has the conceptual artillery to pose the explanatory
question, the conception of subjectively inner states would propose
itself virtually immediately. Now, I notice that in all your examples,
you are concerned with a subject’s understanding/explaining his own
situation. Is the self-reference here essential? Is your complaint
ultimately that Sellars’s myth seems implausible to you because he
has not taken account of the first-person perspective, and that when a
first-person point of view is countenanced, things look entirely
different? If so, we’ve been navigating around another important
issue: the status of the first person.
TT: I disagree with the claim that most people have taken their concept of
thinking as far as you assume they have, but even if they did, I believe
that any Rylean who acquired the merely qualitative conception of
thought based on verbal and visual imagery would have abandoned
behaviorism and already have adopted the key ingredients of Jones’s
insight.
Regarding your question about self-reference as essential to my
argument, ‘essential’ is used in many ways. But if the question is
whether the first-person perspective is an important part of the story
to be told, I would say “yes.” A story about humans that completely
ignores or dismisses the first-person perspective is, in my view, one
 Is Sellars’s Rylean Hypothesis Plausible? A Dialogue 113

that simply fails to account for all that needs to be accounted for.
According to many philosophers, this effectively means that one has
to abandon a physicalist or broadly naturalist account of the mind. I
disagree. But any such account must provide an adequate explanation
for how the world is experienced from the perspective of the subject,
not just from the objective point of view. Such a perspective seems
notably lacking in Sellars’s myth of Jones. And when one tries to fold
that perspective into the myth, it starts looking deeply implausible in
important respects.
WdeV: I see now better than I did before both why you find Sellars’s story
implausible and why you think it is a philosophically troubling, and
not a merely empirical, implausibility. I do not recall any passages
where Sellars explicitly considers the first-person perspective, so
whether he’s got an adequate story about it that could assuage your
concerns is something we’d have to work out. And it wouldn’t be an
easy piece of work, so I think that’s another discussion.

Timm Triplett
Department of Philosophy
University of New Hampshire
Durham, NH 03824
United States of America
e-mail: [email protected]

Willem deVries
Department of Philosophy
University of New Hampshire
Durham, NH 03824
United States of America
e-mail: [email protected]

Visiting Fellow, Philosophy Programme

University of London
London WC1E 7HU
United Kingdom
114 Timm Triplett and Willem deVries

REFERENCES

Chisholm, R.M. and W. Sellars (1957). Intentionality and the Mental. In: H. Feigl, M. Scriven and
G. Maxwell (eds.), Minnesota Studies in the Philosophy of Science, vol. 2, pp. 507-539.
Minneapolis: University of Minnesota Press.
deVries, W.A. and T. Triplett (2000). Knowledge, Mind and the Given: Reading Sellars’s
‘Empiricism and the Philosophy of Mind’. Indianapolis: Hackett.
Gordon, R.M. (2000). Sellars’s Ryleans Revisited. Protosociology 14, 102-114.
Jackson, F. (1982). Epiphenomenal Qualia. Philosophical Quarterly 32, 127-136.
Jaynes, J. (1976). The Origin of Consciousness in the Breakdown of the Bicameral Mind. New
York: Houghton Mifflin.
Lehrer, K. and D.G. Stern (2000). The “Dénouement” of “Empiricism and the Philosophy of
Mind.” History of Philosophy Quarterly 17, 201-216.
Marras, A. (1973a). Sellars on Thought and Language. Nous 7, 152-163.
Marras, A. (1973b). On Sellars’s Linguistic Theory of Conceptual Activity. Canadian Journal of
Philosophy 2, 471-483.
Marras, A. (1977). The Behaviourist Foundation of Sellars’s Semantics. Dialogue 16, 664-675.
Rosenberg, J.F. (2000). Perception vs. Inner Sense: A Problem about Direct Awareness.
Philosophical Studies 101, 143-160.
Ryle, G. (1949). The Concept of Mind. New York: Barnes and Noble.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Young, J. (1973). Intentionality. Review of Metaphysics 26, 696-722.
David Forman

LEARNING AND THE NECESSITY OF

NON-CONCEPTUAL CONTENT IN SELLARS’S
“EMPIRICISM AND THE PHILOSOPHY OF MIND”

ABSTRACT. For Sellars, the possibility of empirical knowledge presupposes the

existence of “sense impressions” in the perceiver, i.e., non-conceptual states of
perceptual consciousness. But this role for sense impressions does not implicate Sellars’
account in the Myth of the Given: sense impressions do not stand in a justificatory
relation to instances of perceptual knowledge; their existence is rather a condition for
the possibility of the acquisition of empirical concepts. Sellars suggests that learning
empirical concepts presupposes that we can remember certain past facts that we could
not conceptualize at the time they obtained. And such memory presupposes, in turn, the
existence of certain (past) non-conceptual sensory states that can be conceptualized.

Parents never teach children language without the children themselves inventing it
simultaneously. The parents simply bring distinctions between things to the attention of
the child by means of certain signifying terms; and so they do not, as it were, put the use
of reason into them, but rather facilitate and promote it for them through language.
J.G. Herder (1985, I.2, p. 727)

Introduction

In Mind and World, John McDowell argues that traditional attempts to explain
our cognitive relation to the world result in an oscillation between two
opposing epistemological pitfalls. Following Sellars, he calls the first pitfall
the “Myth of the Given”: the Myth that cognitive episodes can find
justification outside the realm of the conceptual. According to the most
common version of the Myth, our beliefs about the world are justified not
solely by other beliefs, but ultimately by non-conceptual experiences forming
the interface between mind and world. McDowell claims that Sellars’s
celebrated attack on this Myth in his essay “Empiricism and the Philosophy of
Mind” (1997, henceforth: EPM) leads him to “renounce empiricism” and
instead embrace the opposing epistemological pitfall: “frictionless

In M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 115-145.
Amsterdam/New York: Rodopi 2006.
116 David Forman

coherentism.”1 By renouncing empiricism in favor of coherentism, Sellars

forfeits the ability to account for the “constraint” on our knowledge exerted by
the world itself. Only such a constraint can ensure that our perceptual
knowledge answers to the world around us, that our perceptual knowledge has
authentic empirical content.2
In his more recent Woodbridge Lectures (1998b), McDowell recognizes
Sellars’s attempt to locate the world-guidedness of empirical knowledge and
discusses it in terms of the “transcendental” aspect of Sellars’s philosophy.
Specifically, McDowell discusses how, for Sellars, non-conceptual states of
sensory consciousness – the sensations or impressions associated with our five
senses – are “postulated on general epistemological or, as Kant would say,
transcendental grounds” (1967, I §22/p. 40): their existence is a necessary
condition for the possibility of conceptual perceptual experience having
objective purport (1998b, pp. 444ff.). To say that impressions are postulated on
transcendental grounds is to deny, first of all, that they are discovered (through
induction or direct introspection). In this respect, Sellars treats impressions
much like he treats entities introduced in the course of scientific explanation:
positrons, for example, are not simply discovered by the scientist, but are
rather, like sense impressions, introduced as theoretical entities (see 1997,
§§43-45/pp. 83-88). The postulation of sense impressions is specifically
“transcendental” since impressions serve in an explanation of the possibility of
knowledge rather than in a scientific explanation.
McDowell’s essential claim in the Woodbridge Lectures is that the
Sellarsian account of the logic of our most basic conceptual awareness of the
world (e.g. episodes of seeing) would be sufficient to secure the world-
guidedness of perceptual knowledge if it were combined with a relational
theory of meaning instead of Sellars’s own inferentialist semantics (the view
that propositions, and ultimately concepts, derive their meaning solely from
inferential relations among propositions). By failing to see that veridical cases
of seeing link up with the real world all by themselves, Sellars is forced to
contrive a “transcendental” account of the relation of perceptual knowledge to

1 McDowell engages more directly with Donald Davidson’s attack on the “dualism of scheme and
content” than with Sellars’s attack on the “Myth of the Given” but he considers the two more or
less interchangeable (see 1994, pp. xv-xviii).
2 Sellars himself envisions the threat of such an oscillation and finds the cause of the threat in the

assumption that any non-inferential knowledge that could serve as foundation for knowledge
would have to consist of self-authenticating perceptual episodes: “One seems forced to choose
between the picture of an elephant which rests on a tortoise (What supports the tortoise?) and the
picture of a great Hegelian serpent of knowledge with its tail in its mouth (Where does it begin?)”
(1997, §38/p. 78-79).
 Learning and the Necessity of Non-Conceptual Content in Sellars 117

the “real order” that involves the mediation of non-conceptual states of

consciousness. The claim is that since Sellars’s inferentialist semantics blinds
him to the possibility that conceptual episodes have a relation to objects in the
real world just by being the conceptual episodes that they are, the link with the
world must be effected by non-conceptual states of consciousness. This view is
unsatisfactory, according to McDowell, because it disconnects our conceptual
experiences from the world. In order to explain the possibility of a constraint
that could reconnect our conceptual experiences to the world, Sellars is
ultimately forced back into a version of the Myth of the Given. But McDowell
also emphasizes that Sellars is not committed to this unsatisfactory position by
his basic insights about the logic of our most basic perceptual episodes (e.g.
“ostensible seeings”). McDowell wishes rather to appropriate these insights
and reject only the inferentialism that keeps the world outside of view.
According to McDowell, however, the “transcendental” function of
non-conceptual states of consciousness was not a part of Sellars’s original
doctrine in EPM – even though his inferentialism would require the
transcendental story to secure world-guidedness for empirical knowledge.
McDowell’s view is that Sellars came to see the necessity of the transcendental
story only after EPM, specifically, in the later work Science and Metaphysics.
Consequently, McDowell does not think that the non-conceptual states
introduced in EPM are part of any attempt to secure world-guidedness for our
empirical concepts. Instead, he understands the explanatory function of
impressions or sensations in that work as merely “scientific”; the account of
impressions in EPM is gratuitous, the result of a desire to offer a “scientific
style” of explanation. Hence they cannot rescue the account from the
“frictionless coherentism” implied by his rejection of the Myth of the Given.
Indeed, he says of the account in EPM: “The sensations look like idle wheels”
(1998b, p. 444).
In this essay, I will argue that McDowell does not adequately consider
Sellars’s motivation for introducing sense impressions in EPM. An analysis of
the logic of perceptual knowledge reveals a “transcendental role” for sensory
impressions already in EPM. Even in that earlier work, impressions play the
role of securing the objective purport of perceptual knowledge. However, the
transcendental role for impressions in EPM is not the same one that McDowell
finds in Science and Metaphysics. The problem that prompts Sellars to posit
sense impressions in EPM is not that he lacks the resources to say that
conceptual episodes have a relation to their objects just by virtue of being the
conceptual episodes they are. The problem is rather the problematic status of
our acquisition of the conceptual abilities necessary to have such conceptual
episodes. The existence of impressions emerges in EPM as a necessary
condition for the possibility of concept acquisition (or language learning). Like
118 David Forman

the transcendental role for impressions that McDowell finds in Science and
Metaphysics, the role of impressions in concept acquisition is to link mind and
world. In this case, the link between mind and world is necessary because only
a being that can have impressions brought about by the world around it can
come to have knowledge of that world; for only such a being could learn to use
empirical concepts. However, such a link is necessitated not by Sellars’s
distinctive inferentialism, but rather by three features of his account adopted
McDowell and others: (1) “psychological nominalism”: the view that all
cognitive awareness, even the most primitive, presupposes conceptual abilities;
(2) an opposition to innatism and thus an insistence that all these conceptual
abilities must be acquired in the process of learning a language; and (3) an
opposition to “bald naturalism” and thus an insistence on a qualitative
distinction between these conceptual abilities and other sorts of discrimination
behavior. Consequently, if Sellars is right that the process of concept
acquisition presupposes the existence of sense impressions, then rejecting
inferentialism does not obviate the need for impressions in an account of
perceptual knowledge.
The first section of this essay, by clarifying the nature of non-inferential
perceptual knowledge in Sellars’s EPM, sets the stage for a discussion of the
role impressions play in the acquisition of empirical concepts. Sellars makes
clear that even the non-inferential knowledge gained in experience always
presupposes other knowledge, and I explain the nature of this holism
concerning empirical knowledge. After pointing out the difference between
knowledge of sentence tokens of non-inferential perceptual claims and
knowledge of sentence types of perceptual claims, I go on to argue that Sellars
claims (i) that knowledge of the sentence types one is using must be internal to
the observer for a particular non-inferential perceptual claim to be a candidate
for knowledge (to be true or false) and (ii) that it is only knowledge of
sentence types that must remain internal to the observer: one does not need to
know that one’s own tokening is true in order for one’s tokening to be an
instance of perceptual knowledge. To put the same point differently: although
knowledge of the truth of the perceptual claims associated with one’s
experiences need not be internal to the observer for his perceptions to count as
instances of knowledge (to that extent Sellars can be read as an externalist), the
knowledge necessary for understanding the perceptual claims associated with
one’s experiences must, indeed, be internal to the perceiver. This is the
knowledge that is presupposed even by non-inferential, perceptual knowledge.
In the second section, I proceed to point out that even this limited
internalism creates a potential problem about how one could come to acquire
the knowledge necessary for such an understanding. This is a problem that
Sellars fully acknowledges and discusses in what I call his “regress of
 Learning and the Necessity of Non-Conceptual Content in Sellars 119

learning” argument. I suggest that Sellars sees non-conceptual states of

sensory consciousness, or impressions, as a necessary part of a solution to this
problem. A pre-conceptual, causal interaction with the world ensures that our
subsequent cognitive, conceptual awareness answers to the nature of the world
itself. Impressions can in this way be seen as playing the (“transcendental”)
role of linking us to the world, of securing the “world-guidedness” necessary
for our cognitions to contain genuine empirical content.

1. The Understanding and Veridicality Conditions on Perceptual

Knowledge

McDowell is correct to note that in EPM the difference between an ostensible

seeing and an actual seeing is a difference of facts, not a difference of
justifications at one’s disposal: one’s experience can be described as a case of
seeing even if one does not have a good reason for thinking it is more than an
ostensible seeing.
Sellars tells us that to say that Jones sees a green tree is to attribute to
Jones’s experience a certain claim, and then to endorse his experience’s claim
(as being true) (1997, §16/p. 39). Thus, were Jones to deny that he sees a tree,
it can still be the case that he sees the tree, as long as two conditions obtain: (1)
his experience is indistinguishable from an experience of seeing a tree (the
strongest evidence for this would be if he admitted that there looked to be a
tree there); and (2) there is, as a matter of fact, a tree there (that is related to
him in the proper way).3 It is only this fact which allows Sellars to offer the
following scenario: “If I make at one time the report ‘X looks to be green’ –
which is not only a report, but the withholding of an endorsement – I may
later, when the original reasons for withholding endorsement have been
rebutted, endorse the original claim by saying ‘I saw that it was green, though
at the time I was only sure that it looked green’” (1997, §16/pp. 40-41). As far
as “seeings” are concerned, Sellars can be read as an externalist about
justification.4

3 Of course, condition (2) needs to be given the proper gloss to avoid the accusation of looking at
perceptual knowledge from “sideways on.” We can start such a gloss by noting that we can assess
whether condition (2) is fulfilled only in the same way that we can assess whether there is, in fact,
a tree there (and an observer related to the tree in the proper way). See note 12.
4 Thus: “Both existential and qualitative lookings are experiences that would be seeings if their

propositional contents were true” (§22, p. 51). Hence Sellars says of an ostensible seeing of
something red: “The experience is intrinsically like that of seeing an object to be red in the sense
120 David Forman

Sellars’s thematic treatment of perceptual knowledge in the chapter “Does

Empirical Knowledge Have a Foundation?” in EPM (§§32-38/pp. 68-79) is a
continuation of this earlier account of “seeing”: in this thematic treatment, he
explains in more detail how cases of perceptual knowledge are perceptual
claims that can be endorsed. His account does not require, however, that the
perceiver be able to justify the truth of his own perceptual claims.
Sellars discusses the conceptual aspect of perceptual knowledge in terms of
how an observation report, or “Konstatierung,”5 differs from a mere reliable
response. At times he seems to suggest that the knowing observer must know
that the present conditions are those in which his perceptions would be true.6
But, as McDowell insists, this does not seem to be essential to his view.7

that if certain additional conditions were realized the experience would in fact be one in which
[the subject] sees an object to be red. Among these conditions are (a) that the object be in fact red;
(b) that the object be appropriately responsible for the experience” (1981, I §70/p. 16). But see
note 6, below.
5 Sellars apparently borrows this term from Schlick. Literally, it means “noticing” or even

“seeing.” In Schlick’s account, Konstatierungen are the basic conceptual episodes of observation
that serve to confirm scientific theories; they constitute the “unshakable point of contact between
knowledge and reality” (1959, p. 226; where Konstatierung is translated as “confirmation”).
Brandom says that Sellars gets the term from Carnap, but unfortunately gives no reference (see
1994, p. 215). Brandom may well be right, but Carnap’s preferred term seems to be ‘Protokolsatz’,
whereas Schlick elevates the ordinary word ‘Konstatierung’ to a term of art (see Schlick’s 1986,
esp. pp. 234ff.; see also Louise Röska-Hardy 1998).
6 Two footnotes added to section §22 in the 1963 edition of EPM claim that the subject who has

perceptual knowledge must know that the circumstances he is in are normal. This claim may seem
to contradict my suggestion that Sellars is an externalist about the justification of perceptual
knowledge. (McDowell’s view seems to be that these footnotes suggest a rejection by Sellars of
the earlier externalist implications of his own account.) But the claim might be simply an attempt
to rule out cases where what looks to the subject to be the case merely happens to be the case –
due to certain non-standard circumstances (this can happen in certain kinds of optical illusion).
Another passage, present in the original edition of the text, may also seem to advocate an
internalism of justification. Sellars writes concerning standard conditions of perception: “Not only
must the conditions be of a sort that is appropriate for determining the color of an object by
looking, the subject must know that conditions of this sort are appropriate” (1997, §19/pp. 44-45).
But Sellars’s meaning here is ambiguous. The context is a discussion about the requirements for
having the concept of green (for knowing “what it is for something to be green”), not a discussion
about the requirements for a particular experience to be justified and thus to count as knowledge.
Sellars may simply be pointing out that a knowledge of the difference between standard and non-
standard conditions is necessary for an understanding of concepts like the concept of being green.
7 Although Sellars says that our concept of green requires knowing what standard conditions are,

he clearly does not think that such knowledge is a necessary condition for perceptual knowledge
per se. In a footnote added to the 1963 edition of EPM, Sellars says that his argument does not
rule out the existence of “a rudimentary concept of ‘green’ which can be learned without learning
the logical space of looks talks . . . ” (1997, §19/pp. 44-45). Such a rudimentary concept thus does
 Learning and the Necessity of Non-Conceptual Content in Sellars 121

Rather, what is essential is that the perceiver recognize that, “in the economy
of the language,” his non-inferential claim is a token of a type used to infer the
fact claimed. Sellars writes:
In other words, for a Konstatierung “This is green” to “express observational
knowledge” not only [1] must it be a symptom or sign of the presence of a
green object in standard conditions, but [2] the perceiver must know that tokens
of “This is green” are symptoms of the presence of green objects. (1997,
§35/p. 75; my emphasis)
It is important to notice that Sellars does not say (here anyway) that the
perceiver must know that his token indicates the presence of a green object in
standard conditions. All that Sellars says is necessary is
(1) that the tokening express the fact that occasioned it; and
(2) that the perceiver know that, in his community, tokens of “This is green”
reliably indicate the presence of green objects; that is, that the perceiver
understand the function of the sentence “This is green.”
I will call these, respectively, (1) the veridicality condition and (2) the
understanding condition on non-inferential perceptual knowledge.8

not require knowing anything about “standard conditions,” for “standard conditions” are defined
in terms of the ‘looks talk’ (“‘standard conditions’ means the conditions in which things look what
they are” (§18/pp. 43-44). Nor should we think that the rudimentary concept is a mere ability for
differential response, an ability that a photocell could have; for Sellars continues his note: “The
essential point is that even to have the rudimentary concept of green presupposes having a battery
other concepts.” Hence Sellars remarks: “The grownup’s language expresses concepts which
distinguish between cases in which the object is really red and cases in which it merely looks red,
and between normal and abnormal circumstances . . . Junior does not yet conceptualize his own
experience in these terms. Yet he does make use of some related, if more primitive, concepts
(1981, I §17ff./p. 6ff.; see also §25/p. 8). (Sellars’s account leaves open the possibility that our
own linguistic ancestors could have passed along such rudimentary concepts – although it is pretty
clear why they did not.) Accordingly, the story of John the haberdasher is the story of the
enrichment of his color concepts (see 1981, §§30-35/pp. 8ff.), but not the story of the initial
acquisition of his color concepts; it is a story of a kind of mental awakening, but not the story of
how he is first able to perceive objects as colored. In this paper, I am concerned with what is
common to all perceptual experiences, and so the question of standard conditions is not central for
me. My paper will, however, be concerned with the fact that John, at some point prior to his
employment in the tie shop, must have made the transition from a merely natural being to a being
who possesses rudimentary concepts.
8 One might wish to add a third condition on non-inferential perceptual knowledge, namely that

the claim be made in “standard conditions.” But if “This is green” expresses a fact in non-standard
conditions, then “This is green” would be the conclusion of an inference and not a candidate for
non-inferential knowledge (not a Konstatierung) (see Brandom’s “Study Guide,” p. 138; see 1997,
§14/p. 48). Moreover, as discussed in note 7 above, there could be a system of color concepts
(different from ours) in which standard conditions play no role.
122 David Forman

Now, a talking photocell could produce the sound “This is green” as a

symptom of the presence of a green object in standard conditions. The
photocell’s utterances thus fulfill the veridicality condition. Hence it is clearly
the understanding condition that distinguishes perceptual knowledge from the
mere reliable responses of thermometers and photocells. For a tokening of
“This is green” to be an expression of perceptual knowledge (as opposed to
being a mere reliable response), the perceiver must know that indicating the
presence of green objects is the function of the sentence type “This is green”:
knowing this means mastering the “inferential articulation”9 of the claim,
knowing in what situations one may infer “This is green” and what one may
infer from “This is green.”10
Clearly, Sellars wishes to say that this knowledge expressed by the
understanding condition is possessed by all English-speakers (and that
analogous knowledge is possessed by speakers of all languages). Accordingly,
Sellars’s primary objective is to show how our reliable responses differ from
those of a thermometer or a photocell and not to show how we might infer our
perceptual knowledge claims were they brought into doubt. Sellars’s reflection
on perceptual knowledge here is thus not a reflection on what makes claims
veridical or non-veridical, but rather on what makes them epistemic: i.e., what
makes them “concept-involving” (as McDowell aptly describes the Sellarsian
moniker “epistemic” in his Woodbridge Lectures (1998b)) and thus bearers of
intentionality.
This emphasis on epistemic character, and thus on the understanding
condition rather than on the veridicality condition, holds even in his discussion
of the authority of perceptual sentences. The understanding condition on
knowledge demands that we know both (a) which sentences imply a
conclusion of the same type as my current tokening and (b) which inferences
would be warranted by a veridical tokening of that type. This second piece of
knowledge is equivalent to recognizing the authority of the sentence TYPE. It is
important that this kind of knowledge be distinguished from knowing that
one’s report TOKEN has authority, for this latter kind of knowledge implies
not only that one recognizes the authority of the sentence type, but also that
one knows that one’s tokening expresses a fact and thus may be endorsed. We

9 Brandom’s term. See his “Study Guide” to Sellars (1997, p. 147) and his (1994).
10 See Sellars’s remark in a later article: “One is not a full-fledged member of the linguistic
community until one not only conforms to linguistic ought-to-be’s (and may-be’s) by exhibiting
the required uniformities, but grasps these ought-to-be’s and may-be’s themselves (i.e., knows the
rules of the language . . . To be a language user is to conceive oneself as an agent subject to rules”
(1969, p. 513).
 Learning and the Necessity of Non-Conceptual Content in Sellars 123

need to distinguish these two kinds of knowledge in order to maintain a

distinction between the understanding and veridicality conditions on non-
inferential knowledge.
Sellars explains the requirement that the perceiver recognize the authority
of his own report in terms of understanding the normative purport of that
report type, in terms of that report type’s inferential articulation:
[I]f the authority of the report “This is green” lies in the fact that the existence
of green items appropriately related to the perceiver can be inferred from the
occurrence of such reports, it follows that only a person who is able to draw this
inference, and therefore who has not only the concept green but also the
concept of uttering “This is green” . . . 11 could be in a position to token “This is
green” in recognition of its authority. (1997, §35/pp. 74-75; my emphasis)
I want to emphasize that Sellars is not saying here (here anyway) that the
perceiver must be able to vindicate his report by justifying it inferentially: he is
not saying that the perceiver must be able to make an inference to the report he
is making from other facts. Sellars is saying rather that the perceiver must be
able to make the inference from reports of that type (of which his own report is
one token) to a fact. If he cannot make this inference, he could scarcely be said
to understand what he is doing in making the report; if he cannot infer the
existence of green items from such reports when he hears them, then he does
not have an understanding of the report type “This is green.” For a report to
count as knowledge (as opposed to mere noise, not as opposed to something
false), the perceiver must “know that overt verbal episodes of this kind are
reliable indicators of the existence, suitably related to the speaker, of green
objects” (1997, §36/p. 75; my emphasis; see note 21 below). More generally
speaking, the requirement that the perceiver recognize the authority of the
reports he makes concerns the question whether the perceiver possesses the
requisite conceptual abilities and not the question whether he has information
available to him that could justify a perceptual claim in the face of doubt.
Recognizing the authority of one’s report, then, involves knowing
something about the report type as tokened by any speaker and not something
peculiar about one’s own tokening of the that type. The understanding
condition on knowledge requires recognizing the authority of types,

11I omit here a parenthetical remark in which Sellars says that perceiver must also possess “the
concept of certain conditions of perception, those which would correctly be called ‘standard
conditions’.” This remark shows that Sellars is talking about a “rich” and not a “rudimentary”
concept of green and of uttering “This is green.” But even here, Sellars is not suggesting that the
subject knows whether he is actually “in standard conditions” in this particular case. See note 7,
above.
124 David Forman

recognizing which report types are “reliable indicators” of what. (Notice that
while it makes sense for a report type to be a reliable indicator of something, it
does not make sense for a report token to be a reliable indicator of something.
Smoke may be a reliable indicator of fire, but a particular puff of smoke
cannot be said to be – or not to be – a reliable indicator of fire: it either
indicates fire or it does not.)
Indeed, this “recognition” of the authority of perceptual report types is only
the second of two “hurdles” that Sellars says must be overcome before we can
move from a merely causal account of reporting to one that adequately
explains perceptual knowledge. The first hurdle is to explain what it is for a
report token to have authority. The first hurdle is thus equivalent to the
veridicality condition: “Clearly, on this account, the only thing that can
remotely be supposed to constitute such authority is the fact that one can infer
the presence of a green object from the fact that someone makes this report”
(1997, §35/p. 74; my emphasis). To say that someone “can” make this
inference presumably means that one would be justified in doing so, i.e., that
the report expresses a fact. It is worth repeating that Sellars offers a different
sort of explanation for each of the two “hurdles,” the first in terms of report
tokens and the second in terms of report types. There is therefore no reason to
suppose that the second (recognition of authority hurdle) obviates the first
(possession of authority hurdle). (We shall see, however, that classical
empiricism conflates the two, and is therefore properly called
“foundationalism.”) Seeing how the understanding and veridicality conditions
are distinct but related shows how, for Sellars, perceptual knowledge is a
communal project. The same conceptual capacity – inferring the existence of a
green item from the report “This is green” – is common to the two conditions.
One cannot have the ability to make a report with understanding unless one
can also consider other people’s reports to be veridical – and vice versa.12
I think it is clear by now that Sellars’s point is not that for one’s report to
count as non-inferential knowledge one must be able to infer from one’s own
report (along with some collateral bits of knowledge) a piece of inferential
knowledge with essentially the same content as the original report. That would
be an odd test; for how could a new, justified, and therefore inferential
tokening of “This is green” (or, perhaps better, “That was green”) show that a
previous non-inferential tokening of “This is green” was an expression of
knowledge? How is this scenario different from saying that we have an

12 “We can ascertain, for example, that a person does in point of fact respond as he ought to red

objects in sunlight by uttering or being disposed to utter ‘this is red’” (1974, p. 59).
 Learning and the Necessity of Non-Conceptual Content in Sellars 125

internal mechanical response which we can read in the same way in which we
read a thermometer? (Remember that reading a thermometer is itself a
perceptual activity: we are in regress territory here.) This seems to be the view
that Robert Brandom attributes to Sellars. But I think Sellars’s view is rather
that we demonstrate our recognition of the authority the report type “This is
green” primarily when we infer the presence of green items from other
people’s tokenings of “This is green”; inferences from other people’s reports
are not trivial in the way that inferences from one’s own report would be. Not
only are interpersonal inferences of this sort non-trivial, but learning to draw
them is an essential part of our coming to inhabit the tradition of a particular
language, which in this case can be considered a communal project of
perceptual knowledge. It is also important to remember that it would be
incorrect to assume that since only someone who can draw such an inference
can be in a position to recognize the authority of the report, drawing the
inference is therefore coextensive with recognizing the report’s authority.
According to Sellars, the classical empiricist does not maintain this
distinction between the understanding and veridicality conditions – at least
when it comes to Konstatierungen. For the classical empiricist, merely
following the rules for the use of the report guarantees that the report token is
an expression of non-inferential knowledge. On this view, Konstatierungen are
“self-authenticating” or “intrinsically authoritative” (1997, §32/pp. 71-73): if
the terms in the report “This is green” are used properly, then one may
correctly infer from its tokening the presence of something green (in the
standard version of this view: a green sensation). For the classical empiricist,
then, the understanding and veridicality conditions on non-inferential
knowledge coincide: to make an observation report with understanding is
simultaneously to establish the truth of that report.13 It is for this reason that
the classical empiricist account of perceptual knowledge can properly be called
“foundationalist.”
Sellars expresses the extent of his sympathy with this view in the following
remark: “[W]e have seen that to be an expression of knowledge, a report must
not only have authority, this authority must in some sense be recognized by the
person whose report it is” (1997, §35/p. 74). I take Sellars to mean that the
authority of the report must be recognized in the following sense: one must
understand its authority; one must recognize the normative purport the report

13 Schlick writes: “However different therefore ‘confirmations’ [Konstatierungen] are from

analytic statements, they have in common that the occasion of understanding them is at the same
time that of verifying them: I grasp their meaning at the same time as I grasp their truth” (1959,
p. 225). Also see Schlick (1986, pp. 234ff.).
126 David Forman

has within language. This is what distinguishes Konstatierungen from the

utterances of a talking photocell or thermometer (and from puffs of smoke).
On this point, Sellars agrees with the foundationalists (he mentions H.H. Price)
against the logical behaviorists (such as B.F. Skinner). Sellars differs from the
foundationalists only in denying that understanding the authority of one’s
report implies that one’s report also has authority; he differs from the
foundationalists only in denying that one’s report can be endorsed as
expressing a fact simply by virtue of its being made with understanding.
The requirement that one recognize the authority of one’s report thus
serves the same function in both the foundationalist account of non-inferential,
perceptual knowledge and in Sellars’s own account. In both accounts,
recognizing the authority of a report consists in understanding how to use the
sentence expressed by the report: recognizing authority consists in knowledge
of the sentence type, whereas having authority is a property of a sentence
token. For the foundationalist, making a report with understanding already
implies, of course, that the report expresses a fact: observation reports are
intrinsically credible. In Sellars’s account, on the other hand, making a report
with understanding carries no such implication. Yet Sellars is clear that the
requirement that the authority of the report be recognized is a matter of
showing the report is made with understanding and not a matter of showing
that the report has authority.
We therefore cannot conclude that Sellars’s account differs from the
foundationalist account by virtue of an additional requirement that the
perceiver be able to justify his report. After all, perceptual knowledge, for
Sellars, is supposed to be non-inferential, even if the meaning of a
non-inferential claim can be expressed in terms of the claim’s “inferential
articulation”: the inferences that are typically drawn from, and can lead to, a
claim of the same type. It is difficult to see what it would mean to make a
non-inferential report “in the recognition” of the report’s authority if that
recognition meant having inferential reasons for tokening this particular
report at one’s disposal (that is, it would be difficult to see how, in that case,
there could be such a thing as non-inferential knowledge).14 This is not what

14 “It might be thought that this is a redundancy, that knowledge (not belief or conviction, but

knowledge) which logically presupposes knowledge of other facts must be inferential. This,
however, I hope to show, is itself an episode of the Myth” (1997, §32/p. 69). This illustrates why
it is essential to clearly distinguish the verticality and understanding conditions on perceptual
knowledge. Although perceptual knowledge presupposes that the perceiver knows the facts
associated with the understanding condition, these facts do not serve in the justification of
perceptual claims (see note 2, above).
 Learning and the Necessity of Non-Conceptual Content in Sellars 127

“recognizing the authority of the report” means for the foundationalist, and I
am suggesting that it is not what it means for Sellars either.
To sum up: in the foundationalist account, to recognize the authority of the
report type one is tokening is already to vindicate the credibility of that
tokening: the tokening is intrinsically credible and therefore an expression of
non-inferential knowledge. But in Sellars’s account, recognizing the authority
of the report type one is tokening shows only that the tokening is made with
understanding: the tokening is an expression of a candidate for non-inferential
knowledge. Since, in Sellars’s account, reports are not intrinsically credible,
other knowledge that pertains to this particular tokening would be necessary
to show that the report is indeed credible, that it can be endorsed and thus that
it is an expression of non-inferential knowledge.
We can therefore conclude that Sellars is an externalist concerning the
veridicality condition, but an internalist concerning the understanding
condition: the knowledge needed to vindicate an observation report (to show it
expresses a fact) need not be internal to the perceiver for the report to be an
expression of non-inferential, perceptual knowledge; yet the knowledge
needed to understand one’s own report (the knowledge needed for one’s report
to be even a candidate for an expression of non-inferential knowledge) does
need to be internal to the perceiver.
In the next section, I will examine the implications of this internalist
requirement for Sellars’s account of concept acquisition.

2. Sense Impressions and Language Learning

In order to be comfortable with the idea that we have the conceptual abilities
necessary for knowledge of the world, we must first be comfortable with the
idea that we have acquired such abilities. We must therefore have some
account of the possibility language learning. Such an account may seem
unnecessary: what could be more natural than the fact that human beings learn
languages from their parents?15 However, to treat the inheritance of a linguistic
tradition as unproblematic by simply reminding us that it must be “natural” is
to fall back into a different kind of the Myth of the Given: the Myth that such a
tradition is simply given to us.

15 This seems to be McDowell’s attitude. See lecture IV of Mind and World (esp. p. 84) and his

remark on p. 433 of the Woodbridge Lectures (1998b) regarding the “central thesis” of Sellars’s
account of knowledge.
128 David Forman

In EPM, Sellars undertakes the challenge of explaining how we could

come to possess conceptual abilities. He seeks explain how a creature born of
nature can inhabit and build upon a world opened to him by traditions created
and passed on through language and how a species composed of such creatures
can make the “journey from the grunts and groans of the cave to the subtle and
polydimensional discourse of the drawing room” (1997, §63/p. 117). Put more
starkly, Sellars undertakes the challenge of explaining how a mind could
emerge from mere nature.
According to one possible strategy for undertaking this challenge, the
emergence of linguistic and other rational behavior is to be explained in the
same way that we explain the emergence of other complex natural phenomena.
Man is born a natural being, and rational behavior is just a natural expression
of his congenital capacities: “Nature, who made the mason, made the house.”16
The assumption behind this strategy is that the relations among cognitive states
and between cognitive states and their objects ultimately belong to the class of
relations constituting the domain of natural science. (And if cognitive relations
do not belong to such a class, then so much the worse for the idea of cognitive
relations.) Following McDowell, we can call this assumption “bald
naturalism” (e.g. 1994, pp. xviii, xxff., 73).
This strategy is not open to Sellars since he sharply distinguishes cognitive
or epistemic facts from natural facts (1997, §17/p. 42; see McDowell (1998b,
p. 433, n. 5)). Underlying his commitment to this distinction is the idea that
epistemic facts are essentially and ineliminably normative: “in characterizing
an episode or a state as that of knowing, we are not giving an empirical
description of that episode or state; we are placing it in the logical space of
reasons, of justifying and being able to justify what one says” (§36/p. 76). No
amount of sophistication in our empirical descriptions can overcome this
fundamental divide between the normative and the non-normative (1997,
§5/p. 19).17 Consequently, a person – a being of whom cognitive facts obtain –
cannot be exhaustively described in any empirical language. That is, Sellars is
committed to the idea that a person is more than a merely natural being. He is
therefore committed to the idea that each person individually, as well as the

16 Emerson, (1903, p. 183) quoted in Turner (2003, p. 109).

17 “Now the idea that epistemic facts can be analyzed without remainder – even ‘in principle’ –
into non-epistemic facts, whether phenomenological or behavioral, public or private, with no
matter how lavish a sprinkling of subjunctives and hypotheticals is, I believe, a radical mistake – a
mistake of a piece with the so-called ‘naturalistic fallacy’ in ethics” (1997, §5/p. 19).
 Learning and the Necessity of Non-Conceptual Content in Sellars 129

species collectively, has at some point made the transition from being a merely
natural being to being a rational one.18
If we take this to mean that a being governed by cognitive norms must be a
supernatural being, then the transition from a merely natural being to a
rational being will seem impossible (or at least miraculous). In that case, the
impossibility of such a transition will serve as evidence that we have been
rational beings all along; it will serve as evidence for innatism. Sellars, of
course, has no patience for this kind of thinking.19 But he also does not have
recourse to the bald naturalist solution: to explain the emergence of norm-
governed beings in wholly non-normative terms.
We should not conclude that Sellars’s sharp distinction between the
normative and the natural commits him to the idea that the capacities that
make this transition possible – principally, the capacity for language – are not
natural to Homo sapiens. That would seem to commit Sellars to some form of
supernaturalism. Indeed, Sellars never denies the naturalness of this capacity
for language; he denies only the naturalness of any particular expression of this
capacity: the tradition into which a merely natural infant can be initiated by
means of its natural capacity for language and reason, and the rational relations
access to which is possible only within such a tradition. For Sellars, history is
not a branch of natural science: nature may have made the mason, but the
mason made the house.
It is for this reason that learning is an important theme in EPM. An
adequate account of language learning must explain how we can become
rational beings without thereby becoming supernatural beings. Conversely, the
account must explain language learning as a process natural for human beings
without denying that authentic language learning and concept acquisition
differ qualitatively from mere behavioral conditioning.
Sellars recognizes that his non-traditional empiricism both requires
learning language prior to the most basic forms of cognitive awareness and
threatens to make this process of language acquisition seem impossible. For
example, coming to understand the authority of a report, e.g. of the type “This

18 “[T]he members of a linguistic community are first language learners and only potentially

‘people’, but subsequently language teachers, possessed of the rich conceptual framework this
implies. They start out by being the subject-matter subjects of the ought-to-be’s [i.e., they are
trained to respond appropriately to their environment] and graduate to the status of agent subjects
of ought-to-do’s [i.e., they are subject to the requirement of making sure others respond
appropriately to their environment]” (1969, p. 512).
19 See, e.g. (1997, §6/p. 21) and (1967, I §§46-47/pp. 18-19). I discuss the latter passage in more

detail below.
130 David Forman

is green,” by observing others use that report seems to presuppose that one can
already make the connection between the report and the presence of something
green. The ability to make this connection was not so difficult for the classical
empiricists and other foundationalists to explain. They encountered no
problem about knowing what kind of experiences one has; one simply attached
a particular word to experiences one could already non-conceptually
distinguish and classify as being of a certain sort (1997, §26-28/pp. 57-62; esp.
(1) on p. 58). In this kind of account, it is precisely the recourse to a “given”
element in experience that makes language learning seem unproblematic.20
It might seem possible to avoid this difficulty by claiming one can be
pre-linguistically aware of the fact that two or more experiences resemble one
another even if one cannot yet pre-linguistically classify them. But Sellars
points out that such an account still presupposes the ability to have a
pre-conceptual awareness of facts. Such an account therefore relies on a
subtler form of the same mythical “given” element in experience: “the
givenness of determinate kinds of repeatables, say crimson, is merely being
replaced by the givenness of facts of the form x resembles y, and we are back
with an unacquired ability to be aware of repeatables, in this case the
repeatable resemblance” (1997, §29/pp. 62-64). It is at this point in the
argument that Sellars introduces his own view that “all awareness of sorts,
resemblances, facts, etc., in short, all awareness of abstract entities – indeed,
all awareness even of particulars – is a linguistic affair” (1997, §29/p. 63). The
question thus becomes how an association of a word with an experience or
with a characteristic of a physical object is possible when no repeatable
resemblances can be pre-linguistically noted.
The problem unfolds as follows: learning to recognize the authority of the
report such as “This is green” means learning that “utterances of ‘This is
green’ are reliable indicators of the presence of green objects in standard
conditions of perception” (1997, §37/pp. 76-77; my emphasis). Before a
tokening of X (e.g., “This is green”) could express knowledge of Y (e.g., that
there is something green there), the subject “would have to know that overt

20 “Thus, we conceive [the infant] as a person (or at least a potential person) in a world of physical

objects, colored, producing sounds, existing in Space and Time. But though it is we who are
familiar with this logical space, we run the danger, if we are not careful, of picturing the language
learner as having ab initio some degree of awareness – ‘pre-analytical’, limited and fragmentary
though it may be – of this same logical space . . . In other words, unless we are careful, we can
easily take for granted that the process of teaching a child to use a language is that of teaching it to
discriminate elements with a logical space of particulars, universals, facts, etc., of which it is
already undiscriminatingly aware, and to associate these discriminated elements with verbal
symbols” (1997, §30/p. 65).
 Learning and the Necessity of Non-Conceptual Content in Sellars 131

verbal episodes of this kind are reliable indicators” of Y; he would have to

know “general facts of the form X is a reliable symptom of Y” (1997,
§35/p. 75; my emphasis); for example, he would have to know that in English
the report “This is green” is a reliable symptom of the presence of green
things.21 This introduces a potential regress:
Does [this account] not tell us that observational knowledge at time t
presupposes knowledge of the form X is a reliable symptom of Y, which
presupposes other knowledge of the form X is a reliable symptom of Y, which
presupposes still other, and prior, observational knowledge, and so on? (1997,
§36/p. 76)22
In order to learn to say “This is green,” one must have observed people saying
“This is green” when they were near something green, so it seems impossible
that one could ever learn to understand the report without already knowing
something at least analogous to “This is green,” namely that this is green.
To be sure, one could learn to recognize the authority of “This is green” in
some other, more indirect way, especially if one already speaks another
language. But Sellars’s regress is designed to show that, in the end, even if not
in each particular case, learning to recognize the authority of perceptual reports
such as “This is green” rests on the application of concepts that one learns
from hearing the reports themselves. It seems impossible to ever start
collecting the evidence – in the form of “inductive reasons” (1997, §38/p. 78)
– needed to come to know that the report type “This is green” is a reliable
symptom of something being green, since one cannot have a pre-linguistic
awareness of a repeated item as being repeated. In other words, one does not
seem to be in a position to have the necessary awareness that a repetition has
taken place.23

21 See Sellars’s remark in a later paper: “An essential requirement of the transmission of a

language from generation to generation is that its mature users be able to identify both extra-
linguistic items and the utterances that are correct responses to them. This mobilizes the fact . . .
that, in addition to their logical powers, linguistic expressions have an empirical character as items
in the world” (1974, p. 59).
22 I assume that it is consistent with the way the regress is supposed to work, namely as a regress

of the understanding condition, to suppose that ‘X’ represents a report type, not a report token.
This assumption is consistent with the previous two quotes from Sellars (see my emphasis), which
describe the problem in terms of the “general facts” one must know. (Sellars discusses a different
but related potential regress in an account of language learning in the opening of his earlier essay
“Some Reflections on Language Games,” 1954.)
23 The same paradox seems to make an appearance in Science and Metaphysics: “The view that

before we can have representations of the form x is a cube we must have representations of the
form this-cube is a puzzling one” (1967, I §13/pp. 5-6). See also EPM §10.
132 David Forman

