Gestalt research on problem-solving and today’s Gestalt

This article explores Gestalt Theory's approach to problem-solving, focusing on Köhler’s studies on tool use in chimpanzees and Duncker’s “think aloud” problem-solving research. Both works exemplify productive thinking as a form of restructuring behavior. The article links Gestalt Theory to modern research, highlighting its comprehensive explanatory power in understanding perception, memory, and thinking, particularly through concepts like figure-ground separation and grouping. Gestalt Theory's broad applicability in cognitive psychology is emphasized, advocating for its continued relevance and study.

Esra Mungan

22 Sep, 2023
Gestalt research on problem-solving and today’s Gestalt


This article will focus on two pioneering scientific works in problem-solving, or per Gestalt theory, “productive thinking”. One of them is Köhler’s research on goal-directed tool use and overcoming obstacles in chimpanzees, the other is Duncker’s studies on problem-solving using a “think aloud” technique. In both Köhler’s and Duncker’s work, productive thinking is linked to a restructuring behavior, such as tearing off a thin, long branch from a tree to serve as a stick or removing various objects in a matchbox to transform it from a container to a base on which to mount a candle. In the final section, I will look at how today’s research trends might and should connect to Gestalt theory. In conclusion, just as in memory, the main tenets of Gestalt theory of figure-ground segregation and grouping play a critical role also in thinking. Thus, Gestalt theory seems to be able to come up with a common way of understanding perception, memory, as well as thinking. To my knowledge there is no other single theory within cognitive psychology that has such broad an explanatory power. This in itself is one more reason why Gestalt theory deserves crucial attention across all of cognition, even all of psychology and possibly even beyond.

Gestalt Research on Problem-Solving and Today’s Gestalt


Maryam Mirzakhani (Winner of 2014 Field Medal)

1. Introduction

This article is the third in a series of articles I wrote on Gestalt theory with the purpose of sorting out many misunderstandings as well as “missed” understandings about this astonishing, strongly inspiring and rather grand theory. The first article focused on how the theory came about, its founders, the main tenets and its more or less familiar works on perception (Mungan, 2020). In the second article I examined their completely unknown yet highly interesting propositions and works on memory (Mungan, 2021a). In this article, I will look at their studies and conceptualizations within the field of thinking and problem-solving. I will cover two lines of their research, Wolfgang Köhler’s at its time (and even today) ground-breaking work on chimpanzee problem-solving and Karl Duncker’s highly creative, impressive work on human problem-solving. Duncker, a second-generation Gestaltist who received his graduate training next to the Gestalt founders back in Germany, enjoyed a lot of attention by the followers of the “cognitive revolution” after his emigration to the United States to escape Nazi persecution as the son of an activist socialist family. Despite the American cognitivists’ interest in his work, his most cited study has been misrepresented in the majority of these citations, and even worse, presented completely stripped of its theoretical Gestaltist grounding. Finally, in the last section of this article, I will touch on some studies from different fields of psychology and computer science and look at their possible relation to the propositions of Gestalt theory. Table 1 presents a bird’s eye view of the article’s subsections.

Table 1

Outline of the Article


1. Introduction 5

Table 1 6

2. Thinking: Problem-Solving, ‘Productive Thinking’, Creativity 6

2.1 Wolfgang Köhler’s 1921 “The mentality of apes” book 6

2.1.1 The ability to use roundabout paths: About the haphazard versus the insightful 7

2.1.2 Interacting with tools 11

2.2 Karl Duncker’s 1935 “The psychology of productive thinking” book 17

2.2.1 Restructuring the problem 18

2.2.2 Heuristic methods 19

2.2.3 Styles of thinking: On analytic and synthetic “reading” (or explicability) 20

2.2.4 Learning and partial insight 22

2.2.5 Seek and find 25

2.2.6 Constructing the image 27

2.2.7 An example of Duncker’s experiments: The box problem 27

2.3 The mind of a mathematician: Maryam Mirzakhani (1977-2017) 32

2.4 Gestalt in thinking: A short evaluation 35

3. Today’s Gestalt 35

3.1 Cognitive psychology 35

3.2 Developmental psychology 37

3.3 Neuroscience 39

3.4 Computer science 41

4. In Conclusion 43

5. References 46

2. Thinking: Problem-Solving, ‘Productive Thinking’, Creativity

2.1 Wolfgang Köhler’s 1921 “The mentality of apes[1]” book

This extensive monograph by Köhler came out of his 1913-1917 investigations on chimpanzee problem-solving at the Prussian Academy of Sciences anthropoid research station on the Canary island Tenerife. Immediately on the second page he states that the goal of this book is to put into question the mechanistic, associationistic view of the empirists[2] which was then (and for that matter is still today) the de facto understanding that dominated all of science. Köhler draws attention to the difference between random, non-intelligent and non-random, intelligent behaviors and states that even an empirist would not deny this difference by simple observation. He remarks that his book will focus on these nonrandom behaviors with the aim of understanding their common characteristics. He criticizes Thorndike’s famous studies with cats and dogs for creating setups that make impossible any behavior other than trial-and-error. For instance, placing a cat into a claustrophobic cage with strings, spools, and hinges all around makes it impossible for the cat to understand anything about the setting. Probably, if you put a human in there, they would also have to resort to trial-and-error behavior. But according to Köhler, the problem was not only Thorndike’s bizarre cages. He likewise criticizes American animal psychology research for mostly using uniform mazes that did not provide any opportunities for the animals to discover meaningful connections. In contrast, he says, when presenting animals with settings that allow for a whole-scene inspection, hence for the detection of tools and their possible relations to goals via perceptual principles of grouping and figure-ground, we will start observing non-randomness, that is meaning, not only in animals’ correct behaviors but also in their mistakes.

