History and Theory 51 (October 2012), 313-339

Wesleyan University 2012 ISSN: 0018-2656

The Pasts
Paul A. Roth1
ABSTRACT

This essay offers a reconfiguration of the possibility-space of positions regarding the metaphysics and epistemology associated with historical knowledge. A tradition within analytic
philosophy from Danto to Dummett attempts to answer questions about the reality of the
past on the basis of two shared assumptions. The first takes individual statements as the
relevant unit of semantic and philosophical analysis. The second presumes that variants of
realism and antirealism about the past exhaust the metaphysical options (and so shape the
epistemology as well). This essay argues that both of these assumptions should be rejected.
It develops as an alternative an irrealist account of history, a view based in part on work
by Leon Goldstein and Ian Hacking. On an irrealist view, historical claims ought to be
treated as subject to the same conditions and caveats that apply to any theory of empirical
or scientific knowledge. Irrealism argues for pasts as made and not found. The argument
emphasizes the priority of classification over perception in the order of understanding and
so verification. Because nothing a priori anchors practices of classification, no sense can be
attached to claims that some single structure must or does determine what events take place
in human history. Irrealism denies to realism the very intelligibility of any imagined view
from nowhere, that is, a determinately configured past subsisting sub specie aeternitatis.
A plurality of pasts exists because constituting a past always depends to some degree on
socially mediated negotiations of a fit between descriptions and experience.
Keywords: historical realism, antirealism, irrealism, Danto, Goldstein, Hacking, Mink
We choose our past in the same way that we choose our futureHayden White2

In his justly celebrated analysis of narrative sentences, Arthur Danto establishes

that later events inform on earlier ones so as to add to the list of statements true at
the earlier time t under descriptions not available at t. Narrative sentences establish that the list of statements true of what happens at t does not close at t. Danto
characterizes narrative sentences as follows: Narrative sentences refer to at least
two time-separated events, and describe the earlier event. To cite one of Dantos
1. Many individuals and many audiences have generously shared their comments with me on
various versions of this essay. I benefited greatly from lively discussions when I presented this paper
to audiences in 2010 and 2011 at the University of Lausanne, Etvs Lornd University (ELTE),
the University of Hradec Kralove, Catholic University (Ruomberok, Slovakia), University College,
Cork, and the Society for the Philosophy of History. I would particularly like to thank Kevin Cahill,
Larry Davis, David Hoy, Jon McGinnis, John Zammito, and Eugen Zelenak. Each kindly took the
time to read a full draft of this work at different stages in its evolution. I alone bear responsibility for
remaining errors and infelicities.
2. Hayden White, The Burden of History, History and Theory 5, no. 2 (1966), 123.

314

paul a. roth

canonical examples, The Thirty Years War began in 1618.3 Indeed, not all the
statements true of a time t will be knowable or even statable at t.4 Danto thus
demonstrates a surprising fact about the semantics of such narrative sentences,
namely, they reveal past times as dynamic and not static at least with regard to
an ongoing accretion of truths about happenings at t.
Yet Dantos analysis of narrative sentences might appear to leave undisturbed
a broader commitment to a type of realist metaphysics with regard to past states.
The imagined realist holds that although new descriptions of the past may later
become available, there can exist exactly one immutably real past. From the
standpoint of subsequent times, all statements about the past, even if they are
evidence-transcendent, have a fixed truth-value. The past so conceived must
be perfectly staticnothing can change. Otherwise, truth-values would not be
timeless.5
Debate within analytic philosophy post-Danto takes the two metaphysical
options regarding the reality of the past to be realism and antirealism. Realism
makes historical knowledge investigation independent, that is, that what
judgments are correct in particular circumstances is something determined quite
independently of human reaction to those circumstances.6 As Crispin Wright
observes, But this natural thought is simply tantamount to the assumption that
the passage of time should have no part to play in determining our conception
of what states of affairs may coherently be conceived as possible. . . . And this
assumption, of course, is here at issue.7 Antirealism takes more seriously the
3. Arthur Danto, Narrative Sentences, History and Theory 2, no. 2 (1962), 161/159; example
at 155/152. This article appears, with very slight (stylistic only, so far as I can discern) changes, as
chapter 8 of his Analytical Philosophy of History (New York: Columbia University Press, 1964). Analytical Philosophy of History in turn has been reprinted as part of Dantos Narration and Knowledge
(New York: Columbia University Press, 1985). The reprint preserves the pagination of the original
book. Page references are to the original article, followed by the corresponding book page(s).
4. Consider another of Dantos examples, Aristarchus anticipated in 270 B.C. the theory which
Copernicus published in 1543 A.D.a (158/156). In this case, what Aristarchus did in a very definite
sense caused Copernicus to re-discover the heliocentric theory (159/156).
5. In lectures, I offer a characterization of historical realism as the woolly mammoth view of
the past. This stems from an article that I once read concerning how explorers in some Arctic region
found an entire woolly mammoth frozen, embedded in the ice. Realists view past events on analogy
with such a discovery. As past, events become forever locked into some fixed configuration, awaiting
a historian to come along and chip away the excrescences of time so that the past can stand revealed
in all of its original glory.
Danto too uses such imagery to attack the conception of the past as static. But anyway, there,
in the Past, are situated all the events which ever have happened, like frozen tableaux (Danto, Narrative Sentences, 151/148; see generally 146-151/143-148). Analogously, see Hayden White, The
Politics of Contemporary Philosophy of History, in The Fiction of Narrative: Essays on History,
Literature, and Theory 19572007, edited with an introduction by Robert Doran (Baltimore: Johns
Hopkins University Press, 2010), 136, and Richard T. Vann, Louis Minks Linguistic Turn, History
and Theory 26, no. 1 (1987), esp. 12-13.
6. Crispin Wright, Strict Finitism, in Crispin Wright, Realism, Meaning, and Truth, 2nd ed.
(Cambridge, MA: Blackwell, 1993), 148-149.
7. Wright, Anti-Realism, Timeless Truth and 1984, in Wright, Realism, Meaning, and Truth,
187. Wright takes as an implication of the realist view the claim that the objectivity of truth and the
objectivity of judgment about history cannot come apart (Introduction, in Wright, Realism, Meaning, & Truth, 7). As cases where the two do come apart Wright suggests Hume on causation and, more
generally, inferences from facts to states of affairs that have a clear normative component, as in moral
and aesthetic judgments. Hence, one fundamental philosophical question that lurks here concerns

the pasts

315

fact that we are immersed in time; being so immersed, we cannot frame any
description of the world as it would appear to one who was not in time, but we
can only describe it as it is, i.e., as it is now.8 Each position captures opposed
intuitions about the past; each offends against intuition in its own way. The problem here lies in the implication that variants of realism and antirealism about the
past constitute the only metaphysical options, and so one has to make a forced
choice between them.
But much rides on this metaphysical debate. For one, it subserves a theory of
understanding, that is, an account of how sentences (including tensed ones) could
be learned and shared. Epistemology and logic have traditionally featured in this
debate only insofar as certain types of statements represent a canonical form of
verification (for example, perception and implication). Implication flows on current views from observational statements taken as semantic atoms to statements
about unobservables. Dilemmas generated by current metaphysical debates about
the reality of the past reflect, on my view, a misplaced emphasis on the nature of
canonical verification, a notion shared by realists and antirealists alike. A fundamental aspect of my critique focuses on the assumed canonicity of observational
statements as a prototype of knowledge and associated views regarding the type
of logic needed to account for how language functions.9
The philosophical critique of current metaphysical views about historical
reality that animates this essay emphasizes instead how holism and naturalism
reconfigure the issues regarding the epistemology and metaphysics associated
with historical knowledge. The reality of the past, I argue, proves to be no
more problematic than our account of any other aspect of reality, and so historical claims ought to be treated as subject to the same conditions and caveats that
apply to any theory of empirical knowledge.10 Empirical knowledge, in turn, on
the view defended here requires some general beliefs about the worlda theory
in an extended sense of that termin order for anything to emerge as an event
from the flux of experience.11
where or whether a line gets drawn between the factual/descriptive and the normative, that is, what
makes for the supposed difference in types of judgments about the past.
8. Michael Dummett, The Reality of the Past, in Michael Dummett, Truth and Other Enigmas
(Cambridge, MA: Harvard University Press, 1978), 369.
9. Michael Williams terms the epistemological position that I find shared by realism and antirealism alike epistemological realism. A characteristic of epistemological realism involves a view of
a type of natural ordering of justification between basic and nonbasic beliefs. Williams rejects epistemological realism and its associated metaphysics; he has advanced his views in a number of books
and papers. For a representative statement of the position, see Michael Williams, The Problems of
Knowledge (New York: Oxford University Press, 2001), 170-172.
10. The past as it actually was is not open to our observation, and there is no reason to think that
any remains we now have of it constitute in themselves what might be termed unvarnished transcripts
of past reality. Historical conclusions must accord with the evidence; but evidence, too, is not something that is fixed, finished, and uncontroversial in its meaning and implications. Evidence has to
be authenticated, and again evidence has to be assessed (W. H. Walsh, Truth and Fact in History
Reconsidered, History and Theory, Beiheft 16 [1977], 54). Likewise, Danto remarks, Not to have a
criterion for picking out some happenings as relevant and others as irrelevant is simply not to be in a
position to write history at all (Danto, Narrative Sentences, 167/167).
11. I argue this point in Narrative Explanation: The Case of History, History and Theory 27, no.
1 (1988), 1-13. Indeed, an alternative to an ahistorical notion of observation can in fact be gleaned
from Dummetts writings. For Dummett notes that nominalism, while usually contrasted with real-

