The Role of the Observer

BY

NORBERT WIENER
HE distinction between psychology, logic, and
epistemology is a commonplace. The first
twlf'Noi11i�

treats of experience as an act, experience in its

relation to the individual observer. The sec­
ond concerns itself with the internal marks by
which truth may be distinguished from error,
and in so far as it deals with experience, has to do with some sort
of validity of the experience as evidence of a truth. The third
discusses in a general way all the elements-observer, object,
immediate presentation-which enter into the experience.
There is clearly a rough practical distinction between these
three ways of regarding experience. The interests of an inves­
tigator of colour vision are certainly different from those of an
investigator of mathematical technique, and both are different
from those of a speculator in the theory of knowledge. However,
the assertion that psychology and logic are two different sciences
may mean something much deeper. There are those who make
the distinction absolute, and who assert dogmatically that the
logical value of a proposition is entirely unconnected with the
relation of this proposition to any observer. The Platonist be­
lieves in a world of essence, of cleanly defined Ideas and cleanly
defined propositions concerning these Ideas, into which we may
enter as spectators, but never as participants. They are out
of time, and time is irrelevant to them.
This is pure dogma, and does not check with what we should
naively expect. Of course, our experiences must have some ref-
307
 The Role of the 0 bserver
erence outside themselves, in the sense that they cannot be con­
sidered as completely closed and isolated. Otherwise there could
be no knowledge at all. This by no means asserts that the ex­
perience has a reference entirely unaltered by our participation.
How hard it is to minimize this participation, is shown by the
conventions of the representative arts.
When a dramatist wishes to portray for our amusement or
edification some scene of human intercourse, he places it in a
room with three sides at oblique angles to one another, with an
abundance of doors in utterly abnormal positions, and an entirely
unusual illumination. The actors speak the most personal solilo­
quies, the most confidential whispers, and the best-mannered
drawing-room conversation, in tones that carry a hundred feet.
A dinner-table will have one empty side, so that everyone may
sit facing the audience.
We are used to these conventions, and they do not trouble us.
To a first approximation, they are unimportant. This is largely
due to the technical skill of the dramatist, who deliberately
avoids shocking us by attempts to express the dramatically in­
expressible. The conventions of the drama do not interfere
with his message because he has the sense not to try to impart
any message with which they would interfere. Nevertheless,
when the dramatist invades recent history, as is the case with
Drinkwater's Lincoln, his dramatic groupings and the exigencies
of the stage must conflict with the recollections of contempo­
raries. Small as his alterations may be, they are real, and im­
portant for some purposes.
What is true of the playwright and actor is true of the painter,
the sculptor, the author. To discern this, we need not go back
to the historical tableaux of David or the be-togaed quasi­
Roman busts of his contemporaries. Ruskin has familiarized all
of us with the limitations of the painters' palette as against na­
ture. The older painters, by their very painstaking attention to
detail, presupposed an eye deliberately searching their picture
from spot to spot, and forewent some of the truth of the imme­
diate impression. The impressionists, in declaring the new
truth, prevented themselves from expressing the old one. If the
 N. Wiener
sculptor represents eyes and hair true to their form, his repre­
sentation belies their appearance; if true to their appearance, it
belies their form.
It is however in literature and in the kindred art of illustra­
tion that the observer intrudes himself most insidiously. When
I was a boy, I possessed a copy of "Treasure Island" with an
illustration which troubled me not a little. It showed the hero
being chased through the rigging of the ship by one of the muti­
neers. The point of view from which it was supposed to be seen
was about ten feet to the side of one of the masts. "This pic­
ture,'' I said to myself, "represents the scene as somebody sees
it. Now, it is explicitly stated that only the two characters
shown in the picture are aboard. Moreover, if a third person
were aboard and were to see the scene as it is represented, there
would be no place for him to stand." In literature, the author
also constitutes an extra character, who sees the invisible and
divines the unknowable. The more "psychological" a novel is,
the more this extra character introdes.
In psychological works of the imagination, and above all in
that great work of the imagination known as psychoanalysis, the
author can only reveal by falsifying. It is his task to bring the
subconscious motives of his characters to light, but a subcon­
scious motive brought to light is no longer a subconscious motive.
It has an outline and a body which are foreign to its original
nature. Without doubt the methods of psychoanalysis have a
real validity of some sort. Equally without doubt, this validity
does not consist in merely tearing aside a veil behind which our
original unconscious motives remain unaltered.
Psychoanalysis is not the only branch of medicine in which the
examiner has at his disposal only methods which are to some
degree destructive. In an abdominal emergency, the surgeon
may have to perform an exploratory laparotomy, but an ex­
ploratory laparotomy is itself a somewhat drastic surgical pro­
cedure. A proper examination of the eye demands the use of
mydriatics, which however have blinded many a glaucomatous
eye. The physician learns his trade on a dead, preserved
cadaver, which death and the use of fixatives have altered in
310 The Role of the 0 bserver
texture, form, and chemical composition. He can only examine
the deeper tissues by the removal of those more superficial.
In his histology, he can only discern microscopic structure by
the use of dyes which tend to alter microscopic structure. His
physiological experiments on dogs often involve the alteration
of the course of a salivary duct or the formation of a fistulous
entrance to the stomach, or some change more radical still.
These difficulties, of which the biologist is very well conscious,
were long strange to the physicist. For centuries it did not seem
that the inspection of a physical experiment was in any way
relevant to its outcome. The effect of light on ordinary gross
mechanical occurrences is so slight that for long it lay far below
the threshold of accuracy of our measurements. Nevertheless,
with the gradual improvement of scientific technique, the time
came when light was found to exert a pressure. This pressure
actually alters the mechanical phenomena observed, but it might
be thought and was at first thought· that any such alteration
may be reduced below any desired level by decreasing the in­
tensity of the illumination.
Modern physics has found that a reduction of the intensity
of illumination does not furnish a practical way of minimizing
its effects on the phenomena observed. Light is radiated in
quanta, in units, the size of which bears a relation to the fre­
quency of the light, but not to its intensity. It is impossible to
observe an occurrence unless we illuminate it by at least one
quantum of light, but the effect of this quantum is definite, and
cannot be reduced save by replacing the source of illumination by
one of lower wave-length,
We are now brought up against a dilemma which leads to
Heisenberg's famous principle of indetermination. If we ex­
amine the position of a particle with light of low frequency, our
microscope will have a relatively low power of resolution: that
is, it will be impossible to determine the position of the particle
with any high degree of accuracy. On the other hand, if we
examine the particle by light of high frequency, we shall give the
particle such an impulse as to mask its original momentum very
completely. Every gain in the accuracy with which we deter-
 N. Wiener 311