In his “Study Guide” to EPM, Brandom, I think, misrepresents the point of

the regress (see pp. 156-160). Brandom does not specify what he objects to in
Sellars’s solution to the regress, but he does suggest that the regress can be
avoided by appealing to a “modified externalism,” i.e., by appealing to the
view expressed in his Making It Explicit. But such a “modified externalism”
presents a potential solution to the regress only insofar as the regress is taken
to be one of vindication, and thus one which concerns the veridicality
condition.
Brandom’s solution will not work if (as I have argued) Sellars’s regress
problem is concerns a regress in the account of learning to be able to use, and
thus understand, sentences with empirical content. In that case, shifting the
justificatory burden to another perspective would be a non sequitur.24 The
perceiver must have these inductive reasons at his disposal even to have, in
Brandom’s terminology, a commitment to the report “This is green”; his
entitlement to that commitment is another matter: it is a matter of whether in
addition to being made with understanding, his report token is assessed to be
(a) an instance of a report type that is a reliable symptom for what the report
claims, and (b) made in standard conditions and in an appropriate relation to
what the report is about. His entitlement to the report is thus based on the fact
that there is, as a matter of fact, something green there.
McDowell apparently interprets Sellars’s regress of learning in the same
way as Brandom. In discussing Brandom’s book, McDowell writes:
Strong inferentialism aims to capture observational authority in terms of a
reliability inference. For Sellars, the inference must be endorsed by the reporter,
and this leads him to suppose a warrant for the inference – an accumulation of
cases in which it would have lead to a true conclusion – must be at the
reporter’s disposal. Reasonably, Brandom recoils. (1997, p. 161)
But in contrast to this statement, we have seen that, for Sellars, inductive
reasons are needed not to vindicate the report’s veridicality (or “warrant”), but
rather for the application of the relevant empirical concepts in a way that
exhibits understanding. The regress is about understanding, not about a
skepticism concerning reliability.25 As such, it is not obvious that one can
recoil without settling for semantic nihilism – or at least for bald naturalism.26

24 Perhaps something similar could be said about Davidson’s account of “triangulation”

(introduced in “Rational Animals” and developed in several other essays collected in Subjective,
Intersubjective, Objective, 2001).
25 Following Conant (2004), we might say that the regress is an expression of a Kantian rather

than a Cartesian skeptical worry. This is another way of bringing out a potential objection to
chapter 4 of Brandom’s Making It Explicit: Brandom tries to answer a Kantian worry with
 Learning and the Necessity of Non-Conceptual Content in Sellars 133

Sellars’s solution to this regress has important consequences for his

account of sense impressions. His solution is to say that although a perceiver,
call him Jones, must indeed, at the time he makes his report, have this
conceptual knowledge of general facts of the form “X is a reliable symptom of
Y”; and although he must have these prior facts at his disposal as inductive
reasons; still, these need to be taken as facts only at the time he makes the
report:
And while the correctness of this statement about Jones [viz., that he knows that
among English speakers “utterances of ‘This is green’ are reliable indicators of
the presence green objects”] requires that Jones could now cite prior particular
facts as evidence for the idea that these utterances are reliable indicators, it
requires only that it is correct to say that Jones now knows, thus remembers,
that these particular facts did obtain. It does not require that it be correct to say
that at the time these facts did obtain he then knew them to obtain. And the
regress disappears. (1997, §37/p. 77; my emphasis)
Sellars supplemented this argument with the following explanation in a
footnote added to the 1963 edition of EPM: “My thought was that one can
have direct (non-inferential) knowledge of a past fact which one did not or
even (as the case envisaged) could not conceptualize at the time it was
present” (my emphasis). And we must remember that the “case envisaged” is
not a fancy of the skeptical imagination: it is rather the case of every child
learning a language; it is the case we must explain in order to explain our
initiation into a language.
The solution to the regress seems to require that we presuppose the
existence, in the perceiver, of some sort of non-conceptual sensory states
whose qualities systematically correspond to qualities one is learning to
identify (e.g., the color green). (Only this systematic correspondence could
ensure that our pre-conceptual encounters could pave the way for a later
conceptual awareness of that quality.) But the solution to the regress requires

solution suited only for a Cartesian worry. (Given McDowell’s new reading of Sellars in the
Woodbridge Lectures (1998b), it is possible that he would no longer give a Cartesian reading of
Sellars’s regress argument.)
26 Sellars mentions a different but related attempt to avoid the problem: “And let me emphasize

that the point is not taken care of by distinguishing between knowing how and knowing that, and
admitting that observational knowledge requires a lot of ‘know how’. For the point is specifically
that observational knowledge of any particular fact, e.g. that this is green, presupposes that one
knows general facts of the form X is a reliable symptom of Y. And to admit this requires an
abandonment of the traditional empiricist idea that observational knowledge ‘stands on its own
feet’” (1997, §36/pp. 75-76).
134 David Forman

more than just the existence of such non-conceptual states. It also requires that
there be a sense in which these non-conceptual states can be conceptualized.
Only thereby can past non-conceptual states (e.g. those caused by green
things) play any role in the collection of the inductive reasons necessary for
recognizing the authority of a perceptual report type (e.g. the report “This is
green”).
It is natural to suppose that these non-conceptual states must be sense
impressions, that is, non-conceptual states of sensory consciousness that bear
properties systematically analogous to the physical objects that ordinarily
cause them (see 1997, §61/p. 110, where he calls impressions “inner replicas”
of the physical objects that are their normal occasions). Granting such a role to
sense impressions in stopping the regress of learning would confirm Sellars’s
earlier suggestion that having impressions “may be a necessary condition, even
a logically necessary condition, of non-inferential knowledge” (1997,
§6/p. 21): the regress argument shows how perceptual knowledge logically
presupposes the existence of sense impressions.
The suggestion that sense impressions are a “logically necessary condition”
for having perceptual knowledge belies McDowell’s contention that Sellars
posits impressions in EPM because he wants to “explain the facts of sense
perception in scientific style” (1997, §7/p. 22; see McDowell 1998b, p. 444).27
On McDowell’s reading, impressions are not posited in order to explain the
possibility of non-inferential knowledge; they are introduced instead merely to
tie up a loose end in the account by locating the common cause of veridical
and merely ostensible seeings. Given this reading, McDowell is probably right
to conclude (along with Brandom) that impressions are “idle wheels” in the
account, that the role that Sellars gives to sense impressions in EPM could just
as easily be fulfilled by merely physiological states of the perceiver
(pp. 443ff.). But would the existence of such a mere physiological
commonality also suffice to stop the regress of language-learning? For
example, would the neurophysiological28 commonality among different pre-
conceptual encounters with green objects (identified by others as such) suffice
to allow one to collect the “inductive reasons” needed to learn the meaning of

27 I do not mean to deny the fact that in positing the existence sense impressions, Sellars is also
encouraged by “our ordinary talk about sensations, feelings, afterimages, tickles and itches, etc.”
and by the ordinary view that our experiences have something in common with the experiences of
brutes (1997, §6, item 2 and §7, item 1/pp. 21ff.).
28 McDowell suggests that we could find the commonality at the level of retinal images. But the

possibility of hallucinations suggests that neurophysiological states would better serve

McDowell’s purposes.
 Learning and the Necessity of Non-Conceptual Content in Sellars 135

the concept “green”? Or would the commonality have to consist of common

sense impressions, that is, common non-conceptual states of sensory
consciousness?
This origin of our conceptual capacities in the collection of “inductive
reasons” differentiates concept acquisition from the acquisition of other sorts
of discrimination behavior: it ensures that we understand the concepts we use
and are not simply conditioned to respond to our environment in a reliable
way; it differentiates our responses to the environment from those of
thermometers and photocells. Collecting these reasons, Sellars tells us, takes
the form of recollecting that certain facts obtained in the past. And the
recollecting of a past fact must, in some sense, be responsive to the subjective
state of the perceiver at the time the fact obtained. (Otherwise, it would not
make sense to speak of “remembering” as opposed to just “knowing” that a
past fact obtained.) Since this past subjective state cannot, ex hypothesi, be a
conceptual state, it must instead be a non-conceptual state. And although this
past state must be non-conceptual, it must nonetheless be a state of
consciousness rather than a physical state such as a neurophysiological state.
Our method of access to our past neurophysiological states does not differ
essentially from our method of access to other peoples’ neurophysiological
states; indeed it does not differ essentially from our method of access to
physical states and objects more generally. This is because neurophysiological
states are not themselves subjective states (even if they do correlate with
subjective states): they do not themselves constitute a part of person’s
subjective condition. This contrasts with non-conceptual states of sensory
consciousness, that is, with sensations or sense impressions, which belong
essentially to a person.29 A person can recollect a past fact only if, at the time
the fact obtained, the person himself (and not merely his brain) was somehow
affected by his environment in the appropriate way.
Before returning to Sellars’s account in EPM, I want to point out that
McDowell claims that Sellars offers a completely new justification in Science
and Metaphysics for positing impressions: “the explanatory need that
sensations are supposed to satisfy is not a need for scientific understanding, as

29 See Science and Metaphysics VI, §50 (1967, p. 169). If a connection between subjective states

and neurophysiological states could be established, one could perhaps learn to report one’s
neurophysiological states in a way analogous to the way, on Sellars account, we learn to report
inner episodes. But that does not make neurophysiological states into subjective states. Indeed, it
presupposes that access to neurophysiological states is temporally posterior to and dependent on
access to their corresponding subjective states (access to which, if these are non-conceptual states,
is, in turn, temporally posterior to and dependent on certain conceptual states).
136 David Forman

it seemed in “Empiricism and the Philosophy of Mind”; rather, it is

transcendental” (1998b, p. 444; see Sellars 1967, I §§22, 28/pp. 9, 11).
Impressions play a “transcendental” role because their existence is, as Kant
would say, a condition of the possibility of experience: without impressions,
conceptual episodes would be cut off from any guidance from the world and
would lack genuine empirical content.30 To put this point in McDowell’s
parlance: Sellars needs impressions to avoid the pitfall of “frictionless
coherentism.” And Sellars makes clear here that impressions are posited not to
explain the origin of behavioral discrimination as such, but rather to explain
the possibility of conceptual awareness. If behavioral discrimination were the
explanandum, then presumably neurophysiological states would be adequate
play the relevant explanatory role (1967, I §§41-44/pp. 16ff.).
Now, McDowell’s own view is of course that conceptual episodes have
empirical content without any guidance from impressions: on this view, Sellars
needs to posit impressions linking conceptual episodes to the physical world
only on account of his idiosyncratic inferentialism about the conceptual. But a
closer look at Sellars’s argument in Science and Metaphysics shows that his
conviction that impressions are needed to guide our experiences is based,
ultimately, not on an abstract consideration of how the nature of concepts
allows for the possibility of reference, but rather on a confrontation with the
possibility of learning a language.
After introducing the idea that non-conceptual states of sensory
consciousness are needed to guide conceptual representations, Sellars
immediately introduces two alternative suggestions that might seem to
preserve empirical conceptual content without relying on the guidance of
impressions (1967, I §46/p. 18ff.). His discussion of these two suggestions is
revealing.
The first suggestion is that certain empirical concepts might be innate. In
keeping with his empiricism, Sellars simply dismisses this suggestion without
argument (1967, I §47/p. 19). But what is significant here is that Sellars even
considers innatism to be an alternative to an account that relies on the
existence of impressions. By doing so, Sellars presumes that if some critical set
of empirical concepts were innate, then there would be no need to posit the
existence of impressions. Hence, Sellars never suggests (contrary to what we
might expect) that innate empirical concepts would themselves need

30 Sellars’s appeal here to “transcendental grounds” of justification is therefore equivalent to his

later claim that some facts that cannot be established through induction can still be “warranted as
an essential component of the conceptual framework that defines what it is to be a mind that gains
knowledge of a world to which it belongs” (1970, pp. 624ff.; see 1974, p. 47).
 Learning and the Necessity of Non-Conceptual Content in Sellars 137

impressions to guide their application. Innatism is an alternative to Sellars’s

account precisely because it avoids the whole problem of concept acquisition.
The centrality of the problem of concept acquisition becomes even clearer
in his discussion of the second suggested alternative:
One is taught by one’s linguistic peers, who already have the relevant concepts
and propensities, to play the color-shape language game and, by so doing,
acquire these concepts and propensities. (1967 I, §46/p. 18ff.)
Sellars does not dismiss this suggestion out of hand, as he does with the
suggestion of innatism. But his reaction to this suggestion parallels his reaction
to innatism in one crucial respect: he does not object that even the moves of a
fully competent player would require guidance from the non-conceptual cues
given to him by impressions. Instead, each of the two suggestions stands as an
alternative to Sellars’s account precisely because each denies the need for
non-conceptual sensory impressions in the process of acquiring conceptual or
linguistic competence: innatism avoids the question of concept acquisition
completely; and the language-game account explains language acquisition
solely at the behavioral or linguistic level, forbidding the appeal to the
existence of inner episodes of any kind. Hence, Sellars tells us that the
language-game account fails as an alternative to his account of impressions
because it cannot, by itself, account for the transmission of a language game
from one player to the next:
As for [this suggestion], I do not wish to deny the insights it contains.
Nevertheless, the ability to teach a child the color-shape language game seems
to imply the existence of cues which systematically correspond, in the manner
adumbrated above, to the color and shape attribute families, and are also
causally connected with combinations of variously colored and shaped objects
in various circumstances of perception. If so, the account in terms of the
transmission of the color-shape language game supplements, but does not
replace, the original suggestion. (1967, I §47/p. 19)
Apparently satisfied with the validity of the inference to sense impressions,
Sellars announces at precisely this point in the text that he is beginning a new
stage in his argument: explaining what it means to say that these impressions
have intrinsic properties analogous to the properties of the physical objects that
normally cause them. In other words, his discussion of the necessity of
impressions for language acquisition is his final word on the validity of the
“sense impression inference.”
Positing impressions on the grounds that their existence is a necessary
condition for language acquisition may seem at odds with Sellars’s
characterization of the nature of the sense impression inference: impressions
are to be posited on the “transcendental” grounds that they are needed to
138 David Forman

“guide” our conceptual states. But it would actually be quite appropriate to

consider the conditions for language acquisition as providing such
“transcendental grounds”: to posit something on transcendental (general
epistemological) grounds is to posit it as a necessary condition for the
possibility knowledge of the world; and since knowledge of the world is
possible only if one can acquire the requisite conceptual abilities, the
conditions for the possibility of such an acquisition can surely provide
“transcendental grounds” for positing sense impressions.31
For the same reason, it would also be appropriate to consider the role that
impressions play in language acquisition to be a “guiding” role. The thought
here need not be that conceptual states reach beyond the conceptual to find
their justification in particular non-conceptual sensory episodes (that would be
a version of the Myth of the Given). The thought could be merely that it is
insufficient to say that our conceptual abilities are responsive directly to cues
from physical objects: that would not leave room for the possibility of
learning, since learning presupposes a responsiveness to cues from our own
(past) non-conceptual sensory states. By enabling our acquisition of a
perceptual language, impressions ensure that we possess the conceptual
abilities necessary to represent the world as it is. Impressions thereby ensure
that our thoughts can be guided by the world itself:32 they ensure that our
conceptual episodes have empirical content.
Since this is just the role marked out for impressions by the regress
argument in EPM, it makes sense to understand impressions, even in that
earlier work, as playing this “transcendental” role of “guiding” our perception.
In fact, this characterization of the role of impressions may suit the account in
EPM even better than the account in Science and Metaphysics: in Science and
Metaphysics, Sellars assumes that our conceptual states of perception are
always occasioned by corresponding non-conceptual states; that is, he assumes
that even the moves of a fully competent language-game player have to be
understood as responses to non-conceptual states of consciousness (1967, I
§§37, 44, 78/pp. 15, 18, 30). This assumption is not warranted by a reflection

31 In “Some Remarks on Kant’s Theory of Experience” (1974), Sellars implicitly considers the
conditions for language acquisition to be “transcendental.” After mentioning that his own
“transcendental linguistics” could be considered a linguistic version of Kant’s search for “the
general features any conceptual system must have in order to generate knowledge of a world to
which it belongs,” Sellars immediately proceeds to identify an “essential requirement of the
transmission of a language from generation to generation” (pp. 58ff.). See note 21, above.
32 “To know the language of perception is to be in a position to let one’s thoughts be guided by the

world” (1974, p. 48).

 Learning and the Necessity of Non-Conceptual Content in Sellars 139

on the requirements for concept acquisition, and Sellars offers no other

argument supporting it.33
However, this assumption does not yet implicate his account in the Myth of
the Given (even if the assumption is objectionable for other reasons). To say
that our conceptual episodes of perception are responsive, in the first instance,
to our own non-conceptual states is not to say that these non-conceptual states
stand in a justificatory relation to conceptual states. It is rather to say we can
be conceptually responsive to the world only if that responsiveness is in some
sense simultaneously a responsiveness to our own non-conceptual states (the
latter occurring unnoticed and, as it were, “behind our backs”).34
Reading the account this way acknowledges the fact that Sellars
understands his project in Science and Metaphysics as a continuation of the
project in EPM; it acknowledges the fact that Sellars never moderates his
opposition to all forms of “givenness.”35 Nevertheless, it cannot be denied that
in Science and Metaphysics he sometimes does seem to commit himself to
giving impressions a more radical guiding role, a role that does seem to run
afoul of Sellars’s own warnings in EPM against the Myth of the Given. For
example, Sellars claims that in Kant’s account (which he takes as a model)
“sheer receptivity” provides “the ‘brute fact’ or constraining element of
perceptual experience” (1967, I §21/p. 9), such that the cues it provides are “an
independent factor which has strong voice in the outcome [viz., in a conceptual
representation],” and such that “it can only guide [conceptual activity] ‘from
without’” (1967, §39/pp. 45-46; see §78/p. 59). These claims certainly appear
to give impressions a far more radical guiding role: a role in justifying our
conceptual states. However, these seemingly radical claims are largely
metaphorical, and their meaning is by no means clear. Indeed, such claims are
ultimately compatible with the view that impressions guide our conceptual
activity only in the sense of providing the cues to which our conceptual
abilities are immediately responsive: learning to be responsive the world

33 Perhaps the fact that the acquisition and refinement of conceptual abilities does not stop with
childhood could, in some measure, justify the idea that the kind of guidance from impressions
displayed in language acquisition is a general feature of perception.
34 See McDowell’s Woodbridge Lectures (1998b, p. 468). This can be compared to Kant’s

characterization of the imagination as a “blind but indispensable function of the soul, without
which we should have no knowledge whatsoever, but of which we are scarcely ever conscious”
(1997, A78/B103). Also see Anthropology §5 and Longuenesse’s discussion at (1998, p. 65).
35 E.g. (1967), p. viii. In his Carus Lectures (1981), Sellars remarks that the views introduced in

EPM regarding the Myth of the Given “are so central to my way of thinking that if they were to
fall apart the result would be a shambles” (1967 I, §3/p. 3).
140 David Forman

involves learning how to be responsive to our own non-conceptual states (even

though this responsiveness is known only to the philosopher).36
The temptation to read Sellars as granting an illicit justificatory role to
impressions is diminished by recalling the centrality of the problem of learning
in Sellars’s account and by considering how impressions are meant to solve
that problem. Impressions are not needed because our conceptual states of
perception would otherwise be systematically unjustified; they are not needed
because our conceptual episodes fail to relate to the world simply by being the
conceptual episodes that they are. Impressions are needed rather because a
responsiveness directly to the world itself cannot result in the knowledge of
past resemblances required for learning a perceptual language, the knowledge
required for learning how to make empirical claims with the understanding
distinctive of cognitive beings.
However, I do not wish to insist that this understanding of the guiding role
of impressions is adequate as an interpretation of Science and Metaphysics.
My aim has rather been to point out that the avowedly transcendental
aspirations of the account of impressions in Science and Metaphysics can be
seen as a continuation of the thinking behind the regress argument in EPM. In
Science and Metaphysics, Sellars more or less explicitly associates the
problem of language learning with the “transcendental” question of how
conceptual episodes can represent the world as it is. And the account in EPM
shows that such transcendental thinking need not lead one back into the Myth
of the Given. Regardless of our stance on the interpretation of Science and
Metaphysics, we can be confident that in EPM impressions are not meant to
play an illicit justificatory role. It is this fact about EPM that leads McDowell
to insist that impressions must therefore be “idle wheels” in that work. But in
contrast to McDowell’s reading, I have tried to suggest above that the regress
of learning depicted in EPM shows that Sellars found a substantive role for
impressions even in EPM, a role that makes their existence a “logically
necessary condition” for the existence of perceptual knowledge. I also

36 It might also be possible to distance such a radical guiding role for receptivity from Sellars’s
own considered view by pointing out that Sellars makes these claims in a context more or less
narrowly concerned with the role of receptivity presupposed by Kant’s account of empirical
knowledge. Sellars does, of course, want to appropriate the Kantian idea of a “manifold of sense,”
but he does not think Kant’s own argument for the existence of such a manifold is sound
(§28/p. 11; see §19/pp. 7-8), and we can infer that he would not take himself to be bound to a
Kantian account of the way impressions guide conceptual activity. I have suggested that the
argument for the existence of impressions that Sellars actually puts forward in his own name does
not presume the radical guiding role for impressions that these comments about Kant’s account
might seem to endorse (see §22/p. 40 but also §78/p. 59).
 Learning and the Necessity of Non-Conceptual Content in Sellars 141

suggested that looking to the account in Science and Metaphysics can help us
understand how the role impressions play in concept acquisition, even in EPM,
can accurately be described as a providing a kind of guidance for conceptual
episodes. This is not the extra-conceptual guidance that traditional empiricism
often attributes to impressions, sensations, or sense data. In Sellars’s account,
impressions do not stand between mind and world as the results of the
impingement from the world that, at the same time, are the only true objects of
our conceptual representations of perception; they are not the objects that yield
knowledge justifying all other empirical claims. To the contrary, impressions
contribute to establishing the link with the world that ensures our conceptual
episodes can be true or false of the world itself (rather than merely of our own
private impressions and only indirectly of the world): learning a perceptual
language is learning to make claims about the world and not (in the first place)
learning to make claims about our own impressions. In short, through their role
in the process of learning a language, impressions provide a merely causal
mediation between mind and world that ensures that our linguistic or
conceptual competences are responsive to the world itself.37
It would be somewhat remarkable if, as McDowell suggests, Sellars found
an essential role for impressions only in Science and Metaphysics and thus
only after he went to great lengths in EPM to establish their possibility and
explore their logical status. Indeed, this suggestion ignores the fact that in
EPM Sellars explicitly proposes the possibility that impressions make non-
inferential knowledge possible by virtue of the causal role they play in
language acquisition (the possibility that seems confirmed in the regress
argument):
Once sensations and images have been purged of epistemic aboutness, the
primary reason for supposing that the fundamental associative tie between
language and the world must be between words and “immediate experiences”
[viz., impressions considered as the immediate objects of perceptual
consciousness] has disappeared, and the way is cleared to recognizing that basic
word-world associations hold, for example, between ‘red’ and red physical
objects, rather than between ‘red’ and a supposed class of private red

37 “A language entry transition is an evoking, for example, of the response ‘this is red’ by a red
object in sunlight from a person who knows the language to which this sentence belongs. As an
element in a rule-governed system the utterance is no mere conditioned response to the
environment. Its occurrence is a function not only of the environment but of the conceptual set of
the perceiver. To know the language of perception is to be in a position to let one’s thought be
guided by the world in a way that contrasts with free association, with day-dreaming, and, more
interestingly, with the coherent imaginings of the storyteller” (1974, p. 48).
142 David Forman

particulars. [But this] remark, it should be emphasized, does not imply that
private sensations or impressions may not be essential to the formation of these
associative connections. For one can certainly admit that the tie between ‘red’
and red physical objects – which tie makes it possible for ‘red’ to mean the
quality red – is causally mediated by sensations of red without being committed
to the mistaken idea that it is “really” sensations of red, rather than red physical
objects, which are the primary denotation of the word ‘red’. (1997, §29/p. 64;
my emphasis)
In this passage, the idea that the connection between the concept “red” and the
redness of red physical objects is causally mediated by red impressions is
almost explicitly equated with the idea that red impressions play an essential
role in the acquisition of the concept “red.” This equation shows that in EPM
Sellars invites us to see the role impressions play in language acquisition as
securing the empirical content of conceptual episodes, as securing the link
between empirical concepts and the physical qualities to which they refer.

Conclusion

A solution to the regress of learning in EPM presupposes that our conceptual

abilities are subject at least to the modest form of extra-conceptual guidance
impressions provide in the process of language acquisition. The need to posit
the existence of sense impressions arises from Sellars’s insistence that a
perceiver must know the “general facts” that satisfy the understanding
condition on perceptual knowledge. That is, in order to learn how to make a
particular empirical claim (e.g. to token “This is green”), a language learner
must be responsive to his own past non-conceptual sensory states (e.g. green
impressions); only thereby can he come to know the general facts needed for
an understanding of the claim he is making (ultimately, the fact that in his
community “This is green” is a symptom of the presence of green objects).
Possessing this background knowledge associated with the understanding
condition is what distinguishes a person’s cognitive responses to his
environment from the differential responses of a thermometer or photocell.38

38 Therefore, only if we posit sense impressions can we capture what is distinctive about our

cognitive relation to the world. “Nor, at least primarily, is [the inference to impressions] designed
to explain ‘discrimination behavior’ of the type that can be acquired by flat worms and
encouraged in white rats; ‘consecutiveness’ which in Leibniz’ phrase ‘apes reason’” (1967, I
§42/p. 17). If that were the goal, then impressions could be dispensed with.
 Learning and the Necessity of Non-Conceptual Content in Sellars 143

Sellars is right to insist that preserving this sui generis character of the
cognitive (what stands in the “space of reasons”) with respect to the
non-cognitive (what stands in the “space of nature,” or, if one prefers, the
“space of law”) requires that even the most basic conceptual episodes of
perception (e.g., ostensible seeings) presuppose the satisfaction of the
understanding condition on perceptual knowledge. And he also seems right
that knowledge of the general facts associated with this understanding
condition requires, in turn, the ability to be responsive to our own past non-
conceptual sensory states (if we are to account for the possibility of acquiring
empirical concepts). Hence the argument for the existence of sense
impressions relies not on the details of Sellars’s inferentialism, but rather on a
feature of his account with a much broader appeal: the insistence on the sui
generis character of the “space of reasons.” This, of course, is the feature of
Sellars’s account that McDowell, for one, most wants to appropriate.
Nevertheless, even the modest guiding role outlined for impressions in the
acquisition of a perceptual language is, for all that, no less controversial. I
have argued against one obvious objection to the account: that it reverts to a
version of the Myth of the Given. The regress of learning shows not that our
conceptual states rely on a justification from non-conceptual states of
consciousness, but rather that learning empirical concepts presupposes that our
conceptual abilities can be responsive to our own (past) non-conceptual states.
That no more shows impressions to be a given element in experience than our
responsiveness to physical objects shows physical objects to be a given
element in experience. In short, although the argument for impressions
concludes that impressions are partially constitutive of a perceiver’s subjective
position, it does not conclude that impressions are in any way constitutive of
his epistemic position. The solution to the regress requires only that we posit
the existence of impressions; it does not require that we take impressions to
stand in any relation of justification with our conceptual episodes. Only the
latter requirement would implicate the account in the Myth of the Given.
But we might also question the whole idea – central to the account – that a
responsiveness to our own non-conceptual sensory states could lie behind our
responsiveness to a world of physical objects. I have said almost nothing about
how such a responsiveness to our own non-conceptual subjective states is itself
possible.39 But I have tried to show that that if our account of perceptual
knowledge is to preserve the sui generis character of the cognitive (without

39 Sellars fills in some of the details of an account of how our responsiveness to non-conceptual

states is supposed to reveal an intersubjectively accessible objective world in “The Role of

Imagination in Kant’s Theory of Experience” (1978).
144 David Forman

thereby reverting to innatism or supernaturalism), it must leave room for the

possibility of such a responsiveness to the non-conceptual.40

David Forman
Department of Philosophy
The University of Nevada, Las Vegas
Las Vegas, NV 89154
United States of America
e-mail: [email protected]

REFERENCES

Brandom, R. (1994). Making it Explicit. Cambridge, MA: Harvard University Press.

Brandom, R. (1997). Study Guide. In: W. Sellars (1997), pp. 119-181.
Conant, J. (2004). Varieties of Skepticism. In: D. McManus (ed.), Wittgenstein and Skepticism,
pp. 97-136. London: Routledge Press.
Davidson, D. (2001). Rational Animals. In: D. Davidson (ed.), Subjective, Intersubjective,
Objective, pp. 95-106. Oxford: Oxford University Press.
Emerson, R.W. (1903). Nature. In: The Complete Works of Ralph Waldo Emerson, Centenary
Edition, vol. 3 (Essays, Second Series), pp. 167-197. Boston: Houghton Mifflin and
Company.
Herder, G.W. (1985). Abhandlung über die Ursprung der Sprache. In: U. Gaier (ed.) Werke,
vol. 1, pp. 695-767. Frankfurt, a.M.: Deutscher Klassiker Verlag.
Kant, I. (1997). Critique of Pure Reason. Translated and edited by P. Guyer and A. Wood.
Cambridge: Cambridge University Press.
Kant, I. (1974). Anthropology from a Pragmatic Point of View. Translated by M. Gregor. The
Hague: Martinus Nijhof.
Longuenesse, B. (1998). Kant and the Capacity to Judge. Cambridge: Cambridge University
Press.
McDowell, J. (1994). Mind and World. Cambridge, MA: Harvard University Press.
McDowell, J. (1997). Brandom on Representation and Inference. Philosophy and
Phenomenological Research 58, 157-162.
McDowell, J. (1998a). Meaning, Communication, and Knowledge. In: Meaning, Knowledge, and
Reality, pp. 29-50. Cambridge, MA: Harvard University Press.
McDowell, J. (1998b). Having the World in View: Sellars, Kant, and Intentionality. Journal of
Philosophy 95(9), 431-491.
Röska-Hardy, L. (1998). “I” and the First Person Perspective. Presented at the Twentieth World
Congress of Philosophy and published online at http://www.bu.edu/wcp/Papers/Lang/
LangRosk.htm.

40 I thank an audience at the University of Chicago Contemporary Philosophy Workshop and

especially Gideon Manning and Clinton Tolley for very helpful comments on an earlier draft of
this paper.
 Learning and the Necessity of Non-Conceptual Content in Sellars 145

Schlick, M. (1959). The Foundation of Knowledge. In: A.J. Ayer (ed.) Logical Positivism,
pp. 209-227. New York: McMillan Press. First published as “Über das Fundament der
Erkenntnis” in Erkenntnis 4 (1939), pp. 79-99. Reproduced online at
http://www.fh-augsburg.de/~harsch/germanica/Chronologie/20Jh/Schlick/sch_fund.html.
Schlick, M. (1986). Über Konstatierungen. In: Philosophische Logik, pp. 230-237. Frankfurt a.
M.: Suhrkamp.
Sellars, W. (1954). Some Reflections on Language Games. Philosophy of Science 21, 204-228.
Sellars, W. (1967). Science and Metaphysics: Variations on Kantian Themes. London: Routledge
& Kegan Paul.
Sellars, W. (1969). Language as Thought and as Communication. Philosophy and
Phenomenological Research 29, 506-527.
Sellars, W. (1970). Givenness and Explanatory Coherence. Journal of Philosophy 70, 612-624.
Sellars, W. (1974). Some Remarks on Kant’s Theory of Experience. In: Essays in Philosophy and
Its History, pp. 44-61. New York: D. Reidel Press. First published in the Journal of
Philosophy 64 (1967), 633-647.
Sellars, W. (1978). The Role of Imagination in Kant’s Theory of Experience. In: H.W. Johnstone
(ed.), Categories: A Colloquium, pp. 231-245. University Park, PA: Pennsylvania State
University. Available online at http://www.ditext.com/sellars/itke.
Sellars, W. (1981). Foundation for a Metaphysics of Pure Process. The Monist 64, 3-90.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
First published in: H. Feigl & M. Scriven (eds.), The Foundations of Science and the
Concepts of Psychology and Psychoanalysis. (Minnesota Studies in the Philosophy of
Science, vol. 1), pp. 253-329. (Minneapolis: University of Minnesota Press). Reprinted with
additional footnotes in: W. Sellars, Science Perception and Reality, pp. 127-196. (London:
Routledge and Kegan Paul, 1963). The additional footnotes are not reprinted in the 1997
edition, but are available online at http://www.ditext.com/sellars/epm-n.html.
Turner, F. (2003). Ahead of His Time. Smithsonian 34 (2), 107-111.
This page intentionally left blank
Jeremy Randel Koons

SELLARS, GIVENNESS, AND EPISTEMIC PRIORITY

ABSTRACT. Recent critics of Sellars’s argument against the Given attack Sellars’s
(purported) conclusion that sensations cannot play a role in the justification of
observation beliefs. I maintain that Sellars can concede that sensations play a role in
justifying observation reports without being forced to concede that they have the
foundational status of an epistemic Given. However, Sellars’s own arguments that
observation reports rest, in some sense, on other empirical beliefs are not sufficiently
well-developed; nor are his comments concerning internalism, which is crucial to his
attack on the Given. As a result, both of these aspects of Sellars’s epistemology have
been attacked, and their significance has gone unrecognized by many philosophers. In
this paper, I will try to fill in some of the missing pieces, so that we can see that not
only are Sellars’s theses concerning internalism and epistemic priority correct, but they
represent a devastating attack on the Given, even if Sellars concedes that sensation can
play a role in justifying observation beliefs. In short, we will see that these recent
arguments in support of the Given have not succeeded in reviving it. The Given remains
a myth.

Many philosophers take Wilfrid Sellars to have decisively demonstrated the

Given to be an untenable notion, and therefore that the Myth of the Given has
been laid to rest. Of course, there have always been dissenters (see, e.g. Alston
1983 and Chisholm 1986). But recently, there have been a spate of criticisms
aimed at reviving the notion of the Given (see, e.g. Alston 1998, Bonevac
2002, and Vinci 1998).
I will argue that these criticisms share a common thread. Namely, they
attack Sellars’s (purported) conclusion that sensations (or “appearings” or
“lookings,” etc.; for the sake of simplicity, I will refer to all of these as
“sensations” in this introduction) cannot play a role in the justification of
observation beliefs. I will argue that Sellars can concede this point to the
critics (i.e., he can admit that sensations can stand in a justificatory relation to
observational belief) without having to admit that these sensations have the
foundational status of an epistemological Given. Even if sensations play a role
in justification, this does not show that they can serve as a Given to justify
empirical beliefs and vindicate epistemological foundationalism. The further

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 147-172.
Amsterdam/NewYork: Rodopi, 2006.
148 Jeremy Randel Koons

question, often neglected (but not by Sellars), is “What else would need to be
in place before these sensations could serve to justify our empirical
knowledge?” We will see that many other epistemic states must be in place
before sensation can serve a justificatory role, and that crucially, these states
are epistemically prior to sensation.
However, Sellars’s own comments on epistemic priority (his famous
discussion of reliability and “standard viewing conditions”) are not sufficiently
well-developed; nor are his comments concerning internalism, which is crucial
to his attack on the Given. As a result, both of these aspects of Sellars’s
epistemology have been attacked, and their significance has gone
unrecognized by many philosophers. In this paper, I will try to fill in some of
the missing pieces, so that we can see that not only are Sellars’s theses
concerning internalism and epistemic priority correct, but they represent a
devastating attack on the Given, even if Sellars concedes that sensation can
play a role in justifying observation beliefs. Ultimately, we will be in a better
position to see the truth in Sellars’s comment that “if there is a logical
dimension in which other empirical propositions rest on observation reports,
there is another logical dimension in which the latter rest on the former” (1997,
§38/p. 78). In short, this result shows that these recent attacks on the Given
have not succeeded in reviving it. The Given remains a myth.
The paper is structured as follows. First, we will outline Sellars’s
arguments against the epistemic significance of sensations and “lookings” or
“appearings.” Next, we will discuss several authors’ attacks on Sellars’s
position. Finally, we will discuss and extend Sellars’s views on internalism
and epistemic priority, and show how this aspect of Sellars’s philosophy
undermines his critics’ attempts to revive the Given.

1. The Given

In “Empiricism and the Philosophy of Mind” (1997: hereinafter, EPM), Sellars

is attempting to demonstrate how we can do without the Given in a number of
areas in philosophy. He wishes to show that we can explain linguistic
meanings, private episodes (both thoughts and impressions), empirical
“seemings” or “lookings,” and the justification of non-inferential observation
reports without appealing to the Given in any form. Despite the broad range of
targets discussed in Sellars’s essay, most philosophers’ discussion of the Given
focuses on the role the Given is supposed to play in observation. Clearly, this
is an important target for Sellars; consider the opening sentence in his essay:
 Sellars, Givenness, and Epistemic Priority 149

I presume that no philosopher who has attacked the philosophical idea of

givenness or, to use the Hegelian term, immediacy, has intended to deny that
there is a difference between inferring that something is the case and, for
example, seeing it to be the case. (1997, §1/p. 13)
Right from the start, Sellars puts the role played by the Given in
observation in his sights. Two pages later, he writes, “[T]he point of the
epistemological category of the given is, presumably, to explicate the idea that
empirical knowledge rests on a ‘foundation’ of non-inferential knowledge of
matter of fact” (1997, §3/p. 15). The recent criticisms of Sellars on which I
would like to focus are directed toward this notion of the experiential Given,
and the role it is supposed to play in foundationalist epistemology.
What is this notion of the experiential Given? Fundamentally, it is the idea
that any experience or sensation could be epistemically significant merely in
virtue of its occurrence. In the context of foundationalist epistemology, the
Given is an episode (generally, a sensation) which, merely in virtue of
appearing, justifies some belief or other. Thus, according to the Myth of the
Given, having a red sensation directly entails that one knows
(non-inferentially) that one is looking at a red object.1 For example, my belief
that this tie is green is, on this view, wholly justified by the presence of a green
sensation or sensing. It is this view that Sellars attacks.
A common form of the Myth of the Given is involved with sense data
theory, the theory that we sense sense-data, and it is these data that justify our
observation beliefs. Sellars famously presents the sense-data theorist with a
dilemma regarding the epistemic role of sensation. Sellars asks what exactly is
sensed in sensation. The two possible answers (particulars or facts) form a
dilemma:
(a) It is particulars which are sensed. Sensing is not knowing. The existence of
sense-data does not logically imply the existence of knowledge.
(b) Sensing is a form of knowing. It is facts rather than particulars which are
sensed (1997, §3/p. 16).
Neither of these two options, it is thought, allows the Given to play the
epistemic role it is supposed to play in foundationalism. As Sellars notes, on
the first option, sensing does not entail knowing. All knowing, according to
Sellars, involves classification. (In this, Sellars’s debt to Kant is clear, as Kant
held that all judgment involves subsumption under a general rule or category.)

1 Sellars (1997, §6/p. 21). In this section, Sellars spells out the notion in terms of the sense datum
theory. He writes that on the sense datum version of the Myth of the Given, “X senses red sense
content s entails x non-inferentially knows that s is red.”
150 Jeremy Randel Koons

If it is a particular that is sensed, then sensing cannot be knowing – the element

of generality is missing. Since a key element of the Given is that sensing
entails knowing, the first option abandons the Given. But on the second option
– sensing is knowing, and it is facts which are sensed – sensation plays a
justificatory role, but because sensing is knowing, it itself stands in need of
justification and support.
Sellars’s Dilemma has been given a reconstruction which extends its reach
beyond sense-data theories to any theory which tries to give sensation a
justificatory role in observation. Here is BonJour’s reconstruction:
The basic idea of givenness . . . is to distinguish two aspects of ordinary
cognitive states, their capacity to justify other cognitive states and their own
need for justification, and then to try to find a kind of state which possesses
only the former aspect and not the latter – a state of immediate apprehension or
intuition. But we can now see plainly that any such attempt is fundamentally
misguided and intrinsically hopeless. For it is clear on reflection that it is one
and the same feature of a cognitive state, namely, its assertive or at least
representational content, which both enables it to confer justification on other
states and also creates the need for it to be itself justified – thus making it
impossible in principle to separate these two aspects. (1985, p. 78)
Sellars responds to the dilemma by distinguishing between sensing and
non-inferentially knowing. Sensing is non-conceptual and non-epistemic – but
then sensing does not entail knowing. Sensing is a causal prerequisite for
knowledge, but it does not play a justificatory role. On the other hand, there is
non-inferential knowing, such as when I observe that a tie is green. In such
observations, it is facts that are known – but this knowledge of facts stands in
need of justification, and presupposes much other knowledge. And so
proponents of the Given will find no support from Sellars’s view of
observation and non-inferential knowledge.
Sellars’s discussion of the epistemic role of sensation continues with his
well-known discussion of “looks-talk.” Many philosophers have held that
sentences such as “The apple looks red to me” (a) are incorrigible and (b)
represent an autonomous type of discourse, one that (in Brandom’s phrase)
reports “a minimal, non-inferentially ascertainable, foundationally basic fact”
(1997, p. 139). A foundationalist might be tempted to use such “lookings” or
“appearings” as a foundation for our empirical knowledge; that is, she might
try to use how things look or appear as a foundation for building up knowledge
of how things are. Famously, Sellars denies both (a) and (b). Sellars writes,
Now the suggestion I wish to make is, in its simplest terms, that the statement
“X looks green to Jones” differs from “Jones sees that x is green” in that
whereas the latter both ascribes a propositional claim to Jones’s experience and
 Sellars, Givenness, and Epistemic Priority 151

endorses it, the form ascribes the claim but does not endorse it. (1997,
§16/pp. 40-41)
Thus, looks-talk only appears incorrigible; in reality, it is not incorrigible.
Nor is it corrigible: it is neither, since it is does not really assert anything.
Jones cannot be mistaken when he says, “This tie looks green to me,” but only
because he is not endorsing the claim that the tie is green. Brandom explains
the point as follows:
One can be wrong about whether something is green because the claim one
endorses . . . may turn out to be incorrect . . . But in saying that something looks
green, one is not endorsing a claim, but withholding endorsement from one.
Such a reporter is merely evincing a disposition to do something that for other
reasons (e.g., suspicion that the circumstances of observation lead to systematic
error) he is unwilling to do – namely, endorse a claim. Such a reporter cannot
be wrong, because he has held back from making a commitment. (1997,
pp. 141-142)
So looks-talk is not incorrigible; Sellars rejects (a) above. But he rejects
(b), too. Looks-talk does not form an autonomous discourse: facility with
looks-talk presupposes that one can make judgments about how things actually
are (as opposed to how things merely appear). Is-talk is conceptually prior to
looks-talk. One must first learn to make judgments concerning how things are.
Then, as one learns the distinction between standard and non-standard viewing
conditions, one learns to withhold endorsement from certain observational
judgments one is disposed to make. So one cannot master looks-talk until one
masters is-talk. Sellars infers from this that looks-talk cannot form an
autonomous discourse, suitable for serving as an epistemic foundation,
because knowledge of looks-talk presupposes other knowledge (knowledge of
is-talk and knowledge of standard viewing conditions). Brandom sums up
Sellars’s attack on (a) and (b) as follows:
We see . . . why [looks-talk] is precisely unsuited to use as an epistemological
foundation for the rest of our (risky, corrigible) empirical knowledge. For, first,
the incorrigibility of claims about how things merely look simply reflects their
emptiness: the fact that they are not really claims at all. And second, the same
story shows us that ‘looks’ talk is not an autonomous language game – one that
could be played though one played no other. It is entirely parasitic on the
practice of making risky empirical reports of how things actually are. (1997,
pp. 142-143)
Thus, lookings or appearings cannot serve as a foundation for our empirical
knowledge.
Recent criticisms have to a great extent focused on these points about
sensations (and lookings) in Sellars’s philosophy. That is, they have tried to
152 Jeremy Randel Koons

show how sensations or appearances can be conceptual in nature, or can justify

belief even if non-conceptual; or they have argued for the autonomy of
looks-talk; or they have otherwise attacked Sellars’s dilemma. Having attacked
this aspect of Sellars’s argument, these critics often conclude that they have
demonstrated that the Given is not, after all, a myth. We will briefly look at
some of these criticisms. I will not focus on them extensively, because I want
to argue that even if Sellars concedes that his critics are correct (i.e., that
sensation can be conceptual in nature, or that lookings can form an
autonomous discourse that might serve a justificatory role in observation
beliefs, or that the dilemma is a false dilemma), Sellars’s attack on the Given
remains cogent. This is because even if critics of Sellars can show, e.g. that
sensation is conceptual in nature, or that a non-conceptual sensation can still
justify an observation report, they have not addressed the question of what
other cognitive resources must be in place before these sensations (be they
conceptual or non-conceptual) can successfully justify knowledge claims. As I
indicated earlier, the answer to this question proves that the Given is a myth.