316

paul a. roth

Most important, insofar as a primary motivation for exploring the existing

metaphysical options involves determining which offers a workable basis for an
explanatory account of history, neither does. Rather, I use Dantos early insight to
motivate irrealism about the past.12 On my revised view nothing answers to realisms The Past. Yet for its part, antirealism fails to take seriously the challenges
of historical reconstruction. Irrealism results by acknowledging that our own history must play an important and ineliminable role in determining our conception
of what states of affairs may coherently be conceived as possible. Irrealism as I
develop it also implies that how earlier and later times may influence one another
remains at least partially indeterminate. Indeed, a coherent account of why our
future remains undetermined at least in some respects also presumes a past that
remains open.13
Against, then, a tradition within analytic philosophy from Danto to Dummett
that focuses on the analysis of statements, I develop an alternative account of
historical knowledge owing to Leon Goldstein and Ian Hacking. Goldsteins
account of historical knowing utilizes important but overlooked forms of holism
and nominalism. This position has the advantage of taking seriously issues arising from historiographical, scientific, and epistemological considerations of
knowledge about the past.14
ism, does not entail the specific form of antirealism he considers: a nominalist does not seem to be
committed to being an anti-realist in this sense (Dummett, Realism, in Truth and Other Enigmas,
147). The context of Dummetts discussion here proves important. For Dummett notes that nominalism normally would be understood as the philosophical antithesis of realism. But although antithetical
to realism as usually understood, nominalism does not equate to antirealism as Dummett discusses it.
Dummett has a clear antipathy for Goodmans form of nominalism discussed below, but the reasons
for that need not detain us.
12. Richard Rorty remarked to me in conversation that he did not see any difference between
past and present on this point. I agree, but the arguments in this essay concern just those about the
reality of the past. Although Dantos text might sometimes seem to suggest otherwise, he intends no
antirealist or irrealist conclusions. See references provided to correspondence with Danto in David
Weberman, The Nonfixity of the Historical Past, Review of Metaphysics 50, no. 4 (June 1997), 759,
n. 21. See also 751, n. 6.
13. For a defense of a similar position, see G. H. Mead, The Philosophy of the Present (Amherst,
NY: Prometheus Books, 2002), esp. 35-59, The Present as the Locus of Reality. Danto also hints at
this. See Narration and Knowledge, 340-341.
Two other accounts bearing apparent similarities to the one I defend are Weberman, The NonFixity of the Historical Past, and Martin Bunzl, How to Change the Unchanging Past, Clio 25, no.
2 (Winter 1996), 181-193. But both tie their accounts to problematic aspects of particular theories of
action. Weberman invokes the notion of a skeletal event, a sort of minimalist notion of an event that
can be descriptively thickened over time. Yet, he concedes, to make this distinction precise would
require criteria stipulating just what counts as physicalistic and just how thin a description must be to
qualify an event as part of the skeletal past. It is doubtful that such criteria can be found. . . . We might
think of the skeletal past as a sort of artificial or unrealizable limit (Weberman, The Non-Fixity
of the Historical Past, 754). Bunzl attempts to underwrite his account by distinguishing between
events and facts. One of the aims of this paper has been to make sense of Dantos claim that
the past changes while staying true to the intuition that (in some sense) the past is unchanging. The
paradox is solved by distinguishing between events and facts (Bunzl, How to Change the Unchanging Past, 192). I find the distinction as Bunzl draws it highly problematicLet a historical fact be
statable by a proposition that is true only at a particular location or set of locations in spacetime. . . .
The need for the notion of an event drops out of consideration, and we can still have historical subjects
(the Second World War), now understood as sets of facts which hold at different times and places
(192)and in any case Bunzls notion of a fact presumes a realist semantics I reject.
14. Goldsteins work has received little critical attention. For an appreciative exception, see Luke
OSullivan, Leon Goldstein and the Epistemology of Historical Knowing, History and Theory 45,

the pasts

317

However, Goldsteins formulation suffers because it permits an unreasonable

proliferation of historical knowledge. If a historian constitutes a past, how
could a past so constituted fail to represent? What would it be, in other words,
for an act of historical constitution to go wrong? In order to separate Goldsteins
valuable account of historical constitution from this untoward implication
requires developing an aspect of his position at which Goldstein himself only
hints. This aspect of his position I link to Ian Hackings innovative application
of Nelson Goodmans irrealism to historical analysis. In particular, Hacking
develops a Danto-like thesis that past actions may be indeterminate, at least
in the following way: with new forms of description, new kinds of intentional
action came into being, intentional actions that were not open to an agent lacking
something like those descriptions.15 I then explore implications of irrealism for
the epistemology, metaphysics, and explanation of historical events.
I. Historical Constitution and Historical Knowing

Leon Goldstein defines and develops his signature doctrines of historical constitution and historical knowing by contrasting them to a doctrine he labels historical realism. By historical realism I mean that point of view according to which
the real past as it was when it was being lived is the touchstone against which
to test for truth or falsity the products of historical constitution.16 Realism as
Goldstein opposes it treats the historical past on the model of the experienced
present; it is an extension of our everyday attitudes to the world of past events.17
But realism so conceived Goldstein terms an absurdity,18 a doctrine utterly
false19 to those processes that make historical knowledge possible and a subject
of rational evaluation.
Goldstein finds historical realism operating more as an unquestioned assumption in writings about history than as a doctrine explicitly advocated. Historical
realism is a habit of mindnot a refined doctrinewhich inclines those possessed of it simply to assume that the conceptions of factuality, truth, or reference
no. 2 (2006), 204-228. An obstacle to engaging with Goldsteins work arises from his allegiance to
philosophically problematic doctrines. These obscure the general interest of some of his arguments.
For example, he invokes a distinction between knowledge by acquaintance and other forms of knowledge acquisition in order to distinguish historical knowledge from other forms. Yet once relieved of
such quasi-Russellian views of knowledge (perceptual or scientific), Goldsteins account of the constitution of the historical past emerges in a very different and philosophically more illuminating light.
15. Ian Hacking, Rewriting the Soul (Princeton: Princeton University Press, 1995), 237. Danto also
made this point. One conclusion warranted, I believe, by my discussion of narrative sentences, is that
frequently and almost typically, the actions of men are not intentional under those descriptions given
of them by means of narrative sentences (Danto, Narration and Knowledge, 182). Both Danto and
Hacking acknowledge the influence here of G. E. M. Anscombes Intention (Oxford: Oxford University Press, 1957). See, for example, Danto, Narration and Knowledge, 154, n. 14/151, n. 1; Hacking,
Making up People, in Reconstructing Individualism, ed. T. C. Heller, M. Sosna, and D. E. Wellbery
(Stanford: Stanford University Press, 1986), 230. I further note Hackings debt to Anscombe in my
Ways of Pastmaking, History of the Human Sciences 15, no. 4 (November 2002), 125-143.
16. Leon Goldstein, Historical Knowing (Austin: University of Texas Press, 1976), xxii.
17. Ibid., 38. Realists view the past as a place that one just happens now not to be. See, for
example, Alfred Jules Ayer, Language, Truth and Logic (New York: Dover Publications, 1952), 9.
18. Leon Goldstein, The What and the Why of History (New York: Brill, 1996), 154.
19. Goldstein, Historical Knowing, xxii.

318

paul a. roth

which apply when we speak of the natural world in the natural present must apply
when we speak of the historical past, that, indeed, they must apply to any realm of
discourse to which considerations of truth and falsity obtain.20 Goldstein protests
throughout his writings against conceiving of the past as an independently subsisting touchstone, as something fixed and therefore prior to historical research
that true histories represent.21
Writing in a similar vein and at about the same time of the publication of Goldsteins book (1976, 1978), Louis Mink also focuses on the plausibility lent to the
idea of a historical realismthe past as an untold storyby the assumption
that the past, as past, was fixed, immutable, not open to change. Yet the idea that
the past itself is an untold story has retreated from the arena of conscious belief
and controversy to habituate itself as a presupposition in the area of our a priori
conceptual framework which resists explicit statement and examination. . . . [W]e
assume that everything that has happened belongs to a single and determinate realm
of unchanging actuality. (Whats done is done. You cant change the past.)22
Indeed, the core of historical realism consists of the belief that histories are found,
not made. What determines the truth-value of statements about the past does not
depend on available evidence or human judgment.23
Although he claims to find everywhere thinkers implicitly assuming historical
realism, Goldstein remarks that he knows of no attempt to explicate and defend
historical realism.24 I do. William Dray, for one, insists on this view. For Dray,
the past so conceived constitutes a type of permanent possibility of narrationa
tellable.25 Dray succinctly put the case against antirealism, a view he attributes
20. Ibid., xxiv; see also Goldstein, The What and the Why of History, 243.
21. See, for example, Goldstein, The What and the Why of History, 243. Compare to Dummetts
notion of realism.
22. Louis Mink, Narrative Form as a Cognitive Instrument, in Historical Understanding, ed.
Brian Fay, Eugene O. Golob, and Richard T. Vann (Ithaca, NY: Cornell University Press, 1987), 194.
23. For if the past were not fixed, realism would collapse into the position Dummett identifies as
a limited antirealism about the past. Sentences would be true or false relative to a possible model of
the past, and not true or false absolutely, as realism presumably requires. See Dummett, The Reality
of the Past, 367. In a set of articles published subsequent to his book, Goldstein directly confronts a
version of the sort of realism he disavows. For P. H. Nowell-Smith maintains that a plausible realist
is committed only to the thesis that if a historian states truly that such and such happened, it happened whether or not anyone later found out that it happened or proved by constructionist methods
that it must have happened. . . . The less extreme realist thesis is not limited to events of the observable
kind. If Schneider really showed that the urban oligarchy of Metz was transformed into a landowning aristocracy in the period from 1219 to 1324, then this transformation is something that actually
happened, a slice of the real past, even though it was not, when it occurred, something which anyone
could have observed or with which anyone could have been acquainted . . . [T]he less extreme realist holds that the historian constructs an account of the real pastthe only past there wasand that
the real past plays the important role of being that to which historical statements, when true, refer (P.
H. Nowell-Smith, The Constructionist Theory of History, History and Theory, Beiheft 16 [1977],
7). This less extreme realist view coincides with Nowell-Smiths own. See replies to Nowell-Smith
by W. H. Walsh, Truth and Fact in History Reconsidered, and by Leon J. Goldstein, History and
the Primacy of Knowing, History and Theory, Beiheft 16 (1977), 29-52.
24. Goldstein, Historical Knowing, 38.
25. But the comparison between untold stories and unknown knowledge seems to me misleading.
. . . A better parallel would be between untold stories and unstated facts or undiscovered explanations.
. . . It might be preferable, therefore, although in most contexts it would be an unnecessarily technical
way of putting it, to speak of there being unknown narrativizable configurationstellablesalready
there for the discovering. That, at any rate, is all that need be meant, and all that would generally be

the pasts

319

to thinkers as otherwise diverse as Goldstein, Mink, and Hayden White, as follows:

The separation [by White and Mink] of historical discovery from the aesthetic or moral
task of writing up what has been discovered in narrative form is based on a simple but
serious error: it implicitly, but falsely, denies that part of what the historian discovers is the
configuration the narrative displays. . . . But the form, the configuration, is itself the most
important fact that historians discover. And facts can exist unknown.26

