conduction is a matter of each single nerve fiber, occurring in
one definite manner as long as the properties of the conductor
remain the same? Experimental investigation of the all-or-none law
is not yet complete and, so far as I know, we have no evidence as
yet of its application to central fields. Supposing, however, that
it holds in brain tissue as it seems to hold in peripheral nerves,
the alternative between a strict machine theory of distribution
and dynamical conceptions would remain as open as before, because,
then, in each part of the nervous system the number of elementary
organs taking part in any actual process, i.e., the density of the
process, would have to be determined somehow, as would also the
frequency of current in each single organ. Both properties of the
process may either depend upon given local arrangements or upon
dynamical intercourse in the ganglionic tissue. As distribution
of energy is a dynamical problem in physics notwithstanding the
quantum theory, so the all-or-none law, as a sort of quantum theory
in nerve physiology, does not exclude dynamical distribution of
process.
A final argument which maintains that distribution of process by
preëstablished and isolating conductors is the only way to explain
the order of the field in its relation to the external physical
world, seems to have two parts. _First_, it presupposes that
dynamical interaction, not controlled by special arrangements
at each step, must produce chaos and confusion. We have here the
old human prejudice which, in a great many cases of dynamical
interaction, is absolutely unwarranted. _Secondly_, as to the
relation between the external physical world and the sensory field,
the theory of isolated conductors enforcing order would explain a
strict correspondence of sensory facts to _retinal stimulation_.
But in the last chapter we saw that much experience is exiled “into
the dust-cloud” of psychology by introspectionism just because it
does _not_ agree at all with the properties of stimulation. To
be sure, the constancy of size, form, brightness, localization
and speed in the sensory field corresponds fairly well with the
constant properties of _physical objects_; but this agreement is
_not_ explained by the machine theory of vision which refers to
constant relations between local _retinal stimulation_ and local
experience.
If we are not satisfied by the alternative between order enforced
by preëstablished arrangements and order determined by acquired
arrangements, what else can produce order? For the moment we restrict
our discussion to sensory processes and return to our statement (cf. p.
111) that, in physical systems, the relative influence of topographical
conditions, on the one hand, and of the play of actual forces, on the
other, may vary enormously. In typical man-made machines the rôle of
topographical conditions prevails to such an extent that the only
rôle of dynamics is to drive processes along a path laid down by
those conditions. This means that typical machines are essentially a
_special_ type of physical system and that outside the little world
of man-made machines there exists an immense world of other physical
systems, in which the direction of processes is not completely
determined by topographical arrangements.
Let us consider a drop in a current of water which moves through a
narrow pipe. Why does it move? Because, aside from inertia, pressure
is higher on one side of it than on the other. But this difference of
pressure works in one direction only, the walls of the pipe excluding
all its other effects. Now let us suppose that the pipe disappears
and that the drop (and the whole column of water in the pipe) becomes
a part of a larger volume of water. The drop will probably also move
in its new environment. But now it is exposed to forces on all sides,
and its movement will be in the direction of the _resultant_ vector
of force. Obviously this movement is no less necessary than was the
movement in the pipe, but here there is no local arrangement which can
determine a single direction as the only one possible. Therefore the
path of the drop will be determined _dynamically_ in the new case,
i.e., by resultant force at each moment. It follows that in this
situation the path which our drop takes will depend upon the dynamical
situation it encounters at each stage, and that it will change when
this situation changes. This is one simple example out of millions.
_In all of them not only movement, or process as such, but also the
direction and distribution of process is determined_ _dynamically
by interaction_. It is events of this type which are excluded almost
completely from machines, and the same type of process is practically
excluded by standard neurological and psychological theory. _Gestalt_
psychology sees no ground for its exclusion and rather proposes to give
this type of process a fundamental rôle in psychological theory.
In the pipe the drop of water moves because movement, under
differential pressure, is an approach toward equilibrium. Such is
the effect of forces at all points of all systems. When surrounded
by water, the movement of the drop will still be an illustration of
the same rule.[21] The only difference is that now the direction
of movement, according to the rule, will depend upon the actual
_dynamical_ situation. If we consider all the drops in the given volume
of water, we will find the distribution of water gradually changing
from one moment to the next. But whereas in pipes the distribution of
the movement in space depends upon the form and the spatial position of
the pipes, in the “dynamical case” it depends mainly upon the play of
actual forces. In pipes, order is produced by _exclusion_ of dynamical
interaction; whatever distribution may result in the “dynamical case”
is _produced_ by dynamics itself.