2.1.1 The ability to use roundabout paths: About the haphazard versus the insightful


Figure 1 presents Köhler’s sketch of one of his pilot studies to precede his actual series of experiments.

The basket in Figure 1 was one of those used to feed the chimpanzees. The setup was such that the basket, which was hanging 2 meters above the ground, was tied to a string that passed through an iron ring mounted on a wire-roof, and from there to the top branch of a tree. After preparing the setting, Köhler then goes on to observe what the chimpanzee will do when entering the playground. According to Köhler’s description, the chimpanzee first looked at the basket for a period of time, then all of a sudden, after a brief burst of rage (which Köhler interprets as due to having been left on their own on this new site and away from the other animals), started to climb up to the top branch, grabbed the string and pulled it with his eyes fixed on the basket. As the basket would reach the iron-ring he would let go, and watch it fall back to where it was. He would then pull the string much more strongly such that the basket would hit the ring with more force, make the basket skip over and one of the bananas fall down. The chimpanzee would climb down, take the banana, climb all the way up again, pull the string even more fiercely, make the string break off and the entire basket fall down. The chimpanzee would then climb down again and collect the remaining bananas inside the basket. In this entire setup, the only thing that was familiar to the chimpanzee was the basket itself. He did not, for instance, witness the preparation of the setup or anything similar.

Köhler describes another observation, this time with a female dog[3]. This, too, served as a pilot for his subsequent work. Here, Köhler prepared a setup where the short-cut route to a desirable target was blocked. The only way to reach the goal was by taking a longer detour route. The question was whether the dog would be able to suppress the immediate response and do the roundabout which first would take her away from the target. There were two versions of the setup. In the first one, the dog was behind bars with the food placed outside at an unreachable distance (Figure 2a). Köhler noted that the dog hesitated for a moment and then quickly ran in the opposite direction around the 2x2m2 fenced field to finally reach the food. The obstacle was overcome and a happy ending reached[4]. In the second setup (Figure 2b), the same female dog reached the target even more easily. Köhler then introduced a critical change by placing the food immediately behind the grid. This time the dog became completely fixated on the target and repeatedly tried to reach for it through the grid, but in vain. In other words, it would not ‘occur to her’ to first move away from the food which was the necessary first step to be able to run around the grid and get to the food.



Figure 2a ve 2b (Köhler, 1921; red arrows and blue text added by author)

In the setup depicted in Figure 3, Köhler would now observe the behavior of chickens. The behavior would reveal a much more random, zigzagging behavior that was, if not chaotic, at least completely devoid of insightful behavior, i.e., devoid of a {goal --> obstacle --> overcoming the obstacle --> reaching the goal} type of intelligent-like behavior. He remarks that none of the chickens ever showed a behavior as depicted in Figure 3b of smoothly moving away from the target and then around the fence to reach the outside goal. Instead, those who ended up reaching the goal reached it haphazardly through trial-and-error.


Figure 3 (Köhler, 1921)

With these demonstrations Köhler was able to exemplify the difference between haphazard (the chickens’) and insightful (the dog’s) behavior. He then goes on to test this ability to do roundabouts to reach a desired goal with chimpanzees, using a variation of testing setups. The findings were similar: Just as the female dog in Figure 2a, chimpanzees, too, would immediately choose to take the route where they had first to move away from the target to reach it. Köhler emphasizes that what characterizes this “insightful” behavior was its integrated, smooth flow, from start to end. Haphazard behavior, on the other hand, showed characteristics of being a mix of disjointed, incoherent pieces of movements of varying speeds. Here we see again a very meticulous and encompassing way of analyzing events, the sine qua non of Gestalt theory. Whereas the empirist would only look at whether the animal succeeded or not, how many error turns they made or the overall time it took them to reach the goal, Köhler carefully inspects the entire sequence of movements looking at its various qualitative aspects as well. We furthermore see that Gestalt theory does not confine itself to a narrow either-or dichotomy of “trial-and-error learning” versus “insightful learning”. I think the biggest problem with American mainstream psychology of the past decades is its tendency to not let go of dealing with issues in a constant juxtaposition of “two camps[5]”. In Köhler’s example we see both cases, one where the behavior is clearly a trial-and-error type behavior and one where it clearly is not of that kind. Nature hosts both types of learning and it will often depend on multiple factors such as the animal species, age, the clarity or obscurity of a particular setting etc.