2. Some Recent (and Not So Recent) Attacks on Sellars

I want briefly to examine some recent (and some less recent) attacks on the
above aspects of Sellars’s arguments against the Myth of the Given. I will not
dwell too long on these attacks, since (as I have stated above) I think a
Sellarsian can concede the points made in these attacks and still maintain that
EPM demonstrates the Given to be a myth. But it is instructive to examine
these arguments and see where they focus on Sellars’s argument; we can then
see the crucial step which they omit.

2.1. Pryor
James Pryor (forthcoming) analyzes Sellars’s dilemma, and concludes that
Sellars’s argument rests on a crucial premise, which he calls the Premise
Principle. The principle reads as follows:
Premise Principle : The only things that can justify a belief that P are other
states that assertively represent propositions, and those propositions have to be
ones that could be used as premises in an argument for P. They have to stand in
 Sellars, Givenness, and Epistemic Priority 153

some kind of inferential relation to P: they have to imply it or inductively

support it or something like that. (forthcoming)2
First, Pryor argues that the Premise Principle does not in fact stand in the
way of experience justifying belief. Pryor writes,
many philosophers of mind these days think of experiences as having
propositional content . . . Without supplementation, the Premise Principle
would allow experiences to justify beliefs. For instance, an experience as of
having your hands could justify the belief that you have hands . . . [A]ccording
to the Premise Principle, the experience is able to justify that belief because of
the “inferential relations” its content stands into the content of the belief. (In
this case, the “inferential relation” is straightforward: the experience’s content
is the same as the content of the belief). (forthcoming)
Further, Pryor argues that the most plausible arguments in favor of the
Premise Principle fail. Thus, not only does the Premise Principle fail to forbid
experience to justify belief; the Premise Principle is without support. Hence,
Sellars’s argument is without foundation; it does not demonstrate that
experience cannot play a role in the justification of observation reports. Pryor’s
argument is too complex to summarize neatly, and so I am only showing the
strategy he pursues and the conclusion he ultimately wants to reach. The
conclusion is, as we will see, one that Sellars can concede without having to
concede that the Given is real.

2.2. Alston
In his (1998), William Alston criticizes Sellars’s argument from a similar
perspective. Alston wishes to argue that in sensation, we have non-conceptual
but nevertheless cognitive awareness of particulars; and that this cognitive
awareness represents the Given in observation. Alston writes,
[W]here do I dissent from Sellars’s attack on the given? It comes over the
question of whether we have a direct (nonconceptual) awareness of particulars,
one that constitutes a kind of cognition of a nonconceptual, nonpropositional
sort. Sellars, as I read him, is concerned to deny this . . . It is reasonably clear
that [Sellars] reserves the term ‘cognition’ for mental states or activities that are
conceptually, indeed propositionally structured. And it is that with which I take
issue. . . . I [will argue] that our direct awareness of X’s, and I will be thinking

2 I am not reproducing the reasoning that Pryor goes through in order to arrive at the conclusion
that the Premise Principle underlies Sellars’s argument, as I wish only to spend a brief time on
these arguments. Again, I wish to demonstrate that even conceding these critics’ arguments,
Sellars’s attack on the Given is still cogent.
154 Jeremy Randel Koons

primarily of perception here, provides a basis (justification, warrant . . . ) for

beliefs about those X’s. And this is a direct confrontation with Sellars’s
epistemological interest in the “Myth of the Given.” (1998, p. 2)
Alston’s argument focuses primarily on Sellars’s account of looks-talk.
The foundationalist is tempted to treat looks-talk as an autonomous discourse,
and “lookings” as a foundation for empirical knowledge. As Robert Brandom
writes,
Where collateral beliefs indicate that systematic error is likely, the subject
learns not to make the report ‘x is F’, to which his previously inculcated
responsive dispositions incline him, but to make a new kind of claim: ‘x looks
(or seems) F’. Of course it is tempting to take this as a new kind of report,
indeed, a report of a special kind of particular, a sense datum. This report, then,
is naturally thought of as reporting a minimal, noninferentially ascertainable,
foundationally basic fact, about which each subject is incorrigible. (1997,
p. 139)
Remember, Sellars argues that looks-talk cannot form an autonomous
discourse, and that looks-talk presupposes is-talk. On Sellars’s account, one
must first learn to report how things are. Thus, one acquires a disposition to
respond to (e.g.) an apple with the report “The apple is red.” This assertion
involves an element of endorsement: one commits oneself to the redness of the
apple by making this report. However, as one becomes a more sophisticated
perceiver, one learns that one’s dispositions to make color reports can be
misleading, for example in non-standard conditions. Thus, one learns to make
a new report: “The apple looks red.” This report is the same as the report that
the apple is red, except that it lacks the element of endorsement. One is
reporting one’s disposition to judge that the apple is red, but one is
withholding one’s endorsement from the observation report one is disposed to
make, because one is not sure that the disposition is not mistaken in this
instance. Thus, looks-talk and is-talk are conceptually interdependent. In
particular, is-talk is conceptually prior to looks-talk: you can understand the
former without the latter, but not vice versa.
Alston makes several important claims with regard to this argument. First,
Alston argues that looks-talk is not unequivocal, but that there are several
different species of looks-talk. One species is what Alston calls phenomenal
looks-talk, which describes the intrinsic nature of an experience. To say that
something (phenomenally) looks red is to describe one’s experience; it is to
say that one is having an experience of a particular nature. The foundationalist
(Alston implies) might use such a phenomenal looks-talk as a foundation from
which to build up our knowledge of the world.
 Sellars, Givenness, and Epistemic Priority 155

Second, Alston is willing to concede the conceptual interdependence of

looks-talk and is-talk, and that one might have to learn is-talk before learning
phenomenal looks-talk. Alston writes, “Here is one coherent scenario. In order
to acquire the phenomenal concept of looking red a neophyte must first learn
what it is for a physical object to be red and then be told that something looks
red if it looks the way a red object typically looks under standard conditions
(together with some instruction on what makes conditions standard or
nonstandard)” (1998, p. 10). Alston denies, however, that this interdependence
has important epistemological consequences: “The holism about concepts does
not carry implications for the nature of perceptual experience” (1998, p. 10).
Alston argues that this conceptual interdependence of the concept of is-talk
and the phenomenal concept of looks-talk does not show that lookings or
seemings are conceptual:
We cannot in general infer from ‘the concept of a G is dependent on the concept
of an H’ that G is itself conceptual in nature. The concept of a (chemical)
compound presupposes the concept of elements; but that hardly shows that
either a compound or an element contains or uses or is structured by concepts.
To come somewhat closer to home, the concept of a digestive tract presupposes
the concept of digestion, but that hardly shows that digestion is carried out by
the use of concepts or that the digestive tract (while digesting) makes a
“propositional claim.” (1998, pp. 9-10)
So the fact that the concept of looking red presupposes the concept of being
red does not presuppose that looking red is itself a conceptual process.
There is a third important claim Alston (1983) makes (although he makes it
in a different article from the one I have been discussing here). Sellars wants to
argue that if is-talk is conceptually prior to looks-talk, then looks talk is
somehow epistemically dependent on is-talk, and therefore lookings or
seemings cannot form a foundation for empirical knowledge. Alston denies
that conceptual independence (or even priority) entails epistemic dependence
or priority. In his (1983), Alston is discussing what “epistemizes”
observational beliefs, where an epistemizer is whatever, added to true belief,
forms knowledge. Here is Alston:
Unless I know something about the rest of the number system I cannot so much
as form the belief that 2 + 3 = 5, for I lack the requisite concepts. But all this
says nothing as to what epistemizes the belief, once formed, and it is on this that
the classification into immediate or mediate [belief] depends . . . To suppose
that the conditions for forming the belief are themselves conditions of
epistemization, and hence determinative of the choice between mediate and
immediate, is to confuse levels of questioning. It would be like arguing that
since a necessary condition of my making a request (orally) is that I have vocal
156 Jeremy Randel Koons

chords, part of what justified me in making that request is that I have vocal
chords. (1983, pp. 78-79)
Pryor makes a similar point, writing
the fact that you have immediate justification to believe P does not entail that
no other beliefs are required for you to be able to form or entertain the belief
that P. Having the concepts involved in the belief that P may require believing
certain other propositions; it does not follow that any justification you have to
believe P must be mediated by those other propositions. (forthcoming)
Bonevac also argues that the foundationalist “can resist the move from
logical priority to epistemic priority” (2002, p. 9).
To sum up, Alston’s counterattack against Sellars, focusing on Sellars’s
account of looks-talk, consists of three claims or arguments: (1) in addition to
the concept of looks-talk delineated by Sellars, there is also a phenomenal
concept of looking, pertaining to the intrinsic quality of a looking or seeming.
(2) The conceptual interdependence of looks-talk and is-talk (which Alston is
willing to concede) does not show that lookings or seemings are conceptual.
(3) The conceptual priority of is-talk over looks-talk does not show that
lookings are justified or epistemically mediated by beliefs about how things
are. By way of these three moves, Alston is clearing the ground for the
following position: our empirical knowledge is grounded in and justified by
non-conceptual lookings, whose justification is mediate (i.e., does not rest on
any other claim or belief). This position is, of course, the Given.

2.3. Vinci
Thomas Vinci’s argument is rather more easily summed up. We saw above
that Sellars presents a dilemma for the sense-datum theorist: is it facts or
particulars which are sensed? Vinci argues that this is a false dichotomy; he
argues that there is a third viable possibility: it is properties which are sensed.
Vinci thinks this addition allows the sense-datum theorist to escape Sellars’s
argument and revive the Given. Vinci reconstructs Sellars’s argument as
follows:
1) All knowledge of interest to foundationalists is propositional knowledge.
2) Knowledge of subject-predicate propositions is an act of classificatory
cognition.
3) All forms of proposition which can be known (and would on other grounds
be suitable as epistemic foundations) are, or depend on, acts of
classificatory cognition.
4) All classificatory cognition involves “learning, concept formation, even the
use of symbols.”
 Sellars, Givenness, and Epistemic Priority 157

5) [Therefore], all knowledge of interest to foundationalists involves “learning,

concept formation, even the use of symbols.”3 (1998, pp. 2-3)
Vinci proposes to argue that (3) is false. Sellars’s dilemma presupposes
that all acts of knowledge involve classification under a general concept or
rule. Thus, grasping a particular cannot count as knowledge; whereas grasping
a fact can count as knowledge, but this knowledge would stand in need of
justification (according to Sellars), and would therefore not be able to serve as
an epistemic foundation. Vinci, drawing on his understanding of Descartes’
notion of judgment as intellectual intuition, argues that grasp of properties can
be a legitimate act of judgment, but not an instance of classificatory cognition.
Thus, Vinci argues, there is a legitimate form of judgment (licensed by a grasp
of properties which are given to intuition) which does not presuppose
“learning, concept formation, even the use of symbols,” and which can serve
as the foundation for empirical knowledge.

2.4. Chisholm
Chisholm’s article is not recent, but it is one of the better-known attacks on
Sellars, and he follows the same pattern followed by the above critics.
Chisholm notes that statements about how I am appeared to seem to be
incorrigible. For example, if I believe that an apple appears white to me, my
belief cannot be mistaken. Sellars, of course, explains this by saying that I am
withholding endorsement from the claim that the apple is white: since I am not
asserting the proposition, I cannot be asserting anything false. Chisholm,
however, maintains that beliefs about appearances are self-justifying. He
writes, “The philosophers who talked of the ‘empirically given’ were
referring . . . to those [statements and beliefs] pertaining to certain ‘ways of
being appeared to’” (1986, p. 67).
Are such statements self-justifying? Chisholm cites Reichenbach’s
argument to the contrary: if I judge that an object appears white, I am
comparing it to other white objects I have seen. Thus, beliefs about
appearances are only justified by my justified beliefs about other white objects
I have seen and the sensations (appearances) they generated in me. However,
Chisholm argues that in addition to this comparative use of looks- or
appearance-talk, there is a non-comparative use, whereby “the point of
‘appears white’ is not to compare a way of appearing with something else; the
point is to say something about the way of appearing itself” (1986, p. 67). This
is similar to Alston’s phenomenal concept of looks-talk: such talk of looking
3 For purposes of clarity, I have renumbered Vinci’s original premises.
158 Jeremy Randel Koons

or appearing describes the intrinsic nature of the experience. Chisholm’s

defense of the notion of non-comparative looks talk seems to be as follows: if
we are to say, for example, “This object appears white,” then we are using the
word ‘white’. Knowledge of how the word ‘white’ is used in the English
language requires a comparison to white things, how other people use the word
‘white’, and so on. But judging that an object is white need involve no such
comparison. According to Chisholm, we must “distinguish between (a) what it
is that a man means to say when he uses certain words and (b) his assumptions
concerning the adequacy of these words for expressing what it is that he means
to say” (1986, p. 70). One should not “suppose . . . that what justifies (b) must
be included in what justifies (a)” (1986, p. 70).
Of course, there are veridical as well as non-veridical perceivings, and so if
we are to rest empirical knowledge on appearances,4 we must have some way
of determining which appearances are veridical and which ones are mere
appearances. However, Chisholm regards this problem as surmountable: “The
problem is one of formulating rules of evidence – a rule specifying the
conditions under which statements about what we think we [perceive] can
justify statements about what we do [perceive]” (1986, p. 72).5 He continues,
“The problems involved in formulating such rules of evidence, and in
determining the validity of these rules, do not differ in any significant way
from those which arise in connection with the formulation, and validity, of the
rules of logic” (1986, p. 72). Contemporary philosophers are much less
optimistic that this task can be so easily accomplished, but I will pass over this
without further comment. We will see, however, how his point about veridical
and non-veridical appearances, and the need for a logic of evidence, plays into
Sellars’s hands and allows Sellars to argue that appearances cannot be the
Given on which empirical knowledge rests.

2.5. Bonevac
By far the most sophisticated recent critic of Sellars’s attack on the Given is
Daniel Bonevac (2002). Bonevac’s main goal is to defend what he calls the
justification thesis. The justification thesis deals with the relation between

4 Chisholm actually does not think appearances are the sole foundation for our empirical

knowledge. He thinks there are several classes of self-justifying statements, including statements
of memory, reports as to what psychological state one is in, and so on. See Chisholm (1986).
5 Actually, Chisholm is talking about memory, not perception; but he writes later in the page, “If

we substitute ‘perceive’ for ‘remember’ in the foregoing, we can formulate a similar set of
problems about perception” (1986, p. 72). I have accepted his invitation to perform this
substitution.
 Sellars, Givenness, and Epistemic Priority 159

graspings (2002, p. 2: “inner episodes [which] are non-inferential knowings”)

and sensings (2002, p. 2: “inner episodes [which] presuppose no acquired
conceptual capacities [and which] are necessary conditions of graspings”). The
justification thesis states that “Sensings play a role in justifying graspings”
(2002, p. 3). According to Bonevac, the justification thesis is the crucial
element in the notion of the Given, and the primary goal of Bonevac’s paper is
to defend the justification thesis. Thus, we see that Bonevac’s approach to
defending the notion of the Given is similar to that pursued by the previous
authors we have discussed: he wants to argue that sensations (or appearings,
etc.) play a role in justifying observation beliefs. Further, some of the main
strands in his argument which follow lines similar to those followed by Pryor,
Alston, Vinci, and Chisholm. At the beginning of his (2002), Bonevac follows
Alston and others in arguing that looks-talk can form an autonomous
discourse; and that even if looks-talk presupposes is-talk, this conceptual or
logical priority does not translate into epistemic priority. Bonevac employs
other strategies in attacking Sellars; ultimately, he wants to conclude that
sensations (when accompanied by other acquired conceptual capacities and
beliefs) can play a role in the justification of observation reports. Thus, his
conclusion is similar to the conclusion reached by these other critics of Sellars:
sensation can play a justificatory role in observation. Bonevac writes:
Sensings, lacking propositional content, are not conceptual in the sense of
depending on the active use or even the nonoccurrent possession of acquired
conceptual capacities; they need only be conceptualizable in the sense of being
able to combine with acquired conceptual capacities to yield graspings. (2002,
p. 28)
As I noted above, Bonevac’s paper is among the more important of the
recent attacks on Sellars’s argument, and so it needs to be discussed along with
the other criticisms of Sellars. Unfortunately, Bonevac’s arguments are too
lengthy, numerous and sophisticated to summarize neatly. Thus, we must be
content to point out the elements of Bonevac’s argument that are similar to the
arguments we have already discussed, and to point out his conclusion. I stated
at the beginning that this conclusion is one Sellars can concede without having
to concede that the Given plays the role of the epistemological Given in
observational knowledge. Now I must make good on this claim. It is to this
task that we now turn.

3. Foundationalism and Epistemic Priority

What if the above critics are right? What if sensation is conceptual? Can it then
serve as an epistemic foundation? Or what if looks-talk can form an
160 Jeremy Randel Koons

autonomous discourse? Can it serve as an epistemic foundation? The above

critics infer, from the conclusion that sensations or lookings can play a role in
justification, that these sensations or lookings therefore constitute an empirical
Given. They infer, in short, that the Given is not a myth.
I wish to argue that Sellars can concede the above critics’ arguments
without being forced to concede that the Given can play the role in
epistemology that the foundationalists had in mind for it. Even if sensations
are conceptual, and can therefore stand in inferential relations to observational
reports; and even if looks-talk can form an autonomous discourse, and justified
looks-reports do not depend for their justification on other justified beliefs;
these concessions do not force Sellars to concede that empirical knowledge
rests on a foundational Given. For even if these concessions are made, and
even if sensations or looks-talk play a role in justifying observational reports,
there are nevertheless other empirical statements which are epistemically prior
to these sensations or these looks-judgment; and the power these sensations or
looks judgments have to justify observation reports rests in part on these
epistemically prior observation beliefs. Sellars writes,
If I reject the framework of traditional empiricism, it is not because I want to
say that empirical knowledge has no foundation . . . There is clearly some point
to the picture of human knowledge as resting on a level of propositions –
observation reports – which do not rest on other propositions in the same way
as other propositions rest on them. On the other hand, I do wish to insist that the
metaphor of “foundation” is misleading in that it keeps us from seeing that if
there is a logical dimension in which other empirical propositions rest on
observation reports, there is another logical dimension in which the latter rest
on the former. (1997, §38/p. 78)
We must now investigate what Sellars means when he says that there is a
sense in which observation reports rest on other empirical propositions, and
demonstrate that Sellars is correct. As noted in the introduction, Sellars says
too little about this notion of epistemic interdependence, and as a result, many
critics are not convinced that observation reports do rest on other empirical
knowledge. We will try to correct this shortcoming of Sellars’s account.
Let us consider the position of one of our critics. Let us suppose, for
example, that looks-talk is an autonomous discourse, and that statements of the
form “This apple looks red to me” can count as justified, and that their
justification does not rest on the justification of any other proposition or belief,
and that these lookings can in turn play a role in justifying observation reports
about how the objects in the world are (as opposed to how they merely
appear). Does this amount to a concession that lookings serve as a Given
which justifies an empirical foundation? It does not. It may be that sensations
 Sellars, Givenness, and Epistemic Priority 161

or lookings can play a role in the justification of observation reports. But the
crucial question is, “What else needs to be in place before these sensations (or
lookings) can serve such a justificatory role, and what is the relation of
epistemic priority between this ‘something else’ and these sensations (or
lookings)?” Sellars recognizes that we must possess other empirical beliefs
before our observation sentences can be justified, and that these other
empirical beliefs are in an important sense epistemically prior to the
observation reports (and the sensations and lookings on which these
observation reports may be thought to rest).
Let us ask how, for example, lookings are supposed to justify observational
beliefs. As Chisholm recognizes, we must make the leap from the fact that the
apple appears red to the statement that the apple is red. If we rest content with
the former, we have lapsed into solipsism or phenomenalism. If we want to
have justified empirical beliefs, we must make the leap to the latter – we must
be able to justify statements about the world. This is where Sellars can insert
his wedge into the foundationalist account.
Sellars writes that “before a token uttered by, say, Jones could be the
expression of observational knowledge, Jones would have to know that overt
verbal episodes of this kind are reliable indicators of the existence, suitably
related to the speaker, of green objects” (1997, §36/p. 75). There are two
important elements in this statement. First, Jones’s observation reports must be
reliable. Second, Jones must in some sense recognize his own reliability. Let
us focus on these two elements in turn.
First, Jones’s observation reports must be reliable. In what percentage of
circumstances must a belief-forming process produce true beliefs in order to
count as reliable, and indeed sufficiently reliable to be able to justify the
resulting beliefs?6 Suppose I have a disposition to form the judgment, “Lo! A
cat!” whenever confronted by a four-legged animal. Suppose, further, that only
10% of the animals I encounter are cats. In those cases where the animal I see
is in fact a cat, is my observation report justified? Clearly not. The mechanism

6 This talk of reliable belief-forming mechanisms should not be taken to mean I am imputing
reliabilist tendencies to Sellars. Although reliabilism as a theory did not exist when Sellars wrote
EPM (reliabilism was developed primarily in Goldman’s (1976) and (1979), which appeared two
decades after Sellars delivered the lectures constituting EPM), he foresaw such a development and
joined H.H. Price in dismissing it as the “thermometer view” of knowledge. Nevertheless, Sellars
recognized that reliability obviously had some role to play in the justification of observational
knowledge. We will see, though, that reliability must be combined with Sellars’s internalism if it
is to play an epistemic role.
162 Jeremy Randel Koons

which produced this report produces the correct judgment in too small a
percentage of cases. It does not differentiate adequately among different cases.
On the other hand, a belief-forming mechanism need not differentiate with
100% accuracy in order to count as reliable. If my belief-forming mechanism
does not differentiate between cows and yaks, but I am in an environment
where very few (say, 0.1% ) of the cow-like creatures I encounter are yaks,
then it seems we can say that my judgment “Lo! A cow!” is justified, at least in
those cases where I am, in fact, confronted with a cow.
Let us now return to the case of Jones. Consider his judgments as to
whether objects are, say, black. Suppose Jones has no understanding of
standard viewing conditions, and so reports “Lo! A black object!” whenever he
is confronted with a certain sensation (the sensation caused by black objects in
standard viewing conditions).7 Are Jones’s reports justified? It seems that they
are not. For if Jones is like most of us, he spends a substantial proportion of his
time in non-standard viewing conditions (such as at night). By night, for
example, many non-black objects appear black. Most dark colors (dark blue,
etc.) appear black under such conditions. Thus, if Jones has no understanding
of standard viewing conditions, and always judges black-looking objects to be
black, then Jones is never justified in reporting that an object is black, even if
the object is in fact black. As noted above, a belief-forming process must be
reliable a certain percentage of the time in order for it to produce justified
beliefs. It must be sophisticated enough to discriminate, at least in the great
majority of cases actually encountered by the epistemic agent, between the
objects it claims to purport (cats, cows, and black objects) and objects which
do not satisfy this description. Since Jones’s perceptual mechanism does not
discriminate between black objects and objects that merely seem black, and
since Jones spends a good deal of time in circumstances where non-black
objects appear black, Jones’s judgments concerning which objects are black
are not justified.
And of course, it is not merely in the absence of light that our color reports
are unreliable. We spend much of our time in conditions where there is a good
deal of light, but it is of a poor quality. To give a humble example, my wife
and I recently decided to paint a spare bedroom a pale yellow, and painted a
portion of the wall to see how the color looked. We returned a few hours later
to see how the color looked once it had dried. We found ourselves initially

7 Some Sellarsians will no doubt gripe about this talk of “being confronted with a particular
sensation,” arguing that it concedes too much to Sellars’s critics. But the point of this paper is that
even if we concede the legitimacy of such talk to Sellars’s critics, Sellars still has decisive
arguments against the notion of the experiential Given.
 Sellars, Givenness, and Epistemic Priority 163

unable to see where we had painted, and although we several seconds of

staring at the wall revealed where we had painted, we could not adequately
discern the color. Why? Because it was dark outside, and we were relying on
our indoor lighting which, while plentiful, was of a poor quality for
distinguishing colors. Of course, rather than decide that the paint had turned
off-white during the last few hours, or deciding that paint varied its colors,
chameleon-like, during different portions of the day, we merely withheld
judgment. Upon viewing the paint by daylight the next day, we were not only
able to see easily where we had painted, but were able to discern that the
yellow had too much green in it, and was therefore not suitable for the room.
A perceptual judge need understand more than adequate lighting in order to
produce justified perceptual judgments. If we consider the nature of sensation,
and the judgments which these sensations justify, we realize that we must have
a number of justified beliefs if we are to form justified perceptual judgments.
Let us give some examples:
1) A person who did not understand motion and blurring would judge that
the spokes on a car disappeared or radically changed in composition when the
car was in motion, as the spokes would cease to be (individually) visible. The
same considerations would apply to a number of familiar objects, such as the
rotors on a helicopter, or a jump-rope, or a whip, or the limbs of a runner, or a
spinning top, or . . .
2) A person who did not understand light and shadow would think the grass
under a tree was mottled. Indeed, someone who did not possess the concept of
a shadow would be an exceedingly poor judge of color, since everywhere we
look (especially outdoors during the daytime, which is supposed to be the
paradigm of standard viewing conditions!) we see color variations caused by
shadows. A large percentage of such a person’s color judgments would be
unreliable, since the person would suppose objects to be mottled, or striped, or
would assume various objects to be darker in color than they in fact were (due
to their being in shadow), etc.
3) A person who did not understand distance (and the effect distance has
on perception) would assume that people and objects radically change size
inexplicably.
4) A person who did not have at least a rudimentary understanding of
backlighting would assume that people’s features disappeared when indoors
and in front of a window.
5) Many glass surfaces in my environment (including the glass in my office
window, and several home windows) have imperfections and variations that
cause objects outside to appear wavy, etc. Lack of correction for this
phenomenon would result in my judging that the objects I observed actually
were wavy, blurry, etc. Further, the effect is dynamic: if I see Smith through
164 Jeremy Randel Koons

my window, and then a few minutes later he walks into my office, I would
(without the concepts glass, blurry, etc.) assume that Smith had wavy hair and
an oddly-shaped skull, but that his hair and skull returned to “normal”
sometime between my seeing him through the window and his entering my
office.
6) Look at all of the shiny or glossy objects around you, including plastic
bottles, objects painted with glossy paint, glass shelf doors, books, file
cabinets, thumbtacks, and so forth. All of these reflect, to a greater or lesser
degree, other nearby objects. A person who did not understand the notions of
reflection, shiny, etc. would judge that these objects were not uniform in color,
that their color changed as you moved in relation to them, and so forth.
The list could go on indefinitely. But I think the point is clear. In order to
navigate the world, to make reliable perceptual judgments, and so on, one must
possess a wide variety of concepts. Various observational judgments rely
epistemically on the notions of shadow, reflectance, glass, motion, distance,
etc., etc. If you do not understand distance, your judgments about size are
unreliable and hence not justified. There is a relation of epistemic priority.
Without understanding the concept of motion and blurring, your judgments
about the existence and composition of rapidly-moving objects are unreliable
and not justified. Again, certain bits of empirical knowledge (about motion,
etc.) are epistemically prior to certain observation reports: the latter cannot be
justified without the former. So the justification of our observation judgments
relies on much other knowledge. There is a clear sense in which “if there is a
logical dimension in which other empirical propositions rest on observation
reports, there is another logical dimension in which the latter rest on the
former” (1997, §38/p. 78).
One might charge that I am confusing conceptual with epistemic priority, a
confusion which Alston, Pryor, and Bonevac all warn against. But I am not
making this confusion. I am arguing that if you do not possess certain
concepts, your observational mechanism will be insufficiently reliable and its
deliverances will therefore not be justified. The possession of these concepts is
necessary for the reliability of your observational mechanism, and therefore
for the justification of your observational judgments. Thus, there is clearly a
relation of epistemic priority (or at least interdependence): adequate perceptual
discrimination, and therefore observational judgment, presupposes and
requires a body of empirical knowledge, including knowledge of distance,
motion, etc., as the above examples illustrate.
Thus, even if lookings play a role in justifying observation reports, they
can only play this role because we possess a large body of empirical
knowledge and empirical concepts, some of which concepts (such as distance)
are only indirectly related to the notion of observation. A similar point applies
 Sellars, Givenness, and Epistemic Priority 165

to sensations. Alston writes that “beliefs about what is perceived can be

justified by a nonconceptual experience from which they spring” (1998, p. 13).
But again, sensations (such as a sensation of blackness) can only justify
observation reports if the observer possesses other empirical knowledge and
other empirical concepts. Thus, Sellars can concede to his critics that lookings
or sensations play a role in justifying knowledge. But this does not show that
these lookings or sensations play a foundational role, which the Given was
supposed to do. Similarly with Vinci’s point: even if we concede that grasp of
properties is an instance of genuinely non-classificatory judgment, such a
judgment cannot count as knowledge unless many other empirical concepts are
in place.
Sellars (and his critics) are wont to put the point about epistemic priority in
terms of standard viewing conditions. Here is Sellars:
Now, it won’t do to reply that to have the concept of green, to know what it is
for something to be green, it is sufficient to respond, when one is in point of fact
in standard conditions, to green objects with the vocable “This is green.” Not
only must the conditions be of a sort that is appropriate for determining the
color of an object by looking, the subject must know that conditions of this sort
are appropriate. (1997, §19/p. 44)
Critics also restrict their attention to the notion of standard viewing conditions.
Bonevac writes, “Why . . . does this require any more than knowledge that
conditions are normally suitable for color perception, or that, for example, it is
day? In short, one cannot get much from the requirement of knowledge of
standard conditions” (2002, p. 14). But both Sellars and Bonevac understate
the amount of knowledge that is required before one can make justified
perceptual judgments. Certainly, one must know that conditions are standard.
But even within standard viewing conditions, one must have a mastery of
various concepts – motion, reflection, shadow, and so forth – which alter the
normal appearance of perceptible objects. So it is not enough to know that
“conditions are normally suitable for color perception.” For most of the
examples I gave above are of concepts that one must grasp before making
perceptual judgments (and not just color judgments) in standard conditions.
Sellars’s anti-foundationalist argument was that even if we possess
perceptual concepts (such as green), our observation judgments employing
these concepts rely (in part) for their justification on other knowledge we have,
namely, knowledge of standard viewing conditions. Here, I am arguing that
knowledge of standard viewing conditions is not the only knowledge required
for us to render justified observational reports. There are many other concepts
which we must be able to apply before our standard observational judgments
can count as justified. Observation judgments are therefore epistemically
166 Jeremy Randel Koons

dependent not just on knowledge of standard viewing conditions, but also on

knowledge of other factors which bear on the nature of our perceptual
experience. We will be able to say a bit more about this epistemic
interdependence shortly.
I noted above that there were two important elements in Sellars’s account
of perception that we needed to address. First, if Jones’s observation reports
are to be justified, Jones’s observation reports must be reliable. Second, Jones
must in some sense recognize his own reliability. We have addressed the first
issue – reliability – over the past few pages. But we must now turn our
attention to the second issue, the issue of Sellars’s internalism. For remember,
Sellars writes that “the subject must know that conditions of this sort are
appropriate” (1997, §19/p. 44) for making perceptual judgments; and also
“before a token uttered by, say, Jones could be the expression of observational
knowledge, Jones would have to know that overt verbal episodes of this kind
are reliable indicators of the existence, suitably related to the speaker, of green
objects” (1997, §36/p. 75). This internalism is important for Sellars’s
anti-foundationalist argument. Sellars’s argument that observational beliefs are
not foundational amounts to the claim that these observational beliefs are not
justified unless the agent possesses other knowledge, namely, knowledge
regarding standard viewing conditions (and, I would add, the above-listed
perception-altering factors). These observational beliefs therefore depend for
their justification on other knowledge the agent possesses. That is why they
are not foundational, in the full traditional sense of ‘foundational’. So Sellars’s
anti-foundationalist argument relies on Sellars’s internalism. This also
distinguishes Sellars’s focus on reliability from the reliabilism of later authors
such as Alvin Goldman: for Sellars, the reliability must have an internalist
element (the agent must in some sense recognize her reliability) in order to
have any epistemic import.
The requirement that the agent must know that conditions are suitable for
perception has attracted a good deal of criticism, mostly arguing that it
demands too much of agents. Bonevac’s comments on this point are typical:
“Surely a subject does not have to be able to specify standard conditions for
color perception with any precision. (I, at any rate, surely could not)” (2002,
p. 14). However, nothing in Sellars’s view requires that the agent be able to
describe or give an account of standard viewing conditions. In general, people
who have a skill or ability to do something (such as ride a bike, or snow ski,
etc.) cannot translate this skill or ability into a set of propositions. Indeed,
skillful practitioners often find that attempting to break their ability down into
 Sellars, Givenness, and Epistemic Priority 167

propositional knowledge, or attempting to explicitly follow a set of rules in

performing the relevant activity, actually results in a deterioration of their
performance.8 Rather, one must have the ability to adjust one’s perceptual
judgments when one recognizes that conditions are non-standard. It is a
question of know-how, not propositional knowledge. Thus, one need not be
able to list all the relevant factors that need to be taken into account when
forming perceptual judgments, but one must be able to make these
adjustments, and must possess the concepts necessary to make these
adjustments. For example, I do not suppose the average perceiver could
produce, off the top of her head, the list of perception-altering factors I gave
above. It required considerable deliberation for me to produce it, and it is of
course woefully incomplete. But every competent perceiver can alter her
perceptual judgments in accordance with the elements on the list. Every
competent perceiver possesses the concept of reflection, and can judge that
(say) a glossy item is reflecting nearby objects, and is not (necessarily)
irregular in color. Every competent perceiver possesses the concepts of motion
and blurring, and can judge that a helicopter’s rotors do not disappear in flight.
Every competent perceiver possesses the concept of shadow, and knows that
the objects she views outdoors in the daylight (when shadows are everywhere)
are not really mottled or striped in color. If one did not possess these concepts,
and did not possess the dispositions to alter one’s perceptual judgments in
accordance with them, then one would not count as a competent perceiver,
one’s perceptual judgments would be wildly erratic and inaccurate, and one’s
perceptual judgments would not be justified. This is not to say that the
competent perceiver must be able to produce a list of rules she is following,
any more than a competent speaker of a natural language must be able, on
demand, to recite all the grammatical rules of her language. Again, it is a
question of know-how, not a question of explicit propositional knowledge of
standard viewing conditions and the factors that must be taken into account
when forming perceptual judgments. This account is in the spirit of Sellars’s
internalism, since this know-how is practical knowledge that the agent must
possess in order to be a competent perceiver.
Of course, the presence of these factors (shadow, movement, etc.) is not
inferred in an observational judgment, any more than the observational
judgment itself is inferred. As Sellars and subsequent authors claim,
observation reports are non-inferential, in that they are not arrived at via
inference. Thus, one does not infer that an object is in shadow; one sees that it

8 See, for example, Dreyfus and Dreyfus (1986), especially chapter 1.

168 Jeremy Randel Koons

is in shadow. However, if one were unable to see that some objects are in
shadow (or in motion or . . . ), then one’s color (and other perceptual)
judgments would not be justified, because one would form false judgments,
like “The grass under that tree is mottled in color,” or “The roof of the building
I am observing through this window is curved and wavy.” So these perception-
altering factors, such as shadow and motion, play the same non-inferential role
in perceptual judgment as do the perceptual concepts (such as color and shape)
with which they interact. But crucially, one must possess these concepts (like
motion and shadow) if one’s perceptual judgments regarding color, shape, and
so forth are to be justified. These concepts interact in various ways, and it is
only in virtue of possession of the former concepts that we can accurately
apply the latter. Ergo, observation reports involving the latter depend, for their
justification, on our knowledge of and disposition to apply the former, even
though all of these concepts are applied non-inferentially by the skilled
perceiver.
We can now see more clearly the relation between conceptual
interdependence and epistemic priority. As we noted above, Alston, Bonevac,
and Pryor all argue that conceptual interdependence does not entail epistemic
interdependence. But we need to distinguish between varieties of ways in
which concepts can be interdependent on each other. The most familiar way is
conceptual holism: one concept has its meaning only in virtue of being part of
a network of other concepts. Alston, Bonevac, and Pryor all argue that two
concepts can be interdependent in this way without being epistemically
dependent on each other. But the type of interdependence at work in Sellars’s
argument is interdependence in application (not necessarily interdependence in
meaning): certain features of the world (such as color and shadow) cannot be
coinstantiated without altering the manner in which each is instantiated. That
is, shadow cannot appear with color without altering the appearance of the
color. Thus, one who wishes to apply the concept of color in perception must
also be able to apply the concept of shadow, since the presence of shadow
alters the appearance of color. As I have argued above, there are enough such
perceptual factors (motion, shadow, reflection, etc.) which alter the appearance
of our normal perceptual concepts (color, shape, size, etc.) that if we did not
have knowledge (practical, how-to knowledge, not propositional knowledge)
of how to apply the former, the observational judgments we made using the
latter concepts would not be sufficiently reliable and would therefore not be
justified. Thus, whereas one type of conceptual interdependence (the type
embodied by meaning holism) might not entail epistemic interdependence, a
different type of conceptual interdependence – interdependence in application
– does have epistemic consequences. This latter type of interdependence
 Sellars, Givenness, and Epistemic Priority 169

shows that the justification of observational judgments presupposes knowledge

of (and ability to apply) a wide variety of other empirical concepts.

4. Why the Given Cannot Do What the Given Was Supposed to Do

Bonjour (1986) explains the role that the Given was supposed to play in
epistemology. Bonjour describes foundationalism as a proposed solution to the
epistemic regress problem. Briefly, if we ask what justifies a particular belief
A, we might be told that it is justified by B. B, in turn, is justified by C.
Assuming that this regress cannot go on ad infinitum or circle back on itself, it
must terminate with a belief (or set of beliefs) which is self-justified.9 This set
of beliefs is to serve as the foundation on which all of our other beliefs rest.
Crucially, if these foundational beliefs depend, for their justification, on other
empirical beliefs, then they are not really foundational.
Bonjour asks, “In virtue of what features is this foundational belief
justified?” It cannot be justified by some further belief of ours, for that would
mean the “foundational” belief was not, in fact, foundational after all. What
alternatives remain? Bonjour identifies two. The first alternative is
externalism: the belief is justified by something, but the belief’s possessor
need not know or even justifiably belief that this something justifies the belief
in question. The second alternative is that the belief in question is justified by a
Given. This view, Bonjour writes, is “so venerable that it deserves to be called
the standard foundationalist solution to the problem in question” (1986,
p. 106). In describing this second option, Bonjour writes that the
foundationalist
. . . might grant that it is necessary both that such justification [for the basic
belief] exist and that the person for whom the belief is basic be in cognitive
possession of it, but insist that his cognitive grasp of the premises required for
that justification does not involve further empirical beliefs which would then
require justification, but instead involves cognitive states of a more rudimentary
sort which do not themselves require justification: intuitions or immediate
apprehensions. (1986, p. 102)
For foundational beliefs to be genuine “regress stoppers,” they cannot
depend for their justification on any other belief (at least, not on any belief the
agent is required to justifiably believe). For the Given to play its role in this
scheme, it must completely and all on its own justify the observational belief in

9 I am ignoring the possibility that the terminal belief is not justified.

170 Jeremy Randel Koons

question. But we have seen that the Given cannot do this. Intuitions or
immediate apprehensions might play a role in justifying foundational
observational beliefs, but as we saw they can only play this role if the agent
possesses other, epistemically prior concepts – concepts like shadow, motion,
reflection, standard conditions, and so on. Thus, the Given cannot enable
observation beliefs to serve as regress-stoppers, since the Given can only serve
its justificatory role in the presence of other, epistemically prior beliefs and
concepts.
Further, we saw that the agent must actually possess these beliefs and
concepts, and possess (as practical, know-how knowledge) the ability to apply
these concepts in making observation reports. Thus, BonJour’s first option –
externalism – is not a live option, either. The chief conclusion to be noted is
that the Given can only serve the role that foundationalists identified for it if it
could by itself justify foundational beliefs, thereby rendering these beliefs
genuinely foundational – that is, able to justify our other empirical beliefs
without standing in need of justification by these same empirical beliefs. As
we have seen in this paper, the Given cannot fulfill this function. Sellars is thus
pushed back into a more coherentist model of justification. Of course,
coherentism has its own problems. In fairness to Sellars, he claims that he is
neither a foundationalist nor a coherentist.10 Whether his position avoids the
pitfalls of both views is a topic for another paper. But whatever the status of
Sellars’s positive epistemological project, his critical project is, as I have
argued here, a success: the Given cannot play the role it was supposed to play
in traditional foundationalist epistemology.

Conclusion

At the end of his (2002), Bonevac summarizes the picture of perception that
emerges from his attack on Sellars:
The general picture that emerges, then, is this. Sensings by themselves do not
entail graspings,11 since the former require no acquired conceptual capacities
and the latter do. But occurrences of sensings defeasibly imply graspings for

10 “One seems forced to choose between the picture of an elephant which rests on the tortoise
(What supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail
in its mouth (Where does it begin?). Neither will do” (1997, §38/pp. 78-79).
11 As we noted earlier, for Bonevac graspings are “inner episodes [which] are non-inferential

knowings,” and sensings are “inner episodes [which] presuppose no acquired conceptual
capacities [and which] are necessary conditions of graspings” (2002, p. 2).
 Sellars, Givenness, and Epistemic Priority 171

beings who already have the relevant conceptual capacities and notice the
relevant features of the sensings. A sensing plays a role in justifying a grasping
in much the way that Troy plays a role in justifying the claim that Troy is not
mythical, or that my headache plays a role in justifying the claim that my
headache is physical:
My headache exists.
My headache is spatio-temporally located.
Anything that exists and is spatio-temporally located is
physical.
Therefore, my headache is physical.
A sensing plays a role in justifying a grasping without itself having
propositional content by having the proposition that it exists or occurs serve as a
premise from which the grasping can be inferred. (2002, pp. 27-28)
But this does not provide any solace for the foundationalist. For as we have
seen, Sellars’s point about the relative epistemic priority of observation reports
and other empirical beliefs shows that a sensation (or an appearing/looking,
etc.) cannot by itself justify a belief. So when Bonevac concludes his
interesting critique of Sellars by saying that a sensation can, along with other
empirical beliefs and acquired conceptual capacities, justify a belief, that does
not show that the Given is revived. For the Given, if it cannot play the role
epistemologists wanted it to play in foundationalist epistemology, is a
defanged and tame Given. Its epistemic status is more myth and legend than
reality.12

Jeremy Randel Koons

Department of Philosophy
American University of Beirut
Beirut 1107 2020
Lebanon
e-mail: [email protected]

REFERENCES

Alston, W.P. (1983). What’s Wrong with Immediate Knowledge? Synthese 55, 73-95.
Alston, W.P. (1998). Sellars and the “Myth of the Given.” Presented at the 1998 Eastern Division
Meeting of the American Philosophical Association. Reprinted at http://www.ditext.com/
alston/alston2.html.