At this point our interest concentrates upon this question, whether
Aristotelians and modern theorists are right in assuming that
_anything_ may happen in dynamical interaction and that, therefore,
dynamics is to be regarded almost as a synonym for disorder. What we
see around us in inorganic nature seems to corroborate that opinion,
since the blind meeting of forces and processes usually leads to chaos
and destruction. What we have before us in these cases, however, may
be described as follows: There is a thing at rest or a process going
on uniformly; suddenly a new factor impinges upon the first thing or
process from without, and after a short while, another new influence,
again independent, is exerted from without, and so forth. Anything
indeed may happen under these circumstances, and in most cases the end
result of such fortuitous concurrence is disorder and destruction.
This, I think, is the picture most men have in mind when they refer to
dynamics, as though accidental impact were its only form!
There are, however, other cases, much more interesting for our present
discussion. If in a basin, for instance, water is somehow distributed,
perhaps in full movement, at a given moment there is a definite
pressure at each point, and everywhere differences of pressure will
tend to alter the distribution and the direction of local drops of
water. Supposing now that the basin itself does not change and that
no outer influences accidentally impinge upon the system during its
redistribution, what will the result of continual internal interaction
be? If we try to find the answer by imagining an indefinite number
of drops, each moving under the resultant force in its immediate
proximity, and each again influencing its proximate neighbors by
that movement,--if we notice that this picture changes continually,
as the distribution, and, therefore, the actually resultant force at
each point changes in the smallest fraction of a second,--then we are
inclined to dismiss the task as beyond our efforts, and to treat the
events of this field in terms of the confusion or destruction that is
attributed to nature in the case of accidental impacts.
But in this we are wrong. We are projecting our own confusion into
the course of objective events. We become guilty of anthropomorphism.
The physicist has quite a different attitude toward the problem. By
observation, as well as by theoretical calculation, he is led to the
conclusion that, generally, undisturbed dynamical interaction will
produce a definite orderly distribution.
Let us review an example we entertained at the beginning of this
chapter. To Aristotelian theorists the striking order of astronomical
movements appeared inexplicable without the assumption of special
arrangements controlling them. Now in modern times no one believes
in those crystal spheres. But the order is there! And since the
stars did not “_learn_” to move so orderly, some factors other than
preëstablished and acquired topographical arrangements must be able to
produce and to maintain order in distribution and movement. And in the
prevalent conception of the solar system, it is continuous dynamical
interaction, without any topographical arrangements, which produced and
still maintains the order.
Other examples may be found in all parts of physics and chemistry:
when two atoms come into their sphere of mutual influence, the play
of dynamic interaction immediately begins, and, as the case may be,
depending upon their “relative properties,” they either separate again
or they form an orderly molecule,--an architectonic structure,--without
the aid of any arrangements _ad hoc_.
If we suspend a number of straight wires so that they form different
angles with each other, the whole distribution being irregular,
electric current entering the wires turns them into parallel lines.
This is an orderly result of electrodynamic interaction.
Or again, we pour oil into a liquid with which it does not mix. In
spite of the violent interaction of molecules at their common surface,
this surface remains sharply determined, not by any arrangement
enforcing this orderly distribution, but just by the play of surface
dynamics between the oil and the other liquid. If specific density
is the same for both liquids, these surface forces will change the
distribution until the oil forms a regular sphere swimming in the
other liquid.
I might go on to describe hundreds of examples. In all of them the
situation would be the same in principle. Dynamical interaction,
undisturbed by accidental impacts from without, leads to orderly
distribution, though there are no special preëstablished arrangements.
And what is the explanation of this general tendency in undisturbed
dynamics? It is simple enough. In all these systems we have one
resulting force at each point at each instant of time. All the
resultant forces together form one texture of stresses. From the
principles of physics one can deduce, therefore, that, _for the system
as a whole_, the immediate effect of all those forces will have one
definite direction. At each point the forces will produce changes
of movement or process which, when considered in their totality,
bring the system nearer to the balance of the forces themselves. The
factor of inertia may cause the real course of events to deviate from
the ideal exemplification of this principle. But where, as in most
organic systems, inert velocities not corresponding to actual forces
are destroyed by friction, the real distribution of processes will
exhibit the principle perfectly, and will finally reach a state of
stability, of rest or of stationary process. The fact that this state
will be an _orderly_ distribution has been explained by Ernst Mach as
follows: In orderly and regular distributions the totality of internal
stresses will be more balanced than in a state of disorder. Therefore,
by undisturbed interaction a system approaches order. For all details
I must refer the reader to the literature quoted at the end of this
chapter.[22]
_Dynamical self-distribution_ is the third kind of functional concept
which I propose to add to psychological theory, in addition to
distribution enforced by inherited arrangements and order determined
by acquired arrangements. More concretely and for the visual field,
my assumption is that the order and distribution in this field is in
each case the result of dynamical interaction. From this viewpoint the
processes underlying the visual field in a state of _rest_ represent
the equilibrated distribution of sensory dynamics under actually given
conditions. When not at rest, sensory dynamics will be in a state of
_developing dynamical distribution_.