2.1.2 Interacting with tools The use of tools

When analyzing tool use in different settings, Köhler notes that different tools are used for the same purpose. Köhler describes that in order to pull something towards them from behind bars where their arm could not reach it, chimpanzees would first look for a stick. If they could not find one, they would try to pull it with the rim of a hat, for instance. Köhler reports that if there was no object around, they would tear off a branch from a tree to use as a stick. Each of these choices share a given common “Gestalt” where their visual (and tactile?) characteristic is re-envisioned as a tool to reach a certain end. According to Gestalt theory, for problem-solving to occur, one has to grasp the common structure of different objects. In this example it would be “something that is sufficiently solid and long”, hence, a string would not work, and neither would a long yet soft leaf work, yet a stick or a hard-fold hat or a ripped-off branch from a tree would work. As a matter of fact, chimpanzees generally turn to the right tools right away, says Köhler[6]. As a critical manipulation, Köhler would also vary the visibility of the objects. Sometimes the critical tool would be displayed as a single, isolated object in an open field (hence presented with utmost singularity/Prägnanz[7]), sometimes it would rest among a multitude of useful and useless objects and hence be rendered less visible. In the latter case, discovering the critical tool will often require a haphazard shuffling of the things around them. From this shuffling and mixing, new groupings and segregations, new figure-ground constellations will emerge which might cause an earlier “ground” member to come to the forefront as a “figure”. Hence, such planned or unplanned “messing up[8]” behavior is rather useful in finding solutions, says Köhler. Any given object will be noticed as a tool much faster once (or possibly only if) it is a figure rather than an indiscernible part of a ground. In contrast, the case where the chimpanzee casts an eye on one of the branches of a tree and tears it off is something different from a simple act of reshuffling to find a tool. Here, something that originally is an intricate part of an “inseparable” whole (i.e., the tree) is now re-envisioned as a piece in itself that carries the properties of “something that is sufficiently solid and long”. This, according to Köhler, is a more major restructuring that points to a much grander cognitive flexibility and ability for “creative transfiguration”.

In another experimental setup, the chimpanzee would see a banana hanging down from the top at an unreachable height. Around them would be sticks of varying lengths as well as fruit boxes that were empty or filled with things, facing upwards, sideways or downwards. Figures 4a-c show the different solutions of the chimpanzees. Köhler meticulously reports how each of the seven chimpanzees approached the problem, their first acts, which behavior they continued to show for how long, and even their anger fits in between. Some would immediately go for the most salient stick, try it out and then go to a longer stick when they noticed that the first one fell short of reaching the banana. Here Köhler remarks that going for the short stick, albeit a mistaken choice, is a “good mistake”. In other words, he groups the behaviors not simply as right and wrong but as “correct behaviors, “good mistakes”, and “bad (i.e. random) mistakes[9]”. He also carefully notes down how many times and in which ways they tried to put the boxes on top of each other. He would note that while putting the box sideways (as the second and third boxes in Figure 4a) seems illogical, it is however a ‘practical’ way to increase the elevation with a single box (though at the expense of durability). Sometimes the boxes would contain heavy stones, which meant that they first had to be emptied. Such type of “obstacles” would delay the solution yet be overcome sooner or later, notes Köhler. In one part he records that when the chimpanzees would see one of the keepers, they would first thing run to them, take their hand and pull them over towards the fruit, try to climb on them, and get angry when the keeper refused to cooperate.




Figure 4a-c (Köhler, 1921)

Under tool use, Köhler also lists behaviors such as using their drinking straws to catch termites (in their natural habitat chimpanzees would use thin branches whose outer skin they would often first peel before using it or use thin and sufficient sturdy leaves), arranging the hay to make their sleeping place softer and more comfortable (a luxury they do not enjoy in their natural habitat), using sticks as a weapon (to hit someone or break something), shoving away things they are disgusted by (e.g., feces) with utensils so as not to touch them. Transfiguring and manufacturing objects

Köhler not only observed how an object was used as a tool but he also carefully noted the cases where a tool was obtained by transfiguring it or making it from scratch. In one example, one of the chimpanzees would take a long stick and acrobatically use it as a pole (Figure 5). In another example, again where a banana was hanging from high up with various bamboo sticks of different lengths and diameters spread out on the floor, one of the chimpanzees would pick multiple sticks and after a few failures succeed to make a longer stick by nestling one into the other (Figure 6). Köhler also reports of cases where a chimpanzee would use a stick as a lever and even sharpen one of its ends. He would also record other behaviors such as a chimpanzee teasing the chickens by first giving them food only to take it away from them or a chimpanzee watching one of their mates get on top of a pile of boxes to reach out for the banana and topple the pile right where the other was about to get the banana. Köhler also observed how the chimpanzees would solve the problems when together rather than alone. He would stress that each chimpanzee mostly tried to reach the goal on their own with close to no evidence for planned cooperation and division of labor (Figure 7).


