Henle - Some New Gestalt Psychologies
Henle - Some New Gestalt Psychologies
Research
© Springer-Verlag 1989
Summary. The present decade has seen a resurgence of in- new efforts to deal with problems of organization; at some
terest in perceptual phenomena investigated by Gestalt points it will be useful to try to clarify misunderstandings
psychologists. Some of these new psychologies call them- and to consider criticisms of what is now called "classical"
selves Neo-Gestalt psychology. Thus it seems appropriate Gestalt psychologyJ I shall be selective in my treatment,
to examine their relation to Gestalt psychology. confining myself, for example, to visual perception and re-
In addition to the problems they investigate, these new sisting the temptation to go beyond - or even to cover
psychologies usually have in common an information-pro- completely - the two collections in question, since too
cessing approach to perception. Thus information process- many issues arise for adequate treatment in a single paper.
ing is compared with Gestalt psychology; the major issues It may be remarked at the outset that most of the au-
to be considered are mechanism and atomism. In addition, thors discussed here have in common an interest in certain
a brief discussion of phenomenology is seen to be relevant problems investigated by Gestalt psychologists, especially
to these new psychologies, quite apart from their relation grouping and Prdgnanz, and that most of them take an in-
to information processing. After a short discussion of these formation-processing approach to these problems.
major issues, a still briefer examination of a few Neo- Since this is not the place for a detailed analysis of rep-
Gestalt psychologies and of the parallel distributed process- resentative self-styled Neo-Gestalt psychologists, I shall
ing (PDP) approach is undertaken. first state my conclusion, then indicate the kinds of analy-
sis that lead to it. Then we can consider, briefly and par-
tially, a few individuals.
The conclusion: I can find no kinship between these
Introduction new theories and Gestalt psychology.
The present decade has seen revived interest in problems It has just been said that information processing is
raised by Gestalt psychologists some three-quarters of a common to the various recent psychologies to be dis-
century ago. Michael Kubovy and James R. Pomerantz cussed. We can thus make some comparisons of it with
prefaced their edited volume with the remark: "This is the Gestalt psychology in advance of discussion of specific
first volume devoted to perceptual organization since writers.
K6hler's last book, The Task of Gestalt Psychology, was Ulric Neisser, who holds neither to information pro-
published in 1969" (1981, p. X). Soon after the publication cessing nor to Gestalt psychology, refers to the "mechanis-
of Kubovy and Pomerantz's collection, Jacob Beck edited tic information processing models, which treat the mind as
a volume on Organization and Representation in Perception a fixed-capacity device for converting discrete and mean-
(1982). These books may be taken as representative of a ingless inputs into conscious percepts" (1976, p. 10). This
now substantial literature on perceptual organization. In description neatly contains some of the issues that distin-
the collections mentioned, several authors explicitly label guish Gestalt psychology from information processing and
themselves Gestalt psychologists with specific qualifica- therefore from the Neo-Gestalt psychologies to be dis-
tions. In these cases and others, I shall raise the question: cussed. I shall specify three of these issues - although
What is their relation to Gestalt psychology? I shall con- they are not the only relevant ones. Hereafter I can merely
sider a variety of issues that arise in connection with these refer to them.
they insist on the role of dynamic factors that organisms By an analytic method he means the development of con-
and their nervous systems share with all of nature. These Structs and the investigation of the conditions and func-
invariant dynamics keep machines as well as organisms tional consequences of the phenomena he studies. He be-
functioning, and they are capable of producing the order lieves that Gestalt psychology is not analytic in this sense,
that is so conspicuous in perception and other organic but that it merely catalogues more or less holistic phe-
functions. (Cf., e.g., K6hler, 1969.) They have been ne- nomena. It would be a major project to set forth the long
glected by most theorists, who rely on constraints to pro- and distinguished record of experimental research carried
duce order. out by the Gestalt psychologists - no mere catalogues, but
research that is analyti c specifically in Garner's sense.