12 I am grateful to Michael P. Wolf for helpful comments on a draft of this paper.

172 Jeremy Randel Koons

Bonevac, D. (2002). Sellars vs. the Given. Philosophy and Phenomenological Research 64, 1-30.
Bonjour, L. (1985). The Structure of Empirical Knowledge. Cambridge, MA: Harvard University
Press.
Bonjour, L. (1986). Can Empirical Knowledge Have a Foundation? In: P.K. Moser (ed.),
Empirical Knowledge: Readings in Contemporary Epistemology, pp. 95-115. Lanham, MD:
Rowman and Littlefield Publishers, Inc.
Brandom, R. (1997). Study Guide. In: W. Sellars, Empiricism and the Philosophy of Mind,
pp. 119-181. Cambridge, MA: Harvard University Press.
Chisholm, R. (1964). Philosophy. Englewood Cliffs, NJ: Prentice-Hall.
Chisholm, R. (1986). The Myth of the Given. In: P.K. Moser (ed.), Empirical Knowledge:
Readings in Contemporary Epistemology, pp. 55-75. Lanham, MD: Rowman and Littlefield
Publishers, Inc. Reprinted from R. Chisholm (1964), pp. 261-286.
Dreyfus, H.L. and S.E. Dreyfus (1986). Mind over Machine. New York: The Free Press.
Goldman, A.I. (1976). Discrimination and Perceptual Knowledge. The Journal of Philosophy 73,
771-791.
Goldman, A.I. (1979). What is Justified Belief? In: G.S. Pappas (ed.), Justification and
Knowledge, pp. 1-23. Dordrecht: D. Reidel Press.
Pryor, J. (forthcoming). A Defense of Immediate Non-Inferential Justification. In: M. Steup and
E. Sosa (eds.), Contemporary Debates in Epistemology. Malden, MA: Basil Blackwell Pub.
Reynolds, S.L. (1991). Knowing How to Believe with Justification. Philosophical Studies 64,
273-292.
Sellars, W. (1997). Empiricism and the Philosophy of Mind. Cambridge, MA: Harvard University
Press.
Vinci, T. (1998). The Myth of the Myth of the Given. Reprinted at http://www.ditext.com/
vinci/mmg.html.
Susanna Schellenberg

SELLARSIAN PERSPECTIVES ON PERCEPTION

AND NON-CONCEPTUAL CONTENT

ABSTRACT. I argue that a Sellarsian approach to experience allows one to take

seriously the thought that there is something given to us in perception without denying
that we can only be conscious of conceptually structured content. I argue against the
traditional empiricist reading of Sellars, according to which sensations are understood
as epistemically graspable prior to concrete propositional representations, by showing
that it is unclear on such a view why sensations are not just the given as Sellars so
famously criticizes it. I suggest an alternative transcendental reading, according to
which there are two sides to the subject matter of perceptual judgments: The matter
given in perception (sensation), and its form (intuition). I present an account of
sensations and intuitions on which it is unproblematic to see sensations as what is given
in perception: They are not intelligible independently of their role as the matter of
intuitions, the content of which is accessible to us only in the context of a judgment.

It seems plausible that something is given to us in perception that is the basis

for the judgments we form about objects in view.1And it is tempting to say that
what is given in perception has content in some way without being conceptual.
It seems plausible, on the other hand, that only conceptually structured content
is accessible to us. Approaches that assume perception to be providing us with
(non-conceptual) content, which cannot be expressed in a judgment, play down
or circumvent the second of these two suppositions. In contrast, many

1 This paper has gone through several stages and I am indebted to many people for help. I thank

Robert Brandom, Joseph Camp, Philip Catton, Rick Grush, Henry Pickford, Kevin Scharp, and
Lionel Shapiro for helpful comments on an early draft of this paper and Willem deVries, Kieran
Setiya, and Wayne Wu as well as audiences of meetings of the Joint Session of the Aristotelian
Society and Mind Association in Belfast, the American Philosophical Association in Washington,
the Gesellschaft für analytische Philosophy in Bielefeld, and the Canadian Philosophical
Association in Halifax (all in 2003) for comments on a later version. I benefited from comments
by Stephen Engstrom on the most recent version. Finally, I am indebted to Matt Boyle, Anil
Gupta, and John McDowell for invigorating discussions and detailed comments on multiple drafts
of this paper.

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 173-196.
Amsterdam/NewYork: Rodopi, 2006.
174 Susanna Schellenberg

approaches that assume perception to be concept-dependent drop the first

supposition, arguing that our perception of objects must be conceptual from
the outset.
Sellars is normally taken to be a radical supporter of the thesis that our
perception does not involve a given (see most recently Alston 2002). I will
argue, however, that a Sellarsian approach to perception allows one to take
seriously the idea that perception is concept-dependent without denying that
something is given to us in perception. My argument proceeds in two steps.
First, I argue that the concepts that we employ in perceptual judgments are
related to objects in that they involve rules of application specifying the
circumstances under which it is correct to use a word and the practical
consequences of its use. But taking application rules into account is, I will
argue, not sufficient to explain what is involved in directing a thought at this
rather than that particular object in perception. In a second step, I address this
more specific question. I show that we can consider ourselves to be referring to
objects, not by setting up atomistic semantical relations between expressions
and objects, but by picking out particular objects in the context of a full-blown
language. I argue for this second point by distinguishing two possible readings
of Sellars’s interpretation of sensations (sinnliche Eindrücke) and Kantian
intuitions (Anschauungen). On what could be called an empiricist reading,
sensations are understood as the input on the basis of which we form basic
conceptual representations, namely intuitions, and full-blown propositional
representations. I argue that it is unclear on such a reading why sensations and
intuitions are not the given as Sellars so famously criticizes it.
This understanding of the sensible conditions of perception will be
contrasted with what could be called a transcendental understanding of
sensations and intuitions. On such a reading, intuitions and sensations are what
we must assume figure in our perception if we think of our perception as
having a subject matter. On Kant’s picture, there are two sides to the sensible
conditions without which we cannot think of ourselves as having perception of
objective reality: The matter given in perception (sensation) and its form
(intuition). While intuitions are not graspable independently of being
subsumed under concepts in the context of a judgment, sensations are not
graspable independently of being formed into empirical intuitions. Although
sensations can be seen as what is given in perception, they are intelligible only
as the matter of intuitions, the content of which is intelligible only in the
context of a judgment.
The question of what it means to have perceptions of this rather than that
particular object is approached by thinking through Kant’s understanding of
intuitions and sensations because, I believe, we have much to learn from his
insights in these matters. I see Kant as discussing the materials necessary for
 Sellarsian Perspectives on Perception and Non-Conceptual Content 175

the construction of a framework that leaves room for the suppositions

motivating theories that assume the possibility of non-conceptual content,
without thereby falling prey to a crude empiricist account of perception.2
Thinking through Sellarsian and Kantian approaches to perception will serve
as a springboard to bring out a way of making sense of the idea that something
is given in perception that is not as yet conceptually structured. Thus, this
paper meshes historical and systematic interests.
For Kant the fundamental nature of empirical judgments is their
dependence on receptive intuitions. This dependence on intuitions is an insight
shared with the empiricist tradition. But in contrast to philosophers standing in
that tradition, Kant sees the intelligibility of empirical intuitions as necessarily
relying on a conceptual framework. Intuitions provide direct links to objects,
but these links are not atomistic content-generating links, since we can only
grasp the content of intuitions in the context of a judgment. Discussing these
issues quickly leads to the question of how the relation between intuitions and
demonstratives should be understood. In a third part, I will address this
question by discussing Sellars’s interpretation of intuitions as having the form
“that cube.” Again I will distinguish two possible readings of such a view. I
argue that intuitions must not be understood as a separable contribution to
perceptual judgments, but as providing for the possibility of such judgments.

1. Functional Roles and Rules of Application

The question of what it means to direct one’s thought at a particular object has
a semantical as well as an epistemological dimension. From a semantical
perspective, the question is how the content of our thoughts determines what
the thoughts are about. From an epistemological perspective, the question
takes the form of how our thoughts are affected and guided by the objects that
they are about. Sellars approaches the problem of perceptual content from both
angles. He criticizes ambitions to ground the contents of our expressions by
way of atomistic content-generating links to objects and develops his positive
account by arguing that the contents of our expressions must be understood in
terms of the functional roles of these expressions. This view leads him to
maintain that there are no direct semantical relations between words and
objects and, thus, to a radical non-relational semantics. The same thoughts take

2The term ‘non-conceptual content’ is used rather differently in different accounts of perception.
Throughout this paper, it will be understood as content that is accessible to the perceiving subject
without bringing any conceptual capacities into play.
176 Susanna Schellenberg

an epistemological twist when he argues against the possibility that sense data
are given to us in perception as something that we know solely by
acquaintance.
The main difficulties with Sellars’s account are all connected to the
question of how expressions can be seen as referring to particular objects in
perception.3 To bring out the nature of the problem it will be necessary to lay
out the basic semantical ideas that motivate Sellars’s account of perception. In
“Being and Being Known,” Sellars argues that conceptual episodes “differ
intrinsically qua acts in a way which systematically corresponds to what they
are about, i.e., their subject-matter” (1960, p. 43). He insists, however, that this
is not meant to imply that there are semantical relations between the
conceptual and the real orders. He contends that our thoughts can stand in
semantical relations only to elements in the conceptual order and argues in
Science and Metaphysics that this “non-relational character of ‘meaning’ and
‘aboutness’” is the “key to a correct understanding of the place of mind in
nature” (1967a, p. ix).
What does it mean to say that there are no semantical relations involved in
perception? On what one could call a strict non-relational view, perception
does not represent particular objects at all. Such a view is motivated by the
idea that the perception of two indistinguishable objects is exactly the same.
Martin Davies, for example, argues that if two objects are perceptually
indistinguishable for a subject, then the perception of one has the very same
content as the perception of the other (1997, p. 314).
Such a strict non-relational view is counterintuitive: If the content of the
perceptions of two indistinguishable objects is the same, neither perception
seems to represent the particular object in view. Of course, it need not follow
from such a view that perception does not represent the world. On Davies’s
account “[o]ne way to see how perceptual content can be truth conditional,
although not object-involving, is to take perceptual content to be existentially
quantified content” (1997, p. 314). This is just to say that two perceivers
visually confronted with two distinct, but indistinguishable pens on a table
represent that there is an object of a particular type lying on an object of
another particular type. But this still leaves the question of how we can think
of ourselves as perceiving this rather than that object: When a person perceives
a pen, she knows which pen she is looking at, namely the one in view. As
P. F. Strawson argues convincingly in his discussion of the possibility of

3 I thank Lionel Shapiro for long discussions that helped me see more clearly how to think about
the questions addressed in this section.
 Sellarsian Perspectives on Perception and Non-Conceptual Content 177

massive reduplication in the first chapter of his Individuals (1959), knowledge

about mind-independent objects depends upon perceptual representations of
particular objects in the world. Such knowledge is not accounted for on the
strict non-relational view: Perception only provides a perceiver with
knowledge that there is a pen of a certain type lying on a table of a certain
type.
One possible way to account for our ability to perceptually refer to this
rather than that particular object is to acknowledge that demonstrative
elements figure in the content of perception. By involving the reference of a
demonstrative element that is fixed by the context of the perception in which it
occurs, one can take into account that one and the same demonstrative element
refers to different particular objects in different contexts. In contrast to the
strict non-relational view, such a non-relational view involving demonstratives
finds a place for the representation of particular objects in perceptual content.4
Sellars acknowledges that more needs to be in play than the functional
roles that constitute conceptual content to account for the justificatory force of
perceptual judgments and thus rejects a strict non-relational view (see for
instance Sellars 1997, §32/pp. 68-71). The question is what more needs to be
in play on a Sellarsian view and whether it suffices to account for how it is that
we can refer to particular objects in perception while denying any semantical
relations between words and objects.
On Sellars’s view, there is an isomorphism between the conceptual and the
real orders, without which knowledge of the physical world would be
impossible. How can such an isomorphism come about if it does not involve
semantical relations between the conceptual and the real orders? To give the
same question a Kantian twist: how is it guaranteed that the conceptual order is
not arbitrary and haphazard, but in some way constrained by the objects our
thoughts are about?
The clearest account of how Sellars understands the isomorphism between
the two orders can be found in Science and Metaphysics. He argues there that
“expressions are involved in semantical uniformities (actual or potential) with
the appropriate extra-linguistic items” (1967a, I §59/p. 82). He is quick to add,
however, that this does not suggest that such semantical uniformities involve
relations between conceptual and non-conceptual items, arguing that there is a
relation within the conceptual order holding between words and functional

4Tyler Burge (1991) can be seen as having a version of such a view. As John Campbell (2002)
has pointed out, the question of course arises how the demonstrative element can provide the
perceiving subject with knowledge of what the demonstrative term refers to. I will address this
question in the next section.
178 Susanna Schellenberg

roles, which he refers to as a relation of signification.5 Similarly, there is a

relation in the real order between sign-designs and objects, which he refers to
as a relation of picturing. Picturing-relations map configurations of words onto
configurations of natural objects, where the link between the word and the
object is understood to be a causal one.
Sellars is led to such a non-relational semantics through his account of
meaning-statements. In order to accommodate the different roles played in
meaning-statements of the form
(1) ‘dreieckig’ (in German) means ‘triangular’,
one can distinguish shape-focused and meaning-focused quotations. While
*dreieckig* picks out a sign-design, •triangular• picks out a certain conceptual
role played by, for example, *dreieckig*. The words *dreieckig* and
*triangular* are •triangular•s, which is simply to say that *dreieckig* plays
the same functional role in German as *triangular* plays in English. While
‘*dreieckig*’ refers to tokens of the sign design *dreieckig* without
depending on an abstract entity, the shape, ‘•triangular•’ refers to the
functional role •triangular• without depending on an abstract entity, which one
might want to call the “meaning” of the expression. The crucial point is that all
we need to interpret (1) is a notion of two expressions having the same sign-
design and a notion of two expressions having the same functional role. There
is no third thing to which either of the two expressions flanking the semantic
term ‘means’ refer that accounts for the fact that *dreieckig* and *triangular*
have the same functional role. Thus, (1) is not a relational predication. By
reformulating (1) as
(2) *dreieckig*s are •triangular•s.
or alternatively
(2’) A *dreieckig* is a •triangular•.
the sentence can be recognized more clearly as a sortal predication (or
classification). The sortal ‘•triangular•’ classifies expressions that have the
same functional role. Accordingly, the semantic verb ‘means’ in the earlier
sentence no longer appears as denoting a semantical relation, but rather as
expressing a functional classification.6 Sellars argues that if his reconstruction

5 It is misleading to be speaking of signification-relations, since Sellars insists that they are not

relations at all, but just classifications.

6 In his very early papers (see, for example, Sellars 1950), Sellars makes a slightly different

distinction between sign-designs and functional roles, arguing that the pragmatic quotes (star
 Sellarsian Perspectives on Perception and Non-Conceptual Content 179

of meaning-statements is correct, then “it follows at once that semantical

statements of the Tarski-Carnap variety do not assert relations between
linguistic and extra-linguistic items” (1967a, I §59/p. 82). If meaning-
statements express functional classifications rather than semantical relations
between words and objects, the difference between expressions must, on
Sellars’s view, be seen as a difference in their semantic roles, not as a
difference in what they refer to, be it an abstract object or an object in the real
order.7
So, semantical relations holding between the conceptual and the real orders
are avoided by splitting the assumed word-object relations into signifying
relations on the one hand and picturing relations on the other. While the
sentence
Tokens of the word *dreieckig* picture triangles,
means that the word *dreieckig* represents a triangle, the sentence
Tokens of the word *dreieckig* signify •triangular•,
means that the word *dreieckig* is functionally classified as a •triangle•. This
just is to say that there is a language-language relation of signification and a
world-world relation of picturing, but no semantical relation holding between
the conceptual and the real orders. What connect the picturing-relation with the
signification-relation are only the words that play a role in both relations.
The question arises as to how much philosophical weight the picturing
relations can carry. According to Sellars, picturing relations guarantee that the
structure of configurations of objects is preserved in our language by
establishing referential relations between words and objects. Picturing
relations constitute the isomorphism between the picturing language and the
pictured world through a causal link between words and objects. He does not,
however, want such causal relations to figure as content-generating links.
What leads Sellars to his strong non-relational claims is his view that sentences
of the form “a *cat* is a •cat•” do not involve any direct atomistic content-
generating relations between *cat*s and cats. Avoiding such atomistic
relations, however, does not conflict with seeing concepts as involving
semantical relations insofar as they are related to objects through the rules of

quotes) specify token classes of sign-designs, while the syntactical quotes specify functions in a
language. Token classes are taken to be kinds of visual or auditory patterns that embody
conceptual functions in historical languages. Thus *dreieckig* and *triangular* are two token-
classes of the metalinguistic function that the relevant sign-design plays.
7 For the most elaborate presentations of this argument, see Sellars (1963a) and (1983).
180 Susanna Schellenberg

application that specify the circumstances under which it is correct to use the
relevant word and the practical consequences of its use. Such a view does not
imply that functional roles depend on atomistic content-generating links, since
the rules governing a word’s application are only intelligible in the context of
semantical relations within the conceptual order and thus only graspable by
someone who is already a proficient speaker of a language. The suggestion is
that the functional role in virtue of which a *cat* is a •cat• is embedded in the
real order because it involves rules governing the application of the word
*cat*. What makes the sentence “a *cat* is a •cat•“ true is that *cat*s have the
role played by *cat* in a speaker’s language.
It seems that taking such application rules into account involves taking
relations between words and objects into account. But Sellars denies that.
There are at least two ways in which Sellars might respond to the suggestion
that acknowledging the importance of application rules involves
acknowledging semantical relations between language and objective reality.
He might argue that although the sentence “a *cat* is a •cat•“ cannot be true
unless *cat*s refer to actual cats, the sentence itself does not express such a
hook-up. The sentence merely depends on picturing relations between the
conceptual and the real order, which are what bring about application rules. In
this sense, the rules governing *cat* include not only inference rules but also
application rules. Such an argument, however, amounts to introducing a rigid
distinction between the necessary conditions of the validity of a sentence and
its content. It is arguable whether it makes sense to say that the necessary
conditions of a sentence are not themselves expressed by the sentence.
The closest Sellars gets to acknowledging semantical relations is in his
discussion of semantical uniformities:
[I]n the case of expressions which stand for senses which are intensions, it will
also be true (and necessarily so) that these expressions are involved in
semantical uniformities (actual or potential) with the appropriate extra-linguistic
items. Thus in order for it to be true that . . .
‘Dreieckig’s (in German) are •triangular•s
German ‘Dreieckig’s must participate in uniformities with triangular things,
uniformities which parallel those involving our word ‘triangular’. But this does
not mean that these statements themselves have the form
(Linguistic item) R (non-linguistic item).
(1967a, III §59/p. 82; similarly IV §55ff./pp. 111ff.)
But even semantical uniformities can be understood without appeal to
semantical relations between words and objects, if they are understood as
picturing relations, that is, as relations between a word, say *cat*, considered
as an object in the real order, and other objects in the real order, namely actual
 Sellarsian Perspectives on Perception and Non-Conceptual Content 181

cats.8 In this sense, even the quoted passage can be read as expressing Sellars’s
official line of thought, according to which an account of the content of our
thoughts need not be made to depend on semantical relations between words
and objects.9
This argument, however, relies on uniformities already being established in
our language. Nothing has been said about what brings about these
uniformities in the first place. Sellars’s idea is that the uniformity of our
linguistic behavior establishes picturing relations between words and objects
and thereby secures that our language preserves the structure of the real order.
The crucial point of Sellars’s argument is that the picturing relations are only
linked to the signification relations in that our expressions play a role in both
relations. This allows him to write that “the causal aspect of perceptual
takings . . . accounts for the selecting of one world story rather than another
and connects the ‘is’ of this selecting with the rule-governed or ‘ought to be’
character of the language” (1979, p. 110). An obvious line on which to
criticize Sellars would be to argue that his account of picturing depends on
linguistic objects correctly picturing non-linguistic objects. This would involve
regarding the links they bring about from a normative semantical point of view
and not just as elements in the natural order. Sellars might argue that to say
that a linguistic object correctly pictures a nonlinguistic object is not a
normative matter, although normative language is used: corresponding to
every espoused principle of correctness there is a matter-of factual uniformity
in performance which link words with the objects they picture (compare
Sellars (1962a, p. 222).
As Sellars argues in “Truth and Correspondence” (1962b) the uniformity of
our behavior is only brought about because our language use is rule-governed
in the first place and thus only reflects that our linguistic actions are rule-
governed. Since picturing relations are established because of the uniformity
of our linguistic behavior, such a view suggests that picturing relations depend
on our language use being rule-governed. If this is correct, then picturing
relations depend on conceptual capacities. And if an expression pictures an

8 See, in particular, Sellars (1967a, III §66/p. 86).

9 See also Sellars (1967a , IV §56/pp. 111-112): “If we know, for example, that
‘Sage’s (in F) are •wise•s
we know, by virtue of our knowledge of what ‘wise’s do in our language, that the French word
‘sage’ must occur in uniformities involving both tokens of ‘sage’ and extra-linguistic objects. But
the uniformities do not consist of tokens of ‘sage’ standing for wisdom, but are rather to be
characterized as complex uniformities involving many configurations of many French words and
forms of behavior, on the one hand, and, not wisdom, but wise people, on the other.”
182 Susanna Schellenberg

object only because it is embedded in a complex framework of semantical

relations, then picturing relations can secure that our language preserves the
structure of what it is about only by presupposing conceptual capacities. Thus
on Sellars’s view, conceptual capacities are not only dependent on picturing
relations, our capacity to entertain picturing relations is in turn dependent on
our conceptual capacities.
Acknowledging such a mutual dependency between the two capacities fits
smoothly with Sellars’s “psychological nominalism.” As Sellars elaborates in
the so-called Myth of Jones in “Empiricism and the Philosophy of Mind,”
thoughts should be understood on the model of linguistic acts. We acquire
conceptual capacities by acquiring linguistic competence.10 Furthermore, all
exercises of conceptual capacities are a linguistic affair. What the Myth of
Jones aims to show is that the idea of inner episodes is unproblematic, since
we can model it on overt speech. Thus, operations of conceptual capacities are
not limited to overt uses of language, although overt speech might be
considered as the primary use of conceptual capacities. In this sense, Sellars
understands language not as the expression, but as the medium of our thoughts
(see for example Sellars 1979, p. 74).
Sellars is aware that this position involves the danger of equating words
with concepts. What is crucial is “the denial that there is any awareness of
logical space prior to, or independent of, the acquisition of a language” (1997,
§31/p. 66). Everything that is philosophically significant about our capacities
to have thoughts is, in other words, mirrored in what is significant in our
capacity to be competent speakers of a language.
So far I have argued that concepts, such as •cat•, involve not only
inference rules but also application rules for *cat*. Saying that *Katze* is a
•cat• does not only imply that it plays the same role in German that *cat*
plays in English: it is also to say that a person who knows that *Katze* is a
•cat• knows when to use the word to refer to cats. I have argued that even if
one accepts that concepts involve application rules, Sellars’s account of
conceptual roles is successful as relying solely on signification relations and
picturing relations and thus not involving semantical relations between words
and objects. Sellars’s non-atomistic semantics is, therefore, compatible with
the idea that the functional roles determining the use of words are related to
objective reality through application rules. But this still leaves open the

10 Compare Sellars (1997, §58/p. 105): “the ability to have thoughts is acquired in the process of

acquiring overt speech . . . only after overt speech is well established, can “inner speech” occur
without its overt culmination.”
 Sellarsian Perspectives on Perception and Non-Conceptual Content 183

question of how particular objects figure in the content perception. I will

address this more specific question in the next section.

2. Relating Thoughts to Objects

According to Sellars, the most striking change between Kant’s approach and
that of his predecessors is his insistence “that the class of contents should be
expanded to include individual contents, e.g. the content of an intuitive
representing of this-cube,11 or the content Socrates, which, though an
individual content, is not the content of an intuitive representing” (1967a,
p. 60). It is not obvious how in the framework of Sellars’s conceptual role
semantics we can think of ourselves as accommodating this Kantian insight
and picking out particular objects in perception. More needs to be in play than
the rules of application governing our use of expressions to understand what it
means for a person to direct a thought towards this rather than that particular
object. Addressing this question will bring together Sellars’s account of
functional roles with his account of perception.12
It will be helpful to look at some elements of Sellars’s reading of Kant to
get a clearer view of the issue in question. There are two possible readings of
Sellars on Kant that give a very different place to the role of sensations and

11 ‘This-cube’ refers to an intuitive representation. Since these representations are not

straightforwardly conceptual, it would be wrong to put ‘this-cube’ in dot-quotes. This should not
be read as implying that they are non-conceptual representations, whatever that would be.
12 In contrast to Sellars, Robert Brandom (1994, in particular, pp. 199-229) deals with the problem

of singular thought in his version of a conceptual role semantics without appealing to any
conception of experience. The empirical content of our perceptual judgments is explained by
arguing that our responsive dispositions allow us to react to our environment in a reliable and
differential manner. On Brandom’s view, reliable differential responsive dispositions allow us to
make observation reports. Since observation reports are brought about by our dispositions, rather
than being derived from inferences, they present us with non-inferential knowledge. Their content,
however, is determined by their inferential role. Furthermore, for an observation report to count as
knowledge the reporter must endorse the claim contained in the report. By taking these two
additional requirements into account, Brandom argues that our reliable differential responsive
dispositions are a necessary but not a sufficient condition for observational knowledge and, thus,
distances himself from the kind of reliabilist approach taken, among others, by Alvin Goldman
(1976 and 1986). By arguing that the reliability of our differential responsive dispositions is not
assessed by the perceiving subject, but rather in an intersubjective context, Brandom deals with
skeptical worries that Sellars’s account faces. The question remains whether such an account can
explain how we can direct our thought at this rather than that particular object in perception, but it
would lead too far a field to discuss this question here.
184 Susanna Schellenberg

intuitions in perception. I will argue against what I will call an empiricist

reading of the sensible conditions of perception and embrace what I will call a
transcendental reading.
In Science and Metaphysics, Sellars reads Kant as allowing solely for
semantical relations between conceptual episodes, and consequently takes
Kant to be struggling with the problem of how our thoughts are guided by the
objects they are about, ensuring that our conceptual order is not arbitrary or
haphazard. The following passage brings out the elements of Sellars’s reading
of Kant with which I will be concerned:
[W]hen [Kant] speaks of the productive imagination as “taking up” (A 120) the
manifold of outer sense into its activity . . . the metaphor implies . . . that the
manifold is an independent factor which has a strong voice in the outcome. On
the other hand, it is only if the manifold is mistakenly construed as belonging to
the conceptual order that it makes sense to suppose that it, so to speak, bodily or
literally becomes a part of the resulting intuitive representation. If it is, as I take
it to be, non-conceptual, it can only guide “from without” the unique conceptual
activity which is representing of this-suches as subjects of perceptual judgment.
(1967a, I §34/p. 16)
Sellars’s guiding question in the first chapter of Science and Metaphysics is
how perceptions of objects can yield knowledge. He explains how our
thoughts are directed at objects by appealing to sense impressions, that is,
mental states that are not shaped by the understanding. What seems clear is
that perception about objective reality must involve representations that are
brought about in a passive manner by having objects in view. The problem is
that if these representations are considered to be conceptually formed, it is
unclear how they can be thought of as constraining our conceptual framework.
If, on the other hand, they are considered to be non-conceptual representations,
it is not clear how they can guide at all. Such a view seems to imply that they
(causally) determine our responses and determining a response is not the same
thing as guiding a response. Sellars presents Kant as saying that sensibility
guides the conceptual realm by “sheer receptivity,” thereby yielding non-
conceptual sense impressions. Since Sellars argues that there cannot be direct
semantical relations between words and objects, he turns to “purely passive
representations of receptivity” as what constrains the conceptual activity.13

13 I profited from John McDowell’s Woodbridge Lectures (1998b) while thinking through this

part of Sellars’s writing. I am not concerned here with the question of the immediacy of the
presentness of objects to the intuitionally structured consciousness, which is central for
McDowell’s criticism of Sellars’s reading of Kant. McDowell spells out a transcendental role for
sensibility in terms of the immediate presence of objects to the conceptually shaped sensory
 Sellarsian Perspectives on Perception and Non-Conceptual Content 185

On what one could call an empiricist understanding of the sensible

conditions of perception, “sheer receptivity” brings about representations that
are not as yet conceptually shaped on the basis of which we form conceptual
representations. This view can be found in Sellars when he complains that
Kant has two notions of intuition between which he fails to distinguish clearly.
According to Sellars, one notion accounts for the non-conceptual impact that
guides our representations “from without,” while the other accounts for the
role of perception in understanding. He writes
[I]t is clear that Kant applies the term ‘intuition’ to both the representations
which are formed by the synthesizing activity of the productive imagination and
the purely passive representations of receptivity which are the “matter”
(A86/B108) which the productive imagination takes into account.14 (1967a, I
§18/p. 7)
Sometimes, Sellars refers to the former as intuitions and to the latter as sense
or sense impressions, distinguishing the “guidedness of intuition” from the
“receptivity of sense.” In other passages, he speaks of “the conceptual
character of the synthesis of apprehension in intuition” and “the radically
non-conceptual character of sense” (1967a, I §40/p. 16). For reasons that will
become clear in due time, I will refer to the former as intuitions and to the
latter as sensations.
I will argue that Kant’s notion of intuition in no way conflates two different
kinds of representations. Granted, Kant opens the Critique of Pure Reason by
distinguishing two different sources of knowledge – understanding and
sensibility – and then speaks of concepts and intuitions as corresponding to
these two sources of knowledge, thereby suggesting that intuitions are

consciousness of thinkers. On this reading, thoughts are about things in the real order because
objects are immediately present to conceptual consciousness in intuition. By suggesting a
relational understanding of intentionality, he can argue that “nonsheer receptivity is operative in
intuition” and thereby undermine the threat of idealism with which Sellars’s project is confronted.
McDowell goes along with Sellars insofar as he argues that conceptual episodes differ
intrinsically. But by taking the immediacy of intuitional representations into account, he takes a
different approach than Sellars, arguing that conceptual episodes differ intrinsically, not in a way
that systematically corresponds to what they are about, but because the difference in what they are
directed towards is itself an intrinsic difference in conceptual episodes.
14 Shortly before this passage Sellars writes: “Kant’s use of the term ‘intuition’, in connection

with human knowledge, blurs the distinction between a special sub-class of conceptual
representations of individuals which, though in some sense a function of receptivity, belong to a
framework which is in no sense prior to but essentially includes general concepts, and a radically
different kind of representation of an individual which belongs to sheer receptivity and is in no
sense conceptual.”
186 Susanna Schellenberg

non-conceptual. But this reading becomes unintelligible in light of Kant’s

argument in the Transcendental Analytic. Kant there shows how the picture
drawn in the Aesthetic is too simplistic, arguing that intuitions can only be
grasped in the context of a judgment and thus cannot be understood as
non-conceptual representations. The simplistic dualistic picture with which
Kant opens the first Critique can be seen as a rhetorical move that allows him
to gain the sympathy of his contemporary readers. Kant does, however, have a
notion of sensations, which on the reading of Sellars I aim to defend can be
seen as corresponding to what Sellars means when he speaks of sense
impressions.15
According to Sellars, sensations yield non-conceptual representations of
the real order, whereas intuitions conceptually represent the object. The purely
passive manner in which sensations are brought about guarantees that
intuitions are of the objects the perceiving person takes them to be of.
Sensations are thus necessary for guaranteeing the objectivity of our
conceptual representations.
What does it mean to say that non-conceptual sensations guide our
perceptual judgments? On what I have called an empiricist understanding of
the sensible conditions of perception, we first have basic mental
representations that are not as yet apperceived, on the basis of which we arrive
at conceptual representations. The notion of sensations involved in such a view
is problematic in that it raises the question of what representational status
sensations should have. More importantly, such representations are subject to
the very same objections that Sellars raises against the sense-data view.16 On
such a view, sensations are intermediaries that allow us to form conceptual
representations of the objects in view. Sensations and the conceptual capacities
which they feed into are understood on the model of a scheme-content
dualism, meaning that we have conceptual resources that are void of content,
on the one hand, and are able to grasp empirical content that does not have any
conceptual structure, on the other. The conceptual scheme is considered to be
abstracted (or abstractable) from the sensory content, and the sensory content
is considered to be given and in some way accessible to us independently of

15 In the following, I will always speak of sensations to stick to a more standard translation of

Kant’s ‘sinnliche Eindrücke’ or ‘Empfindungen’.

16 I will not recapitulate these objections here. They have been discussed in detail and with

different outlooks by Robert Brandom (1995 and in his study guide to Sellars in 1997), Willem
deVries (2000), John McDowell (1995 and 2002), and Jay Rosenberg (2000, 2002, and 2003),
among others.
 Sellarsian Perspectives on Perception and Non-Conceptual Content 187

the concepts we use to form judgments.17 The picture suggests that we start
with a scheme and a content and only in a second step bring the two together
to form a conceptual representation. It is unclear on such a reading of
sensations, why sensations are not a version of the given that Sellars criticizes.
Whatever Sellars meant with his appeal to sensations, it must have been
something subtler. I will argue that one can find in Sellars a very different
view than the one that emerges from what I have called the empiricist
understanding of sensations. Shortly after introducing his notion of sensations
in Science and Metaphysics, Sellars argues that
[t]heir “receptivity” is a matter of the understanding having to cope with a
manifold of representations characterized by “receptivity” in a more radical
sense, as providing the “brute fact” or constraining element of perceptual
experience.
The latter manifold has the interesting feature that its existence is postulated on
general epistemological or, as Kant would say, transcendental grounds, after
reflection on the concept of human knowledge as based on, though not
constituted by, the impact of independent reality. It is postulated rather than
“found” by careful and discriminating attention. The concept of such a manifold
is, in contemporary terms, a theoretical construct. (p. 9)
Sellars states here that he wants to understand sensations as “theoretical
constructs.” In “Some Reflections on Perceptual Consciousness” Sellars makes
a similar point when he contends that sensations “are not yielded by
phenomenological reduction, but postulated by a proto-(scientific)-theory”
(1977b, p. 179), arguing that to have a sensation of a pink cube “is not to sense
something as a cube of pink, though it is a state postulated by a theory
designed to explain what it is to see (or seem to see) a cube of pink as a cube
of pink” (p. 181). Here sensations are not thought of as unapperceived
representations on which perceptions are based, but rather as that which we
must assume figures in our perception if we think of them as being about
objective reality. Symptomatically, Sellars uses the words ‘epistemological’

17 The scheme-content dualism that Donald Davidson (1974) argues against has been interpreted
in various other ways. The scheme has been understood as a system of concepts, an equivalence
class of intertranslatable languages, a theory, and a worldview. Similarly, the content has been
understood as everything from objects, events, sense data, to Quinean stimulations. The
plausibility of Davidson’s argument rises and falls depending on how these different
interpretations of the scheme and the content are combined. Some combinations do not amount to
a dualism at all. I thank Anil Gupta for making me aware of this fact. The argument I wish to
bring out against the scheme-content dualism is general enough that it will not be necessary to
address the various possible interpretations of Davidson’s argument.
188 Susanna Schellenberg

and ‘transcendental’ synonymously when he discusses the role of sensations in

perception (1967a, I §22/p. 9 and §28/p. 11). The idea suggested is that we
must assume that there is something given that figures in our perception, if we
want to think of our perception as yielding knowledge of objective reality, that
is, of objects as existing independently of our mental capacities.
This is the crucial insight of Sellars that I would like to hold on to.
Understanding sensations in this second way is to see the role of sensations in
perception as a transcendental condition of perception rather than as a
primitive building stone in an empiricist framework. By contrast to the
empiricist understanding, sensations can now be understood as something
other than intermediaries between our conceptual thoughts and the objective
reality we perceive, or alternatively between the causal impact and the
conceptual outcome of perception.
Such an understanding of sensations is much closer to the Kantian notion
of Empfindungen or sinnliche Eindrücke. On a Kantian view, sensations are
the matter of perception that we must assume figure in our perception.18 But
the matter that is given in perception (sensation) is just one side of the
sensibility of perception, the other side being its form (intuition). On such a
view, the distinction between intuitions and sensations is not understood as
one between conceptual and non-conceptual representations or between two
kinds of representation understood in any other way, but rather as a distinction
between form and matter. Kant does not allow for the possibility of sensations
independently of their role in perception and, thus, independently of their role
in empirical intuitions. Sensations and intuitions necessarily come together and
in this sense are just two sides of one coin. In this sense, Sellars’s talk in
Science and Metaphysics of sensations as that which is given, the matter of
experience, does not invoke speaking of the given as Sellars attacks it.
But why are intuitions not candidates for the given?19 No doubt, Kant
understands intuitions as representations that fulfill the role of linking thoughts
to objects, thereby providing a possibility for judgments and, in contrast to
sensations, he thinks of intuitions as mental representations proper.
Nonetheless, intuitions are not intelligible in isolation either. The content of
intuitions is only accessible to us in the context of a judgment. Although
intuitions are related to an object independently of their role in empirical

18 See for instance Kant ([1781] 1965) A20/B34, A42/B59-60, A167/B209 or A218/B266. I
follow the usual practice in referring to passages in the Critique of Pure Reason by citing the
pagination of the original editions, indicated by A for the 1781 edition, and B for the 1787 edition.
19 For the sake of ease of formulation, I will speak just of intuitions in the following passages, but

it should be kept in mind that empirical intuitions and sensations always go together.
 Sellarsian Perspectives on Perception and Non-Conceptual Content 189

judgments (intuitions are related immediately to objects), we can only have a

grasp on their content if they figure in a propositional representation.
How does this bear on the question of how we can direct our thoughts at
particular objects? Kant argues, on the one hand, that we can direct thoughts at
something particular only when an object is given to us in intuition. On the
other hand, he says that intuitions are not enough to have knowledge,
“properly so called,” of objects (A78/B103). In other words, the immediate
representation we have in intuition is not enough to recognize an object. These
two lines of thought can be brought together in light of Kant’s understanding
of concepts as universally applicable rules that can serve as predicates of
possible judgments. As rules of synthesis, concepts determine intuitions.
Accordingly, intuitional representations are what we grasp when a rule of
synthesis is applied to the sensible manifold. Kant’s metaphor of intuitions
falling under concepts or being subsumed under concepts can be understood
along these lines. The talk of subsumption of intuitions under a concept need
not be understood as implying that intuitions are independently intelligible
representations that we in a second step recognize as falling under a particular
concept. Nor should subsumption be interpreted as suggesting that intuitions
are recognized as having a common property that the concept stands for. The
point is rather that we can identify the content of an intuition only as a result of
applying a rule.20
On the reading of Kant I would like to urge, we presuppose intuitive
representations when we make judgments about an object. We thereby
subsume the intuitive representation under an empirical concept and thus bring
about a conceptual representation. In this sense, we can only retrospectively
say (when we analyze a judgment) that the intuition in the judgment is the
immediate representation of an object, but we cannot have knowledge properly
so-called of an object by means of an intuition alone.

20 Thus, I am suggesting to read Kant on different lines than Manley Thompson (1972), who

writes: “It is thus possible to regard an empirical intuition as in a sense knowledge, and to speak of
it as blind without concepts because only as unified through concepts can it become knowledge
properly so called” (p. 323). According to Thompson, we can have blind empirical intuitions that
provide us with a preliminary form of knowledge, yielding knowledge proper only when
combined with concepts. The main problem with this view is that it remains unclear what it is that
we grasp or represent with an intuition if the intuition is not subsumed under concepts. On
Thompson’s view, intuitions are representations the content of which we have access to
independently of their role in judgments. On such a view, intuitions just are a form of the given as
Sellars criticizes it.
190 Susanna Schellenberg

Concepts as general representations become singular in the context of a

judgment by subsuming an intuition under them.21 On such a reading, it
becomes unproblematic to see intuitions as picking out objects immediately
towards which thoughts are directed only mediately. Thus, although in
perception we are affected by an object and, in this sense, it is not up to us
what our perceptions are about, this does not imply that we can grasp the
content of intuitional representations outside of the context of a judgment. In
this sense, Sellars’s talk of intuitions does not involve speaking of the given as
he attacks it any more than his notion of sensations involves speaking of the
given.22
I have argued that the role of sensibility is to account for how it is that we
can understand our perceptions as being about objective reality. On Kant’s
picture these sensible conditions are twofold: The matter given in perception,
sensation, and its form, intuition. On what I have called an empiricist
understanding, intuitions are epistemically graspable prior to being embedded
in propositional representations. By contrast, on a transcendental
understanding, the content of intuitions is only graspable when brought under
concepts in a judgment, even though intuitions are what in a judgment link us
to the object that the judgment is about.23 While intuitions are not graspable
independently of being subsumed under concepts in the context of a judgment,

21 In “Particulars,” Sellars introduces a notion of basic particulars that integrates both a “this-
factor” (accounting for the particularity of a perceptual representations) and a “such-factor”
(accounting for their generality). In Naturalism and Ontology (1979), he uses this same notion of
basic particulars to play the role of the “medium of alteration” in his process ontology. For an
excellent study of Sellars’s process ontology, see Seibt (1990).
22 I am taking a different line than William Alston (2002), who argues that Sellars’s critique of the

“Myth of the Given” is incompatible with the view that there is a non-conceptual mode of
“presentation” or “givenness” of particulars that is the heart of sense perception. By contrast, I am
arguing that a Sellarsian approach is committed to the view that there is something given in
perception, but it is only accessible to us if we bring our rational capacities to bear on what is
given. Speaking of something being given in perception need to involve speaking of non-
conceptual content. I take the view of perception that Alston endorses to be much closer to
Sellars’s own account than Alston makes it out to be. I am arguing, however, that it is misleading
to think of the central distinction for a Sellarsian approach in terms of conceptual and non-
conceptual content. What must be distinguished are rather the elements accounting for the
singularity and generality of perception, on the one hand, and the form and matter involved in
perception, on the other.
23 This way of thinking about intuitions is influenced by Gareth Evans’s (1982) discussion of

information links. On Evans’s view, our ability to relate to an object is dependent not only on our
ability to receive information about the world, but also on our ability to act in the world. Although
I focus only on the thought-dependency of perception in this paper, I take its dependency on
action to be at least as important.
 Sellarsian Perspectives on Perception and Non-Conceptual Content 191

sensations in their role as the matter for perception of objects are not graspable
independently of being formed into empirical intuitions. Intuitions and
sensations just are the conditions without which we cannot think of ourselves
as having perceptions about objective reality. As will become clearer in the
next section, the central distinction for a Sellarsian approach should not be
understood as one between non-conceptual and conceptual representations, but
rather as between the elements of empirical judgments that account for their
generality (concepts) and the elements that account for their singularity, the
latter involving both form (intuition) and matter (sensation).