Now although Goldstein does not subscribe to any view that would make narrative form a defining feature of historical representation, Drays realist point
requires no such commitment. Note just that Dray straightforwardly asserts what
Goldstein denies, namely, that historians discover the past, a configuration that
exists prior to any activity of historical inquiry, a tellable as a fact unknown.
Yet just what metaphysical status could such historical events have? On what
basis could one hope to say that events qua kinds of human activities are found,
not made? In speaking of the sort of events relevant for historical analysis, I
focus on events characterized as intentional or purposive actions. Such events
behavior characterized under a description of a certain sortprove central to
Goldsteins account of what historical knowing constitutes as the historical past.
An event emerges from all that remains available because some elements can
be imagined as instantiating a purpose.
Goldsteins thesis of historical constitution invokes the methods specific to
historical theorizing as simultaneously constitutive of the object of historical
knowledge.27 By historical constitution I mean that set of intellectual procedures
whereby the historical past is reconstructed in the course of historical research.28
An eventful historical past exists only as a result of human theorizing. History
becomes an artifact of a disciplined disciplinary imagination.29
meant, by the claim that there are untold stories in the past (William Dray, Narrative and Historical
Realism, in Dray, On History and Philosophers of History [New York: E. J. Brill, 1989], 162). For
other examples and informed discussions of realism and antirealism in the philosophy of history, see,
in particular, Andrew Norman, Telling it Like it Was: Historical Narratives on Their Own Terms
or Chris Lorenz, Historical Knowledge and Historical Reality: A Plea for Internal Realism, both
reprinted in History and Theory: Contemporary Readings, ed. Brian Fay, Philip Pomper, and Richard
T. Vann (Oxford: Blackwell, 1998). I critique specific accounts of historical realism in The Object
of Understanding, in Empathy and Agency, ed. H. H. Kgler and K. R. Stueber (Boulder, CO: Westview, 2000), 243-269.
26. Dray, On History and Philosophers of History, 162-163. Jonathan Gorman defends a view
that holds that historians must be committed to agreement on what Gorman terms atomic statements (of fact) but not on their configuration in Drays sense. See J. Gorman, Historical Judgement
(Stocksfield, UK: Acumen Publishing Ltd., 2007). For my assessment of Gormans views, see Notre
Dame Philosophical Reviews (August 2008) http://ndpr.nd.edu/news/23690-historical-judgementthe-limits-of-historiographical-choice/ (accessed February 23, 2012).
27. For reasons that will be obvious as the discussion proceeds, Goldsteins doctrine should not be
identified with views that receive the label social constructionist or constructivist.
28. Goldstein, Historical Knowing, xxi-xxii.
29. But there is no gain-saying the fact that we have no access to the historical past except through
its constitution in historical research. Realists may seem to have some arguments against the claim
that the objects of the external world are constituted by consciousness; it is by no means unintelligible
that there are objects independent of consciousness which provide the touchstones to which our conceptions of things much conform. But no past of history exists in that realistic sense (ibid., xxi; see
also Goldstein, The What and the Why of History, 161). I argue below that what holds for history so
conceived holds, a fortiori, for ordinary understanding of the past (or the present).

320

paul a. roth

Goldsteins formulations underline a point noted from Danto, namely, that

historians characterize events at a timefor example, pre-Columbian Nordic
excursionunder descriptions not available at that time although now true of
that time. What historians constitute when constituting a past might be thought
of as a paradigm of a past resulting from present tracesan account that offers
problem-solving potential with regard to what the traces trace.
The claim no constitution, no reference insists that only in the context of a
theory do historical questions have a meaning.30 For example, questions regarding the Dead Sea scrolls or pre-Columbian Nordic excursions in North America
can be asked only within a prior context that provides these phrases with their
meaning.
It seems clear that everything that we can come to say about the historical past emerges
entirely within the framework of historical knowing. Every attempt to subject to verificational test the claims that historians make requires that the procedures which led to the
claims in the first place be repeated. There seems to be no way to the referent of a historical assertion except by means of the procedures of historical constitution themselves.31

The realtruth-makers for statements about the pastemerges from within a

constituted past. Items appear as candidate truth-makers by virtue of their location within a constituted framework. Goldsteins historian, speculating on the
origins of the Dead Sea scrolls or pre-Columbian Nordic excursions into North
America, shapes explanatory events. In sum, the relation of the historical occurrence to the evidence upon which it is based is not one of logical entailment of
the occurrence from the evidence, but the occurrence is offered hypothetically as
what would best make sense of the evidence.32 Historical events emerge abductively, as part of an inference to the best explanation.33
30. The position sketched here has important points of overlap with Lorenz, Historical
Knowledge and Historical Reality.
31. Goldstein, The What and the Why of History, 168-169.
32. Goldstein, Historical Knowing, 127. The relevant sense of best here concerns whatever the
desiderata happen to be for a scientific explanation. In retrospect, early (and neglected) works by
Murray G. Murphey (Our Knowledge of the Historical Past [Indianapolis, IN: Bobbs-Merrill, 1973])
as well as those by Goldstein mark very early attempts to view philosophy of history through the
lenses of an emerging holistic (but still analytical) philosophy of science. Murpheys book develops
a view of history as a type of theory about the past in which people and events have the status of
posits used for purposes of organizing experience. But Goldsteins work in particular stands out as
advocating the position that historians constitute out of whole cloth the events of historical interest.
However, these contributions were largely overlooked. Analytic philosophy of science, identified as
it then was with Hempel, simply comes to be written off as irrelevant to historical practice. In addition, the near simultaneous publication of Hayden Whites theoretical masterwork (Metahistory: The
Historical Imagination in Nineteenth-century Europe [Baltimore: Johns Hopkins University Press,
1973]) and the emergence of the Foucauldian paradigm for (re)doing history effectively swamped any
influence that analytic philosophy of science might have hoped to exercise. For an elegiac assessment
of the fate of analytic philosophy of history, see Arthur Danto, The Decline and Fall of the Analytical Philosophy of History, in A New Philosophy of History, ed. Frank Ankersmit and Hans Kellner
(Chicago: University of Chicago Press, 1995), 70-85.
33. The view of historical constitution, in Goldsteins sense, as abductive inference finds confirmation in many remarks by Goldstein. See, in particular, The What and the Why of History, for example,
216, 221, 225, 334-335. Goldstein distinguishes between explanations of evidenceexplainingwhatand explanations of eventsexplaining-why. Historical constitution concerns explainingwhat, not explaining-why.

the pasts

321

Again, the inert and fragmentary remains of a second-century construction can

be characterized under multiple descriptions. Even to know that it was built during the reign and, presumably, upon the instruction of Antoninus Pius is not really
to know what it is. . . . To know what it is as something historical is to know what
purpose it served, what thoughtspoliciesit embodies.34 In order for remains
(traces of the past) to be evidence for something, they must be categorized in a
certain way. Categorization will often require attribution of a certain purpose.
The assumed purpose configures artifacts as instantiating a kind or an event.
Similarly, the notion of a career represents a constituted category. What makes
for a career, and where is it located? Pressing these questions, Goldstein maintains, reveals once again the deeply problematic assumptions made by historical
realism.
The absurdity [of realism] emerges from the view that the events of human history are
located in the past. It depends on taking literally the metaphor of temporal location. . . .
Franklin-Roosevelt-being-elected-in-1932 occupies one span of time; Franklin-Rooseveltbeing-inaugurated-in-1933, another span of time; and so on until Franklin-Rooseveltdying-in-1945, which occupies still another. . . . In my view, the unity of all these
disparate Franklin Roosevelts is simply a consequence of the fact that it is one career
which emerges from the attempt of historians to deal with the relevant period of American
history. In fact, there is no problem of unity. . . . What is closer to the truth is that they
constitute a course of events or the course of a life. The continuities are built into the
historical constitution itself.35

Roosevelts career does not exist until constituted by a historian. The grouping
represents an artifact, a colligation by historians studying a particular person or
period.
Goldstein registers an appreciation of the seemingly ineliminable tension
between a sort of common-sense realism about the past as opposed to the ways
of pastmaking that his own account of historical knowing allows. He quotes
with hearty approval, in this regard, the following remark by G. H. Mead:
[T]he estimate and import of all histories lies in the interpretation and control of the
present; that as ideational structures they always arise from change, which is as essential
a part of reality as the permanent, and the problems which change entails; and that the
metaphysical demand for a set of events which is unalterably there in an irrevocable past,
to which these histories seek a constantly approaching agreement, comes back to motives
other than those at work in the most exact scientific research.36

Goldstein then remarks, in keeping with what I earlier termed his conventionalist
view of knowledge, that the quotation from Mead makes its point with respect
to the past, but his point is quite general. Any attempt to take ones stand on reals
which are alleged to be independent of inquiry is motivated by commitments
which are independent of the systematic quest for knowledge.37 He goes on to
34. Goldstein, The What and the Why of History, 319-320. Although he speaks here of Collingwood, I take Goldstein to be expressing his own view as well.
35. Ibid., 154. Danto also suggests and rejects the notion of the past as a container of prefigured
events. Danto, Narration and Knowledge, 149/146.
36. Goldstein, The What and the Why of History, 245 (quoting Mead, The Philosophy of the Present, 28).
37. Ibid., 246.