Atomism. A mechanistic model necessarily precludes free In another sense, Garner believes (1981, p. 119) that
interaction among the relevant processes. Connections and "when we want to understand a phenomenon, we must
even networks of fibers are no substitute for interactions. somehow analyze it, take it apart, see what its components
Such a system of "discrete inputs" is therefore atomistic. are, and otherwise subdue the holistic percept into ana-
On the psychological side, the elements of contemporary lyzed components." This kind of analysis, we have seen, is
theorizing are likely to be features, dimensions, and at- exactly the opposite of that of Gestalt psychology, which
tributes rather than the sensations of earlier times. starts with wholes, respects them, and does not subdue
An atomistic system proceeds in perception by piecing them. Wholes may be analyzed, but not into predeter'
wholes together from parts or pieces; it is contrasted with mined attributes, rather into their natural parts and sub-
an approach in which the role and function of a part de- structures.
pends upon its whole. In its elementarism, too, informa-
tion processing contradicts the basic propositions of Ge- Information processing. Although he does not call himself a
stalt psychology. Wertheimer (1944, p. 84) has summarized Neo-Gestaltist, James R. Pomerantz applies the concepts
in another connection: of information processing to Gestalt problems. He
The basic thesis of gestalt theory might be formulated thus: there criticizes Gestalt psychology for the use of subjective phe-
are contexts in which what is happening in the whole cannot be nomenological methods, not seeing that his own perfor-
deduced from the characteristics of the separate pieces, but con- mance methods depend equally upon the experience of the
versely; what happens to a part of the whole is, in clear-cut cases, experimenter and are in this sense equally subjective:
determined by the laws of the inner structure of its whole. somebody, after all, must read the reaction times and other
measures. That Gestalt psychologists go beyond phenom-
Phenomenology. Another idea shared by many of the Neo- enology need not be repeated.
Gestaltists to be considered, though it has nothing to do In connection with the pervasive ignorance of the Ge-
with information processing, is that Gestalt psychologists stalt research literature, it is interesting to note that Pomer-
have confined themselves to the phenomenological meth- antz stumbles on some of the distinctions made long ago in
od or, as Attneave is cited (Pomerantz, 1981, p. 143) as Gestalt experimentation. For example, he comes close to
calling it, the "look-at-the-figure-and-see-for-yourselF' K6hler and yon Restorff's distinction between monoto-
method. It is true that Gestalt psychologists regard phe- nous and distinctive similarity (1935; cf. K6hler, 1940),
nomenology as preliminary to experimentation; they and, also in connection with perceptual similarity, he ap-
want to make sure what it is that they are studying. But it is proaches Goldmeier's distinction between mode and form
simply not true that they stop with demonstration. Even (Goldmeier, 1937, 1972). Pomerantz does not know what
superficial acquaintance with the research literature of to do with either distinction because he is not acquainted
Gestalt psychology will show an impressive record of ele- with their theoretical context.
gant experimentation. Pomerantz assigns the main work of perceptual group-
Now we may consider a few of the new perceptual psy- ing to a murky, preattentive level below the level of con-
chologies that give themselves some kind of Gestalt label. sciousness, probably presided over by a gatekeeper. 3
Each will be taken up briefly to bring out one or two Grouping is made a matter of failure of selective attention
issues. - a failure to attend selectively to a single part of a dis-
play. It seems strange not to see it, rather, as a positive
Analytic Gestalt psychology. W.R. Garner considers that achievement of the nervous system. Some grouping (by
his interest in Gestalt phenomena 2 and his analytic experi- proximity, similarity, common fate) is made a matter of
mental method make him an analytic Gestalt psychologist. bottom-up processing - that is, data driven, beginning
with the "raw stimulus," while good figure, good continua-
tion, and Pr~ignanzseem to depend on top-down process-
2There are no Gestalt phenomena. I suppose that Garner (and ing, starting with a hypothesis and its application to the
other authors to be discussed) is referring to psychological phe- data. 4 But Gestalt psychology is seen not to fit: wholes and
nomena to which Gestalt psychologists have called attention and whole properties, says Pomerantz (1981, p. 171), are "not
which they have investigated. derived from the output of line detectors," but are a "pure
3It must be added that Pomerantz sees some difficulties with the or unmediated perceptual experience." If these are direct-
gatekeeper analogy. Nor does he regard it as completely adequate, ly perceived, we may ask, why not other aspects of per-
since perception also has "strategic" (i.e., hypothesis-testing) ception?
properties.
4As Michael Wertheimer (1985, p. 26) has pointed out in a It is not clear to me whether Pomerantz is moving to-
comparison between Gestalt psychology and contemporary cogni- ward a more adequate approach to his material, departing
tive psychology, "the similarity between 'from above down' [an from information processing, or whether he is merely con-
expression of the Gestalt psychologists] and 'top-down process- signing the interesting questions to an unknown realm
ing'.., is purely verbal." where they need not disturb us.