3. Intuitions and Demonstratives

Now, Sellars interprets intuitions on the model of demonstratives, and thus

understands locutions such as ‘that cube’ as providing the linkage to objects
perceived. ‘That cube’ is a way of picking up Kant’s phrasing that intuitions
are immediately of objects (A68/B93). Here again I would like to distinguish
two different ways of reading Sellars. On the face of it, understanding
intuitions on the model of demonstratives should strike one as puzzling: ‘That
cube’ is a singular use of concepts, however, empirical intuitions are not a
basic kind of conceptual activity, but rather the sensible condition for the
empirical use of concepts.
No doubt, when our thought is directed at a particular object in perception,
we give a concept a singular use. But intuitional content is not a fragment of
the content we form in perceptual judgments, but rather a necessary condition
for such judgments. For whatever it is worth, Kant is very explicit that it is
concepts that are used in a singular, general, or universal manner, and that a
singular use of concepts is not the same as an intuition.24 While it is the
intuition that picks out the object in view, it is the concept that determines our
understanding of the object. The content of intuitional representations must
therefore not be understood in analogy with the content of conceptual
representations. Indeed, the distinction between intuitions and concepts need
not be understood in terms of their content at all. What are distinguished are
rather the manners in which the representations refer to what they are about.
While intuitions refer immediately and account for the singularity of

24 See Kant ([1800] 1992, §1, Note 2): “It is a mere tautology to speak of concepts as general or
universal; a mistake that rests on an improper division of concepts into general, particular, and
singular. Not concepts themselves but only their use [Gebrauch] can be so divided.”
192 Susanna Schellenberg

perceptual judgments, concepts refer mediately by subsuming an intuition

under them.
Intuitions are involved whenever thought is directed at objective reality.
Understanding intuitions as fragments of judgments would just be to
understand intuitions as a product of receptivity that makes a separable
contribution to the perceptual judgments we form when having an object in
view. Such a view would fall prey to the very same objections that Sellars
raises against the sense-data view. Therefore, Sellars must be after something
else when he understands intuitions as having a demonstrative form.
If we think of intuitions as being a necessary condition for any perceptual
judgment (demonstrative or not), we can understand intuitions as providing for
the possibility of perceptual judgment. Another way of expressing this idea is
that intuitions provide us with invitations to judgments or that intuitions are
“petitions for judgments” – to use Robert Brandom’s (2002) wording. On such
a view, intuitions are conceptual representations that potentially find verbal
expression in perceptual statements when subsumed under concepts. When
looking at a white wall, we can abstract from unimportant details and can
correctly judge the wall as white. Given the circumstances, it might also be
correct to say that parts of the wall appear green and pink and that the corners
have a brownish tint. The two judgments only seem to be incompatible:
although they are about the same object, they refer to different levels of detail.
In an analogous way every situation of perception brings about different
possible perceptual judgments. Indeed, one might want to say that when a
perceiver has something in view there is an open-ended string of possible
demonstrative propositions that she is invited to form.
One might object that such a view invokes a superfluous and potentially
problematic intermediary stage in perception, namely, a stage in which we
make a choice of what we direct our attention to. When a person perceives a
tree swaying in the wind, she might not be aware of its many leaves fluttering
in the wind: It is tempting to say that she is not perceptually aware of the
leaves fluttering in the wind, because her attention is not directed at the leaves.
But saying that intuitions only provide for the possibility for judgments need
not mean that in perception we make a choice of what we direct our attention
to. When we perceive a tree swaying in the wind, we might not be aware of its
many leaves fluttering in the wind, but that is not to say that the content of our
 Sellarsian Perspectives on Perception and Non-Conceptual Content 193

perception does not represent the leaves fluttering in the wind. One might want
to say that we have not made this content explicit.25
It certainly is not the case that we actively decide to direct our attention to
this or that in perception: We do not look out of the window, assess the
situation, the many possible judgments we could form, and decide to direct our
attention to this or that. We do not decide on what we pay attention to, rather
we find ourselves directing our attention to this or that. Only when we have
already distinguished possible different ways of perceiving an object (say
when we have become aware that it is both correct to say that the wall is white
and to say that it appears to have patches of pink and green) that we can think
of ourselves as choosing to perceive it in this or that way. But to say that there
is a choice about what we direct our attention to, is not to say that in
perception we make a choice about what we direct our attention to. It is in this
sense that perception is passive: We find ourselves perceiving this or that. This
is how one can read Sellars interpretation of intuitions as having the form
“that . . . ” without thinking of intuitions as being primitive mental
representations that are accessible to us prior to conceptual representations
proper.

Conclusion

Sellars understands the word *cube* to be at the same time semantically

classified by the functional role in virtue of which a *cube* is a •cube• and to
be picturing an object, namely a cube, at least in the veridical case. As I argued
in the previous section, the application rules involved in •cube•, can be seen as
setting up the mapping relation between *cube*s and actual cubes. This is not
enough, however, to explain how it is that we can direct our thought at this
rather than that particular cube in perception. On Sellars’s view, our ability to
refer to particular objects depends on individual links between judgments and
the objects the judgments are about. There are two sides to such a subject
matter: The matter given in perception (sensation), and its form (intuition). On

25 Indeed we might want to say be that we perceive the tree swaying in the wind, because we see

the leaves fluttering in the wind. The idea underlying this thought is the same idea that motivates
Leibniz’s distinction between grandes and petites representings. On Leibniz’s view our perception
of a the ocean roaring, to use his example, is constituted of a multitude of micro perceptions of the
noise that a grain of sand makes when water crashes on it. We are not aware of the noise that
every single grain of sand makes when listening to the roaring of the ocean. Nonetheless, we hear
the roaring of the ocean because we hear the noise of many grains of sand.
194 Susanna Schellenberg

an empiricist understanding, intuitions and sensations are epistemically

graspable prior to a concrete propositional representation. I argue that it is
unclear on such a view why sensations and intuitions are not just versions of
the given as Sellars criticizes it. By contrast, on a transcendental understanding
the role of sensibility is to account for how it is that we can understand our
perceptions as being about objective reality. Intuitions and sensations are what
we must assume figure in our perception if we think of our perception as
having a subject matter. Intuitions and sensations necessarily go together in
perception of objective reality. While sensations are not graspable
independently of being formed into empirical intuitions, the content of
intuitions is not graspable independently of being subsumed under concepts in
the context of a judgment. On such a view it is unproblematic to see sensations
as what is given in perception: They are not intelligible independently of their
role as the matter of intuitions, the content of which is accessible to us only in
the context of a judgment.
One might want to insist on saying that such sensations have
non-conceptual content. In the contemporary philosophical context, it is,
however, misleading to use the phrase, unless the content in question is
understood to be accessible to the perceiving subject without bringing any
conceptual capacities into play. Although Sellars speaks of sensations as
having non-conceptual content, he is very explicit that the content provided
with sensibility is graspable by the perceiver only in the context of a judgment.
If intuitions can mean nothing to us unless we bring our conceptual capacities
to bear on them, it is just as misleading to say that they are conceptual
representations as it is to see them as non-conceptual representations when
isolated from their role in judgments. The conceptual/non-conceptual
distinction, thus, does not bring us far in a discussion of the sensible conditions
of perception understood in the way I have presented. The crucial distinctions
in play are between the form and matter of the sensible conditions of
perception as well as the manner in which intuitional and conceptual
representations refer to an object: While intuitions account for the singularity
of perceptual judgments, the concepts in play account for their generality. As I
argued, this is compatible with understanding conceptual activity as
constrained by that towards which it is intentionally directed. Indeed, only by
allowing that thoughts are about something can the transcendental requirement
that conceptual activity is intentionally directed towards something be met. In
contrast to the empiricist understanding, sensations and intuitions are not
understood as intermediaries between our conceptual thoughts and the
objective reality we perceive, but rather as the conditions without which we
cannot think of ourselves as having empirical knowledge.
 Sellarsian Perspectives on Perception and Non-Conceptual Content 195

Susanna Schellenberg
Department of Philosophy
University of Pittsburgh
Pittsburgh, PA 15260
United States of America
e-mail: [email protected]

REFERENCES

Alston, W.P. (2002). Sellars and the “Myth of the Given.” Philosophy and Phenomenological
Research 65, 69-86.
Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press.
Brandom, R. (1995). Knowledge and the Social Articulation of the Space of Reasons. Philosophy
and Phenomenological Research 55, 895-908.
Brandom, R. (2002). Non-inferential Knowledge, Perceptual Experience, and Secondary
Qualities. In: N. Smith (ed.), Reading McDowell on Mind and World, pp. 92-105. London &
New York: Routledge.
Burge, T. (1991). Vision and Intentional Content. In: E. LePore and R. Van Gulick (eds.), John
Searle and His Critics, pp. 195-214. Oxford: Basil Blackwell Pub.
Campbell, J. (2002). Reference and Consciousness. Oxford: Oxford University Press.
Davidson, D. (1974). On the Very Idea of a Conceptual Scheme. Reprinted in: Inquiries into Truth
and Interpretation, pp. 183-198. Oxford: Clarendon Press, 1984.
Davies, M. (1997). Externalism, Architecturalism and Epistemic Warrant. In: C. MacDonald,
B. Smith, and C. Wright (eds.), Knowing our own Minds: Essays on Self-Knowledge,
pp. 321-361. Oxford: Oxford University Press.
deVries, W. and T. A. Triplett (2000). Knowledge, Mind, and the Given: Reading Wilfrid Sellars’s
“Empiricism and the Philosophy of Mind.” Cambridge, MA: Hackett Publishing.
Evans, G. (1982). Varieties of Reference. Edited by J. McDowell. Oxford: Oxford Univ. Press.
Goldman, A. (1976). Discrimination and Perceptual Knowledge. The Journal of Philosophy 73,
771-791.
Goldman, A. (1986). Epistemology and Cognition. Cambridge, MA: Harvard University Press.
Hurley, S.L. (1998). Consciousness in Action. Cambridge, MA: Harvard University Press.
Kant, I. ([1781] 1965). Kritik der reinen Vernunft. Critique of Pure Reason. Translated by
N.K. Smith. New York: St. Martin’s Press.
Kant, I. ([1800] 1992). Jäsche Logik. In: J.M. Young (ed. and trans.), Lectures on Logic,
pp. 521-640. Cambridge: Cambridge University Press.
McDowell, J. (1998). Having the World in View: Sellars, Kant, and Intentionality. The Journal of
Philosophy 95, 431-491.
McDowell, J. (1995). Knowledge and the Internal. Philosophy and Phenomenological Research
55, 97-105.
McDowell, J. (2002). Knowledge and the Internal Revisited. Philosophy and Phenomenological
Research 66, 877-893.
Noë, A. (2004). Action in Perception. Cambridge, MA: The MIT Press.
Rosenberg, J. (2000). Perception and the Inner Sense: A Problem about Direct Awareness.
Philosophical Studies 101, 143-160.
196 Susanna Schellenberg

Rosenberg, J. (2002). Thinking about Knowing. Oxford: Oxford University Press.

Rosenberg, J. (2003). Sellarsian Seeing: In Search of Perceptual Authority. In: R. Schuhmacher
(ed.), Perception, pp. 262-285. Paderborn: Mentis.
Seibt, J. (1990). Properties as Processes: A Synoptic Study of Wilfrid Sellars’s Nominalism.
Atascadero, CA: Ridgeview Publishing Company.
Sellars, W. (1950). Quotation Marks, Sentences and Propositions. Philosophy and
Phenomenological Research 10, 515-525.
Sellars, W. (1952). Particulars. In: W. Sellars (1963b), pp. 282-297.
Sellars, W. (1960). Being and Being Known. In: W. Sellars (1963b), pp. 41-59.
Sellars, W. (1962a). Naming and Saying. In: W. Sellars (1963b), pp. 225-246.
Sellars, W. (1962b). Truth and Correspondence. The Journal of Philosophy 59, 29-56.
Sellars, W. (1963a). Abstract Entities. In: W. Sellars (1977a), pp. 49-89.
Sellars, W. (1963b). Science, Perception and Reality. Atascadero, CA: Ridgeview Publishing
Company.
Sellars, W. (1967a). Science and Metaphysics. Atascadero, CA: Ridgeview Publishing Company.
Sellars, W. (1967b). Some Remarks on Kant’s Theory of Experience. The Journal of Philosophy
64, 633-647.
Sellars, W. (1977a). Philosophical Perspectives: Metaphysics and Epistemology. Atascadero, CA:
Ridgeview Publishing Company.
Sellars, W. (1977b). Some Reflections on Perceptual Consciousness. In: R. Bruzina and
B. Wilshire (eds.), Selected Studies in Phenomenology and Existential Philosophy,
pp. 169-185. The Hague: Martinus Nijhoff.
Sellars, W. (1979). Naturalism and Ontology. Atascadero, CA: Ridgeview Publishing Company.
Sellars, W. (1983). Towards a Theory of Predication. In: J. Bogen and J. McGuire (eds.), How
Things Are, pp. 281-318. Dordrecht: D. Reidel Press.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Strawson, P.F. (1959). Individuals. London: Methuen.
Thompson, M. (1972). Singular Terms and Intuitions in Kant. The Review of Metaphysics 26,
314-343.
Matthew Burstein

PRODIGAL EPISTEMOLOGY:
COHERENCE, HOLISM, AND THE SELLARSIAN TRADITION

ABSTRACT. Many philosophers have equated the denial of foundationalism with a

call for coherentist approaches to epistemology. I think such equations are spurious, and
to show why this is so I contrast the views of a paradigmatic coherentist with an
antifoundationalist alternative. This article examines the coherentism of Laurence
BonJour with an eye toward the way in which BonJour’s views fail to fully adopt the
insights of their Sellarsian roots. In particular, I argue that BonJour’s view endorses the
philosophy of mind that Wilfrid Sellars criticized in “Empiricism and the Philosophy of
Mind” (EPM) – which itself may be no sin (though I believe it is), but is a problem
given BonJour’s express sympathies with Sellars and EPM. I begin by contrasting the
approaches taken by BonJour in The Structure of Empirical Knowledge and by Sellars
in EPM, arguing that one could deny foundationalism without, thereby, becoming a
coherentist.

Philosophers holding various positions have been called coherentists,

including Hans Reichenbach, Donald Davidson, Wilfrid Sellars, Richard
Rorty, and Laurence BonJour. I believe that coherentism is a mistake, and I
believe that one can defend positions staked out by Wilfrid Sellars’s in
“Empiricism and the Philosophy of Mind” (henceforth: EPM) without being a
coherentist. This position is itself not necessarily a novel one; John
McDowell’s (1996) elaboration of a Sellarsian epistemology in Mind and
World explicitly criticizes coherentism, and Michael Williams’s (1999)
additions to the most recent version of Groundless Belief corrects his
endorsement of the name ‘coherentist’ in the first edition. However, few have
attempted to show precisely why it is that Sellarsian antifoundationalism is not
coherentist, and I hope to undertake that task here. My work will begin with an
examination of Laurence BonJour’s approach to coherentism by examining his
arguments for the view; BonJour’s The Structure of Empirical Knowledge
(1995) is an important attempt to work out and defend a coherentist
epistemology, and BonJour takes himself to be defending a Sellarsian view in
the book. I intend to develop an analysis of BonJour’s defenses of and
subsequent attacks on coherentism that reveals just how wrong BonJour has

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 197-216.
Amsterdam/NewYork: Rodopi, 2006.
198 Matthew Burstein

been in interpreting the criticisms leveled by Sellars against foundationalism.1

BonJour’s arguments in either direction turn on a unified field of problematic
assumptions about the nature of the mind; these assumptions, it is worth
noting, are not ones shared by Sellars in EPM.
Specifically, BonJour’s view endorses the philosophy of mind that Sellars
criticizes in EPM. Broadly speaking, we might call such views “Cartesian,”
though not because they share rationalist assumptions with the works of
Descartes. Rather, these views proceed in ways that have affinities with the
picture of the relationship between mind and world that Descartes (along with
many of his British Empiricist analogues) adopted; namely, that we are
immediately acquainted with the subjective contents of our minds but only
mediately acquainted with the objective contents of the external world.2
Insofar as BonJour attempts to situate Sellarsian antifoundationalist arguments
within a Cartesian understanding of the mind he distorts the careful work done
in EPM. Situating EPM in this way is a serious mistake;3 to situate Sellarsian
arguments as BonJour does is to ignore the upshot of the closing sections of
EPM (where Sellars develops his methodological behaviorism). The point of
those later passages is to liberate us from being compelled to conceive of
mental episodes in a way that is Cartesian without falling into the behaviorist
trap of having to dispense with a large portion of our common sense beliefs
about the nature of the mental.
The position is a methodological, rather than philosophical, behaviorism
because it claims that we ought to avoid the excesses of more radical forms of
behaviorism (esp. philosophical or analytical behaviorism) by preserving a
space for what we normally might call the “mental life” of agents. What all
forms of behaviorism have in common is the impulse to analyze mentalistic
concepts in terms overt behavior. For an analytical behaviorist, this impulse
manifests itself in the attempt to analyze existing mentalistic concepts in terms

1 Given that I do not have any particular theological commitments along these lines, I am not

interested in providing the final word on Sellars exegesis. Rather, I would just like to disqualify
BonJour’s reading. I am perfectly happy to leave open the issue of who among other philosophers
(e.g., Davidson, Rorty, Putnam, McDowell or Brandom) are correct in their interpretations of
Sellars – though my McDowellian sympathies should be apparent by the end of this work.
2 Another version of Cartesianism is the idea that justificatory entities are exclusively mental

states. This other form is related to the first in that, with the first on board, we only come to have
evidence of the external world based on out knowledge of the internal world – so the things that
count as full-blooded are the entities of the internal world.
3 It is, in fact, one on BonJour’s central errors. In so situating EPM, he robs Sellars of all the

important philosophical tools he lays out for himself, and then attempts to apply this tool-less view
to problems for which it is ill-suited.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 199

of overt behaviors; such views serve to do away with the mentalistic discourse,
or only leave it in place as a mode of speaking. In contrast, the methodological
behaviorist only requires that mentalistic concepts be introduced in terms of
overt behavior. Mentalistic concepts are to be introduced as theoretical posits
of a certain sort – as explanations for certain kinds of behavior – but can come
to be used in an observing way (i.e., by picking out mental objects). One of the
central morals of the Myth of Jones – the myth which Sellars uses to motivate
the possibility of a methodological behaviorism – is that we can deny the
Cartesian thesis that we are directly acquainted with the entities of our mind
(be they sense data, Humean impressions, mental states, or the like) while only
mediately acquainted with the entities of the external world without having to
deny the reality of mental entities altogether.4 Instead, mentalistic concepts can
be introduced as theoretical posits explaining certain behaviors, and through
practice, come to be observational items. The reality of such entities is not in
question for the methodological behaviorist as it is the analytic; rather it is the
question of their role in explanation that is in question.
The goal of this article will be to show how BonJour’s failure to
incorporate the insights of the closing sections of EPM sends him out to
explore the barren fields of coherentism. I will lay out BonJour’s arguments
for coherentism and assess them in light of Sellars’s rather different grounds
for dispensing with foundationalism. Many have equated critiques of
foundationalism with coherentism (and thus have wrongly accused some of the
authors I mention in the opening of this paper as being coherentist), this
project will show how one can deny foundationalism without, thereby, being a
coherentist.

1. BonJour’s Reply to the Epistemic Regress Argument

In The Structure of Empirical Knowledge (SEK), BonJour attempts to

undermine foundationalism so as to show the need for a coherence theory of

4 It is true that my anticartesianism commits me to a certain view about mental entities – a view

that seems to be more like philosophical behaviorism than its methodological sibling. Two things
are worth noting here. First, the claim here is that there is, in fact, room for methodological
behaviorism. One need not, by commitment to methodological behaviorism, be committed to the
more radical claims of the philosophical behaviorist. Second, the claims I have made about
anticartesianism do not, by themselves, require that we deny the reality of various classical mental
entities (e.g. qualia). Philosophical behaviorism’s excesses included the idea that all there is to
such entities is dispositions to behave, but nothing I have said commits the anticartesian to this.
200 Matthew Burstein

justification. (More precisely, he attacks both foundationalism and

epistemological externalism in SEK.) While he takes himself to be largely
influenced by Sellars, BonJour’s work does not take up the same kind of attack
on foundationalism as Sellars. Rather, he takes up related (but rather different)
issues: the epistemic regress argument and what he calls his “basic
antifoundationalist argument.” In this section, I examine the arguments that
BonJour utilized to undermine foundationalism; in the next section, I will look
at what he says more specifically about Sellars.
Bonjour first takes up the foundationalist response to the epistemic regress
argument. This argument – one of the basic arguments marshaled in
epistemology – tries to assess the abstract structure of justification. The
argument may be reviewed as follows: imagine that Haritini is queried about
one of her beliefs – say, her belief that the Library of Congress is located east
of the White House (call this P). In order to justify her belief about the location
of the Library of Congress, Haritini might appeal to some other belief of hers
that entails P – for example, her belief that the Congress is located to the west
of the Library of Congress, between the Library and the White House. This
seems to be natural enough, but it raises a concern: just how long must this
discussion go on? That is, can we say ahead of time whether there is some
stopping point in Haritini’s justification of P – some point beyond which no
call for justification is itself justified?5 There seem to be four candidate chains
of justification: (1) the regress of reasons might terminate in beliefs which are
themselves unjustified; (2) the regress might be linearly infinite (with no
repeat of premises and no conclusion reached); (3) the regress might be non-
linearly infinite; or (4) the regress might terminate in beliefs which are
justified but which do not themselves rely on other beliefs for justification. For
various reasons,6 foundationalists think that the only acceptable solution to the
regress is (4): there are some foundational beliefs that convey justification but

5 Perhaps it would pay to say more here. On some accounts, it is not that a call for further

justification is not justified, it is just that it does not “make sense” (both in the colloquial and
philosophical understandings of this phrase).
6 The most basic reason that foundationalists use to ground (4) is that we cannot understand our

knowledge as being in contact with the world if the regress does not end in a way specified by (4).
However, there are probably two different versions of the epistemic regress argument,
differentiated by their target. One version just claims that there can be knowledge if we cannot
stem the regress of reasons; the other claims that we cannot adequately reply to the global skeptic
if we cannot stem the regress. Why we must endorse (4) may depend on which version of the
regress is being addressed.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 201

themselves do not require similar justification. We shall call these beliefs basic
empirical beliefs.7
In reply to the foundationalist, BonJour poses his “basic
antifoundationalist” challenge (BAC); this reductio ad absurdum of
foundationalism goes as follows:
1. Suppose that there are basic empirical beliefs, that is empirical beliefs (a)
which are epistemically justified, and (b) whose justification does not
depend on that of any further empirical beliefs.
2. For a belief to be epistemically justified requires that there be a reason why
it is likely to be true.
3. For a belief to be epistemically justified for a particular person requires that
this person be himself in cognitive possession of such a reason.
4. The only way to be in cognitive possession of such a reason is to believe
with justification the premises from which it follows that the belief is likely
to be true.
5 The premises of such a justifying argument for an empirical belief cannot
be entirely a priori; at least one such premise must be empirical.
________________________________________________________________________________________

The justification of a supposed basic empirical belief must depend on the

justification of at least one other empirical belief, contradicting (1); it
follows that there can be no basic empirical beliefs. (1985, p. 32)
So, what is the argument here? Premise one is essentially the foundationalist’s
solution to the epistemic regress argument: there must be some kind of belief
(basic empirical ones) which is justified but whose justification does not
depend upon further empirical beliefs. Premise two seems to be nearly
tautologous; epistemically justified beliefs just are those for which there is a
reason to believe.8 The third premise is where some foundationalists will start
to balk – any externalist about justification will claim that this premise is false.
Given that BonJour argues elsewhere against externalism – and that taking and

7 Non-empiricist foundationalist alternatives would use non-empirical basic beliefs, but I believe
that the BAC can be run against such beliefs. BonJour’s argument is, contrary to its formulation,
innocent of questions about the debate between empiricism and rationalism.
8 BonJour’ s formulation here differs a bit from mine, but I think there is not much of

philosophical import that hangs on it. My locution of choice makes it clear that justification is a
practice “all the way down” – that there is no point beyond discourse and practices from which we
can assess these matters, whereas BonJour’s leaves open readings which make the likelihood of
truth a belief something which is fairly radically externalist. I think he could profitably take this
on as a friendly amendment, given his extensive arguments against externalism.
202 Matthew Burstein

defending a position in the internalism/externalism would take me too far

afield – I shall move along on the assumption that premise three is true.9
It is at premise (4) that things really become of interest. The first thing we
ought to notice is that this is not Sellars’s argument against the Given, and in at
least one way it flouts one of the central distinctions of EPM. Sellars’s
argument against the Given yielded a distinction between “being inferable
from” and “being dependent upon.” More concretely, the distinction is
between a background of epistemic practices being required for the possibility
of my knowing that something is red and my inferring that it is so from a
belief about those background practices. On the Sellarsian picture of things,
any piece of knowledge may depend on a whole host of background conditions
(e.g., an agent’s perceptual skills, her knowledge of generalities, her linguistic
skills, and the like); it does not follow from this, though, that the kind of
dependence we need to understand this point is inferential in nature. Surely we
do infer things, but often times we see them straight off as well. In premise
four, BonJour commits himself to an inferential picture of dependence (one on
which there is no distinction between inferable from and is dependent upon);
as a result he cannot countenance non-inferential knowledge.10 Were we to
acknowledge the distinction here, premise (4) could be recast in terms of being
situated within justificatory practices and being in possession of the relevant
epistemic skills (rather than casting the premise in terms of believing with
justification the premises from which it follows). While this recasting of the
premise can countenance the role of inference – due to the fact that making an
inference is a skill within a justificatory practice – it does not limit the picture
of justification to merely making inferences from premises to conclusions.
Now, one might worry that this is, in fact, quite in line with BonJour’s
view. After all, BonJour does not deny that there are experiences (rather than
inferences) which give rise to beliefs; after all, “a belief arrived [at through]

9 This is a bit dodgy, but given space constraints, there is little else to be done. However, part of

the purpose of this work is to make room for a non-coherentist holism; insofar as that is the goal, it
seems as though questions of externalism are orthogonal to my concerns. Moreover, Sellars’s
arguments against foundationalism do not, themselves, turn on a specific position in the
internalist/externalist debate, even if they might give aid and comfort to one position or another.
Indeed, the Sellarsian position may give us a way to eschew the dualism altogether, let alone
choosing a side in the debate.
10 One way of putting the Sellarsian project is that it shows us (or, less tendentiously, attempts to

show us) that, on the one hand, coherentism is wrong because it cannot account for non-inferential
knowledge, and on the other foundationalism is wrong because it fails to adequately account for
the role of background knowledge and skills required by epistemic agents; not view falter because
they fail to take up the varieties of dependence.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 203

inference would not qualify as observational from the standpoint of common

sense” (1985, p. 113). BonJour calls this sort of psychologically direct beliefs
“cognitively spontaneous beliefs.” BonJour wants to admit the possibility of
this sort of psychological directness while wondering why it is “a belief which
originated in some noninferential way [could not] be justified or warranted
only by appeal to coherence with the rest of the system of beliefs” (1985,
p. 113). One way to clarify what is at stake here is by considering the idea that
account of epistemic dependence that I have been stumping for (one which is
broader than inference) essentially boils down to something like “being
inferable, under the right circumstances, from φ.”11 BonJour himself seems to
suggest this sort of analysis when he describes the way he might justify such a
belief (1985, p. 118):
1. I have a cognitively spontaneous belief of kind K that there is a red book on
the desk.
2. Conditions C [of the epistemologically proper sort to make K observations]
obtain.
3. Cognitively spontaneous visual beliefs of kind K in conditions C are very
likely to be true.
________________________________________________________________________________________

Therefore, my belief that there is a red book on the desk is very likely to be
true.
Therefore, (probably) there is a red book on the desk.
Here we see that BonJour’s picture is one on which inference (and explicit
argumentation) does the epistemological work; this is to be expected, given
premise four of the BAC. Clearly, for any purported case of noninferential
knowledge, we could reconstruct such an argument. The question to is whether
we must. If the sorts of epistemic dependence I described above really do boil
down to “being inferable, under the right circumstances, from φ,” then it would
seem that we such reconstructions are the coin of the realm for justification;
the broader notion of epistemic dependence would be a sort of practical
shorthand for this coin.
To help decide between these two options, we will have to draw more out
of the BAC, and there is something about the ways in which we can read
premise four that ought to make us a bit cautious as we proceed. There are a
group of expressions in this argument that can be read ambiguously: ‘having
cognitive possession’ and ‘believe with justification’, for example. These can
be read psychologistically and anti-psychologistically, and we need to be sure

11 I am grateful to an anonymous reviewer for raising this concern.

204 Matthew Burstein

which we are talking about in any given context; is BonJour talking about
psychological entities (believings) or about the things believed (which may or
may not be psychological entities)? Since it is my own concern not to allow a
slide into Cartesianism (at least not an unargued for slide), I want to direct our
attention to the main place where we are likely to give in to such a slide. The
central problem is with premise four: “the only way to be in cognitive
possession of such a reason is to believe with justification the premises from
which it follows that the belief is likely to be true.” This premise is the one
which drives us from giving an internalist account of justification (e.g.,
“having a justified belief” requires, or just is, having a reason to believe) in
premise two to claims about the “premises of such a justifying argument” in
premise five. But, given the discussion here, this slide is too quick, and it
seems to involve two separate problems. The first of these problems is that
having a reason for belief is not necessarily having a belief about a reason to
believe, while the second is that it is not necessary for one to have an argument
for one’s beliefs in order to have a reason to believe.12 I shall take these up in
turn.
In order to get from (3) to (5), BonJour interprets justification in terms of
having an argument – e.g. in terms of being aware of premises which logically
entail a conclusion – but this need not be the case, and, moreover, it is often
not the case. Suppose I claim that there is an ugli fruit in my kitchen. Now, if
one were to ask what evidence I have for my statement, I could just point to
the fruit and say “There is an ugli fruit on the counter” (or, even better, I could
say nothing at all and point out the fruit on the table!). I might, in the absence
of this possibility (e.g., my interlocutor and I are not able to get to the counter
to check), appeal to my having seen it on the counter (and my having been in
standard ugli fruit viewing conditions and the reliability of my memory and the
like); however, if the fact is present and obtains, it would seem that the thing
that provides the justification is the relevant fact.13 If we suspect that adding a
pointing performance to my utterance would be sneaking in another belief or
other linguistic entity, my reply to my interlocutor could just be “There is an
ugli fruit on the counter – go and look for yourself.” While this does provide

12 Perhaps the latter bit is false – that is, maybe having reason to believe means that one does have

an argument for one’s beliefs. However, this is only plausible if ‘having an argument’ is read in a
dispositional way or, perhaps. The notion of argument is extended to include less formal
argumentative approaches (e.g. those we might find in the arts).
13 I should do some clearing up of my own view here. To say that I can “appeal to the fact” is not

to suggest that there are just facts in the world that we could be aware of independent our being
placed within the logical space of reasons.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 205

additional linguistic content, it does not provide any additional descriptive

content; rather it adds attitude (in the tone) and a command to check the
evidence, but it does not add another belief. In such a case, I would not provide
argumentation that the fruit was there; instead, what I would provide is
evidence that the fruit was there – namely, by calling attention to the fact that
the fruit is present. Our justificatory repertoire is much broader than organizing
premises into arguments; while argumentation is a form of evidentiary
procedure, not all evidentiary procedures are (strictly) argumentative.14 In
addition to making arguments, we can (as I suggest here) present evidence to
others and we can (as I will suggest in a moment) demonstrate our skills.
Certain claims we make are about objects in the world, and the presence of
these objects can provide justification for our claims.15
Yet another non-argumentative evidentiary procedure involves the
mobilization of one’s skills. We make claims about our ability to act
skillfully.16 That this is so should be fairly intuitive; there are, after all, many
kinds of claims we already acknowledge as being justified by skillful
activities. For example, claims about one’s acumen at building a house or
riding a bicycle can be proven by appealing to a skillful demonstration of just
that very skill.
However, redemption of one’s commitments by skillful demonstration is
not limited to the realm of statements expressly about skills. Language use is a
way of skillfully navigating the world, and in particular the part of the world
that is more formally known as the space of reasons. Some of one’s
justifications will just have to be the demonstration of one’s competence in
using concepts in the world. That is, in the course of a discussion, my acumen
at applying concepts might be the source of a criticism. The most intuitive case
with regard to skillful application of concepts is the defense of an observation

14 Again, see the caveat in the earlier footnote about expanding the scope of our notion of
argument. Notice, though, that if we do expand the notion of argumentation to be co-extensive
with evidence, then there is not reason to think (as coherentists do) that beliefs (in the sense of
“believing” rather than “thing believed”) are the only thing that can justify a belief. That is, once
the move is made, non-belief based entities (the things we provide as evidence) can justify beliefs.
Indeed, this seems to be very much like the view I wish to endorse.
15 For those so inclined: it seems satisfactory to claim that it is the fact that these objects are

present which provides justification in such cases.

16 By ‘skill’ I mean what Hubert Dreyfus (in his reading of Heidegger) means: roughly, a way of

coping with the world which is in principle uncodifiable, but which is sill a participation in a
practice (and thus a form of rule following). However, the force of the point is not, in fact,
dependent upon a Dreyfusard reading of ‘skill’, even in everyday understanding of what we mean
by skill should be sufficient for the points I make in the present context.
206 Matthew Burstein

statement. Often times, in order to respond to an objection about an

observation statement, one must show that she can use the concept as it ought
to be used. To illustrate this point: imagine the putative objector pointing out
to other candidates for observation and asking if they have the property in
question as well. Being able to pick out the proper objects would be a sign,
though of course a defeasible sign, that one does understand how to apply the
concept in question.17 However, the assertions that would be made in such a
context – for example, “x is red,” “y is not red,” and “z is red” – are not
premises from which “w is red” is inferred; these premises do not entail “w is
red.”
So, our justificatory repertoire surpasses the bounds of argumentation;
argumentation, while important, does not exhaust the domain of possibility
when it comes to justification. Moreover, there is a stronger case to be made:
there may be contexts in which argumentation fails to be the relevant form of
justification: (a) in the ugli fruit case, had one replied to me with an argument
that there could not be such a fruit in the kitchen (instead of checking), this
would have been a peculiar way of attempting to justify his belief;18 and (b)
were I to claim the ability to perform a fantastic stunt on a skateboard, one
would be surprised if I further claim that my best proof for this claim are
further premises (rather than the demonstration of the stunt). Indeed, in these
cases, the inability to provide evidence or to demonstrate the skill are
constitutive of the failure to justify your claim.19
This brings us back to the central problem here, which, as I have said, is
that one need not have an argument for one’s belief in order to have a reason to
believe – that there is a real difference between what I have been calling broad
epistemic dependence and “being inferable, under the right circumstances,

17 It might be that, in any given case, we may have to move beyond such skillful demonstration;
my point, though, is that there are times when such demonstration would be sufficient. In such
circumstances, the assertions made are not further premises from which the conclusion is inferred.
18 One way of reading this point is that not all methodologies are philosophical methodologies.

Argumentation may have an important place in discourse, but to make it all of justification is to
distort and impoverish our justificatory practices.
19 To prevent misreadings, I add the following caveat: I should not be read as saying that one

performance can justify these kinds of claims, or that one failure renders them unjustified, or
(worse) to say that there just are “results” of this variety (evidence or demonstrations). All of
justification is defeasible, so there justifications can be defeated as well. Moreover, there may be
much to be said with regard to demonstrations: about why they are not indicative of a skill, for
example. (Consider: even if I am really capable of fantastic skateboarding stunts, that does not
mean that I always succeed in carrying them out.) Finally, these other forms of evidence can only
serve as evidence given up-and-running justificatory practices – to say otherwise is to turn these
entities into the Given.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 207

from φ.” There is a straightforward application of the foregoing discussion to

the present case: if I have appropriate evidence that something is the case, then
I have a reason to believe it. Sometimes evidence is not argument from further
premises; sometimes my evidence is my skill at performing an activity, and
sometimes my evidence is the fact that is present. Even if it is often the case
that my believing in a particular case might come under the kind of scrutiny
which requires further argumentation, it does not follow from this that all
challenges of my reports are like that. While we can always give some sort of
rational reconstruction – perhaps we can, in any particular case, give many
different adequate rational reconstructions – we do not require their
justificatory services in all contexts. The only reason to think that we need rely
on such rational reconstructions is if we believe that the only way for
understanding our justificatory repertoire exclusively in terms of
argumentation. Given that our justificatory repertoire has a broader array of
resources than argumentation, we ought not assume that argumentation is the
sole source of reasons for belief.20

2. The Central Problem for BonJour’s View

To recap: the suggestion that I have been develop is that least one source of
BonJour’s mistakes comes from a misreading of Sellars; namely, his
conflation of “being dependent upon” and “being inferable from.”21 One of the
central lessons of EPM is that a belief can be dependent upon other beliefs
without thereby, in any given context, being inferable from them. This
conflation pushes BonJour to endorse premise 4 of the BAC – that “only way
to be in cognitive possession of [a justificatory] reason is to believe with
justification the premises from which it follows that the belief is likely to be
true” (1985, p. 32). Thus, BonJour commits himself to the view that beliefs are
the only justificatory entities, and that justification involves all and only the
marshalling of beliefs into arguments. In this way, we are forced quickly from
premise (4) to coherentism.
However, one need not take the Sellarsian call for holism about reasons to
entail coherentism. Coherentism is a thesis which claims that all entitlement to
belief is conveyed inferentially. Holism about reasons only claims that

20Elements of this are drawn from Burstein (2002).

21 I should hasten to add that another of his errors (one that results from this conflation) is thinking
that the only justificatory entity that a belief can be dependent upon is a belief – rather than, say, a
background of justificatory practices, perceptual skills, and a shared environment.
208 Matthew Burstein

something can function as a reason only in light of what we might call

“holistic considerations”: being situated in discursive practices, being a part of
the shared environment, or having an epistemic skill, for example. That is,
something can be a reason only in light of being a part of the space of reasons;
this, however, does not mean that being so situated is a variety of inferential
articulation of beliefs or that it conveys its justification via inference. This
point is just is a recasting of a central lesson of EPM: that there is a difference
between ‘dependent upon’ and ‘inferable from’. BonJour simply fails to take
up this distinction – for the coherentist, justification comes only in one flavor:
inferential. However, a Sellarsian can countenance the other varieties of
justificatory practices that I have highlighted above: redemption by skillful
demonstration and production of evidence. The true consequence of Sellarsian
anti-foundationalist arguments is that nothing can serve as a reason except
within ongoing justificatory practices; by contrast, BonJour’s coherentist brand
of anti-foundationalism makes the stronger claim that the only thing that can
serve as a reason for belief is another belief. Alternatively – to appropriate
McDowelllian locutions – we might say that, for BonJour, the space of reasons
is constituted exclusively by beliefs, but for a Sellarsian, the space of reasons
is constituted by anything that can be a reason: perhaps beliefs, but maybe also
facts, skills, or social practices.22
One of the reasons why the Sellarsian story is much more plausible than
either the coherentist’s or the foundationalist’s is that it is a permissive
account; whereas both the foundationalist and the coherentist trace
justificatory legitimacy through to a single touchstone (e.g., basic beliefs or the
coherence of beliefs), the Sellarsian approach allows us to keep our intuitions
that there are a variety of justificatory practices whose legitimacy does not
trace back inferentially to such a touchstone. While there are still strains of
foundationalism within the coherentist’s account (given that all justification
comes from reference to a single entity), the Sellarsian position is thoroughly
antifoundationalist, both with regard to epistemology and with regard to
meta-epistemological considerations (e.g., how ought we conceive of the
structure of justification?).
As a result of the residual foundationalist impulses found in coherentism, it
is no surprise that BonJour’s tools for dismantling foundationalism are quite
different from those of Sellars. Sellars specifically takes aim at the claim that
some thing could convey justification without itself requiring justification; the

22 I take this list to be illustrative rather than exhaustive. Since my point here is only to show that

BonJour’s view does not countenance a broad enough view of reasons, it would be fortuitous that
there be even more reasons than I have listed here.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 209

problem with this claim is that anything that can provide justification (e.g., be
a reason) must be in the logical space of reasons. However, for a thing to be
located in the space of reasons means that it can (perhaps must) be querriable.
Whereas for BonJour such queries must have been settled ahead of time by an
agent (see premises 3 and 4 of the BAC), the Sellarsian view holds that
discourse has a “default/challenge” structure. The essential idea of such a
structure is that often, when we attribute a belief to someone (ourselves
included),
entitlement [to that belief] is attributed as well, by default. The prima facie
status of the commitment as one the interlocutor is entitled to is not permanent
or unshakeable; entitlement to assertional commitment can be challenged.
When it is appropriately challenged (when the challenger is entitled to the
challenge), the effect is to void the inferential and communicative authority of
the corresponding assertions (their capacity to transmit entitlement) unless the
asserter can vindicate the commitment by demonstrating entitlement to it.
(Brandom 1994, pp. 177-178)
The default/challenge structure of justification provides a way of
understanding how there might be direct knowledge or a place for challenges
to stop even when “is no point fixed in advance where demands for
justification or demonstration of entitlement come to an end” – just as the
antifoundationalist would have it (Brandom 1994, p. 178). Justification comes
to an end for an agent when the legitimacy of a challenge (i.e., the challenger’s
entitlement to make it) comes to an end; this legitimacy might (indeed, most
likely does) vary from context to context. This approach would allow the
antifoundationalist to head off concerns about regresses of discourse without
having to rely on a unified solution to such a regress.23
By formulating the BAC in foundationalist terms – that is, by focusing on
whether there is a layer basic beliefs that, in a unified manner, stems the
epistemic regress – BonJour saddles himself with views about the discursive
situation, argumentative strategies, and philosophies of mind that he need not
have – and, given his Sellarsian sympathies, that he should not have. The
Sellarsian challenge to foundationalism has many strands, none of which
address the epistemic regress argument. By contrast, BonJour’s attack on
foundationalism focuses on the regress argument and the relationship between
beliefs. By taking this as a starting place, BonJour covertly imports many

23 I use ‘would allow’ here because, for current purposes, it is sufficient to show that there are

ways of conceiving of discourse which are antifoundationalist through and through. What
differentiates a Sellarsian view is that it is thoroughly antifoundationalist; BonJour’s view is not.
210 Matthew Burstein

foundationalist assumptions about the nature of justification – in particular, he

imports the problematic Cartesian assumption I have been discussing: that only
justified beliefs can convey justification.24 BonJour is preoccupied with the
relationship between beliefs, rather than reasons for belief. If we have this
anticartesian reading of the mind, some of the reasons I marshal are not my
beliefs, they are reasons.
It is clear what BonJour’s response will be here: the only way one can have
access to reasons is via beliefs. That is, direct awareness or immediate
perception is impossible just because the only justificatory elements are
beliefs. However, this line of argumentation will only work if my arguments in
terms of a permissive account of the space of reasons fails; if my argument,
above, succeeds, then it is simply a mistake to think that we need beliefs to
constitute the space of reasons (to the exclusion of everything else). Moreover,
coherentism is not an option if one denies this variety of Cartesianism.
Consider a permissive coherentism, one which allowed beliefs and facts (or
whatnot) into the space of reasons. In what way could we construe trees as
cohering with numbers or with beliefs? The content of belief – the propositions
that beliefs are attitudes toward – is the kind of thing that can cohere;
coherence is a kind of relationship between discursive entities. It is difficult to
see how the entities in the world could come to stand in such a relationship
with beliefs and justificatory practices. Furthermore, there is another difficulty
for such an anticartesian coherentism: the lack of an argument that the world
itself is coherent.25 That is, we would require an a priori argument that the
world only comes in cohering ways; otherwise, there could be no demand for
coherence to be the source of justification – the (incoherent) world itself would
undermine the very possibility of knowledge.26 I admit that this is probably a
peculiar suggestion; however, if someone were to develop an anti-Cartesian
coherentism (i.e., one in which non-belief elements of the justificatory
repertoire had to cohere with beliefs), she would have to provide some

24 I take it that this is the source for his taking on the antifoundationalist’s picture of the discursive

situation, rather than something like a default/challenge structure.

25 This makes BonJour quite unlike Bradley, whose coherentism was a theory of truth, not

justification. Bradley’s coherence theory of truth had as part of its defense an argument for the
coherence of the world. This only heightens the tension we see in the absence of such a defense,
and accentuates my point about the Cartesian underpinnings of BonJour’s view.
26 It has been suggested in response to these claims that this view presupposes a correspondence

theory of truth. This suggestion is mistaken however, and it turns on taking it that the only entities
that play a role in justification are beliefs. If our justificatory repertoire is expanded in the way I
suggest, then non-beliefs must cohere with each other as well as beliefs; this problem arises for
such a coherentism regardless of theory of truth – excepting a coherence theory of truth, of course.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 211

argumentation that rules out this (admittedly implausible) possibility in order

to provide an adequate accounting of the nature of justification.
Thus, there is a kind of category error in suggesting that anything other
than beliefs can cohere, and so the possibility of an anti-Cartesian coherentism
is a bit of an absurdity. And, if as the discussion here has shown, a Cartesian
coherentism fails to have the resources required for an account of our entire
justificatory repertoire – not only does it fail, as is commonly noted, to make
adequate sense of perceptual deliverances from the world, but it more
generally fails (as all Cartesian approaches to epistemology) to make sense of
the broad array of justificatory resources we have available. So, BonJour had
ample reason to give up on his coherentism; in the next section, I intend to
show that he has equal (and similar) reason to give up his foundationalism as
well.