Figure 5 (Köhler, 1921)


Figure 6 (Köhler, 1921)


Figure 7 (Köhler, 1921)

1.2.3 General evaluation

Due to lack of space, it is impossible to list all the meticulous and ingenious experimental setups and observations that Köhler describes across his 207-page long monograph. His most critical conclusion is that an animal’s way of solving a problem cannot be boiled down to past experience only as the same animal would behave differently (more or less successfully) depending on how the visual layout was changed as this instantly affected their “phenomenal field”, similar to humans. This is something that has been utterly ignored or denied (and, in my opinion, still is) by what he then called “association psychologists”, or to use his later terminology, by empirists, such as the Behaviorists. Köhler emphasizes that chimpanzees would demonstrate clear evidence of non-random, intelligent behavior even in situations where they might fail the test. But most important of all, Köhler’s tour de force work shows that problem-solving skills in animals can be investigated using creative, non-reductionist methods, but at the same time, not fall into the trap of anthropomorphism. We see such rigorous, comprehensive and contextualized investigations of animal cognition only four to five decades later in the 1960s and 70s. To me, Jane Goodall is one of those long-ignored pioneers who came up with a completely different perspective to looking at animal behavior where she rejected all kinds of anthropocentric approaches that dominated animal research then (and, I believe, even today). In that sense, I think her humbleness in trying to understand chimpanzees from their perspective is reminiscent of what Köhler did and what led him to courageously call those behaviors insightful, and this at a time when Behaviorism was dominating all of psychology[10].

2.2 Karl Duncker’s 1935 “The psychology of productive thinking” book

Karl Duncker (1903-1940) is typically known as the first person to systematically and scientifically research human problem-solving and productive thinking[11]. In his 1935 book titled The Psychology of Productive Thinking, Duncker reports his rigorous post-doctoral work on productive thinking where he uses creative experimental setups that include mathematical problems as well as more everyday problems. His goal throughout is to carefully understand how the mind thinks, how a person reconstructs or fails to reconstruct, to re-envision a problem, what their thought processes are as they try to tackle these and how certain manipulations facilitate or block the solving process. For this, he used a method called “thinking aloud” where participants were encouraged to say everything that passes through their minds aloud while working on the problem, including the ones they might find absurd. He notes that this is clearly different from introspection in the Wundtian and post-Wundtian tradition. Whereas in introspection the person is turned into a self-observer, in Duncker’s setups they were only asked to speak out loud what crossed their minds without an “about” evaluation of their thinking. In his studies, Duncker was also careful to only include the kind of problems which would not require any pre-knowledge. His interest was in the process of acquiring insight and the kind of mistakes people would make, which of these could be seen as full mistakes, which ones as “good” mistakes (and here he refers to Köhler’s terminology to mean mistakes that show evidence of insight). Duncker would then classify participants based on the kinds of solutions they produced. In a setup where their task was to think of a method of destroying an inoperable tumor using radiation in such a way as to prevent damage to surrounding healthy tissue (“the radiation problem”, see Figure 8), he would identify three different approaches (Figure 9): (1) solutions that avoid any damage to surrounding tissue; (2) solutions that focus on decreasing the sensitivity of the surrounding tissue to radiation; (3) solutions aiming to repair the radiation damage to the healthy tissue after the tumor has be destroyed. Thereafter, Duncker would scrutinize the different lines of reasoning and the kind of errors within each group.


Figure 8 (Duncker, 1935)


Figure 9 (Duncker, 1935)

2.2.1 Restructuring the problem

The most critical thing he noticed was that those who were able to produce a meaningful solution were those who could restructure the problem. He notes that the process of reformulating the problem and coming up with a solution were two processes that were strongly intertwined. This restructuring required a breaking away of the readily given grouping and figure-ground constellation of its visual and textual content. As such, just like Köhler, Duncker, too, links this new way of looking at the same material, i.e., its regrouping and re-envisioning processes as being intricately related to Gestalt principles.

On the other hand, Duncker notes that some problems turned out to be very simple ones, i.e., problems that almost everyone was able to solve. Common to those problems was that they were presented as something that was almost complete and strongly suggestive of what the missing part could be based on the whole constellation. A puzzle would be a good example for that. Let’s think of a big puzzle where a given region is all complete except for one piece. That gap will cause a strong affordance[12] that will facilitate the finding of the missing piece. Duncker mentions that the closure principle of grouping is indeed a critical mechanism in problem-solving[13].

For more complex problems, Duncker notes that the restructuring of the problem is rarely a single step process. Instead, what is observed is a stepwise and repeated process of restructuring. During this repetitive process, both bottom-up (“von unten her”) and top-down (“von oben her”) processes are at work where the given information in the display is the bottom-up component and one’s general knowledge about the requirements for the solution the top-down one. To go back to our puzzle example: The particular characteristics of the one-piece gap is afforded from bottom up whereas finding that piece in a pile of pieces is driven from top down where now the knowledge of that missing shape and the retained image of it, drives the visual search as it “knows” to directly disregard certain types of pieces and as such narrows down the search.

2.2.2 Heuristic methods

Duncker talks about different degrees of insight. He remarks that insight does not have to be full-fledged, but that, instead, there can also be partial insights (and certainly a complete lack thereof). In partial insights we see the use of heuristics, says Duncker. For example, a person may see a given problem as similar to a problem they know of and thus simply apply or adapt their earlier solution to the new one. This he analyzes using two concepts of Gestalt theory. One is the principle of sameness because it is the perceived sameness between an older problem (Problem A) and the given one (Problem B) which summons the old method as a heuristic for the new setting. The other is the phenomenal environment concept, which emphasizes that this perception of sameness is idiosyncratic. In other words, another person might also be familiar with Problem A but not perceive a similarity between Problem A and Problem B. To put it differently, two persons may look at the same problem, Problem B, but perceive it differently due to their differences in perception and past history.