83
Neo-Gestalt psychology. Calling his theory Neo-Gestalt one can apply to it the assumption of rectangularity rather
psychology, Frank Restle (1982, p. 39) admits that "any than, say, that of circularity? The proximal stimulation it-
theory given the name 'Neo-something' is not well self does not possess this organization. This is a question
defined." He relates to Coding Theory theoretical ideas that Perkins himself sees and, of course, cannot answer.
that he feels "owe heavy debts to Gestalt theory" (1982, Again, how is geometrical knowledge possible at all -
p. 39). His aim is "to integrate parts of Gestalt theory with knowledge of rectangularity, symmetry, parallelism, etc.
information processing theories through the use of coding - in advance of actual perception of rectangles and the
theory" (1982, p. 31). Since Restle believes that "informa- others? In both respects, organized perceptions are pre-
tion processing models of p e r c e p t i o n . . , do not seem to be supposed, not explained; the problem of organization is
apposite to Gestalt Theory" (1982, p. 47), it is hard to see not addressed.
what is gained by such an integration. Perkins makes use of a minimum-energy state which he
In applying coding theory to the description of a fig- believes to be related to the Gestalt principle of Prdgnanz.
ure, Restle enumerates parts of the figure - lines, angles, But he regards it as dealing with the behavior of a quasi-
and other features that, he believes, permit "reconstruction mechanical system, which immediately distinguishes it
of all the essential characteristics of the display" (1982, from the thinking of the Gestalt psychologists. It is the
p. 31). The next step is to simplify the code by the use of now familiar difference between machine theory and a dy-
the rules developed by Emanuel Leeuwenberg (1971). namic theory. Why, according to the Gestalt psychologists,
Restle makes it clear that "coding theory begins with a do systems achieve increasing regularity, symmetry, and
primitive code that consists entirely of a close analysis of simplicity in the distribution of their material and forces as
local details of the display" (1982, pp. 49-50). Again, "the they approach equilibrium? Referring to Mach, K6hler
'earliest stage' of coding of a line drawing or figure is the points out (1969, p. 59):
extraction of lines and angles" (1982, p. 47). Thus "a
When such regular distributions are being established, more and
square may be coded as continued repetition of an angle t~ more components of the acting forces are likely to balance each
and a length L" (1982, p. 48). The procedure, in short, is to other, which means that under these circumstances the equilib-
start from the parts in order to arrive at a whole. The rium or a steady state is quickly or gradually approached... It is
procedure of Gestalt psychology, it has been seen, is just therefore not surprising that during this operation the distribu-
the opposite. tions within the system become more regular, symmetrical and
simple.
D. N. Perkins. A different Neo-Gestalt theory of percep- Given this state of affairs, it takes only the hypothesis
tion is put forth by Perkins. Since given proximal stimuli of psychophysical isomorphism to show the relevance to
might be projections of any number of objects in space, perceptual processes.
Perkins argues that the observer "adds something" - most The process of achieving Priignanz here envisaged is
likely geometrical assumptions - to produce the percept. thus seen to be entirely different from Perkins's own mini-
"The perceiver often resolves the projective ambiguity of mum energy principle or from any of the other mechanis-
stimuli by 'reading in' certain geometric regularities such tic theories (which I have omitted) that he finds accept-
as rectangularity or symmetry and deriving a three-dimen- able.
sional interpretation, one both exhibiting those regularities
and consistent with the stimulus" (Perkins, 1982, p. 73). We have now looked briefly at a few of the so-called
There have been numerous theories in the history of Neo-Gestalt psychologies of perception recently devel-
psychology in which some external process is thought to oped. There is more to say about each of them, and there
have been imposed on, or added to, retinal stimulation (or, are others. I could, for example, have discussed Fred Att-
in some cases, sensory data) to yield a percept. Of these the neave's soap bubble theory (1982), which he considers to
most famous is Helmholtz's; he held that from physiologi- be in the best Gestalt tradition, but which could easily be
cal stimulation (or sensory data) inferences are drawn as to shown not even to be a proper analogy for a soap bubble.
the external object that gave rise to the stimulation. Since In Attneave's system, the interactions occur piecemeal, by
no evidence of such inferences can be found in conscious- way of dipoles, or points in space considered two at a time
ness, the inferences, which derive from past experience, - not the result of forces within the system as a whole.
were said to be unconscious. Similar theories have become They depend on neural machinery; and so much of it is re-
popular in the field of perception today; and there have quired that Attneave himself wonders whether there are
been a number of variants of Helmholtz's theory. enough neurons in the visual cortex to account for the very
It can be seen that Perkins's theory is of the same type: numerous interactions among dipoles. Both his atomism
retinal stimulation whose ambiguity is resolved by the im- and his mechanism distinguish Attneave from the Gestalt
position on it of geometric assumptions. Perkins states that tradition.