3. Stanching the Slide from Anti-Foundationalism to Coherentism

Now that BonJour’s arguments against foundationalism are on the table and I
have shown how it is that he moves from the denial of foundationalism to
coherentism, I want to discuss the ways in which anti-foundationalists need not
feel forced to follow BonJour down this path. This is of interest for two
reasons: first, because he has, since writing SEK, endorsed foundationalism,
and second, because he cites Sellars as the inspiration for his coherentism. As I
suggest above in my discussion of the basic antifoundational argument, there
are reasons to think that a Cartesianism mindset is BonJour’s motivation for
moving from arguments against foundationalism to conclusions endorsing
coherentism. I think that the move from the former to the latter is too quick, as
there are positions in between – indeed, I think the correct one falls between
the two.
In fact, the arguments above suggest a two-fold motivation for BonJour’s
slide from anti-foundationalism to coherentism: his Cartesianism and his
eliding of the distinction between “dependence upon” and “inference from.”
That is, what ultimately motivates BonJour’s coherentism is his view that the
only entities that can serve as justification are beliefs (and the arrangements of
beliefs that are arguments) and that the only way in which justification is
conveyed is via logical inference. While I have thoroughly critiqued the view
that only beliefs can serve justificatory roles, I have yet to take up explicitly
the way in which non-belief entities come to have justificatory status, and how
such non-logical entities can convey justification. I intend to do so now.
In my earlier discussions, I highlighted ways in which our justificatory
repertoire is broader than mere marshaling of beliefs in logical argumentation:
212 Matthew Burstein

in my discussion of the ugli fruit I introduced a notion of evidence, and in my

discussion of the BAC I introduced the notion of skillful demonstration.
(Perhaps ‘introduced’ is a bit strong – it is akin to calling various European
explorers’ landings on the North American continent the “discovery” of the so-
called New World.) In each instance, I claimed that each of these varieties of
justification carry their own justificatory force in a way that does not allow for
reduction to belief/argumentation models of justification. The account I had in
mind is the view that, while argumentation is a form of justification, our
justificatory repertoire extends further than argumentation – in much the same
way that a square is a kind of quadrilateral, even if it is not exhaustive of the
set of things which are quadrilateral.
However, this may seem to be at odds with my endorsement of Sellarsian
arguments against the Given. One might be tempted to see these appeals as
either appeals to other foundations, or, perhaps worse, as to be endorsing an
even more pernicious version of the Given. In what follows, I hope to show
that neither of these claims is correct, and in so doing I shall be able to better
elaborate what I take to be the correct picture of justification.
So, let us turn to the first concern: that I have merely expanded the varieties
of foundations rather than undermined the possibility of them. Given what I
say about evidence and epistemic skills, it might be said to follow, thereby,
that evidence and skills are themselves the foundations of knowledge –
because, on my own account, perceptual statements are founded on them. This,
however, is to miss the force of my argument, for two reasons. First, I do not –
nor ought one – claim that all justification must be articulated back to skills
(rather than, say, basic beliefs). For skills to be foundational in the way that the
objection suggests, I would have to claim that the legitimacy of belief must be
inferentially descendent from skills. I do not think that this is so – because
legitimacy of belief need not be traced back to any kind (or sets of kinds) of
entity. This is fundamentally the synchronic picture of knowledge, one which
Sellars explicitly rejects in EPM (1997, §38/pp. 77-79). The rationality of
belief, for Sellars, comes not from permanent lineages of justificatory
legitimacy, but rather from its being “a self-correcting enterprise which can put
any claim in jeopardy, though not all at once.”
Further, this Sellarsian picture of knowledge is one of being responsible for
one’s judgments – that is, to be a participant in up-and-running justificatory
practices. The upshot of the arguments in EPM is that we never have
knowledge in virtue of being merely in a causal state; we must be causal
beings within the space of reasons in order to be said to know that such-and-
such is the case. The difference between we epistemic agents and
thermometers is not a causal difference, but rather a discursive difference – we
are responsible for our judgments about our causal connections with the world.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 213

So, merely being in possession of a perceptual skill (for example) is not yet to
know that something is the case; instead, it makes possible (a) the navigation
of the world (e.g., without running into things) and (b) perceptual knowledge.
Having and exercising a perceptual skill is not sufficient for knowledge in just
the same way that being properly causally related to phenomena is not –
because the central enabling condition of knowledge (e.g., being in an ongoing
justificatory practice) is still lacking.27
Now, let us turn to the second concern: that my use of ‘evidence’ is an
invocation of the Given. The inclination to raise this objection comes from the
fact that I have claimed that facts can come to justify our beliefs, and this
makes it sounds like the facts themselves are Given to us. Or, to put it another
way: that there just are facts which we can directly know, independent of
anything else. And, indeed, if I were claiming this, I would be indulging the
mythological Given. But, I do not claim this to be the case. What I claim is
that there are facts that we directly know. However, this knowing is only direct
because it is neither inferential nor need it be mediated by mental states. Given
the distinction between ‘dependent upon’ and ‘inferable from’, there is room
for knowledge that is dependent upon justificatory practices, perceptual skills,
theoretical paradigms, and the like, but which is direct knowledge nonetheless.
Evidence is only evidence in light of its being available to play a role in
justificatory practices, after all. So, while I claim that we do have direct access
to facts, it is only in light of being properly situated in the space of reasons that
we can have such access. Given a default/challenge discursive structure, we
can maintain the possibility of such direct knowledge without requiring it to be
unchallengeable; all we might need in such circumstances is that it is
unchallenged (in the sense that such a challenge would be illegitimate).

27 One reason we might be tempted to draw the line here is because of “animal perception.” That
is, animals clearly perceive their environments, and sophisticated animals make plans of action
based on these perceptions of their world. (Indeed, perhaps it is just this ability to make plans of
action that inclines us to call them sophisticated.) So, clearly these are not mere thermometers;
they do not merely reliably respond to their surroundings – they act so as to engage their
environments and to change them. However, it would be a mistake to think that cognition is all or
nothing – that either one is a thermometer or one is a cognizer. What we have in common with
them in perception is not knowledge but rather is perceptual skills. The mere fact that non-human
animals can comport themselves to an environment does not thereby entail that they have beliefs
about their environment – unless, of course, the only way of relating to an environment is in a
Cartesian mode (e.g., mediated by mental states such as beliefs). Since I have shown that there is
some reason to doubt Cartesianism, we can see that we have an option between the all-or-nothing
positions (in deed, it strikes me as yet another virtue of anti-Cartesian approaches to the mind that
it allows for such nuanced views of the world).
214 Matthew Burstein

Moreover, another piece of the argument from EPM will bolster our
arguments against BonJour here. Sellars speaks of the “logical space of
reasons” in Section VII, §36 (1997, p. 76):
It is not just that the objection supposes that knowing is an episode; for clearly
there are episodes which we can correctly characterize as knowings, in
particular, observings. The essential point is that in characterizing an episode or
a state as that of knowing, we are not giving an empirical description of that
episode or state; we are placing it in the logical space of reasons, of justifying
and being able to justify what one says.
This logical space of reasons is of central importance to the Sellarsian project,
and in many ways BonJour’s work overlooks the role it plays in a Sellarsian
antifoundationalism. In particular, he fails to adequately distinguish two very
different ways of understanding what it means for a bit of experience to be
“conceptualized” – namely between “being mediated by mental states” and
“being in the space of reasons.” By following the Sellarsian idea that
conceptualization marks off a logical space rather than a psychological one,
we have way of conceiving of these objects as being proper justificatory
entities; it allows us to understand them as being situated within a justificatory
practice without, thereby, also being limited to being situated within the
domain of mental objects. The reason one might suppose that evidence has to
be mediated by mental states is if the only way for an object to be
conceptualized – to be taken into the space of reasons – is by its being dealt
with in a cognitive way. However, if our comportment to the world is practical
as well as cognitive, then there is a mode of conceptualization, which is itself
practical. That is, we need not think of “conceptualization” as a variety of
voluntary cognition, but rather an exercise of practical comportment toward
the things in the world. To understand what role something has in the space of
reasons is to know how to use it – and this need not (given that we are talking
about “knowing how” rather than “knowing that”) require belief at all.
So, mediation by mental states cannot serve as the mode of
conceptualization in the space of reasons, for the simple fact that the space of
reasons – and our activities within it – outstrips the resources of the domain of
mental states. And, insofar as the foregoing is the case, evidence, as I use it (as
an appeal to facts), is not Given. My account of evidence does not serve as an
end point to inquiry (e.g., it is not foundational in the way that foundationalists
use the term), nor does it – merely in virtue of its being the kind of thing it is –
trump other kinds of explanations. The role it plays is an important one, no
doubt, but it is not the whole story of justification.
 Prodigal Epistemology: Coherence, Holism, and the Sellarsian Tradition 215

4. Coherentism as a Prodigal Epistemology

Unlike the holism I have attempted to sketch throughout this work,

coherentism is a prodigal epistemology; it squanders the valuable
(justificatory) resources available to it in favor of a rarefied sort of
justification: argumentation. Without a doubt, argumentation is a deeply
important part of justificatory practices – surely an epistemologist who does
not want to talk herself out of a job will develop a keen appreciation for the
justificatory role of argumentation! What I have shown, however, is that there
is more to justification than argumentation, and, as a result, that there is more
to epistemic dependence than inference. Moreover, coherentism wastes the
resources of the one of the views from which it draws inspiration and
sustenance: the arguments of EPM. By limiting itself to an account of reasons
that is psychological – by adopting a Cartesian rather than methodologically
behaviorist philosophy of mind – BonJour has needlessly wasted valuable
tools for understanding the relationship between mind and world.28

Matthew Burstein
Department of Philosophy
Southern Methodist University
Dallas, TX 75205
United States of America
e-mail: [email protected]

REFERENCES

BonJour, L. (1985). The Structure of Empirical Knowledge. Cambridge, MA: Harvard University
Press.
BonJour, L. (1999). Foundationalism and the External World. Philosophical Perspectives
(Epistemology) 13, 229-249.
Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press.
Burstein, M. (2002). Epistemology and the Philosophy of Cinema. In: K. Stoehr (ed.), Film and
Knowledge: Integrating Images and Ideas. Jefferson, NC: McFarland and Company, Inc.
Publishers.
McDowell, J. (1996). Mind and World. Cambridge, MA: Harvard University Press.

28I am indebted to the following for assistance with various versions of this piece: Mark Lance,
Wayne Davis, Bill Blattner, Dan Levine, James Mattingly, Karen Stohr, Maggie Little, Nathaniel
Goldberg, and Michael P. Wolf.
216 Matthew Burstein

Rorty, R. (1979). Philosophy and the Mirror of Nature. Princeton, NJ: Princeton University Press.
Sellars, W. (1963). Science, Perception, and Reality. Atascadero, CA: Ridgeview Publishing
Company.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Williams, M. (1999). Groundless Belief: An Essay on the Possibility of Epistemology. Princeton,
NJ: Princeton University Press.
Wittgenstein, L. (1969). On Certainty. New York: Basil Blackwell Pub.
Mark Owen Webb

MEETING OTHERS IN THE SPACE OF REASONS:

FALLIBILISM FOR SELLARSIANS

ABSTRACT. Certainty has proved to be a troublesome epistemological concept, which

motivates many philosophers to be fallibilists. But fallibilism proves troublesome, too,
as it is hard to state in a way that does not either imply skepticism, or deny that there
are necessary truths. The Sellarsian idea of a space of reasons in which there are
normative proprieties attached to epistemic positions allows for an understanding of
fallibilism which allows that there is knowledge, there are necessary truths, and yet we
can be wrong about anything. This result is accomplished by taking the fallibilist claim
as a normative recommendation, not a descriptive thesis.

The quest for certainty has gotten epistemology into a lot of hot water, and we
should give it up as a mistake. We should freely admit we cannot be certain of
anything, and move on. It is, of course, a reasonable question whether we can
consistently get along without certainty, and even if it is possible, whether
there is some terrible price to be paid if we do. I will argue, using resources
from Sellars and Brandom, that it is indeed both possible and salutary to do
without any claims to certainty. I will also argue that, while there may be
prices to be paid, they are by no means as high as the prices to paid by those
who wish to keep certainty as a viable epistemological concept with any actual
application. The motivation for these claims rests on the Sellarsian/
Brandomian interpretation of epistemic principles as normative rather than
descriptive.
But first, let me make good my claim that a lot of trouble in epistemology
has been caused by the quest for certainty. First, consider the case of Plato. In
the Republic, Plato argues that since knowledge requires certainty, and
certainty requires an unchanging subject matter, true knowledge can be only of
unchanging forms. There can be no true knowledge of the changing physical
world, which then becomes the realm of changing, uncertain doxa, mere
opinion. In a stroke, all the kinds of knowledge we take to be constitutive of
science are demoted. It is true that Plato has other reasons for espousing his
theory of forms, and even if he had himself followed my advice and given up

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 217-231.
Amsterdam/NewYork: Rodopi, 2006.
218 Mark Owen Webb

on certainty, he would have had ample reason to retain the theory.

Nevertheless, some of his most perplexing epistemological pronouncements
would have been left entirely unmotivated. For example, the doctrine of
Recollection espoused in the Meno rests on the claim that the only possible
explanation for geometrical knowledge, given its apparent innateness and
certainty, is that everyone was in contact with the forms before they were born.
No certainty, no problem, or at least a different problem. It would have been
open to him then to embrace the justified true belief analysis of knowledge –
more accurately, true belief with a logos – that he ultimately rejects in the
Theaetetus. Epistemology could have leaped ahead two millennia, and some
Hellenistic Greek could have played the functional role of Ed Gettier.
Perhaps a more famous victim of the quest for certainty is Descartes. He
noticed that he was often wrong about things he thought he knew, even about
some things he believed with a great deal of confidence. Not only had some of
his well-grounded theories turned out to be wrong, but even the most certain of
things, like his mathematical calculations (who has not made a mathematical
mistake?). Even his senses sometimes deceive him, when he experienced an
illusion or hallucination. Because of this, he came to realize he was not at all
sure what he knew and what he did not, so he tried to sort out the real
knowledge from everything else. The only way to do this, he thought, is to try
to doubt everything, to see if there is anything that cannot be doubted. In other
words, he applied a very strong test of certainty; all knowledge is certain, so if
there is any real knowledge, it will have to be something that is absolutely
immune to doubt. To check his beliefs to see if they were certain, he devised
his famous two tests: the Dream test and the Evil Genius test.
Everyone who has passed Beginning Philosophy knows how it goes from
here. The only things that are absolutely immune to doubt are that I exist and
that the current contents of my consciousness are whatever they are. A strong
requirement for certainty – namely indubitability – seems to leave us trapped
in our own heads, unable to acquire knowledge of an external world at all.
Indeed the straits are so desperate that Descartes has to enlist God’s help in
getting out, by way of one of the most dubious arguments for the existence of
God ever offered in the history of philosophy.
But even empiricists are not immune to the charms of certainty. Like
Descartes, Hume is interested in discovering the extent of human knowledge.
Unlike Descartes, Hume does not demand absolute indubitability for
knowledge; he is content to settle for a lesser degree of certainty (he uses the
word ‘assurance’). For Relations of Ideas, it is enough that they should be
demonstrable. For Matters of Fact, it is enough that they should be either the
deliverances of sensation or memories of such. The senses and memory, while
not guaranteeing certainty, are good enough to rest on. But again, those who
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 219

have studied Hume know how the story ends. It turns out that if all knowledge
must be of one of those two types, there can be no non-circularly grounded
knowledge of anything outside one’s own experience. After all, all such
grounding would have to be in the form of reasoning, which would have to rest
on knowledge of cause and effect. Reasoning about cause and effect is always
based on past experience, as experience alone can teach us what causes go
with what effects. But such reasoning can tell us about things outside our
experience only on the assumption that things outside our experience are
relevantly like things in our experience, and this assumption is itself neither a
demonstrable Relation of Ideas nor a thing learned from experience. So the
problem of induction is born, and even the limited kind of certainty Hume
sought led to trouble. In fact, it led directly to Kant’s transcendental idealism,
and from there to Hegelian idealism, not to mention Schopenhauer and Fichte.
In the end, we owe to the quest for certainty many kinds of skepticism, the
problem of induction, and a raft of unattractive idealisms; if that is not trouble,
I do not know what is.
But can we give up certainty? Surely we can easily give up indubitability in
the Cartesian style as criterial for knowledge, but can we cut ourselves
completely loose from Hume’s much more innocuous demand that knowledge
be based on some kind of reasonable, certifying ground? In order to answer
that question, we need to get clear on what it is we would be giving up. The
word ‘certainty’ is used in a lot of different ways, and we must be sure we are
not indicting some for the sins of others. To begin with, we can dispense with
psychological certainty without another thought. A feeling of certainty is
clearly not necessary for knowledge, and bitter experience has taught us all
that it is not sufficient. We should also distinguish certainty, whatever it is,
from two less problematic notions: knowledge and necessity. The kind of
certainty I am concerned with is not simply identifiable with knowledge, and
therefore giving up on certainty is not simply embracing skepticism. As we
will see, it is an open question whether we can give up on certainty without
embracing skepticism, but if fallibilism entails skepticism, then that is
sufficient reason to give up on fallibilism. Similarly, fallibilism is too
expensive if its price is that there are no necessary truths. To say that we
cannot be certain of a proposition’s truth should not be taken to entail that the
proposition in question in contingent. Again, it is an open question whether it
is possible to split these notions apart, but if fallibilism entails that there are no
necessary truths, we should give up fallibilism. Finally, we should not have to
give up the use of the word ‘certain’ and its cognates. Lots of words have
natural uses that do not lend themselves to substantive epistemology. Feel free,
under appropriate conditions, to claim certainty for yourself as to where you
220 Mark Owen Webb

parked your car, or what your birthday is, as long as you do not use that claim
as warrant to do any epistemology.
You may reasonably wonder if there is anything left that I can ask you to
give up. It turns out to be difficult to spell out.1 Fallibilism, if true, demands
we give up any claim to infallibility, with respect to any proposition, in any
situation. There is no epistemically privileged person or realm. Absolutely
anyone can be wrong about absolutely anything. There are excellent reasons to
suspect that such a fallibilism is true. Certainly we can be wrong about most
logical and mathematical truths, be they ever so necessary. Though we may be
subjectively certain they are wrong, even dialethic logicians (those who, like
Graham Priest 1979, believe that there are true contradictions), should not be
dismissed without a hearing; after all, they may be right. There is even good
reason to suppose that we can be wrong about our own mental states. Perhaps,
as Bertrand Russell noted (1959, p. 19), even the belief that I exist can turn out
to be wrong, since such a belief involves supposing that there is an enduring
substance underlying my thoughts. The Buddha thought it was a mistake to
believe in your own existence, and a great many Buddhist thinkers after him
have found no problem in that doctrine.2 Eliminative materialism and other
movements in the philosophy of mind have given us some reason to think we
can be wrong even in the positing of our own thought or mental states, insofar
as these involve applying predicates that have no application.3 It is not far to
see that the same sort of thing could be said for pains and other qualia. So it is
not entirely unreasonable to say that we have reason to believe that human
beings are fallible in every epistemic realm. In other words, we can be wrong
about anything. The above insights gleaned from recent philosophy afford us a
general argument:
(1) Every assertion involves, at a minimum, identifying some property as either
instantiated or related to some other property.
(2) Whatever a person’s evidence, she may have misidentified the relevant
property or properties.4

1 It is so difficult that some people have given up entirely, and have changed the subject. See, e.g.
Stephen Heatherington (1999, pp. 565-587).
2 See, for example, the Nikaya (1995, X.60); Harvey (1995), Collins (1990).
3 See Churchland (1988) and Stich (1983).
4 This premise rests on securely Sellarsian ground. In his attack on what he called “the Myth of

the Given,” he had a lot to say about how allegedly foundational observation-statements needed to
be both conceptually contentful (to do the justificatory work necessary for foundational beliefs)
and intrinsically credible (to enjoy the authority necessary to have justification to transmit). But
the application of concepts necessary to secure the former undermines the latter. See Sellars
(1997).
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 221

Therefore,
(3) Anyone may be wrong about anything.
A full-fledged defense of premise 1 is hard to mount, but given the examples
above, it has some prima facie plausibility. Premise two seems to be an
obvious consequence of the human condition. Moreover, it seems to be
supported by the fact that past claims of infallibility for some realm or other
have always quickly been followed by doubts being raised about that very
thing. Think of the various Kantian proposals for synthetic a priori
propositions. All but the mathematical examples are now seen to be
controversial metaphysical claims, if not downright false. Or consider the
notion of absolute simultaneity, once thought to be entirely unproblematic, but
now known to be untenable. There is a strong inductive case, at least, for our
universal fallibility.
A problem arises when we try to formulate explicitly this doctrine about
the human condition. Any adequate formulation of the thesis should fulfill two
conditions. The first can be expressed as a dilemma: the formulation should
neither entail skepticism, nor entail that there are no necessary truths. Some
theorists try to dodge this requirement by restricting their statement of the
thesis to cover only empirical truths, but there is ample reason to think we can
be wrong about the a priori, too.5 The second requirement is that accepting the
formulation should be, in some yet-to-be-analyzed way, an antidote to
dogmatism. The great fallibilist C. S. Peirce thought that espousing fallibilism
is the only way to insure that the road of inquiry was not blocked, and that this
is an excellent reason to be a fallibilist. After all, one of the points of
espousing and arguing for fallibilism has always been to inculcate intellectual
humility as ultimately beneficial to the process of inquiry. In other words,
whatever our fallibilism amounts to, it should have normative bite, counting
certain kinds of epistemic behavior as wrong.
‘Dogmatism’ has been the name of a lot of things in the history of
philosophy, from an epistemic vice to the denial of skepticism. For the
purposes of this inquiry, being dogmatic is not to be understood as having an
attitude of confidence toward one’s belief, even if that confidence is extremely
high, or even higher than the subject’s evidence warrants. Instead, we should
understand dogmatism as a disposition to behave in a certain way in the face of
challenges to one’s beliefs. In particular, we should understand dogmatism as
the disposition to ignore arguments or evidence that one ought to consider as
having a bearing on one’s beliefs. Note that one might have such a disposition

5 For example, see Lehrer and Kim (1990) and Meerbote (1977).
222 Mark Owen Webb

for a great many reasons; the disposition need not spring from overconfidence,
though that will frequently be its source. A person might be disposed to ignore
relevant evidence because of bad training or habits. It is also worth noting that
this understanding of dogmatism makes it an objective matter whether one is
dogmatic; anyone who ignores relevant evidence, whether she appreciates that
she is doing so or not, is dogmatic.6 On this understanding of dogmatism, it is
clearly the case that dogmatism is a bad thing.
The preceding paragraph leaves us with a problem in understanding the
second adequacy condition, since it is difficult to see how any thesis can be
inconsistent with a disposition – or, indeed, with anything other than another
thesis. Obviously, we cannot be looking for logical inconsistency here. The
kind of inconsistency we need is the kind we find in appraising actions by
comparison to a moral code. For example, greed (a disposition) can be said to
be inconsistent with the Golden Rule. The Golden Rule is not a thesis, but it is
a propositional entity of some sort. Greed is inconsistent with it in that an
appreciation of the Golden Rule gives one excellent reason to try not to be
greedy. In the same way, any adequate formulation of fallibilism should be
such as to give people excellent reason to try not to be dogmatic.
The problem of formulating fallibilism can now be reformulated; it is the
problem of capturing the insight that we ought always to be prepared to
consider evidence against any of our beliefs, but not in a way that entails either
that we do not know anything or that there are no necessary truths. Many have
recognized the second danger; that I could be wrong in any of my beliefs
seems to entail straightforwardly that any of my beliefs could be false. Since
some of my beliefs are of tautologies, it seems to follow that tautologies are
possibly false. This shows that it is inadequate to identify fallibilism with the
thesis that each of my beliefs is possibly false, for any non-epistemic notion of
possibility. The thesis can be understood as the claim that it is epistemically
possible for any of my beliefs to be false, but that tack substitutes for one
unanalyzed notion another equally in need of analysis. Once we have an
adequate semantics for epistemic modal operators, formulating global
fallibilism is a trivial exercise.
Here is one try at formulating fallibilism. Susan Haack tries to avoid the
problem of meeting the first adequacy condition by understanding the thesis of
fallibilism as a disjunction: For any of my beliefs p, either my believing p does
not guarantee its truth, or I might have believed something inconsistent with p
(Haack 1979). The first disjunct amounts to what it is to be fallible with

6 For a good discussion of the notion of dogmatism, see Audi (1998, p. 207).
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 223

respect to the empirical, the second what it is to fallible with respect to the a
priori. This is expressed formally as
(p) [ ~(Bp ⊃ p) ∨ ◊B~p]
In other words, my fallibility is captured by my not tracking the truth in the
right way. As far as contingent a posteriori beliefs are concerned, I might
believe any of them without their thereby being true, so they present no
problem for the fallibilist. As far as my a priori beliefs are concerned, some of
them are necessary truths, and some are such that my believing them entails
their truth, but in each of those cases I might have had a different, false belief
in its stead. The same can be said for necessary a posteriori truths. This
formulation makes the thesis of fallibilism consistent with the claim that there
are necessary truths and a priori truths.
Haack’s formulation, as elegant as it is, is not without its own difficulties.
Peter Mott points out that Haack’s formulation of fallibilism (combined with
some plausible assumptions) entails the claim that it is possible for us to
believe anything at all.7 Even if we understand possibility here in its broadest
metaphysical sense, so that we need not hold the laws of human or non-human
nature constant, Haack’s principle still has the consequence that there are no
limits to our doxastic abilities. Finding this consequence unacceptable (as it
certainly is), Mott tries another tack. Noting that the locus of fallibility is to be
found in us and our believings, not in the content of our beliefs, he suggests
that fallibilism is best understood as the claim that we have no guaranteed
routes to the truth, “that there are no infallible methods” (Mott 1980, p. 179).
Even this formulation, Mott realizes, does not completely capture fallibilism,
because it is compatible with believing, on any given instance, that we are
certainly right, and so gives no particular reason to listen to contrary
arguments (Mott 1980, p. 182). In other words, although it correctly describes
our epistemic condition, it does not sufficiently underwrite epistemic caution
to be what fallibilists want. As a curative to this failing, Mott finally suggests

7 See Mott (1980). Here is a brief statement of his argument:

1. (p)[~ (Bp ⊃ p) ∨ ◊B~p] Haack’s formulation of fallibilism
2. p ⊃ (Bp ⊃ p) A theorem of T
3. p ⊃ ◊B ~p from 1 and 2
4. (p) ( ~p ⊃ ◊Bp) substitution of ~p for p; closure of belief under DN
5. ~◊Bq assumption for reductio
6. ~◊ (Bq & q) from 5
7. ◊B(Bq & q) from 6 and 4
8. ◊Bq closure of belief under Simplification
9. (p) ◊Bp reductio of 5
224 Mark Owen Webb

“a more sweeping admission the spirit of which is that we are always wrong”
(p. 182). This suggestion, of course, cannot be taken to be universally and
literally true, or paradox threatens immediately; the idea is that in all of our
high-level theorizing about things, we always have some degree of inaccuracy,
and we never know beforehand where that inaccuracy is. Very probably, then
(according to Mott), all our high-level theories are false. It is reasonably clear
that this does not escape paradox, either.
Another attempt to capture the idea of fallibilism has been offered by L. S.
Carrier (1974 and 1993). He starts from the idea of fallibility as proneness to
error, and defines the thesis as the claim that you never know that you are not
mistaken with respect to any of your beliefs. This is stated formally as
(a)(e)~Ka~Mae
where ‘Mxy’ means ‘x is mistaken with respect to y’. Being mistaken with
respect to a proposition is defined as believing it when it is false. Since you
never know, for any of your beliefs, whether you are mistaken in holding
them, you also never know whether your beliefs are false. Surely this
formulation meets the second requirement of underwriting epistemic caution
and counseling against dogmatism. It remains to be seen whether it meets the
first adequacy condition.
George Schlesinger (1985), in the process of developing his own view,
makes some serious criticisms of both Mott’s and Carrier’s views. He finds
Mott’s final version of the fallibilism thesis inadequate, even while agreeing
with Mott’s rejection of Haack’s version. Perhaps the most telling criticism is
that Mott’s statement of fallibilism – roughly, that there are no infallible
methods – means that there are no methods that always lead to the right answer
with respect to the truth value of a given proposition. But this is a much
weaker claim than fallibilist typically intend to make; they want to say
certainty about any proposition is never one of our attainments, whereas on
Mott’s principle, we could be certain much of the time in the use of a
particular method, as long as it was not the case that we were always certain in
the use of that method. So even though Mott’s thesis meets the first adequacy
condition (it entails neither that we know nothing nor that there are no
necessary truths), it founders on the second by being compatible with a
dogmatic attitude.
Carrier’s thesis is in even deeper trouble, according to Schlesinger. It is
inconsistent with the assertion that we do know some things, and therefore
fails to meet the first adequacy criterion. The proof is as follows. For any
proposition p, if I know it is true, then I know that any disjunction that has it as
a disjunct is also true. In particular, I know the disjunction of p with the claim
that I do not believe that p. That is equivalent by Carrier’s definition of the
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 225

being-mistaken operator to the claim that I know I am not mistaken with

respect to p. Therefore, since the process could be repeated with respect to any
proposition p, whenever we know that p, we can also come to know that we
are not mistaken with respect to p (1985, p. 25). Carrier objects that
Schlesinger needs a controversial principle of inference to make this argument,
namely, the Principle of Knowledge by Entailment (PKE): If you know that p,
and you know that p entails q, then you know that q (Carrier 1993). However,
it should be obvious that Schlesinger need not endorse anything like so strong
a principle. First of all, he only needs it to be the case that when you know that
p, then you know that (p or q); he need say nothing more general about
entailment. Second, he need not endorse the claim that you always know the
logical consequences of your beliefs, but only that you can come to know
some of them, and surely that is uncontroversial. If I know that p, and can
come to know thereby that I am not mistaken with respect to p, then it is
possible for me to have infallible beliefs, and so fallibilism is false, even on
Carrier’s own understanding of it.
Schlesinger’s positive account involves distinguishing several grades of
justification; he then defines several matching grades of epistemic possibility
as the absence of the corresponding degree of justification for the negation of
the proposition in question. This yields the result that there are several
corresponding grades of fallibilism, each defined as the claim that no one is
ever, with respect to any proposition, justified to that degree in believing it
(Schlesinger 1985, pp. 34-36). It is difficult to say what to make of this version
of the thesis, though, since its plausibility relies entirely on undefined
operators standing in for degrees of justification. We can partially define a
maximum and a minimum degree of justification in terms of their entailment
relations with other notions of justification; Schlesinger adopts this strategy,
and offers as the strongest kind of justification JFap, which we are to
understand as “a is justified in having full conviction that p” (1985, p. 33). We
can define it such a way that having this degree of justification entails having
every other kind, and there is no other degree of justification the having of
which entails it. The corresponding weakest form of fallibilism is the claim
that no one is ever justified to that degree, or
(e)~JFae
This certainly sounds like the kind of thing we were hoping to get at, but what
does it mean for someone to be justified in having full conviction that
something is true? The same problem we had formulating fallibilism in the
first place returns in another guise.
The deep problem with all of these views is the same: they either lack
normative bite, or acquire normative bite at the cost of skepticism. They are all
226 Mark Owen Webb

trying to define fallibilism as a description of our actual epistemic state, and

derive therefrom a normative propriety about how we should respond to other
inquirers. Most are dimly aware of the twofold nature of the beast they are
trailing. Haack says “What Peirce calls ‘fallibilism’ is, in part, an
epistemological thesis – a thesis about our propensity to hold false beliefs –
and, in part, an epistemological recommendation – that we should always be
willing to revise our beliefs in the light of new evidence” (1979, p. 41). She
then goes on to say that she will be concerned with the thesis, and only then
turn to the recommendation. Schlesinger says that “Fallibilism – unlike many
forms of skepticism – is to most people who subscribe to it a ‘practical’, as
distinct form a mere ‘metaphysical’, thesis” (Schlesinger 1984, p. 263). He
also criticizes Haack’s formulation on the grounds that it defines fallibilism in
terms of logical possibility, and so is completely compatible with total actual
infallibility. If so, then it is surely compatible with total dogmatism about all
my beliefs (1984, p. 268). Carrier says that “Fallibilism should not be defined
so as to be compatible with universal omniscience. To say that, as a matter of
fact, one is both omniscient and fallible appears to be a contradiction in terms”
(1993, pp. 364-365). Of course, this is mere appearance; there is no
contradiction. Only if you think of omniscience as an essential property – not
just actually knowing all the truths, but necessarily knowing all the truths –
does it entail infallibility. Nevertheless, Carrier is on to something.
What they are all noticing is the same thing: no matter how the descriptive
thesis is worked out, recognizing it to be true is perfectly compatible with a
dogmatic attitude in any particular case. I can grant that I am often wrong, or
that I am mostly wrong, or that none of my methods is infallible, or that all my
cognitive apparatus are prone to error, and still add, quite consistently, “but in
this case I am quite sure I got it right, in spite of your holding up apparently
relevant counterevidence, so I need not listen to you.” No descriptive thesis is
going to do the anti-dogmatic job fallibilists want done.
What will do the job, then? We have two choices here. The first, and, I
predict, more popular response will be to say, “So much the worse for
fallibilism; certainty is attainable after all.” My only complaints are a weak
one that believing we can be certain is tactically dangerous, as it leads to
theories of forms, problems of induction, and idealism, and an apparently only
slightly less weak complaint that allowing people to claim certainty makes for
bad epistemic practice. However, there are excellent reasons to believe that
this second objection is less weak than it appears. Assume that it is true that it
would be best if we all behaved in a non-dogmatic way, taking no position as
so well justified that we can safely ignore challenges to it. We are partly
justified in taking this to be true on the strong inductive evidence that for any
possible claim, there are imaginable challenges to that claim that are not
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 227

obviously crazy. As we saw earlier, even the law of non-contradiction has been
challenged in such a way that we cannot simply dismiss the challenge as crazy.
To understand what underlies this normative propriety, we should first
understand in what way descriptive facts underlie normative proprieties. One
thing we clearly do not need is for the description of the underlying facts to
entail that the normative propriety holds. No description of a football game in
the language of physics will entail that movements of a certain kind should be
treated as instances of defensive pass interference. Only the description of the
disposition of objects in space plus a description of the rules of football –
themselves normative proprieties of a sort – will have that entailment (if even
that is enough).
This is precisely where Sellars, and those who have developed his views on
language and knowledge, can help us. We are comparing two epistemological
pictures. One picture of the function of epistemic principles is this:
justification and knowledge are completely reducible to natural facts about
knowers and their contexts, and so true epistemic principles are true
descriptions of the conditions under which knowers are well-placed with
respect to their subject matter. This has been the standard view in western
epistemology in the contemporary era.
Here is another picture: justification and knowledge are, at least partly,
irreducibly normative, and so epistemic principles are endorsements of
strategies for acquiring information. The move here is from thinking of
knowledge as having a kind of quasi-logical status (signaled by terms like
‘epistemic possibility’) to thinking of it as having a social-normative status. To
say that someone knows something, or is justified in believing something, is in
part to endorse her cognitive state, or how she got into her cognitive state; it is
to place yourself alongside her in the “space of reasons.” It is to attribute to her
both commitment to a claim (by recognizing that she believes it) and
entitlement to that view (by allowing that she holds that belief rightly), and
also to undertake that same commitment yourself (by sharing that belief).8 One
can respond to another’s commitment-takings by challenging them; if the
challenge is a reasonable one, the original claimant now has a responsibility to
respond to the challenge by giving reasons for her commitment. Certainty, on
this view, corresponds to a “safe” position in the game; a commitment-taking
that is immune to challenge. Such a position would have to be one to which no
reasonable challenge can be made.

8 See Brandom (2000, p. 168). There may be more to knowledge attribution than Brandom says

here. So far one may be attributing only justified true belief.

228 Mark Owen Webb

Take one candidate for such a safe commitment, dear to the hearts of
foundationalists of all stripes: the observation sentence. Surely, the reasoning
goes, a person cannot be wrong about that sort of thing. So such a
commitment-taking has to be immune to reasonable criticism. Sellars gives us
excellent reason to think otherwise. Consider the following passage, from a
brief discussion in the Review of Metaphysics:
Even if there were statements whose manner of causation entailed their truth, so
that they would be in one sense “incorrigible,” this would by no means involve
that they were not “subject to revision,” nor do the philosophers who postulate
them believe they are not subject to revision. (Sellars 1954, p. 623)
The key here is to remember the difference between an observation statement
and an observed fact. Whereas it makes some sense to say that the mental state
caused by some fact in the world (via sensory apparatus) is unassailable, the
thing of which one is certain is not the fact, but a claim about the fact. Thus
Sellars’s well-known complaint about foundationalism’s use of the ‘given’ as
something that simultaneously provides a reason (and reasons are claims) and
is immune to doubt (by virtue of being a natural fact).9
On this view, epistemic principles are all about what commitments people
ought to make, which of others’ commitments they ought to challenge, and
what sorts of other commitments help buttress or undermine those
commitments. These principles, like all rules, take the form of hypothetical
imperatives.10 The proper way to evaluate an epistemic principle, then, is to
ask whether people should follow it. The principle of fallibilism is no different.
This second picture of the function of epistemic principles, which we owe to
Wittgenstein and Sellars, and now Brandom,11 sees epistemic talk as a
sub-practice of linguistic practice, concerned with the asking for, giving, and
evaluating of reasons. The practice is a rule-governed language game. On this
account of epistemic principles, for example, Cartesian foundationalist
principles can be seen as a set of recommendations for which assertions to
allow to be unchallenged and which must be defended if challenged.12 The
claim that I exist is always immune to challenge, as are claims about the
current contents of my consciousness. Further claims must be defended (if

9 I take it that uncovering this tension is the aim of “Empiricism and the Philosophy of Mind.”
10 See Sellars (1991, p. 323). These imperatives are grounded, of course, in natural facts, but that
does not undermine their imperative nature, especially since being committed to obeying an
imperative will amount to taking on commitment to those facts, which in turn is a matter of
obeying certain imperatives.
11 See Sellars (1963); Wittgenstein (1969); Brandom (1994).
12 See Kaplan (1991) for an exposition of what happens if we take foundationalism that way.
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 229

challenged) by being deductively derived from these immune claims. Further,

any possible doubt, no matter how “hyperbolic,” counts as a legitimate
challenge.
On the naturalistic picture of epistemic principles, fallibilism is probably
untenable. But it fares much better on the normative understanding of
epistemic principles; on that sort of view, fallibilism is best understood as a
proposal for a normative propriety in epistemic language-games. The idea is
that it is best if we treat each other in such a way that no statement is de jure
immune from criticism; any claim, no matter whether logically necessary or
obtained by privileged access, must be defended if it is met with a challenge
that is itself within the rules of the game.13 We may be subjectively certain that
Graham Priest is wrong when he issues a challenge to the law of
non-contradiction, and the law of non-contradiction can be necessarily true
(after all, it is), but we may not simply ignore what he has to say. The fact that
he challenges the law of non-contradiction from premises that he himself is
entitled to hold, by methods that we all endorse as reasonable, makes it the
case that his challenge is a fair one we must be prepared to answer. It may be
that there are claims we cannot be wrong about, that is, there may be utterance
tokens which cannot in fact be successfully challenged, but even about those it
is best if we follow the fallibilist rule. After all, we cannot be sure that the
sentence token we are about to utter is one of those de facto safe ones. The
reason this rule binds us is not strictly entailed by any facts about us. The
various attempts to make such underlying facts explicit have all collapsed into
skepticism, or entailed that there are no necessary truths, or had some other
bad consequence. Instead, we should realize that the fact that we are frequently
wrong, and that we are sometimes wrong about the things about which we feel
the most certainty, and that we are in general bad judges of the success of our
own epistemic endeavors, is enough to show that we ought to be very cautious
indeed. Just as a judge may not rule on a case in which he has an interest,
because people in that situation are prone to bias, so we ought to behave as if
we could be wrong in any particular case, because we are wrong in so many
cases. We think it the best procedure in general if the particular judge is not
permitted to preside in that case, even if he is in fact immune to bias. In the
same way, we should think it right to treat every one of our beliefs as
potentially in error, even if in fact, in some cases, we cannot be wrong. The
question fallibilism answers is the question of what rules are best for

13 This is the view espoused by Robert Brandom in Making it Explicit, chapters 3 and 4. Which
challenges are within the rules of the game and which are ruled out is a context-sensitive matter,
as contextualist epistemologists have long known. See, e.g. Annis (1978).
230 Mark Owen Webb

knowledge-seeking as an endeavor. So fallibilism is best understood, not as a

thesis about our abilities, but as a rule to guide our epistemic endeavors.14

Mark Owen Webb

Department of Philosophy
Texas Tech University
Lubbock, TX 79409
United States of America
e-mail: [email protected]

REFERENCES

Annis, D. (1978). A Contextualist Theory of Justification. American Philosophical Quarterly 15,

312-319
Audi, R. (1998). Epistemology: A Contemporary Introduction to the Theory of Knowledge. New
York: Routledge.
Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press.
Brandom, R. (2000). Articulating Reasons: An Introduction to Inferentialism. Cambridge, MA:
Harvard University Press.
Carrier, L. S. (1974). Skepticism Made Certain. Journal of Philosophy 71, 140-150.
Carrier, L. S (1993). How to Define a Nonskeptical Fallibilism. Philosophia 22, 361-372.
Churchland, P. (1998). Matter and Consciousness. Cambridge, MA: The MIT Press.
Collins, S. (1990). Selfless Persons: Imagery and Thought in Theravada Buddhism. New York:
Cambridge University Press.
Haack, S. (1979). Fallibilism and Necessity. Synthese 41, 37-63.
Harvey, P. (1995). The Selfless Mind: Personality, Consciousness and Nirvana in Early
Buddhism. New York: Routledge.
Heatherington, S. (1999). Knowing Fallibly. Journal of Philosophy 96, 565-587
Kaplan, M. (1991). Epistemology on Holiday. Journal of Philosophy 88, 132-154
Lehrer, K. and K. Kim (1990). The Fallibility Paradox. Philosophy and Phenomenological
Research 50 supp., 99-107.
Meerbote, R. (1977). Fallibilism and the Possibility of Being Mistaken. Philosophical Studies 32,
143-154.
Mott, P.L. (1980). Haack on Fallibilism. Analysis 40 (4), 177-183.
Priest, G. (1979). The Logic of Paradox. Journal of Philosophical Logic 8, 219-241.
Russell, B. (1959). The Problems of Philosophy. Oxford: Oxford University Press.

14 Various versions of this paper have benefited from comments by William Alston, Robert Audi,

Patrick Beach, Jonathan Kvanvig, Mark Lance, Alastair Norcross, and Michael P. Wolf. This
paper is a revised and expanded version of my presidential address at the North Texas
Philosophical Association meeting (1999). One section of it is adapted from my (1996).
 Meeting Others in the Space of Reasons: Fallibilism for Sellarsians 231

Nikaya, A. (1995). Girimananda Sutta. Translated from the Pali by Vhikka Nananoli. Onalaska,
WA: Pariyatti Pub.
Schlesinger, G. (1984). Possibilities and Fallibilism. Erkenntnis 21, 263-278.
Schlesinger, G. (1985). The Range of Epistemic Logic. Aberdeen: Aberdeen University Press.
Sellars, W. (1954). Comments on Mr. Hempel’s Thesis. Review of Metaphysics 5, 623-625.
Sellars, W. (1963). Some Reflections of Language Games. In: W. Sellars Science, Perception and
Reality, pp. 321-358 Atascadero, CA: Ridgeview Press.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Stich, S. (1983). From Folk Psychology to Cognitive Science: The Case against Belief.
Cambridge, MA: The MIT Press.
Webb, M. (1999). Presidential Address. Colloquia Manilana 7, 86-97.
Webb, M. (1996). Fallibilism Is Not a Thesis. Southwest Philosophy Review 12, 45-51.
Wittgenstein, L. (1969). On Certainty. Edited by G.E.M. Anscombe and G.H. von Wright, New
York: Harper and Row.
This page intentionally left blank
Michael P. Wolf

SELLARS ON THE REVISION OF THEORETICAL COMMITMENTS

ABSTRACT. This paper addresses some of Sellars’s views on conceptual change and
revision, spread across several books and articles. It begins with Sellars’s distinction
between rules of criticism and rules of action. I argue that Sellars’s distinction here
actually sheds light on the epistemology of theoretical revision. Many (if not all)
revisions of theoretical commitments can be motivated by the force of rules of action
that govern the maintenance of our theories. I offer a partial account of this, positing
two rules of action. A rule of maximizing explanatory strength compels us to refine and
revise theories to improve their performance, while a rule of conservatism puts
boundaries on the acceptable pursuit of those goals. The dynamic between these two
guides revision, narrowing the range of acceptable options. Sellars’s conception of the
roles of analogy and affinity are considered, as well as an elaboration of a Sellarsian
point about meaning change.