Though the direction of local process in a system is not altogether
determined by local arrangement, the result of dynamical
self-distribution as a whole may still depend upon topographical
conditions (cf. pp. 112-113). Thus the electric current in a network
of wires is distributed dynamically; yet the actual distribution
as a whole depends upon the position of the electrodes and upon
the conductivities in all the conductors. Similarly, the totality
of processes in the optical part of the nervous system will depend
upon given conditions in each case. For the moment, though this will
prove to be an inexact assumption when it is examined more closely, I
shall suppose that in the interior of the optical network the general
conditions of conduction remain constant. But, then, as a set of
peripheral conditions, we have the patterns of different chemical
reactions on the retina, as they are produced in each case by actual
stimulation. Upon these varying conditions the self-distribution
of process will depend primarily. If neurologists are correct when
they assert that between the retina and the _area striata_ of both
hemispheres, conduction is a matter of isolated pathways, then the
_area striata_ will be a sort of “central retina,” in which the pattern
of retinal stimulation is copied by a pattern of central processes. In
this case _dynamical_ distribution will begin here, depending upon the
actual pattern of processes in the occipital lobes.
It will be evident that the task such a theory must face is enormously
more difficult than that of machine theory. Where arrangements are
altogether responsible for distribution, the concrete dynamical
properties of process are of little concern. If we do not know about
them, it does not matter very much. In dynamical theory, on the
contrary, the whole development of the theory requires that we know
those properties of the process. And since as yet physiology is not
very instructive concerning them, the extension of the theory depends
instead upon the assumptions which we make. Experimentation on visual
experiences will have to give us the necessary hints for making our
hypotheses, and the consequences of our hypotheses will then have to be
tested by experiments of the same sort. At the present time only the
first steps have been made in this direction, and it will take us a
long time before we feel firm ground under our feet. Let us remember,
however, that all the perplexities we may find on our way, and all
the mistakes we make in its course, are not to be referred to the
fundamental concept of self-distribution by interaction; they should be
referred to the particular assumptions concerning process, force and
interaction, which we have made.
As an example which will make our analysis more concrete, I shall
introduce one special problem. If from a given pattern, either on
the retina or on the “central retina,” processes start into the
conducting network beyond, and if their distribution is determined
dynamically, why should differences among the processes, corresponding
to differences in color, be preserved on the way? This is generally
the case: to the accurate retinal contour of the image of a letter
on a white page corresponds the sharp outline of the letter in my
visual field. If the totality of processes issuing from the area of
the letter remains detached somehow from the processes surrounding it,
we may understand, perhaps, how even in dynamical distribution the
most essential and crude geometrical properties of retinal pattern may
be sustained. But how can one set of processes remain detached from
the rest in dynamical theory? This embarrassing question certainly
could not occur for the assumption of isolated pathways. My answer
is based on the fundamental principle of the theory. If under these
circumstances one set of processes remains detached from the rest there
must be dynamical factors at work. Generally, processes corresponding
to a definitely colored area will have definite properties as a class
of processes, different from the properties of a class of surrounding
processes which corresponds to another color. They will remain
segregated in the nervous network if we suppose that in the ganglionic
fields, where they “touch” each other, their differential properties
provide _separating forces of contact_, so that they mutually exclude
each other. Take as an example the contact of oil and water. Here
interaction is so strong that the form of the surface is determined
by it; but this surface as such remains a sharp boundary and the
drop of oil remains detached from the water by those same molecular
forces which, at the time, mold the form of the drop. I shall assume,
then, that, in optical processes, contours are preserved by similar
forces of antagonistic contact, depending upon differences in the
properties on the two sides of the contour. From this viewpoint,
interaction will thus determine general distribution; it will also
have an influence upon size and form, and so forth, but it will not
generally “dissolve” processes, corresponding to a homogeneous area of
color, in the surrounding processes. This may seem a bold hypothesis.
But recent experiments have shown that this assumption is in the
right direction. It would be surprising if _all_ process differences
should act as separating surface forces of the kind just described in
the case of color, for in physics such forces exist between oil and
water, for instance, but not between water and alcohol, or a great
many other pairs of different materials. Indeed, when reëxamining the
results of older experimental investigations, Liebmann[23] proved that,
whereas very slight differences of brightness gave sharp contours,
neighboring areas of different color, even with maximal differences
of chroma _alone_, will show mutual diffusion.[24] We may conclude,
then, that mainly differences of brightness provide effective forces
of separation. Under normal conditions neighboring areas of different
chroma will almost always be different in brightness. Therefore, sharp
contours on the retina will in most cases give accurate contours in the
visual field.