322

paul a. roth

reflect on what his position implies with regard to, for example, conflicting interpretations of the Holocaust (a favorite bogeyman for all those who view failure
to subscribe to historical realism as tantamount to a moral failing). Although he
recognizes and sympathizes with the realist desire to have a metaphysical club
with which to beat down revisionists and others, he observes that however worthy such a goal, in the end it cannot be realized. . . . The only past we can talk
about is the past as it is known to us.38
The contrast here of Goldsteins view and Dummetts reflections on realism
and antirealism proves instructive. All the metaphysical options as Dummett
conceives of them assume observation-like sentences as a model of verification.
All statements about nonobservables, including statements about the past, build
inferentially on these. But Goldsteins considerations bring to the fore how radically nave and inappropriate this model turns out to be for statements about the
past, and especially ones at any significant historical remove. Goldsteins discussion has then the virtue of highlighting what sorts of inferential practices actually
come into play in constituting the past. Better actual practices than philosophical
fictions to the same effect.39
The perspective from which a historian makes statements such as those found
in histories typically does not consist of a perspective that could have been had by
any observer at that time. Even if what a historian reports appears to be a matter
of fact, no observer at that time could likely have described the event in that way.
The structure given to time and memory reflects not mere strings of observations,
but a significance that emerges regarding what happened when viewed looking
backwards.40 The logic used to constitute the past resembles not a recursive
structure built on observation. Rather, Goldstein emphasizes the prominence of
abductive inferenceinference to the best explanation.41 Relatedly, appreciating
that observability itself becomes identified intratheoretically, Goldsteins account
of the interpretive element in the constitution of evidence ceases to mark history
off from other forms of inquiry.42 Further issues regarding evidence and inference
38. Ibid., 252.
39. One might easily miss what remains of live interest in Goldsteins philosophy just because the
account of confirmation on which he relies ignores the holist constraint on which he otherwise insists.
40. As Danto puts this point, narrative structures penetrate our consciousness of events in ways
parallel to those in which, in Hansons view, theories penetrate observations in science (Narration
and Knowledge, xii). The key point here bears primarily on what might be termed the theoretical
structure of the past and the model that results. The point emphasized above involves the past as a
theoretical construction. The theory accounts, among other things, for who we take ourselves to be
and why. Any discussion of the reality of the past constituting human history must then appreciate
that the narrative determines the significance to assign observations as well as (often enough) what
was in fact observed (under what description to characterize an action).
41. See also Danto, Narration and Knowledge, 122. The general point to keep in mind involves
the fact that scientific reasoning begins as parasitic on common sense, that is, whatever passes
as received knowledge for a time and place. Science certainly refines such views and may in time
transcend and transform common sense. But reasoning begins from within some set of received
viewsQuines web of belief.
42. Goldstein throughout his writings presumes that historical knowledge cannot claim the perceptual base that scientific knowledge can. But his particular inference from a lack of knowledge by
acquaintance of historical events to the lack of relevance of perceptual knowledge (since he takes the
two notions to be equivalent) simply does not hold. What we come to believe about the human past
can never be confirmed by observationcan never be known by acquaintanceand so can never be

the pasts

323

emerge below. But for now the point of note concerns how a prior theoretical
structure determines the semantics of statements about the past.
Yet Goldsteins version of antirealism has its costs. If the activity of historical
knowing constitutes the very objects of historical knowledge independently of
perception,43 then Goldstein leaves unclear just how, on his account, any activity of historical knowing could fail to produce knowledge. Since Goldsteins
antirealist constitutes the past, how can there be any error in representation?
There seems no way for a historian to go wrong. Thus, an ironic consequence of
Goldsteins antirealism would appear to be not a lack of historical knowledge,
but its proliferation. Historical knowledge so conceived seems to be knowledge
too easily had. Knowledge proliferates because nothing on this account appears
to remain by which to drive a wedge between representations of the past and its
putative objectthe past.
Interestingly, a very different way of answering this vexed questionwhat
limits the process of historical constructioncan also be found in Goldsteins
writing. On the alternative formulation Goldstein offers, historical knowledge
stands as prototypical of empirical knowledge. All knowledge, Goldstein says
in this other mood, turns out to be constitutional in something like this broadly
naturalistic way. Knowledge becomes understood as an artifact of a particular
approach within which interrogation of nature proceeds and through which one
interprets its answers.
In this regard, nothing marks off the intratheoretical methods and practices
constitutive of what Goldstein terms historical truth or historical objectivity44
from any other form of scientific inquiry. I want my remarks to be general,
since I believe that the primacy of knowing is a generally sound epistemological stance, though I do not want to stray too far from philosophy of history.45
Historical constitution preserves bivalence just as Dummetts limited antirealism
did, by making sentences true or false relative to a model. The description is historically true not because it corresponds to an actual event as a witness may have
observed it, but rather because given the evidence in hand and the ways in which
historians deal with and think about such evidence it is reasonable to believe that
some part of the human past had such-and-such characteristics.46 This emphasis
on the role of prior beliefs and a disciplinary matrix has the great merit of bringing into high relief a feature common to both historical realism and antirealism.
Each explicitly utilizes current habits of categorization in its characterization of
past events and actions.47 Absent some magical ability to reproduce a bygone
put to the test of observation, the method of confirmation which is virtually the only one explicitly
recognized by science and philosophy (Goldstein, Historical Knowing, xii).
43. The historical way of knowing in no way involves seeing or any other of the senses. . . . He
does not have sensory experiences of the events he attempts to construct. The very point of history is
to provide knowledge of past events that cannot be had in the sensory way (ibid., 11).
44. I have tried to emphasize that while historical knowledge is relative to the discipline of
history, in the same way that any sort of knowledge is relative to the disciplined way in which it is
produced, it is not relative to the subjectivity of historians (Goldstein, The What and the Why of
History, 161).
45. Ibid., 163; see also Historical Knowing, 89-91.
46. Goldstein, The What and the Why of History, 117-118.
47. Goldstein, Historical Knowing, 38. According to the realist view, one would expect that the
evidence would be grouped together according to what it was evidence for. All evidence concerning

324

paul a. roth

Weltanschauung (with no small part of the magic residing in the belief that there
exist such determinate and shared mindsets to reproduce), what historians do
involves using resources available now in an effort to reconstruct prior patterns
of categorization.
One final consideration here concerns how realism and antirealism have been
treated as the only available options for conceiving of the reality of the past. Realism demands that all sentences about the past now have a determinate truth-value.
Dummetts global antirealism allows sentences to have a truth-value based only
on what can be known now; consequently, there will be truth-value gaps. Some
sentences about the past will be judged neither true nor false. Finally, what Dummett terms limited antirealism and Goldstein dubs historical constructivism
have truth or falsity relative to a model of historical knowledge. In this regard,
Goldsteins animus toward realism appears tempered by a type of Peircian faith
in the convergence of inquiry. That is, both Goldstein and Dummett hold out for
notions of truth that transcend relativity to a model. To the extent that each does,
each remains committed to a traditional metaphysical picture of a structured past
prior to any constitution by human categories.48
Danto, however, offers a sophisticated analysis of the problem of attaching
truth-values to sentences about the past that shows that the variants canvassed by
Dummett and Goldstein do not capture the full complexities. The complexity I
want to bring into view concerns a sense in which knowledge of the past remains
contingent, but a contingency that does not arise because of any lack of evidence
about the past. Contingency so conceived offers a counter to an antirealist view
of the past without yet being realism.
I begin by developing an example from Danto.
(1) Talleyrand begat Delacroix and Delacroix painted the Mort de
Sardanapale.49
This sentence has the following interesting logical feature: although both
its conjuncts are now true, they were not always simultaneously true. Some
years passed between the state of affairs described in the first conjunct and that
described in the second conjunct. This generates the following puzzle. Sentence
(1) is a conjunction, and so formally its truth-value should be a function of the
truth-values of its conjuncts. But the conjuncts are indexed to different times. So
depending on the time of the statement, the first conjunct may be true and the
second false. In order to capture the cases when just (1) would be true (Danto
calls this time-true), Danto maintains that the time-true version of (1) is:
the Essenes, for example, would be placed in one intellectual pile. . . . To say that the evidence is to
be grouped according to what it is evidence of is to make that which it is evidence of the criterion for
the grouping. But in point of the actual practice of history, this is not the case at all. The criteria for
the grouping are drawn entirely from those intellectual operations which are the practice of history
itself (ibid., 132; see also 131).
48. Like any intellectual enterprise, history is carried on collectively and self-correctively. . . .
[Historians may find] new ways of dealing with a further enriched body of evidence and, arriving at
what one may expect will be increasingly agreed to, historical truth (Goldstein, Historical Knowing,
90). Dummett privileges statement meaning over that of meaning-relative-to a-model, and so does not
take limited antirealism as a serious candidate for a theory of understanding. Dummett, The Reality
of the Past, 367.
49. Danto, Narration and Knowledge, 195; see generally his discussion on 193-196.

the pasts

325

(2) Talleyrand begat Delacroix and Delacroix will paint the Mort de Sardanapale.
But although (2) may be time-true, is it true? If true, (2) entails
(3) Delacroix will paint the Mort de Sardanapale.
Yet (3) is a paradigm instance of a sentence without a truth-value, since it
speaks of what will be, not what is or was. Danto takes this to show the nonequivalence of time-true and true. Dantos narrative sentencessentences
that mention events standing in a determinate relation in time but that utilize
a later event to describe the earlier (for example, Pier da Vinci begat a great
genius)will typically be analyzable as containing a time-true part. This creates
a logical puzzle. Future tense statements like (3) conjoined with any true statement should yield a statement without a truth-value, but (2) proves otherwise.
Moreover, since we take the logical relation of time-dependent events such as
those in (2) to be contingent, Danto notes that when any such compound proposition also contains a time-true, past referring, singular proposition, the entire
compound proposition [such as (2)] expresses a past contingency. So not every
time-true sentence about the past is true or false.50 This analysis of narrative sentences therefore yields time-true sentences, not sentences true or false absolutely.
Simply put, narrative sentences that have determinate truth-values relative to a
model (that is, are time-true) do not allow for the usual inferences regarding the
truth-values of their constituent statements.
Call such sentences inferentially opaque, meaning that without the relevant
model, uncertainty exists regarding whether or not the usual deductive inferences
can be applied. The source of opacity resides in what Danto terms their pastcontingency; some passage of time must be assumed in order for both conjuncts
to be true. How matters turned out for this child of Talleyrand illustrates in turn
how future events lead to a redescribing of a past event. 51
It is important to note that Danto recognizes that insofar as what happens later
leads to redescriptions of what happened earlier, changing the past can change
the present as well.
But for the rest, I think, it may be said that to the degree that our past is in doubt, our
presentthe way we live in the worldis no less in question. And indeed, our very
actions inherit these margins of incertitude, for what we do can only have the meanings
we suppose it to have if is located in a history we believe real. . . . The present is cleared
of indeterminacy only when history has had its say; but then, as we have seen, history
never completely has its say. So life is open to constant re-interpretation and assessment.52

But now tensions within Dantos position emerge full flower. On the one hand,
Dantos account of narrative sentences denies a key realist doctrine regarding
all sentences having their truth-values timelessly. But, on the other hand, he
endorses the arch-realist doctrine that true sentences are, in Wrights earlier
50. Ibid., 196.
51. In effect, so far as the future is open, the past is so as well; and insofar as we cannot tell what
events will someday be seen as connected with the past, the past is always going to be differently
described (ibid., 340; see also 196). This echoes remarks already found in Narrative Sentences, for
example, The Past doesnt change, perhaps, but our manner of organizing it does (Danto, Narrative
Sentences, 167/166-167). See also the final paragraph of that piece.
52. Danto, Narration and Knowledge, 341.