2.2.3 Styles of thinking: On analytic and synthetic “reading” (or explicability)

We typically use the terms analytic and synthetic with respect to ways of thinking. Duncker uses them as an act of reading, actually, reading from, i.e., ablesen in German. In productive thinking, that is, in a type of thinking where something is turned into something else, be it in mathematics, logic or scientific research, a synthetic insight is a sine qua non, according to Duncker. He defines synthetic insight as looking at events (1) from different viewpoints, (2) in new directions (say, different intermediary goals, just as Köhler’s dog who would run in the opposite direction from the food to ultimately reach it), (3) in new reorganizations, (4) for the first time as a whole. Synthetic insight

As an example for synthetic insight Duncker presents the mathematical rule of transitivity: if a > b and b > c, then a > c. What is critical here are not the individual values that a, b, and c carry but their relative relation and position within the entire set. In contrast, the fact that a participant in a psychophysical study would notice the difference between a given brightness P and P+2x but not between P and P+x might be an important finding in itself but it would not contain any synthetic insight. In Duncker’s words, as much as the participant might be surprised about this “irrationality”, they would blame their nervous system for that. In such a setting, there is no room for an a > c type of synthesis. Here, Duncker makes another interesting proposal. Whereas analysis contains a Kantian type of examination and reason, synthesis, according Duncker, arises directly from the appearance of the external stimuli (hence his use of the German word “ablesen”). And this, he remarks, is in line with Husserl’s phenomenology (as well as James J Gibson’s ecological theory of perception for that matter).

For Duncker, this ability to directly see/intuit a synthesis is what gives it the quality of generality and steadiness. This he explains with a beautiful example. When we see a house, a tree and a person in natural space we immediately understand that the house is larger than the tree and the tree is larger than the person. But what is more interesting is that when we then look at a vase, an inkwell and an eraser we, again, instantaneously and directly perceive the same relationality (vase > inkwell > eraser). Such syntheses are sometimes, as in these two examples, easy and accessible to everyone, but sometimes they are less so. For instance, how does one “know” that the sum of the inner angles of any triangle sum up to 180 degrees? What is critical here is whether the participant will be able to look at the triangle beyond the immediately afforded constellation, i.e., its shape as is. Instead, to understand why the sum of the angles add up to 180 degrees, the participant needs to restructure the layout, maybe add new elements to it to make the solution emerge (Figure 10). Once they succeed in doing this the answer can now be simply and directly “read off”[14].


Figure 10 (Duncker, 1935)

Duncker emphasizes that what he presented for mathematics, with its exactness and precision, also applies in daily life situations. As an example, he asks his reader to think of a room. On one end of the room there would be a stove and straight across it a door with a nightstand right next to it. In such a setting, each time a person would walk from the stove to the door, they would necessarily pass near the nightstand. This, too, according to Duncker would be a direct and holistic perception (that includes a perception, an experiencing of the entire layout), and as such, a type of immediate synthetic insight. Naturally, this layout would not be absolute and unchanging as in the geometry example. The synthetic insight regarding the room emerges as a result of how the things happened to be placed in the room[15].

2.2.4 Learning and partial insight

Duncker remarks that the learning organism has to be able to differentiate relations that are stable from those that are not. If, for instance, certain causes and effects have a given persistence we can talk of such stability. Yet, it is possible that such relational networks are not necessarily understood as a whole. Here, Duncker gives an example from medicine. Then and even today, there are a lot of methods and treatments in medicine that “work” even though we have no clue about the underlying mechanism that make them work.This is also why a lot of unexpected side effects of drugs are observed only after they are being used, and likewise, why certain drugs that were believed to be effective are withdrawn because it turned out that they were useless[16]. These effects cannot be known in advance because we do not yet understand the complex system of interactions a given agent becomes part of once it enters the body.

The discovery of the stable is mostly the discovery of stable relations. Duncker refers to the term “Gestaltqualität” (Gestalt quality) as first coined by philosopher Christian von Ehrenfels in 1890. To illustrate what is meant here, let’s imagine a triangle as in Figure 11a. If that triangle is flipped upside down (Figure 11b), it still retains the same inner relationships, that is, a simple rotation will not change its overall Gestalt. Likewise, nature also has certain properties, certain “interrelations” that are stable (as best evidenced by physical laws).


Figure 11a (Wiki commons)


Figure 11b

The biggest question here is how learning can occur when faced with a problem that lacks transparency with respect to the interrelations it contains. Duncker presents an example from Huang’s 1931 study with children. Children of different age groups are made to watch the placement of different objects in a container filled with water. The metal objects would sink, the nonmetal ones would not. In the critical condition, a child would then watch a careful placement of a metal needle such that it would not sink (”floating needle experiment”). The surprised child, when asked why the needle did not sink, would try to find various reasons in this non-transparent setting. At that point, for one group of children the needle would always be placed such that it would be floating, whereas for another group it would be placed such that it sometimes would float sometimes sink. Duncker states his amazement at how the children in the second group would see the most innocent differences in placing as “the cause” for the sinking or floating of the needle.