Without further discussion we can draw a rather important conclusion.
We are accustomed to regard order enforced by rigid arrangements as
exceptionally secure. If, however, the order of sensory experience
be conceived as the result of sensory dynamics, this will seem to
most people like explaining the quiet life of an orderly citizen as
the outcome of many moral struggles and catastrophes. But, in this
connection, the chief point is that in those cases in which the machine
theory of visual order seems to be most convincing, dynamical theory
gives exactly the same results. As an example we may take the symmetry
of a seen circle, corresponding to the symmetry of its retinal image.
Is it necessary to explain this correspondence by insulated conductors
which maintain the geometrical properties of the retinal image all
along the line of those processes underlying visual experience? By no
means. Assuming that surface forces of separation keep the processes
of the circle distinct from the surrounding processes, the dynamical
influences exerted upon the processes of the circle will be the same in
all directions,[25] if, in the immediate neighborhood of the circle,
surrounding processes are homogeneous. Therefore, if the network has
the same conducting properties throughout that part of the nervous
system, i.e., if it is functionally “homogeneous,” there will be
not the slightest reason why the symmetry of the process should be
disturbed. In this case, dynamical theory gives the same result, then,
which until now has been rather clumsily explained by preëstablished
arrangements _ad hoc_.
Since the rôle dynamics plays is so slight in contemporary theory,
it may seem very strange to some psychologists. Hence I wish to make
the following statement expressly: _these concepts do not contain a
single thought in the direction of vitalism_. It is true that a number
of difficulties may be removed by them, difficulties which vitalism
can still raise quite properly against the claims of unqualified
“mechanistic” ideas. But, in our field of research, “mechanistic” ideas
have almost always been explanations by topographical arrangements,
and “dynamical” ideas, the type of idea prevalent in general physics,
occurred no more to the mechanist than to the vitalist.
I shall not try to develop more detailed theorems about sensory
dynamics until we become further acquainted with the main facts and
their significance. Some consequences of dynamical theory will be
obvious at once. From this viewpoint, sensory experience is allowed to
be as _fluid_ and _manifold_ as, to observation, it reveals itself.
Furthermore, as local processes are not isolated,--their existence
and their actual properties depending upon the dynamical context in
a larger area,--the visual field may prove to be _organized_. In this
case, as everywhere in physics, organization will be found to depend
upon what I have called the “relative properties” of stimulation. We
may eventually find specific properties in experience characteristic
of extended areas and not analyzable into local sensations, just as
existence exclusively as _functional wholes_ is a property of many
dynamical states in physics. The next chapters will be concerned with
the development of these matters.
BIBLIOGRAPHY
K. Koffka: _The Growth of the Mind_. 1924.
W. Köhler: _Die physischen Gestalten in Ruhe und im stationären
Zustand_. 1920.
W. Köhler: _Gestaltprobleme und Anfänge einer Gestalttheorie_. 1924.
W. Köhler: _Komplextheorie und Gestalttheorie_. Psychol. Forsch. 6.
1925.
W. Köhler: _Zur Theorie der Regulation_. Arch. f. Entwicklungsmech.
1927.
M. Wertheimer: _Untersuchungen zur Lehre von der Gestalt, I_.
Psychol. Forsch. 1. 1921.
M. Wertheimer: _Drei Abhandlungen zur Gestalttheorie_. 1925.
V
_Sensory Organization_
A DYNAMICAL distribution will be rightly regarded as a functional
whole. Take, for example, a simple electric circuit: the differences
of potential and the densities of current distribute themselves
along the conductors in such a manner that a stable or stationary
state is produced and maintained. No part of this distribution is
self-sufficient; local processes depend throughout upon the totality of
the distribution.
If a similar conception is to be applied to the processes underlying
sensory experience, we must avoid a mistake. Protesting against the
atomism which had been introduced into the treatment of sensory
experience, William James once said that, in the sensory field, local
experiences are interwoven with their neighbors in a manner which is
beyond the grasp of purely intellectual theory. He seems to think that,
in original sensory experience, there is uniform continuity and that
all cuts and boundaries are introduced later on for pragmatic reasons.
From the viewpoint of _gestalt_ psychology such a statement does
not correspond to the facts. In the last chapter we saw that,
notwithstanding the general dynamical interdependence throughout the
field, there may be boundaries in it where forces of “segregation” and
“separation” take the place of those of “coherence” operating elsewhere.