326

paul a. roth

quoted phrase, investigation-independent.53 He does this because, like Dummett, McDowell, and Wright, he perceives a deep link among meaning, truth, and
logical structure.54 But Dantos own analysis of narrative sentences indicates how
this very structure fails as an analysis of statements about the past.55
More generally, what Goldstein and Danto show each in his own way is that
questions of the reality of the past turn out to be anything but investigation-independent. Goldstein highlights the roles of prior theoretical beliefs and abductive
inference. Danto demonstrates that narrative sentences will generate the type of
truth-value gaps in statements about the past that Dummett takes as a hallmark
of an intuitionist approach to understanding and a type of antirealism about the
past. But these logical problems turn out to connect to a yet more general logical
problem regarding the constitution of kinds, events, and intentional actions. Ian
Hacking develops and exploits this.
II. From Constructivism to Irrealism: The Case of Ian Hacking

Keep in mind that metaphysical issues remain tied to questions of how language
can be learned and sharedwhat Dummett terms a theory of understanding.
Although I have no such theory to propose, my arguments do show that whatever
logic drives such a theory needs to provide an account of how humans agree in
judgment with regard to language use, in particular categorization. Emphasis on
so-called canonical forms of verification presupposes agreement in judgment
with regard to categorization rather than explaining it. By complicating any
account of agreement through discussions of the reality of the past, part of the
goal here concerns showing how much more interpersonal coordination a theory
of understanding involves than is usually acknowledged.
I propose to approach puzzles about how to assign truth-values to statements
about the reality of the past as yet another instance of world-making by kindmaking, that is, as exemplifying Nelson Goodmans new riddle of induction.
Pioneering and innovative applications of Goodmans work to the constitution of
historical events by Hacking help transform questions about what events occurred
historically into ones regarding the projectibility of variant modes of categorization and so offers a novel and insightful way to assess inferential practices, even
abductive ones.
Goodman asks after projectible predicateswhich descriptors or categorizers can reasonably be extended to unobserved or unknown cases. Famously,
53. Their [sentences] being true is not a further bit of description, in virtue of which the reality
described has a special property in addition to those it is described as having. And so, when something
satisfies the truth-conditions of a sentence, there is not some further thing it needs to do to make
the sentence true: being true is not a further truth-condition of the true sentence (ibid., 318). More
generally, Danto fails to notice how uneasily his own account of narrative sentences fits with this
investigation-free notion of truth. For a more extended critique of Dantos notion of evidence and
truth along these lines, see discussion by Louis Mink, Philosophical Analysis and Historical Understanding, in Fay, Golob, and Vann, eds., Historical Understanding, 139-141.
54. Danto, Narration and Knowledge, 320-321.
55. Dummett takes an analysis of tensed statements to be the litmus test for determining whether
an account can function successfully for purposes of a theory of understanding (Michael Dummett,
Truth and the Past [New York: Columbia Universit Press, 2004], 44-46).

the pasts

327

Goodmans new riddle of induction poses a challenge regarding how to logically discriminate between inductive inferences that make appropriate use of
evidence to assimilate new cases to prior classifications. Nothing about the riddle
as formulated makes direction in time a logically critical feature.56 As Hacking
puts the point,
To use a name for any kind is to be willing to make generalizations and form expectations about individuals of that kind. To use a name for any kind is also, of course, to be
prepared to distinguish, to sort, to classify according as things are, or are not, of that kind.
Goodmans riddle arises in full force when we separate classifying from generalizing, and
think of classifying first, and inducing later.57

Application of categories, in this regard, may extend back in time, and, with
regard to inferences about past actions and events, typically does. Questions
about the reality of things past become a subset of a more generic and familiar
logical problem.58 For judgments about the real presuppose a prior categorical
structure on the basis of which one could speak of perceptual verification.
I suggest that Goldsteins historical knowing or constructivism be understood as just a type of Goodmanian exercise, of organizing traces into kinds.59
Goldstein read as a Goodmanian recognizes that criteria for the grouping are
drawn entirely from those intellectual operations which are the practice of history
itself.60 This point applies quite generally. What is the case for historical knowing as a type of constituting extends to all forms of knowledge. What counts as
evidence, and for what it counts, turns out to be a product of practices of inquiry
as informed by the use of predicates (past or present). Training, feedback, and
56. As Goodman states, the problem of prediction from past to future cases is but a narrower
version of the problem of projecting from any set of cases to others (Nelson Goodman, Fact, Fiction, and Forecast, 3rd ed. [Indianapolis, IN: Hackett, 1979], 83). Also, The problem of the validity
of judgments about future or unknown cases arises . . . because such judgments are neither reports
of experience nor logical consequences of it. Predictions, of course, pertain to what has not yet been
observed (ibid., 59).
57. Ian Hacking, Entrenchment, in Grue!: The New Riddle of Induction, ed. Douglas Stalker
(Chicago: Open Court, 1994), 215.
58. Put another way, the difference between making sense of the past differs from the task of
making sense of the here and now only as a matter of degree, not of kind. The primacy of knowing is not a condition peculiar to history; it has relevance to the world of science and even to that of
perception. We think we confront independent reality in perception, but the truth is that, as Goldstein
puts it in one place, there is more to being a delphinium or a garden than the brutely given. To be
either is to realize a concept, and that immediately plunges us into a context of knowing. Walsh,
Truth and Fact in History Reconsidered, 61. The reference to Goldstein is from History and the
Primacy of Knowing, 40, n. 9.
59. Again, Danto anticipates this point. In one sense, if we knew all of a mans behaviour during
a certain interval, we would know everything he was doing. In another sense, however, we should
have only the raw materials for knowing what he was doing. In the one sense, the I.C. tells us everything we want to know, in another sense it doesnt. Not to have the use of project verbs is to lack the
linguistic wherewithal for organizing the various statements of the I.C., but more importantly, for the
I.C. to lack the use of project words is to render it incapable of describing what men are doingand
so disqualifies it from setting down whatever happens, as it happens, the way it happens (Danto,
Narration and Knowledge, 163/162). I take Goldstein to be making essentially the same point. What
happens to be perceptually available does not suffice to inform by direct perception, absent some
classificatory work, what actions occur or items exist.
60. Goldstein, Historical Knowing, 133.

328

paul a. roth

group reinforcement anchor words to the world.61 The features that Goldstein
identifies as central to historical knowing turn out to be generic features of what
passes for empirical knowledge.
This allows Goldsteins account of historical knowing to connect with considerations that Hackings work brings into high relief. Hacking emphasizes two
basic points. Hacking notes, first, that categorization precedes induction and is
required for it. My own way of stating what we learned from Goodmans new
riddle of inductionfor it was in that context that he coined his neologism projectibleis that to use a name for any kind is (among other things) to be willing
to make generalisations and form expectations about individuals of that kind.62
Essences do not account for entrenchment. Essence names a kind of theoretical entrenchment. This leads to Hackings second point: projectibility embraces
community practices. Projectibility cannot name a linguistic habit in a private
language, for the reason that no individual in isolation can possess a criterion
of correctness for the use of a projectible predicate. Individuals use predicates
correctly just when doing so conforms to or wins acceptance by their relevant
communitieslinguistic, disciplinary, and so on.
In this regard, Hackings interpretation of Goodmans riddle bears directly
on issues central to a theory of understanding that so concern Dummett and his
interlocutors.63 For there to be communication, a linguistic community requires,
as Wittgenstein famously remarks, agreement in judgment. This presupposes at
least some agreement on how predicates can be applied or reapplied. So learnability conditions must involve acquiring compatible standards of inductive inference, that is, classification. This suggests that these issues about how language
can be learned and sharedthe concern with a theory of understanding that
motivates the metaphysical debateshould not take observation and deductive
inference as basic. The perceptual and logical operations presuppose an understanding of classification but do not explain it.
This suggests that learning a language has important analogies to learning a
theory or to processes involved in theory change. Hacking takes to heart Kuhns
observations on the training of scientists and applies them to linguistic communities generally. When individuals no longer receive training or guidance in the
use of the discarded predicates, all criteria of correctness for application of them
becomes lost.
61. I discuss what I take to be general limits and constraints on a theory of meaning in Mistakes,
Synthese 136 (September 2003), 389-408 and in Why There is Nothing Rather than Something:
Quine on Behaviorism, Meaning, and Indeterminacy, in Philosophy, Psychology, and Psychologism:
Critical and Historical Essays on the Psychological Turn in Philosophy, ed. D. Jacquette (Kluwer
Academic 2003), 263-287.
62. Ian Hacking, Aristotelian Categories and Cognitive Domains Synthese 126 (2001), 497.
Hacking quotes here from his own Entrenchment, 193. As Hacking also observes, As Nelson
Goodman pointed out long ago, philosophical issues about lawlike regularities, dispositional properties, and projectible classifications go hand in hand (Hacking, Degeneracy, Criminal Behavior, and
Looping, in Genetics and Criminal Behavior, ed. D. Wasserman and R. Wachbroit (Cambridge, UK:
Cambridge University Press, 2001), 155-156.
63. Ian Hacking, Working in a New World: The Taxonomic Solution, in World Changes:
Thomas Kuhn and the Nature of Science, ed. Paul Horwich (Cambridge, MA: MIT Press, 1993), 295.

the pasts

329

[T]he nominalist replies, (a) the world is a world of individuals; the individuals do not
change with a change of paradigm. But a nominalist may add, (b) the world in which
we work is a world of kinds of things. This is because all action, all doing, all working
is under a description. All choice of what to do, what to make, how to interact with the
world, how to predict its motions or explain its vagaries is action under a description; all
these are choices under descriptions current in the community in which we work and act
and speak. Descriptions require classification, the grouping of individuals into kinds. And
that is what changes with a change in paradigm: the world of kinds in which, with which,
and on which the scientist works.64