1. Revising Our Commitments

A funny thing happened somewhere in the middle of the last century; everyone
changed their minds about changing their minds. Many of the major figures in
analytical philosophy turned away from the quest for indubitability that had
been the focus of epistemology. “Everyone” overstates matters a bit of course,
since there were certainly those who stuck to various guns on particular issues,
but the increasingly pervasive sentiment found its best-known manifesto in the
Quine’s “Two Dogmas”:
[I]t becomes folly to seek a boundary between synthetic statements, which hold
contingently on experience, and analytic statements, which hold come what
may. Any statement can be held true, come what may, if we make drastic
enough adjustments elsewhere in the system . . . Conversely, by the same token,
no statement is immune to revision. (Quine 1953, p. 43)
Expanding the set of claims that we stuck to, come what may, caused little
consternation in the philosophical world. Stubbornness is at least as old as
philosophy, so this hardly seemed like much of an innovation. It was the
prospect of challenging that which was to be held to be above challenge that

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 233-255.
Amsterdam/NewYork: Rodopi, 2006.
234 Michael P. Wolf

caused such a stir; Quine himself suggested that the Law of Excluded Middle
might be a casualty of the revolution in quantum physics (1953, p. 43). Surely,
the Law of Non-Contradiction would not (and would) be far behind.
Quine has certainly received the most press for his expression of this idea –
let us call it revisabilism for now1 – but he was far from alone in expressing it.
Both Kuhn and Popper made a vehement case for a pervasive openness to
revision in the philosophy of science, albeit with very different agendas. What
Quine said of individual statements, Kuhn might separate into questions of
revision within a theory (normal science) and rejections of whole theories
(radical science). Kuhn would stress the systematically rational character of
revisions in normal science and contrast it with the arational or at least
incompletely rational character of complete changes of theories or
“revolutions,” but the revisabilist strain of his work remains. Many have
interpreted Wittgenstein’s holism about language and knowledge in On
Certainty as a kind of revisabilism as well, and more philosophers now take
such an approach than we can reasonably recount here.
One name that does not regularly appear in discussions of this matter is
Wilfrid Sellars. This is surprising both because of the broad acceptance of his
critique of foundationalism and the centrality of such a doctrine in his account
of empirical knowledge:
Above all, the [foundationalist] picture is misleading because of its static
character. One seems forced to choose between the picture of an elephant which
rests on a tortoise (What supports the tortoise?) and the picture of a great
Hegelian serpent of knowledge with its tail in its mouth (Where does it begin?).
Neither will do. For empirical knowledge, like its sophisticated extension,
science, is rational, not because it has a foundation, but because it is a
self-correcting enterprise which can put any claim in jeopardy, though not all at
once. (Sellars 1997, §38/pp. 78-79)
Aside from the usual reason for Sellars’s absence from discussion – his dense
and obtuse writing style – he may also be excluded in light of the Kantian
strain in his work and subsequent endorsements of a reformed category of

1 The position I am describing here obviously relates to fallibilism, as espoused by a number of

philosophers, including Mark Webb in this volume. I am introducing this distinction not for the
sheer love of making distinctions, but because the view I am describing stretches to involve
conceptual content, while fallibilism has traditionally been formulated as a theory of justification.
One could be a fallibilist while being a Platonist who believes in eternal, unchanging meanings
waiting to be discovered by our analysis. Obviously, I think that position is untenable even though
I accept fallibilism, because I think the epistemic and semantic issues here simply cannot be
segregated from one another.
 Sellars on the Revision of Theoretical Commitments 235

analytic truths (Sellars 1963a; 1963b; 1974c, pp. 250ff.). If so, his exile is
undue. His twofold distinction demarcates (a) analytic1 claims whose revision
would be a matter of rejecting a theoretical framework altogether from
synthetic1 claims that can be revised within that system, from (b) analytic2
claims whose purport to analyticity is a matter of the deployment of logical
vocabulary. Quine offers a comparable deflationary account of analyticity in
Word and Object and elsewhere (1964, pp. 270ff.; 1974, pp. 78-80) and Sellars
himself stressed that his thoughts on the matter were in keeping with Quine’s:
That the cluster of inference by virtue of which an expression refers to a red
triangle as such (or a bachelor as such) are open-textured and variable, and need
only have a family resemblance from context to context, is the truth contained
in Quine’s attack on the analytic-synthetic distinction. My central concern over
the years in stressing material rule of inference has been not to deny the point
Quine is making – indeed, to a behavioristically oriented anti-Platonist, the
denial would be foolish – but rather to insist that inference patterns other than
those formulated by logical truths are essential to meaning and reference. That
these extra-logical inference patterns do not neatly divide into “explicit” and
“implicit” definitions, and that they trail off into contingent generalizations, are
theses that have emotional charge only for those who are still fighting the battle
of the Museum. (Sellars 1977, p. 309)
My initial contention here is that Sellars should be considered as a figure in the
development of this strain of contemporary epistemology. What we must
consider is what distinctive contribution he can be said to make to it.
So we find ourselves with no shortage of contemporary figures in
epistemology who stress the importance of the de jure revisability of all (or
almost all) of our beliefs and Sellars clearly endorses such a line of thought.
Some have seen this as reason to endorse a revived form of skepticism in the
Cartesian tradition, e.g., Unger (1975), while many others – including the
authors mentioned here – have seen this as part and parcel of a broad rejection
of that tradition. In the modes of analysis that gave rise to analytical
philosophy, many would see this as cause for alarm if not despair. Without
foundational beliefs to anchor us or subsequent certainties to assure us, our
beliefs become a patchwork of probabilities spinning in a Pyrrhonic void. Such
concerns are often overcome rather than refuted; in part, we simply learn to
live with greater contingency. More importantly, we come to see justification
and the rationality of our beliefs as diachronic matters, whereby their
epistemic value lies not in their immunity from present and future challenges
in virtue of some set of intrinsic properties, but rather in their ongoing success
and centrality in theories and practices that survive and evolve over time. As
part of his critique of foundationalism, I see the development and
236 Michael P. Wolf

incorporation of this idea as one of Sellars’s most significant contributions to

contemporary epistemology.
Much of this is nascent in the pragmatist tradition. Dewey, inspired by both
Darwin and Hegel, rejected Descartes’s foundationalism in favor of ongoing
inquiry and pragmatic engagement, and these themes resonate in the
denouement of Quine’s “Two Dogmas.” However, these answers also lean
towards the purely descriptive in many cases. Quine’s efforts to naturalize
epistemology and erase the line between science and philosophy effectively
psychologize the epistemic process of revision and reduce it to one further
array of pattern-governed behavior. He emphasized that where such revisions
were made, they were driven by “pragmatic” considerations. If by “pragmatic”
concerns here we mean nothing more than instrumental concerns about the
simplicity, economy and efficacy of a theory in achieving its assigned tasks,
one could hardly deny their importance. However, one might also feel these
factors underdetermine an account of the processes by which we deliberate and
adopt revisions and legitimately expect a more developed account.
With respect to epistemic matters, Sellars parts ways with many of his
fellow pragmatists, asserting instead that,
[T]he idea that epistemic facts can be analyzed without remainder – even “in
principle” – into non-epistemic facts, whether phenomenological or behavioral,
public or private, with no matter how lavish a sprinkling of subjunctives and
hypotheticals is, I believe, a radical mistake. (Sellars 1997, §5/p. 19)
The crucial implication here would be that our account of revision must also
be a fundamentally normative one. When we revise a theoretical commitment
– whether it takes the form of a claim, a method or anything else that might
come up – legitimate questions arise regarding the propriety of such changes.
Such questions may relate to the demand for such a change (“Is it really
needed in light of the evidence?”) as well as the specific details of the change
made (“Is this really the best choice among the many options available to
us?”). It would certainly seem suspect if working scientists simply chose to
revise a theoretical claim, even though the theory sufficed for our explanatory
purposes and accorded well with observed results. Such changes would prompt
dramatic revisions elsewhere in the theory to maintain bare consistency, but
there is nothing logically illicit about this.
While robustly normative accounts of epistemology – or anything for that
matter – are not the most welcome moves in some quarters, I do think that
Sellars’s insistence on their place in epistemology is especially pertinent here.
Given the enormous number of possible replacements for a commitment to be
revised in any given case, as well as the novelty and complexity of such cases,
some shared sense of what makes for appropriate extensions and revisions of
 Sellars on the Revision of Theoretical Commitments 237

theoretical resources is the only hope for anything approximating a rational

epistemic process.2 In fact, I would argue that the seeds of a more developed
account of revision are present in Sellars and I will set them out here. I cannot
necessarily claim that this is Sellars’s view, since he certainly did not make the
argument I am making here in the form that I am making it, but I do think that
it is a view that grows out of his thinking and one who is drawn to certain
aspects of his work would be drawn to this as well. For the time being, I will
also focus on the revision of empirical claims, particularly those related to the
natural sciences. Other bodies of propositional knowledge – moral claims,
explicit rules, claims about social kinds, etc. – may receive an account
generally similar to the one I offer here, though we may need to tailor the more
motivational aspects of what I will say here to each of these bodies of
knowledge. Establishing this view will require some detours through a theory
of meaning, as one might expect, but the first important insight to take from
Sellars relates to his views on the nature of rule-following in a self-critical
fashion.

2. Rules of Criticism

In some sense, all rules compel us.3 However, as any good therapist will tell
you, not all compulsions are the same. One significant contribution that Sellars
has made to the pragmatist and analytical traditions has been to expand the
idea of rule-following by distinguishing two fundamental categories of rules
based on what they prescribe and require of us. In “Language as Thought and
Communication” (1969), Sellars outlines a distinction between rules of action
and rules of criticism. Rules of action, as the name would suggest, prescribe
that actions be taken in particular sorts of situations. Thus, they take the form,
“If one is in C, one ought to do A,” where ‘C’ refers to a state of affairs and ‘A’
refers to some type of action. Rules of criticism took as their focus obligations
to see that some state of affairs obtains, hence the form, “It ought to be the
case that B,” where ‘B’ refers to some state of affairs. Some comparison with

2 As I mentioned earlier, Kuhnians might have their reservations about calling periods of radical
science rational at all, though that strikes me as a stronger claim than Kuhn’s original argument
would reasonably suggest. In any case, acceptance of this claim restricted to the realm of normal
science should be far less contentious and I shall have more to say about periods of radical science
later in this article.
3 A note of thanks is due to Pedro Amaral, who reminded me in conversation to think about the

rules of criticism when approaching this topic.

238 Michael P. Wolf

the categorical and hypothetical imperatives of Kantian ethics is in order, as

Sellars (1969) acknowledges. It should also be noted that a rule of criticism
implies various rules of action and thus agents may establish what actions are
required of them by the rule of criticism.
In Sellars’s (1969, p. 508) example, the rule of criticism
(1) Clock chimes ought to strike on the quarter hour.
implies, roughly, the rule of action
(2) (Other things being equal and where possible) one ought to bring it about
that clock chimes strike on the quarter hour.
If the example seems dated, we can always replace it with a rule about our
VCRs not blinking “12:00” at us, although that example is quickly becoming
dated, too. The second example (2) does not take the form of a conditional
statement, though we could easily rewrite it as such to emphasize that if clocks
are not chiming on the hour, then one ought to correct this. More pertinently,
Sellars take this as a case of a rule of action because its subjects are agents
who are assigned responsibilities that invoke “recognitional capacities,”
e.g. recognizing clocks and chimes, telling time, etc. The conditional nature of
these rules, even if concealed by surface grammar, lies in their assigning us
obligations if our recognitional capacities tell us that we are in some situation.
Rules of criticism are thus rules for states of affairs and rules of action
presuppose the capacities that belong to rational agents for their satisfaction. A
rule of criticism might be satisfied without our involvement or at least without
much of our involvement – imagine a flawless clock created, set in motion and
never adjusted again – but there would simply be no satisfaction of rules of
action unless there were agents like us to satisfy them. Much of the
philosophical consideration of rules in the last century has focused on what we
are calling rules of action here. They are particularly appetizing to those with
strongly naturalistic and reductionistic views, since they seem amenable to
accounts of conditioning and processing, either by symbol manipulation or
brute conditioning. However, it would be a grave error to leave rules of
criticism out of our account, as they have an indispensable role to play in our
understanding of our own rationality.
As their name would suggest, rules of criticism serve a pivotal role in the
diachronic self-critical evaluation and evolution of our practices. As Sellars
says:
Many of the rules of language which are of special interest to the epistemologist
are ought-to-be’s rather than ought-to-do’s. For only by taking this fact into
account is it possible to carry out a program according to which (a) linguistic
 Sellars on the Revision of Theoretical Commitments 239

activity is, in a primary sense, conceptual activity; (b) linguistic activity is

through and through rule-governed. (1969, p. 510)
Much of Sellars’s concern with rules of criticism there and elsewhere
surrounds the underlying basis of conceptual activity; we begin by obeying
rules of criticism with respect to responses and dispositions to respond and
subsequently “graduate” to become agents who grasp not only the subject
matter of those rules, but also grasp ourselves as the subject matter of further
rules, i.e., rules of action. Becoming a rational agent involves achieving a level
of self-critical capacity whereby we can ask whether our practices and actions
satisfy the “ought-to-be’s” that guide us and adjust our rules of action
accordingly.
A community of speakers who had no rules of action would not yet have
taken up the sort of reflective engagement on their pragmatic possibilities that
we do; they may act in regular ways and features of their environment may
accord with some set of explicable rules, but they do not yet take themselves to
be the subject-matter of prescriptions on their own behavior. The rules of
criticism would serve as a basis for the development of that reflective
engagement on pragmatic possibilities. Sellars suggests that becoming a
teacher of the language involves grasping rules of criticism in such a way that
one gets others to be disposed in certain ways, as when we first teach children
to use a few parts of a language differentially, even though they do not yet
grasp the full conceptual content of what they do. Note here that we are not yet
getting children to take actions in a robust and reflective sense, we are just
getting them to be disposed to certain sorts of things, hence dealing only with
“ought-to-be’s.” Accordance with a set of rules of criticism, notably those
involving differential responses to circumstances, serve as the initial stage in
the development of our conceptual activity. It would be wrong to construe
them as an epistemic foundation though, since it is only in developing rules for
action and an acumen with the inferential role of the words and beliefs
involved that any of this actually becomes genuine conceptual activity, as
Sellars’s critique of foundationalism suggested.
We should not forget how broad the category of “ought-to-be’s” might be.
Sellars’s focus in “Language, Thought and Communication” (1969) was with
dispositions that might serve the role described above, but as the clock
example suggests, the category is in no way limited to agents and their
psychological and behavioral dispositions, nor even to potential agents and
their various states. Once we acquire the sorts of conceptual capacities to
discern, respond, plan and invent, we can recognize and involve ourselves with
a great many more states-of-affairs with a variety of agendas. What I would
suggest here is that there comes a point at which the subject of a rule of
240 Michael P. Wolf

criticism might be a practice or set of practices itself. This much should not be
terribly controversial, since it is essentially an unspoken assumption in a great
deal of ethical and meta-ethical reasoning. We are familiar with claims like, “It
ought to be the case that social practices are just,” which is not yet and not
necessarily a claim about the justness of the particular actions of particular
agents. Actions may be called just in part because of their general
permissibility within a set of practices, and those practices may be evaluated,
wholly or in part, in terms of the consequences that they bring, which are not
in themselves actions taken by agents. The practices themselves are the
subjects of such judgments. The crucial move in tying all of this back to our
original theme of the revision of our commitments lies in seeing that the status
of our theories and practices may be the subject matter of rules of criticism in
various ways and critical evaluation of their statuses may entitle us to make the
case for particular revisions.
Some rules of criticism in this neighborhood should be familiar enough. It
is hardly controversial to assert that, all things being equal, our theories should
be as parsimonious as possible. But, we should also guard against both the
impulse to impose tidy systematizations of the rules of criticism at work here,
lest we smooth over salient details from the actual practice of theoretical
inquiry. A penchant for simple, overarching principles has been the downfall
of many philosophers of science, and it is a tendency that we should resist. A
more discerning epistemology must acknowledge that numerous distinct
normative components shape theories, and that these rarely lend themselves to
neat hierarchies. In presenting rules of criticism that bear on theoretical
revision is thus to offer rules that contribute to the determination of a revision,
but no one of which suffices to determine every case by itself. Each of these
rules of criticism should be seen as defeasible under the right conditions, and
the entire account will have a holistic character. Such judgments are by their
very nature contextual matters and contentious ones at that, and my goal here
is only to characterize in broad terms the sorts of reasons that contribute
significantly to such a judgment. A holistic and decentralized approach to
these epistemic questions is certainly messier and less grand than many past
efforts in the philosophy of science and epistemology, but it does not seem any
less faithful to the actual practices and strategies of those who are building
such theories. The task for the epistemologist here is to explicate what rules of
criticism seem to be operative in many cases and to explain the dynamics of
the process that drive and constrain our efforts and what sorts of conditions
might allow one to trump the other.
In keeping with such an approach, what I would like to do here is to offer
two rules of criticism I believe are particularly important and say how they
come to be obligatory for us and how we might find ourselves entitled to make
 Sellars on the Revision of Theoretical Commitments 241

exceptions to them. This is by no means intended as an exhaustive list of all

such rules, even if we throw in parsimony as mentioned above. It is one piece
of a larger epistemological picture. For now and for the sake of brevity, let me
call these rules by the sort of condition they would ideally produce. I will
speak of explanatory strength and conservatism, for there is an account I
would like to offer about their importance and relation to one another that I see
emerging from Sellars’s work.
The notion of explanatory strength has a long history in epistemology and
the philosophy of science, though not necessarily one of great unanimity.
There is great disagreement about what constitutes explanatory strength and
explanation in general, though there is broad agreement that theories do
explain. For my purposes here, the account of explanation that one endorses is
not terribly important, so long as we accept that it sometimes happens and that
theories are part of doing it. As a rough formulation, let us say that that the rule
of criticism here might take the form:
(3) It ought to be the case that our theory is adequate to explain its phenomena.
By ‘phenomena’ here, I only mean those events, objects, etc. for which a
theory bears explanatory responsibility. So physics is responsible for all the
physical stuff and what it does, biology for the organisms, etc. There are of
course regions of theory that overlap such categories, e.g., biochemistry, and
presumably physics overlaps them all in some sense. Anti-reductionists will
analyze these as cases of one theory borrowing resources from another but not
being replaced by it, while the ardent reductionists among us will no doubt
insist that I am too casually assuming the disunity of the sciences to begin with
and that all will eventually reduce to physics. But this poses no special
problem for the rule as I have stated it. If the ardent reductionists were right, it
would just make physics responsible for everything in the long run. My only
point is that in pursuing a particular theory – whether it is all-encompassing or
one of many – we elaborate what a theory can actually explain, and what it
should be able to explain which will itself adjust as the theory develops.
Theories invariably fall short of completion in this sense, in part because new
developments suggest new possibilities and new problems to be solved, and in
part because some developments undermine our sense that the theory is
sufficient after all.
It is straightforward enough to suggest that theories should explain as large
a number of things as possible within their field given their theorems, though I
doubt we can legitimately quantify the scope of our theories in very concrete
terms. Also, the result of innovations in a theory is sometimes to eliminate
things we thought we needed to explain but which cease to be plausible
concerns after all, e.g., the aftermath of the Michelson-Morley experiment is
242 Michael P. Wolf

simply the abandonment of the ether and the more specific problems of
measuring its properties. Moreover, a demand for greater explanatory strength
cannot guide an account of revision on its own for the simple reason that well-
motivated revisions often lead us to decrease what we take our theories to
explain. Not all progress is progress towards an appropriate goal.4 Finding
dead-ends and trimming back what we claim in light of these is a crucial part
of theory-building, as well. Again, consideration of the Michelson-Morley
experiment may be instructive. In this case, we find ourselves compelled to
revise certain claims by giving up entitlement to them and what they imply,
leaving the most broadly accepted theory greatly impoverished and clearing
the way for a successor. The danger in not heeding this point is a certain sort
of insulation from challenges to a theory – implying that if a suggested
revision would decrease the explanatory strength of a theory it should be
rejected out of hand in favor of revisions that would preserve the status quo.
Tertiary epicycles lie down this path.
So like any rule of criticism, this establishes an ongoing obligation. We do
not necessarily meet this obligation by taking any particular action, as I might
meet an obligation to keep a promise or utter a phrase at the right moment. We
should also clarify that this remains a rule of criticism despite the inclusion of
a verb like ‘explain’. Although this does pertain to resources for practical
engagement, this is a rule about the status of a set of practices (what something
must be), not about our obligations to give explanations (what we must do).
This rule may entail various rules of action with respect to giving explanations
and making revisions, but it is not a call to take an action in its own right.
Given its focus on practices as resources for theoretical explanation, it is more
akin to a rule like “You should have the right tools for the job,” or “You
should have the pantry well-stocked.” Either one may send you off to the store
in the long run, but they are not rules about shopping.
The second condition, conservatism, is not intended to overlap with the
conservatism of philosophical logic, much less that of contemporary political
discourse. Rather, the issue here is the degree to which later stages of a
theory’s development resemble earlier ones. Conservatism as I have called it
here is simply the imperative to change our theories as little as possible at each
stage of their development. All things being equal, we should opt for revisions
that preserve as much of our existing theory in place as possible, particularly

4 I cannot resist a story here. This point came to me most clearly driving from Chicago to
Baltimore with a friend when I was 18. After missing a crucial turn, my friend refused to turn
back, diverting us by hundreds of miles, saying, “All progress is progress towards a goal.”
Bewildered, I eventually managed to yell, “Yeah, but what if it’s not our goal?”.
 Sellars on the Revision of Theoretical Commitments 243

explanatory strategies. Again, some qualifications and elaborations should be

made here. First, the degree to which a theory is “preserved” here is not
necessarily a simple quantitative matter. Theories offer a wealth of resources,
including models and problem solving strategies whose value and importance
to us may outweigh other, more straightforward forms of affinity. We might
happily accept a broad expansion of an existing theory to incorporate new
commitments that allow us to preserve or expand explanatory strategies that
we already endorse. For instance, analytical chemistry adds a great deal of
material to our accepted body of theory, which apparently forces us to accept
tremendous changes, in strictly numerical terms, of commitments. We happily
accept these extensions because they allow us to discount certain sorts of
deviation in our evidence and thus reinforce the acceptability of parts of
theories we already endorse. More extensive sub-theories of gravimetric
analysis or volumetric titrimetry add a lot of details, but pay off in spades in
the long run. So what I am suggesting here is not an enumeration of all of our
theoretical commitments (assuming such a thing were possible) and a
subsequent one-to-one mapping of its items onto counterparts in a potential
replacement theory. Stability and affinity between stages in the development
of a theory are rarely, if ever, so straightforward. Rather, the distinction is
about what theories do and how they do it, and thus is what we are inclined
and obliged to preserve.
The second qualification we must make here is that conservatism is a much
more readily defeasible rule of criticism than explanatory strength. Revision is
so ubiquitous in scientific practice that we cannot reasonably read
conservatism as a demand that we never change our theoretical commitments.
But even if we read it as implying that theories do change and we must opt for
the candidate that most resembles earlier versions, then there will still be a
wide variety of cases and contexts in which we make exceptions to this rule.
Kuhnian revolutions, or at least the periods and events in the history of science
that Kuhn called revolutions, serve as familiar examples of periods in which
important commitments were dropped and new ones taken up, leading to very
different arrays of models, experimental methods and problem-solving
strategies. Even if we reject Kuhn’s bolder claims about “living in different
worlds” from one paradigm to the next, these are certainly historical cases in
which preserving features of a theory is no longer of great importance simply
244 Michael P. Wolf

because the theory no longer does what we expect it to in the face of new
evidence or disappointing developments in the theory.5
It is particularly important to note at this juncture that most of these cases
in which conservatism is defeated are precipitated by the pursuit of greater
explanatory strength for a theory. The function of a rule of conservatism is not
to prevent change or avoid expansions of a theory, but to stabilize it, pending
adequate grounds to point us towards a different set of commitments. I would
argue that this stricture does not arise from an aversion to change, but from a
shared need among builders of theories to make the process of revision public.
What we are looking at here is not merely a change of habit, but also genuine
conceptual change, and this obliges us to judge our own entitlements to our
claims as well as the entitlements of our ancestors and what we anticipate
granting to our progeny. To assert as true is to grant a universal license to a
theoretical claim, and so how we relate that claim to past and future speakers is
of significance in making it. This obligates us to establish conceptual and
semantic content with the greatest degree of affinity between stages possible in
order to avail ourselves of the various entitlements that already go along with
our accepted theory. Explanatory strength thus drives us to revise our claims to
make more powerful theories, while conservatism generally drives us to
choose the least disruptive means of doing so at a given point so that we may
retain as many of the licenses we already have and employ. Philosophers have
mentioned both of these drives in the past, although not always in normative
terms, but it is the interplay between them that is essential in forming a picture
of the rationality of revision. Fleshing out that interplay in a Sellarsian vein
will require some attention to his account of meaning and conceptual change
though, to which we now turn.

3. Meaning and Changing Our Minds

The semantic repercussions of revisabilism have been far more difficult and
controversial than the epistemic ones, perhaps because three centuries of

5 All I mean to suggest here by the last clause is that a theory may be abandoned or radically

revised not because of startling new discoveries, but rather because it offers no plausible means of
improvement, or it fits poorly with theories of related phenomena without hope of reconciliation,
etc. The most commonly cited example of this is Ptolemaic astronomy, which could continue to
account for new evidence if one was just willing to stomach more odd local idiosyncrasies. As
Kuhn (1996a, pp. 154ff.) pointed out in the case of Copernicus, scientists are sometimes even
willing to jump to a theory that seems to have more problems and loose ends in such cases.
 Sellars on the Revision of Theoretical Commitments 245

post-Cartesian skepticism wore down our collective expectations and

ambitions with respect to knowledge. Meaning change remains a concern
because, in the view of some, it threatens to undermine the very intelligibility
and rationality of discourse. If my great-great-grandfather used ‘length’ to
refer to a particular property of physical objects that did not vary with inertial
frame and I use the same expression to refer to a property that does vary in that
way, then we apparently do not share the expressions in our utterances and
inscriptions, despite the superficial similarities. In the most pessimistic views,
this puts my great-great-grandfather (or even my contemporaries) and I at an
impasse, saying entirely different things to one another and thus immune from
one another’s challenges and isolated from agreement or support. Justification
and objectivity become hopeless, theories break down and soon we are all
locked into idiolects, or at least strongly localized and untranslatable
languages. Fodor and LePore (1992) have made arguments against holism and
functional-role semantics based on the assumption that dense interrelation
among pieces of a language would make such incommensurability inevitable,
and Kuhn (1970) suggested a version of this (albeit one with much greater
uniformity enforced by social institutions) that he apparently embraced and
others have incorporated into a picture of the social sciences.6 Although I do
not share such pessimism about what changes of meaning would bring, critics
are quite right to assert that a complete immunity or isolation from one another
would be catastrophic and we would have sufficient grounds to reject any
theory of meaning that had this as a consequence. The key to resolving this
tension is to offer an account that embraces both holism and change of
meaning by explaining how speakers with distinct but similar sets of
commitments surrounding their language could speak to one another in a
productive fashion.

6 I stress that Kuhn has “apparently” endorsed this view because, like many great philosophers, we

get different pictures emerging from his work at different stages of his career. Before 1970, Kuhn
happily spoke of scientists in different paradigms “living in different worlds” and having entirely
incommensurable beliefs. But Kuhn’s choice of the term ‘incommensurability’ from mathematics
and logic reflects a recognition that the terms in different paradigms could not be mapped onto or
substituted for one another through a formal process, as prevailing conceptions of science and
semantics at the time would have demanded. Kuhn (1996b, pp. 220ff.) preserves all of the subtlety
and disparity in describing historical stages of the development of our “disciplinary matrix,” but
also speaks of hermeneutic processes whereby we may visit or recreate the disciplinary matrices
of previous paradigms in search of enlightening affinities. (I suppose past generations could do the
same with future paradigms, though scientists of past paradigms tend to be both dead and short on
time machines.)
246 Michael P. Wolf

Sellars (1997, §30-31/pp. 64-68; 1974b) does offer such a picture, whose
features I generally endorse. As should be familiar to those who know
Sellars’s view on semantics, he argues that meaning claims do not refer to
extralinguistic entities, but rather express a specialized form of the copula that
relates metalinguistic distributive terms with one another. Thus:
(4) ‘Und’ (in German) means ‘and’
should be understood to have the force of
(5) ‘Und’s (in German) are •and•s
where the quoted expression on the left picks out all appropriate past, present
and future inscriptions of that German word and the dot quotes indicate any
item in a language that functions as the embedded word (‘and’) does in ours.
“Roughly, to say what an expression means is to classify it functionally by
means of an illustrating sortal” (1974b, p. 180). If we accept such a functional
account of expressions – often articulated in terms of “inferential roles” – how
are we to account for changes in those functional roles? Presumably a
functional role would be specified in terms of related items and a direction of
inferential fit for the relata, e.g. “Mingus is a cat” implies “Mingus is a
mammal,” but not vice-versa. Assuming there is some finite set for any such
expression, changes of meaning would require some addition to or subtraction
from the elements of that set. We seem to be right back where we started, in
fear that any change of meaning leads to radical incommensurability. In
looking at the analogous problem with the diachronic identity of physical
objects and their properties, Plato was moved to add the forms and Aristotle
added substances to serve as persistent substrates of change. Some sort of
extra-physical bearer of meaning here would be mysterious and decidedly
un-Sellarsian here though, and so Sellars happily bites this bullet; we have as
many different concepts as we have possible inferential characterizations,
though obviously only some very small subset are deployed in the actual
history of our world. What matters here is how we as speakers interpret one
another’s usage and then navigate and negotiate towards a shared conception.
Sellars offers a brief but incisive suggestion on this point in Science and
Metaphysics:
To say that the semantic rules governing ‘f’s in our language could change over
a period of time, and yet that the ‘f’s could all be •f•s is what is meant by saying
that f-ness has changed over this period. Just as we have the concept of a
developing language or conceptual scheme, from which the concept of a
language as studied in current formal semantics is an abstraction, so we have
the concept of a developing linguistic or conceptual role from which the usual
concept of a “sense” or “intension” is also an abstraction. To be an •f• (stand for
 Sellars on the Revision of Theoretical Commitments 247

f-ness) does not require in this context that the expression to be classified plays
the identical determinate role currently played by ‘f’, but that its function in the
earlier stage of the language is sufficiently similar to the current function of ‘f’
to warrant classifying them together. Roughly, ‘f’s at t1 stand to L at t1 as ‘f’s at
t2 stand to L at t2. (Sellars 1967, §47/p. 131)

The way forward from here is to see that in talking about shared meanings
with other speakers, our concern is with the degree of affinity we can find with
other speakers (generally past speakers in our present case) and the epistemic
and practical authority that that affinity can grant to us. I am not sure that there
is an English word for quite what I am suggesting here, and I do not wish to
add a new technical term here. It is not univocality – speaking with one and
only one meaning – that is at stake here, but rather something like a rational
and practical empathy with other speakers. We share something like the
commitments and theories they undertake, and see ourselves in their shoes to a
certain degree, and in many cases we judge this affinity sufficient to take them
as one of us in order to share their reasons and warrants for the claims they
make. We cannot always do so; some speakers genuinely are “worlds apart”
from us in some ways. But it will be something that we can do very effectively
much of the time on issues of great importance to us, especially where the
historical, methodological and cultural gaps between speakers are less
pronounced. (It is a great deal easier to talk about physics with a pre-Feynman
or even pre-Einstein physicist than an Aristotelian physicist, for instance.)
Complete univocality cannot be our goal, for the concepts we embrace simply
do change. Instead, we must search for the affinities between different versions
of those concepts that allows us to draw from past inquiries to guide us in how
to carry on from here. To do so, we need an account of how the rules of
criticism I have described above serve to both propel us forward and keep us
grounded.
The dynamic I wish to suggest here is hopefully becoming clear to the
reader. I suggest that our theoretical commitments evolve in such a way as is
most useful for the goals we have in adopting them – thereby maximizing what
I have called explanatory strength – but that the ongoing mediation of
discrepancies between speakers and motivations for making one revision
rather than another in pursuit of explanatory strength must be constrained by
some sort of buffer – a regimen of conservatism – in order to both facilitate
and stabilize that process. As speakers revise their commitments and lobby
others to do so, each move is one that must be defended as a plausible
recasting of the semantic and epistemic order of things. A first, if defeasible,
condition that good epistemic agents seek in making such judgments is affinity
with prior well-defended sets of commitments, i.e., conservatism as I have
248 Michael P. Wolf

described it. At face value, this may appear to support a mild form of
dogmatism; if we support new options precisely because they do not differ
much from past choices, do we not engage in an undue restriction of our
inquiries in those cases in which past choices were simply misguided?
Defeasibility is sufficient to counter such an objection – those are just the sorts
of cases in which the conservatism condition should be defeated, so the
condition is not de jure dogmatic. But we must also note points in favor of a
reasonable degree of conservatism.
Preserving as many of the commitments surrounding a concept as possible
serves to preserve as much of the work done to entitle us to claims involving
such a concept or other concepts closely related to it as possible, which in turn
allows us to preserve the explanatory strategies in which they figure
prominently to the greatest degree possible. There are good practical reasons
for doing so; no one wants to go back to square one all the time, there is great
sunk cost in any research program and it would be wasteful to abandon it
prematurely, etc. But moreover, there are good epistemic reasons. Significant
success with incomplete versions of a theoretical project at a given point in its
development is the strongest sort of prima facie reason to adhere to such
commitments. If a theoretical project has enjoyed substantial success at its
assigned roles, then moves to dilute and undermine it without the promise of
greater efficacy would be illicit under any reasonable standard. This is not to
say that they may not be the right moves to make at a later stage of the
theory’s development (let us all raise a glass in Aristarchus’s honor), perhaps
even the right moves in a robust scientific realist’s sense, but we cannot
generally endorse them in the absence of further evidence. If what I suggest is
right, then the epistemic burden in any case where we revise our commitments
lies with those who would opt for more radical changes, and the only grounds
for such moves would be problems so deep with the present approach that it
should be abandoned in large part or altogether. The pursuit of explanatory
efficacy pushes revision and conservatism restrains it to preserve what works
well.
A quick note of clarification is in order here. Kuhnians will object that in
some cases there are radical breaks in scientific inquiry, that these are just the
sorts of cases in which the sort of gradualism and piecemeal revision described
here breaks down, and that these are some of the most crucial moments for the
development of a theoretical account. The move away from absolute time in
physics, the move away from geocentrism in astronomy, the move away from
immutability of species in zoology and botany – all of these are radical breaks
and all of them shape the research that follows in profound ways. These are
not exceptions or short-term fixes, these are defining moments. To deny that
there is something to this would be implausible, even if one does not accept
 Sellars on the Revision of Theoretical Commitments 249

the most radical readings of Kuhn’s early work. I think the best answer to this
is that conservatism as a rule of criticism is most clearly defeated in these sorts
of cases. Where a research program runs up against a wall in its efforts to
explain and solve problems and further refinements only serve to illustrate that
these failures are intractable, to be staunchly conservative may be an
indefensible move. This is especially true in those cases where a novel and
radical approach at least promises some kind of future success to many of
those who know its problems well. In such cases, revolutionary moves may be
the only “reasonable” ones, even if there is no systematic, non-contextual
account of what makes, say, the choice of heliocentrism over geocentrism a
rational one. But we also need not think that theories are thrown away as a
whole in scientific revolutions to accept that they can happen and mark turning
points for theories. Rational confrontations between different stages of a
theory are not strictly logical affairs, but as Kuhn later acknowledged (see
note 3), a variety of historical and hermeneutic methods may still be
informative and compelling, and that may include just the sorts of approaches I
have described here.
The key to averting the catastrophic outcomes of revision described at the
beginning of this section is to establish that different stages of a theory’s
development may bear some rational relation to one another despite their
substantial differences. We want to be able to say that something about the
claims made about water, mass, etc. made a couple hundred years ago are
affected by the claims that we make today. They are not necessarily so; it
might well be that what some of our ancestors meant by an etymological
cognate of some term we use is so radically different that we really should just
take them to be talking about something else and take that etymological
lineage to be a curious accident. But we would not want to endorse a position
in a theory of meaning or epistemology that obliged us to say this about every
term in our theories or our language as a whole, as some critics have feared.
Whatever else we might say about the revision, our approach here must permit
such rational confrontation between stages of theoretical development. What
we must spell out to make this of greater epistemological interest is what the
process of making judgments about the affinity between different versions of a
concept involves to emphasize that such moves are (or at least can be) rational
for agents at a stage in their evolution.
As Harold Brown (1986) has noted, Sellars’s answer to this can be traced
through his views on analogy and conceptual change (1974b and 1965).
Well-motivated episodes of conceptual change – including what we have been
discussing as revisions here – can indeed be seen as cases of abandoning one
concept for another, but to leave it there would be to miss the fact that not all
such moves involve the same degree of abandonment of an established set of
250 Michael P. Wolf

concepts and theoretical commitments. Different concepts may bear an affinity

to one another, measured by their contribution to the explanatory theories to
which they belong, and the greater the degree of similarity between two
concepts, the more readily we should accept such a revision when the evidence
appears to motivate it. Perhaps the clearest cases of this are those in which the
role of a concept in a theory does not change at all, aside from a refinement or
correction of some accepted value. If new techniques and equipment allow us
to revise the value of a physical constant by a fraction of a percent (as happens
occasionally in any given field), we have abandoned an old commitment to
embrace a new one and thus changed the details of a theory, but such revisions
are hardly revolutions, nor do they even draw much attention from others
working in the field. For example, initial experimental measurements of
Planck’s Constant put the value at 6.63x10-34 joule-seconds. More precise
measurements subsequently moved the value down to 6.62559x10-34.
Revisions were made when research on superconductors led to the discovery
of the Josephson Effect in the early 1960s, which resulted in a change of the
accepted value from 6.62559x10-34 to 6.626196x10-34. That is an upward
correction of roughly one one-hundredth of a percent. Nothing much changes
about how those constants are used in such cases, or how they relate to other
pieces of the conceptual puzzle, so the adjustments are minimal and
presumably bring slightly greater harmony with observed results.
Brown also notes that judgments about similarity between concepts on
Sellars’s view are based upon a pair of complementary approaches. First, we
have what we might think of as a comparative approach that searches for
explicit (or explicable) similarities that already exist in our formulation and
usage of a concept. For instance, in the example of revising a physical constant
I mentioned above, the constant continues to stand for the same quantity,
appears in the same laws related to the same things in the same ways, etc.
Ample similarity at this level may leave us with the impression that we are not
even changing concepts or commitments, just making minor corrections to an
existing conceptual stock. Second, there is the use of analogy in the change of
concepts and the introduction of new concepts. To deploy a familiar example,
various versions of molecular theory, including the kinetic theory of gases,
introduced molecules by an analogy with billiard balls. The analogy suggested
that they would be governed by the same mechanical laws as billiard balls,
which made various attempts to explain their behavior much more plausible.
The analogy came with a commentary as well, limiting the degree of similarity
and cutting off various paths of inference that might result from taking the
analogy too literally and too far. We should not necessarily think of these two
approaches as exclusive, nor think of either one of them as prior to the other.
In any given case, both of them might be deployed to some effect. In cases
 Sellars on the Revision of Theoretical Commitments 251

where a revision is very minor, the use of analogy may seem artificial, since
analogy serves as a tool to help us imagine possibilities that we cannot yet
fully grasp, and minor revisions are cases where we do not face such
difficulties. But in cases of more radical revision, the comparative search for
similarities may seem to give us something too thin to really inform our
decision, leading us to deploy analogy more fully. The weight given to each of
these approaches and their success in guiding our judgments will vary widely
from context to context, so what more we can say in general terms here is
rather limited. Let us move on to other problems surrounding such changes.

4. One More Objection

One serious concern here is that for all the trappings of revisability I have
offered and all the epistemic machinations Sellars described, there is still just
the sort of disparity between stages of a concept’s evolution that critics had
feared. We now have a view in which there are a great many different versions
of a concept, each fixed in its inferential role, and joined to subsequent stages
only by our judgments of similarity between the stages. By analogy, imagine
the individual frames of a motion picture, either in celluloid or digital form.
Each one is fixed and discrete, with no actual motion present between the
stages; our sense of motion is merely an illusory byproduct of many still
frames being shown to us in a rapid fashion.7 Following the analogy, any sense
of continuity that we have between stages of theoretical development would be
illusory, as each stage is fixed and discrete, with a different concept introduced
at each subsequent stage. If each stage of a concept’s evolution amounts to a
fixed, discrete inferential role and changes amount to adoptions of similar but
distinct inferential roles, then we are apparently committed to the sort of
discontinuity we hoped to avoid.
The first step to neutralizing this criticism is seeing our epistemic duties
surrounding inferential roles more clearly. Once a concept has been accepted
by the relevant community of speakers, we may expect that there will be an
inferential role associated with its articulation in the language. (How and when
that accord comes about are interesting questions, but beyond the scope of this
discussion.) Indeed, even proponents of the Kripke/Putnam view here will
agree that at any given time there is some set of inferences licensed by, or

7 Zeno’s Arrow Paradox makes a similar point bout our understanding of motion. If, at each
instant along its trajectory, the arrow is still, how can those many moments of stillness add up to
motion.
252 Michael P. Wolf

descriptions associated with, our usage and inquiries into the matter. The next
step involves keeping sight of the fact that our adoption of such commitments
is not an arbitrary or impulsive, even if it is voluntary to some degree. We take
up a version of a concept because we expect it to do something and to the
extent we can affirm such things, it apparently succeeds in doing so. (More
hypothetical concepts may be less tightly bound to concrete evidential and
explanatory expectations, but their role is presumably one of probing into
novel possibilities and their utility in doing so is the basis for any sustained
interest in them.) Our adoption of such inferential commitments is predicated
upon their significance and efficacy at their prescribed tasks, and
consequently, those commitments are defeasible in light of new evidence or
other suitable considerations. Thus, our entitlement to make assertions
generally is defeasible and does not grant us the sort of a priori grounds to
assert something as unchallengeable fact that the “cluster concept” view of
which Kripke and Putnam were so critical would have. A particular
articulation of the inferential role played by a term in our language might
prescribe such things, but it will generally be an open question whether that is
the inferential role we should adopt for such a term, given the tasks we ascribe
to it as part of our language. Failure at some portion of those tasks would be
appropriate grounds for a revision.
The final step in resolving this difficulty is seeing that the crucial question
in establishing the requisite sort of affinity among speakers is not an empirical
one of tracing which inferences they did in fact endorse and when they did in
fact stick to their guns, but rather a normative one of determining which
revisions we and they should make. To endorse a revision for ourselves is not
simply to make an impulsive move to a new set of commitments, it is to assert
that the evidence and other factors warrant such a move, and given a suitable
sense of affinity in the sorts of inquiry that we undertake, our ancestors
undertook and future generations will presumably continue, such revisions
would be rationally obligatory for any community of speakers given that sort
of evidence. What makes our claims and their claims relevant to both
communities of speakers is neither a relation to an extra-linguistic ideal, nor
some small core of unchallengeable commitments that we stick to, come what
may. It is the ways in which we share relevantly similar social practices that
we articulate with such terms, including our efforts and obligations to create
theories that accord with the results of our continuing inquiries.8 This implies

8 A slightly less tortured locution here might be “fit the facts,” but I am wary of appealing too
bluntly to such things in offering an account of what theories and empirical inquiry do. Many such
formulations quickly devolve into naïve realism in my experience, and so I am simply attempting
 Sellars on the Revision of Theoretical Commitments 253

that no community acting as it should ever has the sort of a priori authority
about its natural kind terms that this objection decries; what those practices
require and commit us to is informed by an ongoing process of observation,
development and refinement. Those communities can of course stick to their
guns and insist that water is simple (i.e., without a molecular substructure) or
that length is an absolute concept, or any other comparable sort of claim,
regardless of what sort of evidence arises. But in doing so, they grant
themselves a certain sort of license to prescribe their own practices that is not
in keeping with the ones we undertake. To the extent that they do so, they
cease to speak and practice as we do, and any sense that they mean what we do
by similar sets of symbols, noises, etc. is superficial.
So the response to the “frames of a film” objection offered above turns on a
rejection of the analogy, or rather the attribution of a view on meaning that
motivates it. The various commitments associated with a term at different
stages should never have the sort of a priori status that some philosophers have
suggested and the defeasibility of those commitments entails that we never
grasp them in such a static, isolated, immune fashion. It would not be wrong to
suggest that each of these stages could be made explicit as part of a synchronic
picture of the language at that point – like a frame in a film – but to leave the
story there would be a distortion. If any such acceptance of a set of
commitments is subject to the sort of revision and obligations to revise that I
have described, then not one of them stands in the sort of static epistemic or
pragmatic position that we had feared. To think that it might rests on the
assumption that we could grant the semantic a unilateral and pervasive
authority over pragmatic and epistemic matters. This is an assumption that
proponents of the New Theory of Reference have rightly cast aside, but it is
one that the pragmatist tradition, and Sellars in particular, had moved away
from long before.
What I have suggested here is by no means intended to be a comprehensive
account of epistemology or even the subject of revision within epistemology.
Rather, I have attempted to draw on Sellars’s idea of rules of criticism and his
approach to meaning to show how they might jointly inform one another,
giving us one very broad facet of the practices of empirical inquiry. If we
accept conservatism and explanatory strength as rules that guide our

to make my language as neutral as possible here. A more subtle account of how the physical world
and the space of reasons form a certain synthesis, distinguishable but interwoven with one
another, will strike anyone familiar with Sellars as a fine goal, and in my view would make a fine
manifesto for epistemology in the next hundred years, although it is a little bit too expansive to
address here.
254 Michael P. Wolf

self-evaluation of practice, we can see – albeit in very broad ways – how most
revisions of our empirical theories take place. We can then see how the
meanings of our utterances and the need to preserve as much of our semantic
and epistemic inheritance as possible drives us towards an acceptance of those
rules of criticism in the first place. This leaves us with an account grounded in
the pragmatic, with deep connections between the semantic and the epistemic,
that is characterized by its diachronic rational character rather than appeals to
the abstract and ideal or the merely conditioned and accepted.