"no subject ever reported any conscious geometric strategy Or I could have included Stephen Palmer's (1982) at-
for performing the task" (1982, p. 80). Thus his geometric tempt to integrate three approaches to perception: Gestalt
assumptions must be as unconscious as Helmholtz's in- theory, Gibsonian theory, and information processing. But
ferences. having found that Gestalt psychology and information
Although Perkins calls his theory a Neo-Gestalt theory processing are incompatible because of the mechanism
of perception, it is interesting that this kind of theory has and the atomism of information processing, I see no point
many times been criticized by Gestalt psychologists (from in trying to extend a failed integration.
1913 on). Transposing some of the criticisms to the present A more extended discussion would have enumerated
theory, the Gestalt psychologist would ask: How does the some of the old misunderstandings of Gestalt psychology
perceiver know which geometrical assumptions to apply in that persist in the Neo-Gestalt theories and contribute to
a particular case? Must one not perceive a rectangle before their differences from Gestalt psychology. About them I
84
shall merely say that they will not go away unless one ther for an adequate account of perception or for an at-
reads the sources. Restle is probably correct in saying that tempted rapprochement with Gestalt psychology.
anything called "Neo-something" is not well defined. If Another recent attempt to bring together PDP and
the Neo-Gestaltists want to build on Gestalt psychology, Gestalt psychology comes from Cees van Leeuwen (1989).
they would do well first to get it right. Then they would Since the problem of integrating mechanistic information
know what has to be done. processing with nonmechanistic Gestalt theory is not ad-
I do not know whether the misunderstandings arise dressed here either, this effort can be dealt with briefly.
from ignorance of the literature (including the research lit- Van Leeuwen achieves his goal in part by sacrificing iso-
erature), from the inhospitality of the current intellectual morphism and advocating the study of mental phenomena
climate to Gestalt ideas, or from both. But I doubt whether without regard to the corresponding physiological facts.
Kubovy and Pomerantz would feel embarrassed in talking This is a sacrifice that most Gestalt psychologists are un-
about Gestalt psychology (1981, p. ix) if they were talking willing to make. It is not within the scope of this paper to
about the same psychology that Wertheimer, K6hler, and defend isomorphism. But even its opponents would be
Koffka discussed. hard pressed to demand that we throw out a powerful
heuristic for the sake of a problematic systematic rap-
PDP. Although my original intention was to limit myself prochement.
to the volume edited by Kubovy and'Pomerantz and that To make plausible his rejection of isomorphism, van
by Beck, a few words about another recent and increas- Leeuwen takes von Ehrenfels as representative of Gestalt
ingly popular development may not be out of place. psychology without any examination of the differences
The two papers now to be discussed consider a newer between his thinking and that of the subsequent exponents
kind of information processing, parallel distributed pro- of Gestalt theory who have come to represent this theory
cessing (PDP) in relation to Gestalt theory. Unlike the in the minds of contemporary psychologists. Such an ex-
papers previously reviewed, these are not concerned with amination would throw serious doubt on the advisability
specific perceptual phenomena, but rather with what the of this theoretical step.
authors see as the metatheory of Gestalt psychology. As Van Leeuwen sees differences between Gestalt psycho-
with the papers discussed above, only one or two issues logy and PDP: for example, what he calls the empiricistic
will be extracted in each case, although there is more to bias of PDP (which is not the same as the empirism
be said. criticized by Gestalt psychologists). But the core issue,
It must be understood that the following is not a dis- mechanism vs. dynamics, remains untouched and, in my
cussion of PDP, but merely of its use by theorists who view, calls into question van Leeuwen's conclusion that
would like to apply it to Gestalt theory. Gestalt psychology could indeed make use of PDP mo-
William Epstein (1988) believes that, although Gestalt deling.
psychology has not fulfilled its promise (p. 4), there are as-
pects of its metatheoretical commitments that are worth
A final word. All investigators have the privilege, indeed
saving and that PDP may provide the means of doing it.
the obligation, to carry their thinking in any direction that
This model, also referred to as connectionism, is being de-
seems fruitful to them. What they label their approach is
veloped by David E. Rumelhart, James L. McClelland,
likewise their own concern. But to call such incompatible
and the PDP Research Group, and by others as well.
approaches as information processing and Gestalt theory
Since I want to make sure that we are talking about the
by the same name seems to me confusing. Theoretical psy-
same thing, I shall mainly follow Epstein's account of
chology is difficult enough without adding semantic con-
PDP. It arose, he points out, from the dissatisfaction of the
fusion.