An important historicist-like point, moreover, emerges just here. Once again, for
habits of classification at any significant historical remove, claims to be able to
use this language, absent a living core of users, become quite literally meaningless.
Hacking illustrates this point by reference to alchemical theory and Paracelsus.
One can, he notes, read the words Paracelsus wrote. But the challenge of knowing
what Paracelsus meant by the terms comes now in applying those terms to things
in the world as they presumably were when Paracelsus and his contemporaries
were alive. But how can one judge if one has the use right if there now exists
no community to corroborate judgments regarding use? 65 For the reality of the
past construed in terms of witnessing presupposes either magical access to what
now no longer existsa community of users who support and sustain patterns
and habits of applicationor assumes unjustifiably that present patterns of categorization suffice for the witnessing involved. But absent communities of past
speakers or a fact of the matter with regard to meaning, neither assumption can
be said to enjoy even the slightest plausibility.66
Hacking strongly endorses Goodmans riddle and its chief consequences
nature does not dictate any organizing scheme to us, and different schemes need
have no connection to one another. It [Goodmans new riddle] shows that whenever we reach any general conclusion on the basis of evidence about its instances,
we could by the same rules of inference, but with different preferences in classification, reach an opposite conclusion.67 No organizing scheme has primacy;
different organizing schemes need not be compatible with or reducible to one
another. Hence, different worlds come to be.68

64. Ibid., 277-278.

65. Ibid., 297. Hacking makes the intriguing observation that application need not require a long
history of use. Indeed, scientific revolutions appear to be precisely cases where new habits rapidly
trump and replace entrenched predicates. But a community of users does prove indispensable to the
process of having a working scheme of things. If Kuhn is right, a scientific revolution can introduce
a projectible term with no entrenchment. Revolutions override entrenchment. Projectibility does not
need a record of past usage. But it needs something precious close to that. It needs communal usage,
which is brought about by a revolution (ibid., 305).
66. This objection echoes Saul Kripkes reading of Wittgenstein and how Dummett anticipates
Kripkes point and uses it to criticize McDowell. See Saul Kripke, Wittgenstein on Rules and Private
Language (Cambridge, MA: Harvard University Press, 1982).
67. Ian Hacking, World-Making by Kind-Making: Child Abuse for Example, in How Classification Works: Nelson Goodman among the Social Sciences, ed. Mary Douglas and David Hull
(Edinburgh: Edinburgh University Press, 1992), 181.
68. Goodman, Ways of Worldmaking, 4-5.

330

paul a. roth

These points bring together Goldstein, Goodman, and Hacking not only with
regard to constituting a past, but also with the most striking and remarkable
consequence that there must be a sense in which the past can be indeterminate,
open to change. Consider what lends credence to the thought of a human past as
fixed and immutable. The events constituting that history must themselves be of
the right metaphysical stuff, a stuff beyond human reach to change or alter. But
then the notion of a fixed or immutable past requires that of essences and natural
kinds, of grouping not made by humans. Yet entrenchment does not presuppose a
carving of Nature at the joints.69 Nothing intrinsic makes actions be what anyone
or any group takes them to be.70
In particular, skepticism and indeterminacy regarding present kinds applies to
past schemas as well, particularly actions qua kind of behaviora kind distinguished by the presence of intentions.
We can well understand how new kinds create new possibilities for choice and action. But
the past, of course, is fixed! Not so. As Goodman would put it, if new kinds are selected,
then the past can occur in a new world. Events in a life can now be seen as events of a new
kind, a kind that may not have been conceptualized when the event was experienced or the
act performed. What we experienced becomes recollected anew, and thought in terms that
could not have been thought at the time. . . . This adds remarkable depth to Goodmans
vision of world-making by kind-making.71

Goodmans riddle challenges the belief that the categories and classifications
employed to name events also specify metaphysical essences. It suggests that
identifying events proves no more fixed than current habits of classification.
Insofar as actions appear immutable and their effects flow forward from this
nature, the past appears fixed. In this respect, entrenchment goes deep; it fosters
69. Danto can be seen again to anticipate these points. Just which happenings there and then are
to be counted part of the temporal structure denoted by The French Revolution depends very much
on our criteria of relevance. Doubtless there are shared criteria so that no disagreement exists over
certain events. But insofar as there is disagreement over criteria, the disputants will collect different
events and chart the temporal structure differently, and obviously our criteria will be modified in
the light of new sociological and psychological insights. The Past does not change, perhaps, but our
manner of organizing it does. To return to our map making metaphor: there is a sense in which the
territories (read: temporal structures) which historians endeavour to map do change. They change as
our criteria change, and at best our criteria are apt to be flexible . . . (Danto, Narration and Knowledge, 167/166-167).
70. Hacking also celebrates Quines critique of the notion of natural kinds. See his Natural
Kinds, in Perspectives on Quine, ed. R. Barrett and R. Gibson (Cambridge, MA: Blackwell, 1990),
129-141.
71. Ian Hacking, The Social Construction of What? (Cambridge, MA; Harvard University Press,
1999), 130/[191]. Chapter 5 in this book is a lightly revised version of his essay, World-Making by
Kind-Making: Child Abuse for Example, 180-238. The number in brackets provides the citation for
the latter volume. I note in passing that although Hackings use might appear to expand the scope of
Anscombes phrase, it does not do so in a way that undercuts its philosophical purport. The phrase
under a description can apply to a specific bit of behaviorthe raising of an arm, the flicking of a
switchthat can be described in different ways and to different effectssignaling, alerting a burglar,
and so on. But under a description can also be used to characterize or mark out certain happenings
as related, for example, by labeling them as World War II or grue. Discussions of intentional
action focus typically on the former sorts of cases; discussion of events or (natural) kinds focus on the
latter. The philosophical point that links these cases insists that nothing makes something a general
type of thingan action, an event, a kindother than shared practices that result in categorizing these
particulars as instances of the general type.

the pasts

331

the illusion that the past consists of something more, by way of events, than contingent classifications.
The argument has been that antirealism still privileges a naive notion of the
observational, and so creates a false contrast between knowledge in the present and knowledge of the past. Damian Cox suggests that one cannot avoid a
dichotomy between some version of metaphysical realism on the one hand, and
some version of irrealism on the other.72 Cox explores senses in which worlds
can, following Goodman, be said to be made and yet not fashioned from materials that, in the end, appeal to the very sort of metaphysical realism with which
irrealism was to contrast. For example, it poses no particular affront to realism to
suggest that before the stars were mapped in a particular way, the Big Dipper did
not exist. But what of the stars the maps map? If the Big Dipper doesnt predate
our introduction of the Big Dipper Concept, do the stars themselves predate
our development of the concept of a star?73 But Goodman has a response. If the
world made contains stars billions of years old, it poses no problem to the claim
that we made that world that it has features not possessed by the version of the
world we make. For example, one can make a two-dimensional representation
of a three-dimensional object, or a black-and-white version of a colored object.
The making need not have every feature imputed to the made. As Cox comments, We make a starry past, in part, by making the spatiotemporal order of
the past. Since there is no ready-made spatio-temporal order, we make a past by
imposing a temporal order on things. We make stars in the remote past, but we
shouldnt expect this making to have itself occurred in the remote past.74 Once
made, concepts do not remain in control of their makers.75 This implies, among
other things, that whether there exist traces supporting the made-up world cannot be determined except by looking.
III. Irrealism and Explanation

Having affirmed that the kinds of actions/social roles possible for people connects
to the kinds of descriptions available, the question then arises for Hacking of
how this impacts the space of possibilities for accounting for or describing past
behaviors. What is curious about human action is that by and large what I am
deliberately doing depends on the possibilities of description.76 But what sense
can be given to the declaration that the possibilities for what we might have been
are transformed? What transforms the possibilities for what we might have
been lies in the fact that if, as Hacking holds, action just is behavior described
72. Damian Cox, Goodman and Putnam on the Making of Worlds, Erkenntnis 58 (2003), 37.
73. Ibid., 40.
74. Ibid., 41.
75. For Goodman, the fact that the worldly extensions of our concepts are not entirely up to us is
an effect of pragmatic constraints on worldmaking. Worldmaking is constrained by coherence, consistency, fit with intuitive judgement and intelligible purpose. Conceptual work aims at rightness and
the rightness of a version of things is not up to us (ibid., 42-43). For a related argument in support
of Goodman here, see Robert Schwartz, Im Going to Make You a Star, Studies in Essentialism:
Midwest Studies in Philosophy 11 (Minneapolis: University of Minnesota Press, 1986), 427-439.
76. Hacking, Making Up People, 231.

332

paul a. roth

using community approved descriptions, changes in community practices literally change the actions attributable to an individual (or individuals) at a time, by
both addition and deletion.
Child abuse is an example of a kind of action that, Hacking argues, can be used
to retrospectively describe, and a past, so described, often changes.
What happens to the person who now comes to see herself as having been sexually
abused? . . . I am referring to placing oneself in a new world, a world in which one was
formed in ways one had not known. Consciousness is not raised but changed. This is perhaps the strongest and most challenging application of Goodmans dictum, that worlds are
comprised by kinds. Child abuse is a new kind that has changed the past of many people,
and so changed their very sense of who they are and how they have come to be.77

I noted earlier that realism precludes allowing a change in the inventory of events
at t to alter consequences after t. But consideration of cases such as child abuse
show that the openness of the past resides, at least in part, in the fact that given
new categories with which to think about the past, the future can change as well.78
In this respect, a belief in the imperviousness of ones past to alteration or
change can become just a counsel of despair regarding ones future. It suggests
that only one possibility exists for configuring what actions took place, and that
this configuration stands as both determinative and unalterable. But this assumption attributes a special metaphysical glue to behaviors over and above their
categorizations under a description. Absent an account of why behaviors must
adhere one way and not possibly another, however, actions as a kind should be
taken as no more of a given than any other feature of empirical knowledge.
Hacking further illustrates his point here by briefly noting how the use of the
category of suicide evolved over the nineteenth and early twentieth century.79 As
much as almost any other concept of the sort that interests Hackingones that
involve the determination of states of mind for purposes of medicalization
suicide enters the realm of what can become medicalized once statisticians
begin to count and classify which deaths belong to this kind.
Any number of people have over the course of recorded history had a hand in
actively bringing about their own demise. But what interests and concerns Hacking is the emergence of a special notion of suicide, one that classifies a suicide,
any suicide, as a type of insanity.80 Suicide thus becomes one of the indices of
mental health for individuals, and the rate of suicide becomes a corresponding
barometer for the mental health of national groups. This connects, Hacking suggests, Foucaults two poles of development, one centered on how to classify
77. Hacking, World-Making by Kind-Making, 230. See also 223.
78. Hacking, Rewriting the Soul, 249-250.
79. I should note here that Hacking regularly cautions that his own analytic scheme is itself
provisional. It is less a theory than an invitation to pay attention to the historical details of concepts
that interest us. But just because it [his account of dynamic nominalism] invites us to examine the
intricacies of real life, it has little chance of being a general philosophical theory. Although we may
find it useful to arrange influences according to Foucaults poles and my vectors, such metaphors are
mere suggestions of what to look for next (Making Up People, 236). Hacking in this regard as well
proves very Wittgensteinian, that is, more intent upon assembling reminders for particular purposes
than in offering a detailed theory of this or that.
80. Hacking, Making Up People, 234.