Duncker presents another example, this time from Köhler’s work on insightful learning in chimpanzees. Chimpanzees were presented with two pieces of food placed on differently dark gray napkins. Their task was to learn to always pick the food on the napkin that was a lighter-colored gray. The darkness of the napkins was constantly changed, but what was not changed was the fact that one was always a lighter-colored gray than the other. While the error rates would haphazardly go up and down in the first set of trials, a point would then come when they would drastically plummet down to close to no errors. This they called the moment of insight which occurred despite the obscurity of the situation because despite the challenge caused by constant change there was one thing that never changed, and that was the rule itself, i.e., choosing the food on the relatively lighter napkin. The cause-effect relations in nature

According to Duncker, a common aspect of cause-effect relationships in nature is the principle of proximity, one of the grouping principles in Gestalt theory. This principle has two different manifestations, says Duncker. One is a phenomenal manifestation, for instance, when a hand presses the switch and the light goes off. A learning being that observes this a few times will start to form a cause-effect relationship between pressing the switch and the turning off of the light. The other manifestation is a more direct one, such as the knocking sound that results from hitting on wood. A learning being that observes this will also form a causal relationship, and moreover, this time, the cause-effect relation would be of a much more direct kind (as the “knock” sound comes directly from the point where the hand hits the wood). Yet, what is common to both is the proximate succession between what is perceived as cause and what as effect. Maybe this is why we as humans have a difficult time to tie a distant event as a cause to a presently observed effect. Any intervention in nature will show its various effects not immediately but gradually and spread across often very long timescales. And it seems that because of that many people still have a hard time to grasp the causal relation between the distant and the proximate. If one remembers how long it took humans to understand the climate crisis (and many still do not understand it, particularly the middle-aged and older generations), one can see that this changed only after the effects of global warming could partially be modeled and hence made visible. Problem-solving using instinctual behavior

Babies have a tendency to grasp objects, to toss them around, and hit them on other objects. These are instinctual behaviors that are a must for the learning and discovering of the use of tools. Duncker then points to cross-species differences in these instinctual behaviors. While it is easy for crows, for instance, to remove obstacles to reach food, based on Köhler’s findings, this is something where chimpanzees experience serious difficulties. Duncker explains the crows’ ease of tackling such tasks with their instinctual food caching ability, an ability which is not part of chimpanzees’ instinctual repertoire.

2.2.5 Seek and find

Duncker remarks that when we try to find a pen, our eyes immediately fixate at places where it is likely to be, then start a search for “something like a pen” when scanning the surroundings. In this search, objects that do not look similar to a pen are not even “seen” in that sense, whereas objects that look similar are caught by the eye and finally, if, say, the pen is on the nightstand next to the door, a “resonation” occurs between the sought and the found.

Yet sometimes, the image of what is searched for does not exist, as in the case when one searches for a word in vain. Suppose I meant to say “he thinks he is something” but the actual expression I am after just does not come to my mind. Yet I remember its rhythm, for instance, that instead of the ---/--- (long/long) rhythm of “something” it is a -/---/--- (short/long/long) rhythm. There might even be a resounding faint phonotactic flow but too faint to be of any help. Then, after a while, it might pop up just like that, the expression I was after was “he thinks he is the bee’s knees”. The blurred auditory image component is quite interesting and should not be confused with the tip-of-the-tongue phenomenon because in our case the searched words did not feel at all to be about to be retrieved, just the opposite. For instance, there is no first letter ready at hand to help as is often seen in the tip-of-the-tongue phenomenon (likewise, going through the alphabet to find that first letter will also be in vain). The search here is obviously very different from searching a pen. Yet, the general process seems similar with the only difference that in one case, what is searched for is recognized (the pen) and in the other, remembered (“the bee’s knees”).

Duncker provides a nice example for the tip-of-the-tongue phenomenon. Suppose, he says, a person fails to remember the author of the novel “A Hero of Our Time”, yet remembers that the name has three syllables and ends with “-ov”. The correct answer is indeed Lermontov. When American psychology started to become interested in investigating such higher-level cognitive phenomena experimentally they would only care to ask (maybe for lack of time?) whether participants would know the answers to a variety of knowledge questions, if not, to provide a rating on a, say, 7-point Likert scale to indicate the likelihood of recognizing it among a list of options. Hence, we have two very simple quantifications at hand, (1) whether or not the participant knows the correct answer, (2) their feeling or judgment of knowing rating as to whether they would recognize the target answer among a set of options. Someone from a Gestalt perspective, in turn, would also wonder about the phenomenal experience of the participant: how did they do their word search, which clues did they have at hand to help and guide them in their search etc. Naturally, those would vary from participant to participant and one would likely expect those differences to be predictive of how long it would take them to find the right answer and their metacognitive judgment about their possible success in a subsequent multiple-choice test. Such richness is utterly neglected when a researcher sees the participant as a quick data producing being rather than an active, curious, thinking being.