mine the position of a particle is a loss m the accuracy with

which we determine its momentum.
One might suppose that it is still possible to maintain that a
particle such as an electron still has a definite momentum and
a definite position, whether we can measure them simultaneously
or not, and that there are precise laws of motion into which this
position and this momentum enter. Von Neumann has shown
that this is not the case, and that the indeterminacy of the world
is genuine and fundamental. There are no clean-cut laws of
motion which enable us to predict the momentum and position
of the world at future times in any precise way in terms of any
observable data whatever at the present time. In other words,
while it is possible to give an account of the world in terms of
our observations which themselves disturb our world, this ac­
count has only statistical validity, and cannot be brought closer
to precision by any chain of observations.
Thus physics, the most exact of all sciences, has had to have
a thorough logical housecleaning. We no longer conceive the
laws ofphysics to apply to some mystical world of reality behind
our observations and instruments: they merely constitute an
intelligible statement of the manner in which our observations
and the readings of our instruments hang together. About any
proposition of physics, we must ask: does it enable us to predict
the result of an actual or possible experiment. If it does, it
stands or falls with this experiment; if not, it has no meaning
whatever, Physics is merely a coherent way of describing the
readings of physical instruments.
There is no reason why a similar criterion should not be ap­
plied to all branches of knowledge. Biology should be an ac­
count of the outcome of dissections, physiological experiments,
and observations of the behavior of animals. Psychology should
be a reasoned history of introspections and observations of be­
havior, which will allow us to fit in new observations and intro­
spections. Mathematics should be an account of theorems and
their recognized criteria of truth or falsity, which will allow
us to place new theorems in this respect. Whatever view we
have of the "realities" underlying our introspections and experi-
3 12 The Role of the 0 bserver
ments and mathematical truths is quite secondary: any propo­
sition which cannot be translated into a statement concerning
the observable is nugatory. It is perfectly conceivable that we
shall be led back to the same relations between experiments and
observations that we are supposed to obtain at present, but it is
equally conceivable that we are being misled by a preconceived
notion of "reality" into ignoring a much better correlation be­
tween our experiments than we have yet suspected.
Let us apply our new and more stringent logical conscience
to the so-called "universals," the Platonic Ideas. It does not
take much sophistication to recognize that the universals of
science-let us say, "evolution" or "mathematical function"­
have a definite history. "Evolution" does not mean quite the
same thing to us that it meant to Charles Darwin, and it cer­
tainly did not mean the same thing to Charles Darwin that it
meant to Erasmus Darwin. It is possible to envisage this pro­
gressive alteration in a universal as the gradual rendering ex­
plicit of what was already present in esse, or in the replacement
of one universal by another through successive stages. Let it
be noted, either of these views constitutes a prejudging of the
logical situation by a reference to a postulated real world behind
the observations and has no legitimate scientific status.
The only universals with which the logician has any real con­
cern are the universals as they actually appear on an inventory
of our knowledge. These universals, unlike those of Plato, have
various degrees of vagueness and mutability. Let us consider
as an example the notion of a series in mathematics. Euler made
a plentiful use of series, generally with ultimate results capable
(at least in many cases) of a correct interpretation, but with a
total neglect of what we should now consider the most elementary
requirements of mathematical rigor. Convergent series, series
summable in the sense of Abel, asymptotic series, all appear
mixed together, without the slightest semblance of discrimina­
tion; We may say that Euler here applies the theory of conver­
gence, there that of summability, which is a manifest anachro­
nism; or that his work, not being mathematically rigorous,
consists of false propositions, or of no propositions at all, which