the pasts

333

individualshis anatomo-politics of the human bodyand the second that

characterizes the biopolitics of the population.
Durkheims Suicide virtually creates at least one paradigm for a fledgling
discipline by its model of how to forge links between these two poles. Fashioning this link would have been impossible but for the statistics collected over the
previous years, and the statistics required a prior commitment to counting and
classifying an intention leading to death. Like crime, there have been suicides
forever, but the suicide was not thought of as a kind of person, with various subkinds of self-destructiveness, until early in the nineteenth century. . . . Something
is thought of as a scientific kind of person when experts begin to propose laws
about the kind.81 But an intention here can be for Hacking nothing other than a
piece of behavior we choose to describe in a particular way. Even the unmaking
of people has been made up.82
Changing the past by changing the descriptions available works, then, for
Hacking in at least two different ways. Reclassification can change the past
impersonally, that is, in ways regarding others but not oneself, or it can change
ones own past, that is, with regard to oneself. Hackings discussion of the notion
of suicide illustrates the first case. Although brief, his account of suicide shows
how reclassification changes the past because a description of action introduced
laterthe medicalized notion of suicideliterally changes what someone previously did. How could it not? What other kind of thing could it be?
But surely what has been done cannot be undone. That will turn out to depend
on what one takes a doing to be. If what happens in the world is at least in part
a function of human actions, and if what actions are are Goodmanian kinds, that
is, exemplifications of ways a given community descriptively collates behaviors
in particular ways, then when new descriptions, new ways of collating physical
doings, become available, this changes what actions happened, whenever they
happened. Only descriptions create a past in which human actions have meaning.
Central to understanding Hackings position here involves the claim that one
can, in effect, respectively reorganize experience, and so come not only to see
but also to experience ones own past in a different way.83 As Hacking stresses,
attributions of, for example, intentionality do not ultimately rest on or receive
validation by some mental fact of the matter. For this reason, Hacking does not
want to say that attributions of intentions to actions are either correct or incorrect.
Rather, they represent descriptions we can give based on ways we apply such
predicates.84
81. Hacking, Degeneracy, Criminal Behavior, and Looping, 143.
82. Hacking, Making Up People, 235.
83. See my Ways of Pastmaking.
84. As a cautious philosopher, I am inclined to say that many retroactive redescriptions are neither
definitely correct nor definitely incorrect. . . . It is almost as if retroactive redescription changes the
past. That is too paradoxical a turn of phrase, for sure. But if we describe past actions in ways in which
they could not have been described at the time, we derive a curious result. For all intentional actions
are actions under a description. If a description did not exist, or was not available, at an earlier time,
then at that time one could not act intentionally under that description. Only later did it become true
that, at that time, one performed an action under that description. At the very least, we rewrite the
past, not because we find out more about it, but because we present actions under new descriptions.
Perhaps we should best think of past human actions as being to a certain extent indeterminate (Hacking, Rewriting the Soul, 243; see also 249-250).

334

paul a. roth

This point segues rather nicely with Hackings more recent emphasis on what
he terms styles of reasoning. A salient feature of these styles of reasoning
is what Hacking terms their self-authentication of the facts with which they
deal.85 New paradigms or theories introduce, in this context, new predicates that
practitioners need to learn to apply to cases. The issue that most interests and concerns Hacking here involves whether and how a community successfully learns to
manage these predicates, and so whether the (putative) facts/phenomena become
stabilized over some long term, and so become objects around which a science
can grow. Some phenomena, in this regard, he views as primarily the creatures of
new taxonomies, and so lacking in any stability apart from what those who accept
the taxonomies provide it.86
Two ideas loom large here. One is that of normalcy and how this becomes
interpreted when applied to classes of individuals. The other concerns the idea
of the mind or soul as an object of knowledge. What is mind, memory, or character such that it can be an object of a science, something that can be classified,
counted, and subject to discernible laws?
But in my context, talk of politicsa politics of memoryis no metaphor. It is, however,
a politics of a certain type. It is a power struggle built around knowledge, or claims to
knowledge. It takes for granted that a certain sort of knowledge is possible. . . . Underlying
these competing claims to what we might call surface knowledge there is a depth knowledge; that is, a knowledge that there are facts out there about memory, truth-or-falsehoods
to get a fix on. There would not be politics of this sort, if there were not the assumption of
knowledge about memory, known to science.87

Hacking expresses both a profound skepticism with regard to and a deep suspicion of those who would insist that there exists a science of the human soul.
Perhaps the most compelling account of the past as indeterminate results when
one reflects on how memories can be thought of as neither veridical nor nonveridical, but rather as experiences that one shapes and reshapes in light of new
concepts. The result here does not involve in any sense a deepening of knowledge
of ones past, in the sense that one later comes to have, perhaps, details that one
lacked. Memories are no more the unvarnished news of reality than any other
form of experience, including perceptual experience under normal conditions.88
Rather, the very possibility of experiences having a meaning at all depends on
their functional role in ones cognitive economy. The investment one makes in
85. See especially his discussion in The Participant Irrealist at Large in the Laboratory, British
Journal for the Philosophy of Science 39 (1988), 277-294.
86. Hacking explores these issues in detail especially in Working in a New World: The
Taxonomic Solution, in Horwich, ed., World Changes, 275-310, and in Memoro-politics, Trauma
and the Soul, History of the Human Sciences 7 (1994), 29-52.
87. Hacking, Memoro-politics, 31-32; see also 38. He develops similar themes in Normal
People, in Modes of Thought, ed. David R. Olson and Nancy Torrance (Cambridge, UK: Cambridge
University Press, 1996), 59-71, but especially 63. Insofar as normal kinds become a projectible part
of some alleged science of memory, therapists of one description or another become free to rework
the history of their patients. Hacking clearly worries about how this influences those subject to more
questionable taxonomic characterizations of human experience, for example, such as those working
in the Multiple Personality Disorder movement he discusses.
88. Hacking, Rewriting the Soul, 254.

the pasts

335

these experiences can and not uncharacteristically does change.89 Kuhns much
contested remarks about living in a different world post-scientific revolution take
a yet more plausible cast given the structure of historical revolutions.
As a final illustration of an irrealist explanation, I consider an argument by
a historian of science, Gad Prudovsky, concerning the legitimacy of imputing
to historical figures concepts they could not possibly have had.90 The particular
case concerns Koyrs interpretation of Galileo, and specifically the ascription by
Koyr to Galileo of a conception of mass that Galileo did not possess. Prudovsky
asks: what can be the justification of ascribing to Galileo a terminology (mass)
of which he knew nothing? And second, can this type of ascription withstand
the anti-anachronistic critique of recent studies in the methodology of historical
writing?91 Without here examining the full complexity of Prudovskys sophisticated defense of the concept of inertial mass that Koyr reads into his reconstruction of Galileo, the core point that emerges involves a deliberate strategy to make
the historical personage as rational as possible.92 The justification maintains,
unsurprisingly, that the concept or something like it exists already implicitly in
Galileos reasoning. This legitimates, Prudovsky argues, the imputation of the
anachronistic concept.
What makes Prudovskys account of particular interest involves the Goodman
ian account that he (unawares) finds in the implicit yet still unarticulated notion
of inertial mass that Koyr ascribes to Galileo. Koyr wanted to argue that this
is the first step in the development of the concept of inertial mass in the history
of science. Such a step is obviously a preliminary move, not wholly clear to
those who made it, and hence lacking the maturity of the later classical concept.
89. A similar account of the indeterminacy of action can be found in the work of Roy Schafer.
Logically, the idea of multiple and new definitions of individual actions implies multiple and
changeable life histories and multiple and changeable present subjective worlds for one and the same
person. To entertain this consequence is no more complex an intellectual job than it is to entertain,
as psychoanalysts customarily do, multiple and changeable determinants and multiple and changeable self- and object representations (Roy Schafer, The Psychoanalytic Life History, in Schafer,
Language and Insight [New Haven: Yale University Press, 1978], 21). Or again, Put in historical
perspective, there is far more to an action than could have entered into its creation at the moment of
its execution. It is the same as the effect of a new and significant literary work or critical approach on
all previous literature: inevitably, fresh possibilities of understanding and creation alter the literary
past (ibid., 21). Schafer appreciates that from his conception it follows that each person can have
multiple life histories. See, for example, ibid., 10, 19-20. On Schafers debt to Goodman, see Schafer,
The Analytic Attitude (New York: Basic Books, 1983), especially 205, 206, 249, and 276. I offer a
more detailed account of Schafers views in Paul A. Roth, The Cure of Stories: Self-Deception,
Danger Situations, and the Clinical Role of Narratives in Roy Schafers Psychoanalytic Theory, in
Psychoanalytic Versions of the Human Condition, ed. P. Marcus and A. Rosenberg (New York: New
York University Press, 1998), 306-331.
90. Gad Prudovsky, Can We Ascribe to Past Thinkers Concepts They Had No Linguistics Means
to Express? History and Theory 36, no. 1 (1997), 15-31.
91. Ibid., 16.
92. But let me confront the contextualist counterargument on its own ground by showing how
it is possible for someone to have an idea even though he or she has no linguistic means to express
it. I take this challenge to be equivalent to showing just when it is justified to ascribe a concept to a
person who lacks the linguistic means to express it. . . . In effect, one has a concept when others are
justified in ascribing it to one as a way of interpreting ones inferences, and when one engages in such
inference-making in a way that is licensed only by such a concept (ibid., 29; see also Prudovskys
discussion on 18).