2.2.6 Constructing the image

Naturally, the ability to construct images of varying complexity is critical in problem-solving. These images do not have to be well-known images. To give Duncker’s example, when one asks people to picture “a yellow crow on the table with a cigarette in its beak” they typically can do so quite easily. We might expect that the ability to imagine completely novel images is one of the most critical elements for creative thinking. We can see this, for instance, in the inspiring case of Maryam Mirzakhani, the winner of the 2014 Field Medal in mathematics. In a later section we will try to take a glimpse into her way of thinking.

2.2.7 An example of Duncker’s experiments: The box problem

Duncker's most influential work on American psychology in problem-solving was his “box problem” experiment. The problem is structured as follows: The participant is told that three candles need to be mounted on a door at eye level. In the room, there would be a desk with multiple paraphernalia, such as paperclips, pieces of paper, string, pencils, tinfoil, old parts of an apparatus, ashtrays, joints, pieces of wood. In addition, there would be three small, matchbox-like boxes of different form and color and numerous tacks spread around the table. The participant is asked to use any of the materials they wish and to think aloud throughout the session so that the experimenter would know about their ideas and the reasons for each of their actions.

Before proceeding to the list of variables, I want to make note of how this so famous problem, known in the English literature as the “candle problem”, was misrepresented in a series of images (e.g., Figure 12a and 12b) that were constructed to supposedly facilitate an understanding of the general setup but instead completely missed the very essence of the actual setup. What is worse is that these depictions have been copied into some articles and even books (e.g., Ashby, Isen ve Turken, 1999; Halpern, 2013; Isen, Daubman & Nowicki, 1987; Lubarsky & Thomas, 2020). In some articles this was so even though the article’s main focus was that very experiment by Duncker (e.g., Glucksberg & Weisberg, 1966; Weisberg & Suls, 1973; see Figures 12c and 12d)

This is hard to understand given that Duncker’s 1935 book was fully translated into English only ten years later in 1945 with Lynne S. Lees rather meticulous translation. Hence, I think that only scholars who read Duncker’s original work know what this study was really about. An examination of the misrepresentation of his famous box/candle problem in the English scientific literature could be an interesting study in itself.


Figure 12a (Source: )


Figure 12b (Source: Wiki commons)


Figure 12c (Weisberg & Suls, 1973)


Figure 12d (Glucksberg & Weisberg, 1966)

As mentioned, Duncker’s actual setup was such that the critical objects (three little boxes and tacks) were placed inconspicuously amongst all the other paraphernalia laying on the desk. The critical manipulation was similar to the branch that was “tied” to the tree. To remember, in one of Köhler’s setups the chimpanzee would see a banana that was beyond reach. and no sticks to help reach it. There seemed to be no solution up until the chimp would look at one of the branches of the tree that was just of the size and kind needed. At that point, the chimpanzee started to see that particular branch no longer as an inconspicuous part within the wholeness of the tree, but instead as a piece that can be separated from it and function now as a new whole in the form of a stick. Besides this kind of structural “fixedness”, Duncker also points to functional “fixedness”, or as known today as functional fixation. For instance, a ruler’s actual function, i.e., the very function for which it was manufactured, is to measure length. Yet, it can as well be used, say, as a rhythmic instrument. And it is this structural and functional fixedness that Duncker focuses on in his problem-solving experiments.

In his experiments, the box problem is but one in a series of five different problems. Importantly, what is manipulated across all setups is whether the critical objects are presented with pronounced or unpronounced functional fixedness. I will focus on the box problem since it is the most famous one of the five problems[17]. Here, for half of the participants, each of the three little boxes is filled with things. One of them contains the tiny thin candles that they ultimately have to mount on the door. The other one contains tacks and the third one matches. This is the enhanced functional fixedness condition (eF) which he called the “with pre-utilization condition” (i.e., where the boxes were presented as “pre-utilized” containers). For the other group, the boxes were presented empty, with the candles, tacks and matches laying on the desk amongst the other paraphernalia. Since the empty boxes were not “pre-utilized” he called that condition the “w/o pre-utilization condition”, which I will call the reduced functional fixedness condition (⌐eF). The critical question was whether participants in the first group would have a harder time to single out the boxes, change their usage from container to platform, mount on each one of them a candle and pin each box on the door by using the tacks. The other four problems, too, were presented in two versions, eF and ⌐eF, which served as a within-participants factor across five problems. Duncker counterbalanced the conditions such that one group would receive Problem 1 (eF), Problem 2 (⌐eF), Problem 3 (eF), Problem 4 (⌐eF), Problem 5 (eF), and one group Problem 1 (⌐eF), Problem 2 (eF), Problem 3 (⌐eF), Problem 4 (eF), Problem 5 (⌐eF). The main dependent measure was the number of solved problems. If the participant was not able to solve the problem within 3 minutes and was unmotivated to continue or felt stuck, they proceeded to the next problem. Duncker’s findings are shown in Table 2 The qualitative data collected through the “think aloud” protocol and post-experimental inquiry were also analyzed. The participants’ reports confirmed the hypothesis. When the experimenter would ask a participant in the enhanced functional fixedness condition as to why they did not think of using the little boxes as a base the answer was, for instance, “But those were boxes!”. Overall, constellations that enhanced functional fixedness led to approximately 40 percent lower performance.