 N. Wiener
is an even more manifest piece of pedantry. The actual facts
are quite simple. Until the last quarter of the nineteenth
century, mathematical science did not regain the degree of
mathematical rigor demanded by the Greeks, and Euler was the
last man of whom rigor could be expected. There is too much
arduous pioneer work on the frontiers of science to permit all
the niceties of civilized intercourse; and in the middle of the
eighteenth century, scientific as well as colonial frontiers were in
a state of rapid recession as a result of the efforts of hardy pio­
neers. Euler's work was great and crude. His concepts and his
proofs were powerful but not clean-cut. We may neither deny
his achievement nor read into it the subtleties of later genera­
tions.
It might be maintained with some show of plausibility that the
heroic age of mathematics is over, that we have devoted the
effort of two generations to the improvement of our logic, and
that the real timeless universals of elementary mathematics have
at last been uncovered. The relative recency of Zermelo's
doubts as to the multiplicative axiom, of Russell's paradoxes
and of the theory of types by which he avoids them, and of the
denunciation of the law of the excluded middle by Brouwer, all
tell against this supposition. Even more telling is some of the
recent work of Kurt Godel. Godel shows that in any mathe­
matical system there are propositions, the truth or falsity of
which can only be determined by a reference outside the system.
That is, with the best will in the world to determine a mathe­
matical system categorically, in such a way that its propositions
can only have one definite meaning, we can only put off the ques­
tion of its categoricity to that of some larger, more inclusive
system. At some stage, we always run back to that which has
no adequate mathematical definition. In every actual mathe­
matical system, there are always details which (as far as we
know) may be defined in two or more ways, and an imperfect
decision between these ways may precipitate us into new logical
paradoxes such as those of Russell. No matter how far we go,
our logic will always appear Eulerian to some possible scholar
of the future.
 The Role of the 0 bserver
Thus even in mathematics, precise Platonic Ideas and precise
Platonic Truth represent rather asymptotic ideals than anything
directly accessible and significant to us. Any really useful logic
must concern itself with Ideas with a fringe of vagueness and a
Truth that is a matter of degree. A logic which ignores the ac­
tual history of ideas and the limitations of human faculty is a
logic iti vacuo, and is useless. The distinction between logic,
psychology, and epistemology cannot be made absolute.
In science, two opposite motives are always manifesting them­
selves: the urge for generality and the urge for concreteness.
Mathematics is at the same time in the process of extending its
stock of concrete theorems, and of proceeding to more general
theories. The urge for the concrete is perhaps most intense in
history, economics, and the other human sciences, and above all
in psychology. In history, besides the great general principles
which govern all human affairs, we have a definite need to know
the intimate details of various particular occurrences. An eco­
nomic theory of depressions is incomplete without a special
account of the depression of 1929-?, The psychologist must be
a master of the very difficult art of disclosing the concrete con­
tent of the mind, of introspection.
The very fact that this art is difficult and must be learned
should excite in us a doubt as to whether it is a mere process of
unveiling in unaltered form what is already present. Intro­
spection throws immediate light on the content of the mind when
thrown into an introspective attitude:-when we deliberately
refuse to generalize, even to the extent which is normal in every­
day existence, and cast ourselves into a state of the greatest
possible passivity. The example of quantum theory should
make us most sceptical and unready to take the picture we thus
obtain of the content of the mind as representative of what is in
the mind under more normal circumstances. Just as universals
should only be taken to be definite in so far as they have been
through a process of definition, so the data of introspection are
concrete only in so f ar as they have been rendered concrete by a
deliberate process.
To define a universal is to render it definite in meaning at any
 N. Wiener