336

paul a. roth

Thus, the concept does not remain constant: it changes in the transformation
from its implicit phase to its explicit one.93 That is, as Prudovsky reads Koyr,
the concept of inertial mass does interpretive work, but what he employs is not
equivalent to the contemporary concept, but only some indeterminate approximation of it. Like Kripkes plus and quus, one cannot say that Galileo must
have some particular function in mind. This allows Prudovsky to explain how
Galileo applies the concept to the cases without having to claim that Galileo has
a worked out version of the concept and so knows precisely to which cases it
extends and which not.94
One could say here, on Prudovskys reading of Koyr, that by ascribing to
Galileo a concept he could not have possessed Koyr constitutes a Dantoesque
eventthe moment in history where Galileo introduces what will become the
concept of inertial massand then uses this to explain why what Galileo argues
makes sense. No one could say or predict at that moment that this was happening,
or that a certain concept would come to have a settled use in a future scientific
community. The point rather involves illustrating how historical events may be
constituted and explained in terms of concepts in some sense present in but not
known to those to whom they are attributed.
At the most general epistemological and metaphysical level, no principled
distinction emerges between empirical knowledge generally and knowledge of
the past. The forms of inference required to have empirical knowledge at all
inductive, abductive, and deductivearise for all such cases of knowledge. Once
the presumption of givenness with regard to evidence or of shared conceptual
schemes goes, the shape of the past and the shape of the present receive
their form under fundamentally similar holist constraints. Temporal distance
may accentuate problems of making sense of others and what they did, but the
problems posed turn out not to be at all unique. Only in a theory do thingsfor
example, facts, events, kinds, actionsexist and have explanations.
The suggestion that people now decide what traces are traces of proves shocking only if one imagines that this attaches only to attempts to know the past. A
persistent fear post-Kuhn has been that once a clear line between experience and
theory goes, nothing real remains to serve as a check on interpretations. What
people imagined empirical evidence to be turns out to be theorizing by another
name. Excesses of this sort do exist.95 But this fear proves overblown inasmuch as
the position developed here simply makes divides between theory and evidence
93. Ibid., 26.
94. Prudovsky indulges in an unfortunate account of reifying ideas. But this implausible move
becomes unnecessary if one goes Goodmans way. What then determines the application of a concept
requires only community practice (ibid., 20 and 27-28).
95. For a discussion of this problem, see my The Disappearance of the Empirical: Some Reflections on Contemporary Culture Theory and Historiography, Journal of the Philosophy of History 1
(2007), 271-292. See also John Zammito, Reading Experience: The Debate in Intellectual History
among Scott, Toews, and LaCapra, in Reclaiming Identity: Realist Theory and the Predicament of
Postmodernism, ed. Paula M. L. Moya and M. R. Hames-Garca (Berkeley: University of California
Press, 2000), 279-311. Zammito provides an excellent study of how this debate originates in philosophy of science in A Nice Derangement of Epistemes: Post-positivism in the Study of Science from
Quine to Latour (Chicago: University of Chicago Press, 2004).

the pasts

337

or observables and nonobservables into a contingent fact and not a necessary or

conceptual one.
More generally in the philosophy of science, problems arise concerning, for
example, charting progress across incommensurable scientific theories where the
events or facts described in one have no status in the theory that supersedes it.
But for all such cases, no ultimate arbiter for what constitutes the reality of kinds
and events exists.96 Analogous factors in the philosophy of history have not been
given the attention they deserve.
Mink comes closest to making explicit why reference to events in time
Danto-like narrative sentencesmust make a difference to the form that historical representation and explanation take. Mink seizes on Dantos suggestion that
historical events lack, and scientific events have, a standard description as what
separates historical discourse from scientific discourse. A scientific theory specifies what features a description of an event must include in order to be considered
complete; events in historical discourse remain descriptively incompletable.
One might distinguish science and history in this way: a scientific account of an event
determines a standard description of the event. . . . History, on the other hand, reports how
descriptions change over time. . . . Thus there can be a history of science, that is, of the
changes in the kinds of description accepted as standard at different times, but no science
of history, that is, a complete description of events which includes or subsumes all possible descriptions.97

Dantos analysis of narrative sentences simultaneously demonstrates not only

why complete descriptions of the pasta full catalogue of what events the past
containsprove impossible, but also why there exists no standard description.
The salient features of a situation often emerge only retrospectively, so one cannot state (timelessly) what (for a particular time) will be of significance.
Mink published his review of Danto in 1968, when a symmetry of explanation and prediction was widely assumed. He uses this assumption to argue that
Dantos account demonstrates why a lack of a standard description will make
historical events unpredictable, and so inexplicable. Events can be given a
standard description only when proper names can be replaced by the relevant
general information that a standard description requires. But Dantos own analysis demonstrates that historical descriptions of earlier events often incorporate
knowledge that comes later, either because the later event informs on the earlier,
96. This points to the important parallel between the insistence of Goldstein and Mink that historians constitute the events they seek to study and what Donald Davidson refers to as anomalous
monism. There exists no a priori reason to expect that the events and regularities that interest historians should map onto any categories that happen to be those employed by other scientific theories.
The locus classicus for discussion of anomalous monism is Donald Davidson, Mental Events, in
Experience & Theory, ed. L. Foster and J. W. Swanson (Amherst: University of Massachusetts Press,
1970). See also Nancy Cartwright, The Dappled World (New York: Cambridge University Press,
1999). Regarding what prompts a felt need for explanation, see Stephen R. Grimm, Explanatory
Inquiry and the Need for Explanation, British Journal for the Philosophy of Science 59 (September
2008), 481-497. Some of these points receive elaboration in my Varieties and Vagaries of Historical Explanation, Journal of the Philosophy of History 2 (2008), 214-226, with replies by Karsten
Stueber, Stephen Turner, and John Zammito.
97. Mink, Philosophical Analysis and Historical Understanding, 139, fn. 6. Cf. Danto, Narration
and Knowledge, 176/177.

338

paul a. roth

or because the conceptual vocabulary comes later. In such cases, nothing known
by anyone at the time could have been used to predict, and so explain, what lies
ahead. So, Mink concludes, the analysis of descriptions possible only after
the event is also an argument against the possibility of covering-law explanations
in characteristically historical discourse.98 But renouncing as necessary a symmetry between explanation and prediction might appear to deprive this argument
of its force.
I draw a different moral, for Minks contrast between what would be required
of an event in order for it to be fodder for a scientific theorya standard descriptionand the absence of such a description for historical events remains an
important and useful insight. Mink and Danto agree that essentially historical
discourse requires expression through narrative structure. Moreover, that some
events allow a scientific treatment in the most robust sense of the term, Mink
acknowledges, is one thesis on which reasonable men will not disagree. But
he then adds, There is nothing wrong with being wise after the event; it is just
that we cant be wise after the event, before the event.99 To be sure, what comes
to be learned later just might reveal what could not possibly have been known
earlier. Hindsight may teach that nothing could have remedied ignorance of what
was to come.
But about which events can one be wise before the fact, and which only after?
What shows that some historical event could not have been predicted at the earlier time? In order to avoid toy examples, consider the following.
[T]he long Second World War commenced at the moment in which various states
required that their peoples liberties be subordinate to their nationality. No precise definition is possible, but 1922 would be a sensible starting point for Italy, 1931 for Japan, 1933
for Germany and perhaps 1929 for the USSR. . . . [The Munich Agreement] sounded the
final prelude to those actual wars which would break out in Europe and the wider world
between 1939 and 1941 and which are known as the Second World War.100

So, according to Bosworth, the long Second World War begins in the period
19221933. Judgments regarding what could have been known may vary and
change.
Danto at one point speaks of philosophical analysis as revealing a descriptive
metaphysic, by which he means a general description of the world as we are
obliged to conceive of it, given that we think and talk as we do.101 Mink agrees,
but quickly makes the point that any such descriptive metaphysics will itself be
subject to historical influence and change.
But one may still ask: could we think and talk differently? The answer must be yes, by
the witness of history itself. . . . [A]nd to acknowledge this possibility is to bring our
descriptive metaphysic under the category (itself historical) of history. Yet since our
central concepts stand and fall together, change cannot be capricious, or fragmentary, or
idiosyncratic.102
98. Mink, Philosophical Analysis and Historical Understanding, 145.
99. Ibid.
100. R. J. B. Bosworth, Explaining Auschwitz and Hiroshima: History Writing and the Second
World War 19451990 (New York: Routledge, 1993), 6-7.
101. Danto, Narration and Knowledge, xv.
102. Mink, Philosophical Analysis and Historical Understanding, 145-146.

the pasts

339

I do not endorse any suggestion that human cognition embodies functions that
must be or even typically are capricious, or fragmentary, or idiosyncratic. But
the arguments of this paper suggest that any descriptive metaphysic represents
a historically fashioned imposition on the flux of experience and not a discovery
of categories in the mind shared by all who communicate. Claims to conceptual
necessity turn out to be just one more attempt to lay hands on the really real.
Once belief in shared conceptual frameworks goes, so goes the explanatory
utility of appeals to such shared mental structures whether Kantian, Marxist,
Freudian, Carnapian, and so on. As Donald Davidson perceptively notes, For if
we cannot intelligibly say that schemes are different, neither can we intelligibly
say that they are one.103 The extent to which various ways of characterizing the
world stand or fall together remains an open question.
This essay has charted a course that began with Dantos insight that established
that descriptions true of a past time cannot be determined at that time. What
events can justifiably be said to have taken place at a time changes over time.
Using Goldsteins account of historical constitution, I then argued that historical
events said to occur at any particular time must be a product of attributing some
unifying theme or purpose. Events as usually discussed in human histories must
be constituted at least in this sense. Finally, I developed this notion of historical
constitution further by employing Hackings view that what events can be said
to exist depend on the stock of descriptions or categories available. In particular,
I argued, when the stock changes, by addition or deletion, the extant events at a
time do as well.
The overall import of these arguments has been to problematize the notion of
an event in particular, and evidence in general, in relation to the construction of
pasts, of histories. To speak of pasts as constituted and not found emphasizes the
priority of classification over perception in the order of understanding. Because
nothing a priori anchors practices of classification, no sense can be attached to
claims that some single structure must or does determine what events take place
in human history.
Irrealism denies to realism any imagined view from nowhere, a past seen
sub specie aeternitatis. Given alternative modes for structuring what happens,
changes in descriptions can alter relations among events imputed to a past, and
so how a past thus structured impacts what becomes possible going forward.
A plurality of pasts results because constituting a past depends to some degree
on socially mediated negotiations of a fit between descriptions and experience.
Even what we take to mark what can change and what cannot itself depends on
the descriptions deployed. Unless for reasons now unknown there ceases to be a
possibility of descriptive change or reclassification, human histories will continue
to reveal a multiplicity of pasts.
University of CaliforniaSanta Cruz
103. Donald Davidson, On the Very Idea of a Conceptual Scheme, Proceedings and Addresses
of the American Philosophical Association 47 (1973-74), 20.

Copyright of History & Theory is the property of Wiley-Blackwell and its content may not be copied or emailed
to multiple sites or posted to a listserv without the copyright holder's express written permission. However,
users may print, download, or email articles for individual use.