Michael P. Wolf
Department of Philosophy
California State University, Fresno
Fresno, CA 93740
United States of America
e-mail: [email protected]

REFERENCES

Brown, H.I. (1986). Sellars and Conceptual Change. Synthese 68, 275-307.
Crosland, M., ed. (1971). The Science of Matter: A Historical Survey. Harmondsworth, UK:
Penguin Press.
Donovan, A., ed. (1988). The Chemical Revolution: Essays in Reinterpretation. Philadelphia, PA:
History of Science Society.
Fodor, J. and E. LePore (1992). Holism: A Shopper’s Guide. Oxford: Basil Blackwell Pub.
Kripke, S. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press.
Kuhn, T. (1996a). The Structure of Scientific Revolutions. Third edition. Chicago: University of
Chicago Press.
Kuhn, T. (1996b). The Natural and the Human Sciences. In: J. Conant and J. Haugeland (eds.),
The Road Since Structure: Philosophical Essays 1970-1993, pp. 216-223. Chicago:
University of Chicago Press.
Putnam, H. (1975). The Meaning of ‘Meaning’. In: Mind, Language and Reality: Philosophical
Papers, vol. 2, pp. 215-272. New York: Cambridge University Press.
Quine, W. (1953). From a Logical Point of View. New York: Harper and Row.
Quine, W. (1964). Word and Object. Cambridge, MA: The MIT Press.
Quine, W. (1974). The Roots of Reference. LaSalle, IL: Open Court Press.
Sellars, W. (1963a). Science, Perception and Reality. Atascadero, CA: Ridgeview Press
Sellars, W. (1963b). Is There a Synthetic A Priori? In: (1963a), pp. 298-320.
Sellars, W. (1963c). Some Reflections on Language Games. In: (1963a), pp. 321-358.
Sellars, W. (1965). Scientific Realism or Irenic Instrumentalism. In: R. Cohen and M. Wartofsky
(eds.), Proceedings of the Boston Colloquium for the Philosophy of Science, 1963 (Boston
Studies in the Philosophy of Science, vol. 2), pp. 171-204. Dordrecht: D. Reidel Press.
Sellars, W. (1967). Science and Metaphysics: Variations on Kantian Themes. Atascadero, CA:
Ridgeview Press.
 Sellars on the Revision of Theoretical Commitments 255

Sellars, W. (1969). Language as Thought and Communication. Philosophy and Phenomenological

Research 29 (4), 506-527.
Sellars, W. (1974a). Essays in Philosophy and Its History. Dordrecht: D. Reidel Press.
Sellars, W. (1974b). Conceptual Change. In: W. Sellars (1974a), pp. 172-188.
Sellars, W. (1974c). Empiricism and Abstract Entities. In: W. Sellars (1974a), pp. 245-286.
Sellars, W. (1977). Berkeley and Descartes: Reflections on the Theory of Ideas. In: P. Machamer
and R. Turnbull (eds.), Studies in Perception, pp. 298-311. Columbus, OH: Ohio State
University Press,.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Unger, P. (1975). Ignorance. Oxford: Clarendon Press.
This page intentionally left blank
Jaroslav Peregrin

DEVELOPING SELLARS’S SEMANTIC LEGACY:

MEANING AS A ROLE*

ABSTRACT. Wilfrid Sellars’s analysis of the concept of meaning led, in effect, to the
conclusion that the meaning of an expression is its inferential role. This view is often
challenged by the claim that inference is a matter of syntax and syntax can never yield
us semantics. I argue that this challenge is based on the confusion of two senses of
“syntax”; and I try to throw some new light on the concept of inferential role. My
conclusion is that the Sellarsian view that something can have meaning only if it is
subject to inferences is viable, and that inferential role is a plausible explication of
meaning. However I also argue, pace Sellars, that the inferential nature of meaning
does not prevent us from engaging in the enterprise of Carnapian formal semantics.

1. Sellars on Meaning and Abstract Entities

In his seminal article “Abstract entities,” Wilfrid Sellars claims that the
meaning of a linguistic term is best construed as sort of role played by the term
within the drama of language (and that, consequently, abstract entities, which
are normally seen as expressed by such terms are linguistic in nature).
“Redness,” writes Sellars:
[A]s a first approximation, is the word •red• construed as a linguistic kind or
sort which is capable of realization or embodiment in different linguistic
materials, e.g. *red*, *rot*, and *rouge*, to become the English word ‘red’, the
German word ‘rot’, and the French word ‘rouge’ . . . Thus •red• is a type which
is shared by the English word ‘red’, the German word ‘rot’, and the French word
‘rouge’. (Sellars 1963, p. 267)
(Here the asterisks are supposed to form names referring to expressions as
mere shapes, quotation marks form names referring to expressions qua

* Work on this paper was supported by research grant no. A9009204 of the Grant Agency of the
Academy of Sciences of the Czech Republic.

In: M.P. Wolf and M.N. Lance (eds.), The Self-Correcting Enterprise: Essays on Wilfrid Sellars
(Poznań Studies in the Philosophy of the Sciences and the Humanities, vol. 92), pp. 257-274.
Amsterdam/NewYork: Rodopi, 2006.
258 Jaroslav Peregrin

expressions of a particular language, and the names formed by the dots refer to
the roles instantiable by different expressions in different languages.)
Crucial for this conception is the distinction between what can be called
“generics” and “true universals.” When I say “The lion is tawny” and use “the
lion” as expressive of the generic, my claim is reasonable and true. On the
other hand, should I construe it as standing for the universal (“Lionhood is
tawny”), what I say would be false (if not utterly meaningless) – a universal
surely is not tawny (see Sellars 1963, p. 627). Now what Sellars claims is that
universals are in fact a kind of “second-order” generics – they are generics
whose instances are linguistic expressions. Thus the universal lion is in fact the
role played by the word ‘lion’ in English (and by other words in other
languages).
Hence Sellars may be seen as distinguishing between generalization, which
is manifested by the step from particular statements to a general one (“All
lions are . . . “ resp. “The lion is . . . “) and abstraction, which is a matter of a
further step to a particular statement about a universal, which is the notorious
“e pluribus unum” (“Lionhood is . . . ”). And while philosophers had
traditionally wanted to analyze the abstraction as a matter of a peculiar knack
of human mind to put many things into a nutshell of one, Sellars proposes to
analyze it as a capacity into which the capacity of generalization mutates when
language sets in. For language itself provides us with a certain “one over
many” (one word ‘lion’1 over many particular lions); and when we apply our
capacity of generalization to it, we find the abstract universal (lionhood) as the
one over the many words playing the role assumed by ‘lion’ in English within
other – actual or possible – languages. Hence the peculiar ability of our minds
to abstract appears as secondary to its ability to (generalize and) “language
about languagings” (1974, p. 425). This leads Sellars to the conclusion that the
senses of our expressions (or their intensions) are best seen as certain roles of
the expressions.
What is a role? Sellars invokes probably the most frequented simile within
modern philosophy of language and compares language to chess.2 It is rather
clear what the role of, say, pawn is: it is delimited by the rules saying how the
pawn moves, how it interferes with other pieces, etc. Sellars points out that to
be a pawn usually involves two kinds of criteria: descriptive criteria (“to look

1 The word as such is, of course, itself an abstract item, one shape over many particular

inscriptions or sounds. However, it seems that Sellars takes this kind of abstraction as somehow
straightforward and unproblematic.
2 In this way he queues up into the long line of linguists and philosophers already containing

Husserl, Frege, de Saussure, Wittgenstein, Jakobson, etc.

 Developing Sellars’s Semantic Legacy: Meaning as a Role 259

thus and so”) and prescriptive criteria (“to be correctly moved thus and so”).
Now the role in his sense is what remains if we subtract the descriptive part.
Sellars (1963) invites us to imagine an alternative game, “Tess,” which is
played in Texas and which is just like chess, save for the fact that the various
kinds of pieces are replaced by various types of cars and the fields of the
chessboards are replaced by counties. The same role which is assumed within
chess by our pawn is assumed in Tess by, say, their Volkswagen; and the role
is what these two utterly dissimilar objects share.3
Now if we accept that language is something like a “rule-governed
activity” (as urged not only by Sellars, but of course also by the later
Wittgenstein and his followers), the idea that also meanings might be sort of
roles comes quite naturally. But what kind of rules is constitutive of this kind
of roles? There appears to be an abundance of linguistic rules: we have, for
example, the rule that it is correct to change ‘read’ in the third person singular
to ‘reads’, or the rule that claiming that somebody stole something commits
the claimer also to the claim that the very somebody is a thief; or the rule that
it is impolite to say “Ciao” to the Queen of England. Which of them, if any,
represent the ones that are responsible for meaning? Sellars indicates (see esp.
1953; 1992) that the rules which are crucial in this respect are the rules of
inference (Sellars 1992 calls them “consequential rules,” see p. 114), i.e., the
rules exemplified by the middle one of the three examples.

2. What Is an Inferential Rule?

Inferential rules are often taken as kind of prescriptions for passing from an
antecedent to a consequent: perhaps from asserting the former to asserting the
latter, or from believing the one to believing the other. But this is clearly
ill-conceived: when we say that it is correct to infer Fido is a mammal from
Fido is a dog, then we surely do not want to say that whoever asserts the
former is obliged to assert the latter – if we were always to assert all
consequences of everything we assert, we could never stop making assertions.
And also we surely cannot want to say that one is obliged to believe all the
consequences of what she believes – we do not control our own beliefs in such
a way that this kind of obligation would make more sense than, say, the
obligation to have luck.

3 These themes are further elaborated especially in Sellars (1950; 1954; 1963; 1969; 1974; 1992;

1997). See also Harman (1970), Marras (1973) and Seibt (1990).
260 Jaroslav Peregrin

To understand the Sellarsian notion of inference, we consider his

distinction between what he calls the “rules of doing” and the “rules of
criticizing” (1999, p. 76). Inferential rules are basically the latter – they do not
tell us what to do, but rather what to despise – and hence are really more the
rules of what not to do. They delimit a space of what is approvable: if you
assert that Fido is a dog, then you should not deny that Fido is a mammal; and
if you do deny it, you are a legitimate target of criticism.
Hence the rules of inference – which are constitutive of conceptual content
– are not a matter of regularities of linguistic conduct, of “passing over” from
some utterances or beliefs to other ones (whatever the “passing over” might
mean). They are a matter of what Brandom (1994) calls normative attitudes:
attitudes to one’s own and others’ linguistic utterances, of “languaging about
languagings.” This is an important thought: conceptual content is not a matter
of regularities of linguistic behavior, but rather of (regularities of) a
“metalinguistic behavior” (see Sellars’s (1950) criticism of “regulism”).
This is closely related to the Sellarsian “dialectics” of ought-to-do and
ought-to-be: whereas an ought-to-do in a sense directly is also an ought-to-be
(and hence a rule of criticism directly piggybacks on every rule of doing) an
ought-to-be bears an ought-to-do via a kind of a “practical syllogism” (see
Sellars 1969): via the conviction that we ought to do what brings about that
which ought to be. (But of course that if roles of inference are really rules of
what ought not to be, they lead us to what not to do.)

3. Can We “Get Semantics out of Syntax”?

One of the main reasons why many people would reject the proposal to see
semantics as a matter of inferential rules right off stems from the combination
of two “received wisdoms”:
(i) inferential rules are a matter of syntax; and
(ii) syntax is not enough for semantics.
I think that for each of the two claims there is a sense of ‘syntax’ in which it is
true. The trouble is, however, that the two needed senses are different; and it is
impossible to take the two claims to be true jointly (without taking the two
occurrences of ‘syntax’ in them to be mere homonyms). And in particular, it is
impossible to infer that inferential rules are not enough for semantics.
What does the word ‘syntax’ mean? In one sense, syntax is the theory of
well-formedness: it specifies rules for composing more complex expressions
out of simpler ones, thus delimiting the space of all expressions, and especially
 Developing Sellars’s Semantic Legacy: Meaning as a Role 261

all sentences, of a language. It is clear that if we understand syntax in this way,

(ii) will be true, but (i) not.
But syntax is sometimes understood in a wider sense, as amounting to
every property of expressions (and every relation between them) specified
without reference to anything else than the shapes of the expressions (this is
the sense put into circulation by Carnap).4 Taken in this way, (i) becomes true,
but at the same time the falsity of (ii) ceases to be a matter of course. Let us
see why.
Let us consider a list of names, such as a telephone directory. We can say
that the names it contains have two kinds of properties: those which they have
“by themselves” (and which thus can be “read off” them), and those which
they have as a consequence of being names of particular persons (and which
are thus not “visible”). Examples of the former properties are to begin with ‘N’
or to consist of more than ten characters; an example of the latter is to be a
name of a bald man. Let us call the former “syntactic,” and the latter
“semantic.”
Now suppose somebody finds out that, just by chance, all and only bald
men have names beginning with ‘N’. With this knowledge, it becomes possible
to read the property to be a name of a bald man off the name – it is enough to
check whether it starts with ‘N’. Does this mean that we have managed to
change the property in question from a “semantic” to a “syntactic” one? Surely
not, we have only found its “syntactic symptom” and thereby a way to
“represent” it by a “syntactic” property.
Notice that the same holds if we are not so improbably lucky as to find out
that the names of all bald men share such a simple property. We achieve the
same by simply making the list of all the relevant names: then it will be again
possible to find out whether a name is that of a bald man by simply inspecting
its shape – namely by comparing it with a given list of shapes. If our directory
is finite, then we can “syntacticize” any property in this way; if it is infinite,
then perhaps not any, but surely we still can do so with many.
Now it seems to be clear that correct inferability is basically a semantic
relation – of course that we cannot see whether a sentence is correctly
inferable from other sentences by merely inspecting the shapes of the
sentences (i.e., without knowing what they mean). However, we can make a

4 Elsewhere (Peregrin 1999) I argued that once we abandon the view that language is a collection

of labels stuck to things and see it as a means of an activity, the Carnapian concept of
syntax/semantics boundary is no longer available. For then the only perceptible boundary is that
between the question which strings are employed (syntax) and the question how they are
employed (semantics-cum pragmatics).
262 Jaroslav Peregrin

list of correct inferential patterns (such as the one assembled by Frege in his
Begriffsschrift as well as by the host of his followers), and given we know this
is a (complete) list of valid patterns, we can recognize inferability simply from
the shapes.5 (In fact, this is what appears to have been behind the famous
Hilbert’s program: once we are able to show that whenever A follows from X,
it is derivable from it in terms of our rules and vice versa, we may reduce truth
to provability and hence to a “mechanical manipulation” of symbols.)
This means that inference, though its rules are articulated in syntactic
terms, is not syntactic in the other – more important – sense; and this removes
the absurdity from taking inferences as constitutive of meanings.

4. A Digression: Keepers, Joiners, Crashers, and Co.

To throw more light on the nature of roles which we will claim underlie
meanings, let us make a digression. Suppose we have an alphabet and consider
strings which can be formed by its means. Suppose the space of all such
potential strings is divided into two groups: the “good” and the “bad” ones;
and we are interested in the ways of building the good strings.
When we toy with putting such strings together, we may come to notice
that various letters and strings display specific kinds of “behavior.” We may,
for example, discover that some letters/strings have the property that when
concatenated with any strings whatsoever they always produce bad strings; and
we may conceive of calling such rogue strings crashers. Other letters/strings
may do the same thing when they are concatenated with other strings in a
particular way: we may, for example call a string an l-crasher iff it produces a
bad string whenever it is prefixed to a string. Still other strings may have the
property that they produce a bad string whenever they are prefixed to good
strings – we may call such a string l-killers. Or we may come to call a string a
joiner iff it produces a good string if and only if it is put in between two good
strings.
Here are some examples of the basic categories of strings we may discover
when studying buildability of good strings:
S is a crasher: any string containing S as a substring is bad
S is a l-crasher (r-crasher): any string starting (ending) with S is bad

5 Of course that as the number of such patterns is unlimited, the list can be given only
“potentially,” via a finite list of some basic patterns (axioms) and some derivation rules. A valid
proof is then the decomposition of a given inference into a chain of the axiomatic patterns.
 Developing Sellars’s Semantic Legacy: Meaning as a Role 263

S is a l-killer (r-killer): any string consisting of S followed (preceded) by a good

string is bad
S is an l-inverter (r-inverter): a string consisting of S followed (preceded) by
another string is good iff the other string is bad. (Every inverter is obviously
an killer.)
S is a l-keeper (r-keeper): any string consisting of S followed (preceded) by a
good string is good
S is a l-superkeeper (r-superkeeper): any string consisting of S followed
(preceded) by a good string is good iff the string is good. (Every
superkeeper is an keeper.)
S is a joiner: any string consisting of two good strings with S between them is
good
S is a superjoiner: any string consisting of two strings with S between is good
iff the two strings are good. (Every superjoiner is a joiner.)
Consider, for the sake of illustration, what we will call Language 1. The
vocabulary is
a, b, c, d, e
and the delimitation of the good strings over it is as follows:
(i) a, b and c are good strings;
(ii) if s is a good strings, then d ∩s (i.e., the symbol d followed by the string s)
is a good string;
(iii) if s1 and s2 are good strings, then s1∩e∩s2 is a good string;
(iv) no other string is good.
The words of this language can be clearly classified into three categories.
First of them is constituted by the letters a, b and c; the second one by the
letter d, which is an l-superkeeper; and the third by the letter e, which is a
superjoiner.
Now suppose we distinguish a further subset of the set of good strings; call
its elements “perfect” and call those good strings which are not perfect
“imperfect.” We can, of course, again start to classify the roles of letters and
strings with respect to the new set. Returning to Language 1, imagine that a
and b are perfect strings, while c is imperfect. Further imagine that the string
d ∩s is perfect if and only if s is imperfect; and that s1∩e∩s2 is perfect iff both s1
and s2 are. It is easy to see that from the viewpoint of the delimitation of the
perfect strings, then, d comes to be an l-inverter, whereas e is again a
superjoiner.
Now consider a little more sophisticated kind of language, Language 2,
based again on the letters a, b, c, d, and e. Suppose that a string over this
alphabet is good if it is of the shape s1∩e∩s2 where the strings which are usable
in the position of s1 and s2 are defined recursively: a is usable, any usable
264 Jaroslav Peregrin

string preceded by b is again usable, and any string which is of the shape
s1∩c∩s2 is s1∩d ∩s2 with s1 and s2 usable is again usable. Examples of good
strings are
aea
bbaea
baebbbbba
acbaeba
acaebbbadbbbba
To characterize the roles of the strings of this language we do not make do
with the simple concepts of joiners, inverters etc. We first have to characterize
the strings which may surround e in a good string – call them e-joinables. a is
an e-joinable. b is what could be called e-joinable-l-keeper, whereas c is what
we could think of calling e-joinable-superjoiner.
Now imagine that the perfect strings of this language are defined in the
following way:
aea is perfect;
if s1∩e∩s2 is perfect, then so is b ∩s1∩eb ∩s2;
s∩cae∩s is perfect;
if s1∩c∩s2∩e∩s3 is perfect, then so is s1∩cb ∩s2∩eb ∩s3;
s∩daea is perfect;
if s1∩d ∩s2∩e∩s3 is perfect, then so is s1∩db ∩s2∩e∩s3∩c∩s1;
nothing else is a perfect string.
Again, these stipulations furnish the strings with roles, which are now
rather intricate. Call two e-joinable strings s1 and s2 e-comparable, if s1∩e∩s2 is
a perfect string. We may say that a string cancels s from the left in the end
position (in short, is a s-l-e-canceller) if a string ending with c∩s is
e-comparable to a string without this ending. And we may say that it shifts s
from the left to the beginning (is a s-l-b-shifter), if a string containing c∩s is
e-comparable with a string starting with s and continuing by something
e-comparable with the original string with s excerpted. Using this terminology,
c is a a-l-e-canceller and a b-l-b-shifter.

5. Meanings as Roles

We have seen that any delimitation of a subset of the set of strings over an
alphabet furnishes the strings with certain roles, which are a matter of their
“behavior” with respect of the distinguished set and which may be quite
complicated. Especially any language and any axiomatic system (which also
delimits a set of sentences, namely the set of its theorems) does so. Now the
 Developing Sellars’s Semantic Legacy: Meaning as a Role 265

meaning which an axiomatic system confers on an expression of its language

can be seen as precisely this kind of role.
Consider formal languages of the kind that we can formulate within the
framework of the propositional and the predicate calculus, with good strings
being the well-formed formulas. It is, for example, easy to see the usual syntax
of the propositional calculus is only a minor variation on our Language 1. We
also have three categories of signs: the propositional letters, the l-superkeeper
¬ and the superjoiners ∧, ∨ and →.6 More generally, taking the set of good
strings as the set of well-formed formulas, the roles expressions thus acquire
are usually called syntactic categories. (We would not say that, ∧, ∨ and → are
superjoiners, but rather that they are propositional connectives.)
Now, moreover, take the set of perfect strings to be the set of true
sentences (under some interpretation). What, then, do the resulting roles
correspond to? Consider conjunction: we have noticed that with respect to the
perfect (i.e., true) sentences it acts as a superjoiner. This means that it produces
a perfect output iff the two strings which are appended to it from the sides are
both perfect, i.e., it maps perfect and perfect on perfect, whereas perfect and
imperfect on imperfect etc. However, as perfect is true and imperfect is false,
we have the function mapping T and T on T and any other combination of the
truth-values on F. This is to say that the truth table usually thought of as the
meaning of conjunction can be plausibly seen as the recapitulation of the role
of a superjoiner. This indicates that to be the conjunction simply is to be the
superjoiner (with respect to the class of true sentences), to be the negation is to
be an l-inverter, etc. Thus our Language 1 amounts to nothing else than the
classical propositional calculus.
Now look at language Language 2. If we replace the signs a, b, c, d and e
with 0, Succ, +, ×, and =, respectively, we can see that Language 2 constitutes
a kind of rudimentary arithmetic. In this case, the roles are much more
complicated and they are mutually entangled (so that it is not possible to
specify the role of, say, + independently of that of 0), but nevertheless they are
still roles in the very sense as before. Hence, we may say, the meaning
arithmetic confers on the sign + is the role of a “0-l-e-canceller” and a
“Succ-l-b-shifter.” (Expressed in a more prosaic way, this means that to add
zero is to add nothing and to add a successor of a number is to produce the
successor of the addition of the number.)

6 This is an oversimplification for if we consider the standard form of the syntax, we need
brackets; and if we consider the Polish variant, the connectives will not be literally superjoiners
but rather something which could be called, say, l-bikeepers.
266 Jaroslav Peregrin

However, as Lakatos (1976) reminds us, this “Euclidean” (may be we

should better say “Hilbertian”7) way of viewing an axiomatic system
(according to which any meanings the signs have are conferred on them by the
axioms) is only one possible way; and in fact such systems are often
understood in a quite different way, which Lakatos calls “quasi-empirical.”
From this point of view, such a system is seen as codifying some pre-existing,
proto-theoretical knowledge, and hence is responsible to it: if it produces
theorems whose interpretation is at odds with what we take for granted within
the proto-theory, it is revised.
Let us take a formula of PA. The answer to the question whether it is true
can be sought in two quite different ways: we may check whether it is
deducible from the axioms; or we may check whether the “informal”
mathematical sentence which is taken to be “captured” by it (in view of the
fact that the basic vocabulary and grammatical rules of PA more or less copy
the basic vocabulary and rules of our pre-formal arithmetical language) is true.
If we see the system in the Euclidean way, then a possible discrepancy
between the two results need not worry us – for the latter is completely
irrelevant for us. However, if we see it in the quasi-empirical way, then the
discrepancy might lead us to revise the axioms.8
Hence we should qualify the claim we made above: any axiomatic system
furnishes the expressions of the underlying language with Euclidean meanings.
But is it not so that such an Euclidean meaning is restricted to formal systems
only, and that the meaning in normal languages is something quite different
which is perhaps “incommensurable” with it?
Observe, first, that we can reconstruct, and in some special cases perhaps
even replace, parts of natural language by means of formal languages
constituted in the axiomatic way. The axiomatic Peano arithmetic is not only a
commonly accepted codification of pre-formal arithmetic, but in the eyes of
many mathematicians it simply replaces it. This means that Euclidean

7 In a letter to Frege, Hilbert writes: “You write: ‘From the truth of the axioms it follows that they
do not contradict one another’. I was very interested to read this particular sentence of yours,
because for my part, ever since I have been thinking, writing and lecturing about such matters, I
have been accustomed to say just the reverse: if the arbitrarily posited axioms are not in mutual
contradiction with the totality of their consequences, then they are true – the things defined by the
axioms exist.” Hence, according to Hilbert, each delimitation of a set of strings constitutes an
Euclidean semantics and this is the only semantics we could possibly require; whereas according
to Frege, Euclidean semantics of the systems we assemble is not the true semantics, for the true
semantics is what expressions of our natural language have and what we cannot freely create.
8 I have discussed the consequences of these two radically different aspects of logical systems

elsewhere (see Peregrin 2001).

 Developing Sellars’s Semantic Legacy: Meaning as a Role 267

meanings can be not only seen as suitable explications of the corresponding

“natural” meanings, but sometimes even as their descendants. What makes this
possible?
The Sellarsian answer is that natural languages are themselves sort of
Euclidean systems, only with axioms and rules not being explicit in the form
of formulas or statements, but implicit within the praxis of using language. Our
arithmetical praxis is governed by the rule that adding zero to x means leaving
x unchanged just as the axiomatic PA is – though it need not be written
anywhere explicitly. Children learn to add in this way by being corrected – the
“ought-to-do” imposed on them by their teachers thus becoming their
“ought-to-be” (see Sellars 1969). And just like within the Peano arithmetic
there is nothing to semantics save the explicit rules (axioms and derivation
rules), there is, in natural language, nothing to semantics save the implicit rules
of our “language games.” Thus, the claim is that every meaning is a sort of
Euclidean meaning.
This is not to deny that there is an important difference between systems
constituted by explicit and hence rigid rules and those governed by implicit
and hence flexible ones, and so consequently between the respective “formal”
and “natural” meanings. However, let us postpone further discussion of this
difference to the concluding section, and let us now accentuate the common
Euclidean character of both kinds of meaning: the former are explicitly
Euclidean, the latter are implicitly Euclidean and can be explicated in terms of
the explicitly Euclidean ones. Given this, we may want to consider the
following two general questions:
(1) Is every “formal meaning” a meaning worth its name?
(2) Is every “natural meaning” reconstructible as a “formal” one?
The answer to the first question is clearly no. We may have a formal language
with a single sentence; or a language without axioms, or such that any sentence
is inferable from any other. No such system obviously yields the rich variety of
roles characteristic of what we would be willing to call language, and hence
furnishes the expressions with what we would be willing to call meanings.
(The last example covers the case of Prior’s 1960/61, tonk, which is a grand
example of the fact that nontrivial axioms need not lead to a nontrivial
semantics. This is important, but it does not show, pace Prior, that axioms or
inference rules are generally short of conferring meaning. See Peregrin 2001,
§8.5 for more details.)
The other question is more intricate; and prima facie there seem to be at
least two counterexamples. First, meanings of empirical terms; and, second,
meanings of some common non-empirical terms which do not seem to be
conferrable by means of rules. Let us deal with the two topics in turn.
268 Jaroslav Peregrin

6. Empirical Terms

Of course that Sellars does not claim that the meaning of an empirical term is a
matter of nothing else than inferences in the usual sense on the term. What he
claims (1974, pp. 423-424) is that, in general, the ways in which expressions
are used (and hence their meanings) are a matter of three kinds of regularities
or patterns of linguistic activity9:
(1) Language Entry Transitions: The speaker responds to objects in perceptual
situations, and in certain states in himself, with appropriate linguistic
activity.
(2) Intra-linguistic moves: The speaker’s linguistic conceptual episodes tend to
occur in patterns of valid inference (theoretical and practical), and tend not
to occur in patterns which violate logical principles.
(3) Language Exit Transitions: The speaker responds to such linguistic
conceptual episodes as “I will now raise my hand” with an upward motion
of the hand, etc.
Hence if we want to restrict the term ‘inference’ to the language-language
moves, we cannot claim that meaning is in general the inferential role, but
rather a role with respect to more inclusive kinds of rules. (The term inference
is sometime stretched to cover even Sellars’s (1) and (3), which then allows for
equating meaning and inferential role quite generally, but this appears to be at
odds with how the term ‘inference’ is most often used.)
We must, however, notice that it is only the intra-linguistic moves which
we can nontrivially capture when we are doing a theory of semantics. The
trouble with the other two is that they consist of a language-part and a world-
part, and when we want to talk about the latter, we have no choice but to put it
in words, by which it becomes identical with the former. We can practically
teach somebody to recognize the circumstances in which it is appropriate to
assert “A dog has just run by,” but when we want to capture them within a
theory, we end up with the disquotational claim “It is appropriate to assert ‘A
dog has just run by’ iff a dog has just run by.” As Wittgenstein (1977, p. 27)
remarked “The limit of language manifests itself in the impossibility of
describing the fact that corresponds to (is the translation of) a sentence without
simply repeating the sentence.” True, we need not use the very same sentence
to describe the circumstances, but if we do not do so, then the ensuing

9 A more fully elaborated (if not overelaborated) version of this classification is presented by
Sellars (1992, p. 114).
 Developing Sellars’s Semantic Legacy: Meaning as a Role 269

nontriviality can be seen as a matter of the intra-linguistic relationship between

the two sentences. This means that insofar we take semantics to deal
exclusively with language, we cannot expect that it completely elucidates the
meanings of non-empirical terms: part of their meanings is a matter crucially
depends on their relationship to the extralinguistic world. After all, the
delimitation of the meaning of “dog,” i.e., of what a dog is, should not be
expected of a semanticist, but rather of a biologist.
The situation is, however, different, as far as non-empirical words are
concerned. Their meaning can be usually seen as a matter of inferences alone.
There is nothing more to the meaning of ‘and’ than the inferences which
govern the term: the inferences from X and Y to both X and Y, from X and Y to
X and Y and perhaps some more.10 This means that while the meaning of an
expression is, from the current viewpoint, the role of the expression within the
relevant “language game,” the theoretically articulable part of the meaning
(which may in some cases, like that of ‘and’, wholly exhaust it) consists in its
inferential role.
Hence, we may say, syntax, in the form of inferential rules in some cases
can alone be sufficient for semantics. (The meaning of an non-empirical term,
such as ‘and’ can be directly seen as consisting in its inferential role.) And
there is an asymmetry between inferences (the language-language kind of
moves) and the other two kinds of moves in this respect: while inferences
alone are (sometimes) capable of conferring a meaning, neither the world-
language moves, nor the language-world moves alone are. An expression
cannot become truly meaningful by being subjected to merely language entry
rules, i.e., by becoming a correct reaction to something extralinguistic: it must
also be subject to inferences.
The reason is that language is a holistic, interconnected structure – an
expression which would not be connected, by the rules of language, to other
expressions would be the notorious “idle wheel.” (See Wittgenstein 1953,
§271/p. 95e.) Just like there cannot be a chess piece without a potential
interaction with all other pieces, there cannot be an expression which would be
connected only to something extralinguistic and not to other expressions. And
it is only inferences which are capable of interlinking all its elements. Hence
syntax, in the form of inferential rules, is in all cases essential for semantics:
for something can have meaning only if it is subject to inferences, i.e., if it is in
inferential relations to other expressions.

10 The proviso is here because the meaning of the English ‘and’ does not utterly coincide with that
of the classical conjunction operator, it, e.g. often behaves non-symmetrically.
270 Jaroslav Peregrin

7. Classical and “Standard” Semantics

There seem to be some limits to what kinds of meaning can be conferred by

rules; and it is sometimes claimed that we also need meanings of a kind that is
beyond the reach of semantics generated by inferences.
The fact is that if we understand the term inference as we did so far, we
even do not get classical propositional connectives in the inferential way. For
though there is an inferential pattern which inferentially characterizes the
classical conjunction, there is none which would so characterize the classical
disjunction, negation or implication. For though we can inferentially express
that the conjunction is true if one of its conjuncts is true, we cannot express
that it is false whenever both the disjuncts are false.11
However, as I have pointed out elsewhere (see Peregrin forthcoming), this
obstacle can be solved by reconsidering the concept of inferential pattern,
namely by admitting that an inferential pattern can be read as providing a kind
of an exhaustive listing. Claiming
A├ A∨B
B├ A∨B

we claim that A∨B follows from A and from B; but presenting the two
inferences as a self-contained pattern indicates that this list of premises be
exhaustive; i.e., that there is nothing else which would entail A∨B and at the
same time be entailed by both A and B.
This can be accounted for by letting inferential patterns incorporate an
implicit exhaustivity claim: if I give an enumeration of premises each entailing
a given conclusion, then we assume that the list is to be exhaustive. (Why? For
this is how we usually understand enumeration: if I say “I own a VW Beetle
and a Pontiac,” the normal understanding is that these are all my cars.) As
Koslow (1992) has shown, the admission of such “extremality conditions” into
the concept of inferential pattern leads directly to an inferential capturing of all
of classical logic.
And as I have indicated elsewhere (see Peregrin forthcoming), it could
even lead to the inferential accommodation of what is usually called the
standard semantics of higher-order logics (as contrasted to the Henkin one,
see, e.g., Shapiro 1991). Its peculiarity can be located into the incompleteness

11 We must not think that the completeness proof for the classical logic shows that the usual truth-

functional meanings of the logical operators are conferred on them by the standard axioms: the
axioms are compatible even with certain non-truth-functional interpretations.
 Developing Sellars’s Semantic Legacy: Meaning as a Role 271

of arithmetic – whichever axioms and rules satisfiable by the standard number

sequence we accept, they will be also satisfied by some nonstandard ones.
(The existence of the non-standard interpretations of higher-order logics then
can be seen as a matter of the fact that the corresponding semantic structures
are bound to incorporate the natural numbers sequence pattern.) Now if we
construe the “exhaustivity condition” which we take to be implicit in
inferential patterns broadly enough, we might even think of it as blocking the
nonstandard models: the standard axiomatization of Peano arithmetic says that
zero is a number and the successor of a number is a number, and the
exhaustivity says that nothing else is a number, hence that we intend only the
minimal model.
Of course the hints I have just given cannot establish that any kind of
semantics can be seen as ultimately grounded in inferences (They are clearly in
the need of further elaboration, which is, however, beyond the scope of the
paper and which I hope to produce in the future.) What I hope they can do is
shatter the preconception that some common varieties of semantics are, as a
matter of principle, unavailable for the inferentialist. This means that nothing
appears to stand in the way of our embracing Sellars’s role semantics, i.e., the
view that meanings are, as a matter of principle, roles of the expressions which
are said to express them.

8. “Natural” and “Formal” Meanings

To conclude, let us return to the previous claim that natural languages are
themselves one sort of Euclidean system. Of course the difference between a
game like chess or a formal language, the rules of which are explicit, and our
natural language games with their rules implicit in our linguistic conduct is not
unimportant. In fact, Wittgenstein’s Philosophical Investigations (1953) can be
seen to a great deal as wrestling with the question of what it means for a rule to
be implicit in behavior and what it takes to follow such a rule. But precisely
the fact that it is possible to explicate parts of natural language in terms of
formal systems constituted by means of explicit rules gives further credence to
the conviction that the rules are implicitly constitutive of our language
games.12

12 Brandom (1994) claims that this kind of making implicit rules explicit is the crucial force

behind the development of our language, behind logic, and in general behind our reason.
272 Jaroslav Peregrin

As we have also already noted, in the typical case the relation between a
natural meaning and the corresponding formal one is that between an
explicandum and its explicatum. (Only in cases of those parts of natural
language which do not deal with empirical reality, like the language of
arithmetic, might it be possible to directly replace the former by the latter.)
Natural meanings are empirical phenomena; we do not voluntarily constitute
them, we encounter them.13 Formal meanings are, in contrast to this, directly
of our making – constituted by our deliberate definitions. However, just
because of this it is useful to explicate the former by the latter (which then
results in the Lakatosian “quasi-empirical” view of the latter).
As I have discussed in greater detail elsewhere (Peregrin 2000; 2001,
Chapter 9), the way we humans often approach complicated empirical
phenomena consists in isolating their structure which then can be studied by
non-empirical, i.e., mathematical means. Thus we make mathematical models
of the weather, of economic developments or of geological layers; and we
make also such models of language. The point of this kind of mathematization,
which is characteristic of a great deal of modern natural science, is the
resulting beneficial division of labor: the task of the mathematician is to study
the model, whereas that of the natural scientist is to build it and see to it that it
remains adequate.
I think that the situation with logic and semantics is quite similar: as soon
as Boole, Frege and others achieved what they saw as the basic logical
structure of natural language, Hilbert and others subjected the structure to
mathematical treatment, giving birth to what has come to be called
mathematical logic. But the study of mathematical structures cannot tell us
anything about natural phenomena unless we see to it that the structures can
reasonably be seen as structures of these phenomena. (To avoid
misunderstanding, recall that the rules articulated within logic do not merely
reflect regularities of general linguistic traffic. They reflect the regularities of
normative attitudes of participants in the traffic and thereby the rules of
language.)
What is fundamentally important is that this kind of mathematization
always involves (smaller or greater) idealization: as the explicatum (a natural
meaning) and the explicandum (a formal one) are entities of different “orders,”

13 There is an important difference between two possible modes of such an encounter – we can

either view them “from outside” of language, so to speak, where they appear as something directly
on a par of other phenomena of the world, or “from inside” of it, where they appear rather as
“transcendental” presuppositions of our dealing with the world. But this distinction is not
important for us at the moment.
 Developing Sellars’s Semantic Legacy: Meaning as a Role 273

passing from the former to the latter always means a kind of a “bettering”: we
lead sharp boundaries where there are merely fuzzy ones, we extrapolate and
round off, and sometimes we even try to repair or upgrade.
All of this means that, in contrast to Sellars, I do not believe that the
inferential nature of meaning would prevent us from engaging in the enterprise
of Carnapian formal semantics. We only have to realize that the set-theoretical
denotations with which the systems of formal semantics work are not to be
regarded as representations of entities on which expressions of natural
language really stuck, but rather as recapitulations of the inferential roles of
the expressions (see Peregrin 2001; 2003).

Jaroslav Peregrin
Dept. of Logic, Institute of Philosophy
Academy of Sciences of the Czech Republic
Jilská 1, 110 00 Prague
Czech Republic
www.cuni.cz/~peregrin
e-mail: [email protected]

REFERENCES

Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press.

Harman, G. (1970). Sellars’s Semantics. Philosophical Review 79, 404-419.
Koslow, A. (1992). A Structuralist Theory of Logic. Cambridge: Cambridge University Press.
Lakatos, I. (1976). A Renaissance of Empiricism in the Recent Philosophy of Mathematics? In:
I. Lakatos, J. Worrall, and G. Currie (eds.), Mathematics, Science and Epistemology:
Philosophical Papers, vol. 2, pp. 24-42. Cambridge: Cambridge University Press. Reprinted
in Tymoczko (1998), pp. 29-48.
Marras, A. (1973). Sellars on Thought and Language. Nous 7, 152-163.
Peregrin, J. (1999). The Pragmatization of Semantics. In: K. Turner (ed.), The
Semantics/Pragmatics Interface from Different Points of View, pp. 419-442. North-Hollard:
Elsevier Pub.
Peregrin, J. (2000). The “Natural” and the “Formal.” Journal of Philosophical Logic 29, 75-101.
Peregrin, J. (2001). Meaning and Structure. Aldershot: Ashgate Pub.
Peregrin, J. (2003). Pragmatismo e Semantica [Pragmatism and Semantics]. In: P. Ghirardelli, Jr.
and J. Zabatiero (eds.), Revista Filosofia Americana. CEFA & FTSA. English original
available at http://jarda.peregrin.cz/mybibl/PDFTxt/482.pdf.
Peregrin, J. (forthcoming). Semantics as Based on Inference. In: J. van Bentham, G. Heinzmann,
M. Rebuschi, and H. Visser (eds.), The Age Of Alternative Logics. Dordrecht: Kluwer.
Prior, A. N. (1960/61). Roundabout Inference Ticket. Analysis 21, 38-39.
Seibt, J. (1990). Properties and Processes (A Synoptic Study of Wilfrid Sellars’s Nominalism).
Atascadero, CA: Ridgeview Pub.
274 Jaroslav Peregrin

Sellars, W. (1950). Language, Rules and Behavior. In: S. Hook (ed.), John Dewey: Philosopher of
Science and Freedom, pp. 289-315. New York: Diall Press.
Sellars, W. (1953). Inference and Meaning. Mind 62, 313-338.
Sellars, W. (1954). Some Reflections on Language Games. Philosophy of Science 21, 204-228.
Sellars, W. (1963). Abstract Entities. Review of Metaphysics 16, 627-671.
Sellars, W. (1969). Language as Thought and as Communication. Philosophy and
Phenomenological Research 29, 506-527.
Sellars, W. (1974). Meaning as Functional Classification. Synthese 27, 417-437.
Sellars, W. (1992). Science and Metaphysics. Atascadero, CA: Ridgeview Pub.
Sellars, W. (1997). Empiricism and the Philosophy of Mind (with an introduction by Richard
Rorty and a Study Guide by Robert Brandom). Cambridge, MA: Harvard University Press.
Shapiro, S. (1991). Foundations without Foundationalism. Oxford: Clarendon Press.
Tymoczko, T., ed. (1998). New Directions in the Philosophy of Mathematics. Revised edition.
Princeton, NJ: Princeton University Press.
Wittgenstein, L. (1953). Philosophical Investigations. Translated by G.E.M. Anscombe. Oxford:
Basil Blackwell Pub.
Wittgenstein, L. (1977). Vermischte Bemerkungen. Edited by G.H. von Wright. Frankfurt:
Suhrkamp. English translation: Culture and Value (Oxford: Basil Blackwell Pub., 1980).