cost to its definiteness of localisability in time. To place a sense­

datum in our introspection is to orient it as precisely as possible
in the time-scale of our experience, with so much cost to its
logical definition that it becomes a mere "that" instead of a
"what." Neither of these processes is in fact capable of being
carried out to perfection, but they are fundamentally incompati­
ble, and every improvement in the one is at the expense of our
ability to apply the other to the same situation. They have
much the same sort of duality as observation by high- frequency
light and observation by low-frequency light, as observation
disclosing position and observation disclosing momentum. Un­
der normal conditions, the content of our mind is defined in
meaning, but imperfectly defined; it is localizable in time, but
only to a limited extent. It is only under exceptional circum­
stances that we either attempt to view things sub specie aeterni­
tatis, or try to seize the moment in all its immediacy. Ra­
tionalism and empiricism are alike accounts of asymptotic modes
of thought toward which we tend in one mood or another, rather
than adequate epistemologies of our normal experience.
The logic of our normal thinking, in which universals exist,
but have no perfectly clear-cut outlines, and in which the growth
of universals is real, and not a mere fictitious account of our
perception of eternal verities, is inductive logic. Deductive
logic and the canons of pure introspection only exist as ideal
limiting cases of inductive logic. The practicing mathematician
knows very well that mathematics as a living investigation is
inductive and experimental, whatever it may be when stuffed
and mounted in text-books. When I want an auxiliary function
to do a definite job, I try one after another, finding the first
too big here, the second too small there, until by grace of luck
and a familiarity with the habits of the species, I come on an
exact fit. Nine-tenths of the possibilities are eliminated on the
basis of a general feeling for the situation before it comes to a
matter of any real deductive logic whatever. The tenth sug­
gestion slips into place in a way which convinces an old hand
that there is something in it-it resolves the difficulties at just
the right points, but not so readily as to excite suspicions of a
 The Role of the Observer
sheer blunder. Once the key will go into the lock, and the bolt
begins to show signs of turning, it is a matter of mere file-work
and oil to get a perfect fit. The final deductive finish is impor­
tant for a pretty job, but may be left in the hands of any com­
petent apprentice.
The philosophy of Hume furnishes the dreadful example of
what happens to an . empiricism which seeks its fundamental
reality in the fugitive sense-data of immediate experience. If
the raw stuff of our experience does not contain something of a
universal nature, no manipulation can ever evoke anything which
might even be mistaken for a universal. What Hume has parted
asunder, let no God join.
Science is the explanation of process. It is neither possible
under a rationalism, which does not recognize the reality of
process, nor under an empiricism, which does not recognize the
reality of explanation. The problem of the justification of in­
duction, and of the formation of a real inductive logic, which has
been one of the eternal scandals of philosophy, arises from the
attempt to fit induction into a system originally conceived
without reference to induction, in which inductive knowledge is
derivative instead of primary. The difficulty is the same as
that which we incur when we try to force quantum theory into
the frame of classical mechanics.
In Kantian terminology, inductive knowledge is synthetic,
for it gives us actual new information. It is certainly at least in
part a posteriori, but the mind does not receive it purely pas­
sively, so that it may be said to contain an a priori element.
Deductive knowledge is analytic, for the simple reason that no
perfect univers al has been defined until all possible questions
concerning it have been answered, and are consequently con­
tained in its definition.
A particular case of induction which contains much that is
typical of more general cases, and merits detailed discussion, is
given by periodogram analysis. Periodogram analysis is a
method used when it is desired to throw light on the irregular
changes of a measurable quantity, such as the air temperature
at a particular observing station, and to uncover hidden periodic-
 N. Wiener