PDP researchers with "the standard information process-
Needless to say, I welcome the new interest in Gestalt
ing theory of the 1960s and 1970s" (1988, p. 5).
problems and would welcome new work in the spirit of
The fundamental characteristic of the PDP model is the assump- Gestalt theory, work that advances the task of Gestalt psy-
tion that information processing takes place through the interac- chology and corrects the theory where appropriate.
tions of a large number of simple processing units, each sending
excitatory and inhibitory signals to other units ... There are only
simple units operating in parallel in that many units carry out
these computations at the same time. References
Units are connected to one another. It is the pattern of connec-
tivity that constitutes what the system knows and determines how Attneave, F. (1982). Pr~ignanz and soap bubble systems: A theore-
it will respond. (Ibid.) tical exploration. In J. Beck (Ed.), Organization and represen-
tation in perception (pp. 11-29). Hillsdale, NJ, and London:
One kind of information processing, in short, is being sub- Erlbaum.
stituted for another; but the mechanism characteristic of Beck, J. (Ed.) (1982). Organization and representation in percep-
information processing has not been altered. Indeed, Ep- tion. Hillsdale, NJ, and London: Erlbaum.
stein sees no need for doing so: he speaks of the human Epstein, W. (1988). Has the time come to rehabilitate Gestalt
brain as a machine (1988, pp. 2, 4). Such a conception, it theory? Psychological Research, 50, 2-6.
has already been shown, has no resemblance to Gestalt Garner, W. R. (1981). The analysis of unanalyzed perceptions. In
psychology. M. Kubovy & J. R. Pomerantz (Eds.), Perceptual organization
(pp. 119-139). Hillsdale, NJ: Erlbaum.
The information processing units, in the thinking of Goldmeier, E. (1937). Uber .g,hnlichkeit bei gesehenen Figuren.
the PDP group, are models of neural networks. But it must Psychologische Forschung, 21, 146-208.
be added that molar processe s also occur in brains that are Goldmeier, E. (1972). Similarity in visually perceived forms.
not accounted for in terms of the activity of neurons, even Psychological Issues, 8, No. 1, Monograph 29. (First published
in networks. Such molar processes cannot be neglected, ei- in 1937, translated and enlarged by E. Goldmeier.)
85
K6hler, W. (1940). Dynamics in psychology. New York: Liveright. Pomerantz, J. R. (1981). Perceptual organization in information
K6hler, W. (1969). The task of Gestalt psychology. Princeton, NJ: processing. In M. Kubovy & J. R. Pomerantz (Eds.), Perceptu-
Princeton University Press. al organization (pp. 141-180). Hillsdale, N J: Erlbaum.
Kubovy, M., & Pomerantz, J. R. (Eds.) (1981). Perceptual organi- Restle, F. (1982). Coding theory as an integration of Gestalt psy-
zation. Hillsdale, N J: Erlbaum. chology and information processing theory. In J. Beck (Ed.),
Leeuwenberg, E. (1971). A perceptual coding language for visual Organization and representation in perception (pp. 31-56).
and auditory patterns. American Journal of Psychology, 83, Hillsdale, NJ, and London: Erlbaum.
307-349 Van Leeuwen, C. (1989). PDP and Gestalt. An integration?
Neisser, U. (1976). Cognition and reality. San Francisco: W.H. Psychological Research, 50, 199-201.
Freeman. Wertheimer, M. (1944). Gestalt theory. Social Research, 11,
Palmer, S. E. (1982). Symmetry, transformation, and the structure 78-99. (Translation of lecture at the Kant Society, Berlin,
of perceptual systems. In J. Beck (Ed), Organization and 1924).
representation in perception (pp. 95-144). Hillsdale, NJ, and Wertheimer, Michael (1985). A Gestalt perspective on computer
London: Erlbaum. simulations of cognitive processes. Computers in Human
Perkins, D. N. (1982). The perceiver as organizer and geometer. In Behavior, 1, 19-33.
J. Beck (Ed.), Organization and representation in perception
(pp. 73-93). Hillsdale, N J, and London: Erlbaum.