1t1es. It will be seen without much explanation that the direct

record of the temperature as a function ofthe time is very hard
to decipher, and is of very little direct use.. It is desirable to
give a separate analysis of the components of this phenomena,
correlating an amplitude with the period of each component.
Naturally, our data only run for a finite interval of time.
The methods of resolving the fluctuations over a finite time­
interval into components have what we call a finite resolving
power. To distinguish two hidden periods differing in fre­
quency by a quantity a, we need a record running for a length of
time inversely proportional to a. If we are dealing with a
phenomenon whose general nature varies only slowly, we accord­
ingly take the longest period possible for the basis of our har­
monic analysis, and go back as far as our data are reliable.
This is not the right way to handle data subject to profound
outside influences. A run of economic data since the onset of
the depression may show quite different periods from a run be­
tween the war and the depression, and these may again differ
from those of data anterior to the war. Accordingly, while very
long runs may be of interest to a philosopher of history viewing
his subject from the point of view of eternity, they are valueless
to the person with normal human interests. We have not yet
come to the point where the proper view of human history
is that suitable to an observer of the vicissitudes of a population
of ftuit flies. In other words, periodograms that are historically
significant have an intrinsically limited resolving power.
This fact casts a considerable amount of light on the prospects
of the extreme mathematical school of economists. These aspire
to give refined mathematical methods the same rele in economics
that they already play in physics. The example of the periodo­
gram at least suggests that the humanly important phenomena
of economics may be capable of no very precise mathematical
definition. The well-trained scientist finds something morally
abhorrent in writing down five figures when only three have a
meaning. The use of delicate mathematical methods on a very
coarse subject-matter is tainted with the same sort of dishonesty.
This school of Pareto seems to be tarred with this brush.
 The Role of the Observer
It is not merely in the matter of periodograms that the his­
torian finds himself in danger of falling between two stools. The
mere unreflective narrator of events does not fulfill the complete
function of the historian, nor does the philosopher of history
on such a vast scale that individual human lives and motives
lose all significance. The intellectual historian must think in
terms of such generalizations as church and nation, democracy
and fascism, constitution and throne, but he must be aware of
the continually changing content of his generalizations. Just
where to drop the urge to generalize, and where to obey it, just
how much to grant to narration and how much to explanation­
this is the great art of the historian, and is only to be learned
by practicing it. It is not easy to place an individual occurrence
on the canvas of history without so subordinating it that it loses
significance.
The philosophical attitude expressed in the present paper is
derived from an attempt to apply generally the same maxims
that have proved fruitful in the development of quantum theory.
These maxims lead to the following propositions, which may be
called "postulates" in a very loose sense:
(1) Science is an explanation of process in time;
(2) The act of explanation and of the development of concepts
actually consumes a certain minimum of time, and in­
volves a reference to a certain minimum of time;
(3) No concept nor theory is well-defined except in so far as
it has been through an actual process of definition;
{4) The time consumed in attaining to a certain degree of
definition, as well as the time referred to in a definition
of this degree of preciseness, is of an order of magnitude
roughly proportional to the degree of precision attained.
I have made no attempt to apply an actual quantum-theoretic
account of the nature of life and of the self. In this respect,
my method is fundamentally different from that followed by
Professor J. B. S. Haldane.1 He starts with a theory of the
mind as a highly degenerate quantum-theoretic system, and
lQuantum Mechanics as a Basis for Philosophy. Philosophy of Science, Vol. I, No. 1,

January, 1934.
 N. Wiener
proceeds to a theory of universals which leads him to conse­
quences much the same as those of the present paper. His
conjecture is a powerful piece of imaginative work, but it is
imaginative, and the theory of universals at which he arrives is
obtained by the transferrence of a physical theory to a field
where, to say the least, it has not yet been tested. On the other
hand, in this paper, I transfer the criteria and sanctions of quan­
tum theory to new matter, and not its details. My results are
entirely compatible with Haldane's. This may be taken as a
partial indication that his view of the nature of the mind is quite
possibly more than an interesting fairy tale. The dialectical­
materialist-quantum-theoretic view to which he inclines certainly
leads him to a logic in which propositions (1)-(4) of this para­
graph hold good. On the other hand, even if Haldane's account
of life is not found to be tenable, our intrinsic discussion shows
that the universals of logic are subject to limitations much like
those which he ascribes to them.
In proposition (4) of the last paragraph, we have postulated
certain numerical relations between time and accuracy. In these
relations, certain scale-constants must occur. If Haldane's view
is correct, these quantities must bear a definite relation to c, the
velocity of light, and to h, the quantum constant. Haldane's
theory is not yet developed to the point of formulating this
relation, and of course the present paper does not even give an
indication that such a relation exists. If it ever is found, it will
constitute a very substantial confirmation of Haldane's views.
National 7'sing Hua University
Peiping, China