In a different version of the box problem, participants would find totally unrelated objects in those boxes such as buttons. Duncker observed that in this case, performance went even further downhill. This he explains as being possibly due to the enhanced distance (hence reduced proximity to use the Gestalt grouping principle term) of the critical target objects (candles and tacks) from the instrumental object (boxes). He talks about the chances of becoming aware of the possible relation between the boxes and the candles and tacks, if the participant starts playing around with them while trying to solve the problem. Hence such haphazard behavior of playing around with the objects, emptying the boxes and changing their constellations can bring about an insight[18]. In contrast, when the boxes were empty, their “container” function would be less salient, hence making it easier for the participant to “see” its potential to serve as a base for the candle.

In sum, one notices that in Duncker’s experiments, just as we saw in Köhler’s meticulous setups with chimps, the visual configuration of the setup is of utmost importance. By using controlled conditions of layouts with different degrees of functional (and possibly structural) fixedness, Duncker goes about to carefully inspect via different visualizations as to why the functionally fixed versions led to worse performance. He notes that when a given object has to be transformed in order to assume a new function (e.g., in one of the problems, a paper clip had to be turned into a hook so as to serve as a hanger), the task would be even more challenging. He also remarks that the multiplicity of functions a given object may serve in daily routine was also an important factor in breaking functional fixedness. To go back to the example of a ruler, if it is often used not just as a ruler but also as a stick to hit someone, it will be much easier to think of using it as a rhythm instrument[19].

2.3 The mind of a mathematician: Maryam Mirzakhani (1977-2017)


Maryam Mirzakhani

The many issues that Gestalt theorists discussed in respect to what they called productive thinking makes one think of the immense creativity of some people. Indeed, in his 1945 book Productive Thinking, Wertheimer attempted to scrutinize the minds of three such minds, Gauss, Galileo, and his friend Einstein. Here, I want to touch on yet another impressive mind, the mind of the late Iranian mathematician Maryam Mirzakhani who won the 2014 Field Medal[20] for her imaginative work on the dynamics and geometries of Riemann surfaces only to pass away three years later due to breast cancer. I wished a time would come where a Gestaltist working on thinking would analyze Mirzakhani’s creativity. What made me add this section was a recent documentary on her, Secrets of the Surface: The Vision of Maryam Mirzakhani. In this documentary, her friends from school (way back from middle school), PhD students and colleagues talk about how she would relate to the problems she wanted to solve, how she would restructure them to reach the solution. I would like to give some examples from those descriptions that struck me as strongly reminiscent of what the Gestaltists were talking about. Naturally, I will glimpse at the form of thinking rather than its content, which should be left to a mathematician[21].

While working on a problem, a colleague asks Mirzakhani whether she knows what she is doing and she would say “no, I have no idea...”. Her friend, a physics and mathematics teacher, Mahshid Pourmand remarks that “she was thinking and not going anywhere...”. Whereas a lot of people would give up when getting stuck repeatedly, the mind of a person who does not lose their hope or tenacity continues brooding on it. And then, one day, the solution emerges, sometimes just like that at the least expected moment. These words of a PhD student about Mirzakhani I think are quite telling: “Maryam would see influences of all these different fields. And instead of trying to understand just implications she was trying to understand how things would interact.” Maybe the most beautiful description from a Gestalt perspective came from a colleague of hers from Harvard University: “She would start describing sort of elaborate stories or mathematical narratives and these narratives were almost like science fiction. She would sort of think ahead as to what the shape of the theory might be that was yet to be discovered.” When looking at Maryam Mirzakhani’s drawings (Figure 13) it is as if we virtually see the very clues into her way of thinking. Who knows, maybe one day, mainstream psychology on thinking and problem-solving (nowadays fashionably dominated by “decision-making” studies) might consider scrutinizing the thought processes of such unusual, creative minds by looking at their drawings and scribbles as they work on problems. I believe that once cognitive psychology starts using more in-depth methods that include multiple tools both quantitative and qualitative, it will reach to a much richer understanding of creativity, one of the most interesting yet insufficiently understood areas of cognition. After all, believing that one can understand creativity by asking people in how many different ways a brick can be used is a very narrow perspective to say the least. The main reason for this narrowness is a lack of theory, or rather, a good, full-fledged, real theory (also cf. Gigerenzer, 2010), hence it does not come as a surprise that creativity poses the highest challenge to the empiristic conceptualization of linear, mechanistic associationism[22]. Maybe this also explains why creativity is still not a central part of mainstream cognitive psychology as is perception, memory or decision-making. Let us hope that this “stepchild” treatment will come to an end.


Figure 13 (From:

2.4 Gestalt in thinking: A short evaluation

Be it in chimpanzees, in ordinary participants of experiments, or in people who have shown incredible creativity in their fields of work, what is common to productive thinking in all of these cases is a restructuring of the immediately given. Gestalt theory explains restructuring through its principles of organization, i.e., grouping and figure-ground segregation. Hence, from a Gestalt view, productive thinking is, for instance, the segregation of something that originally was an inseparable part of a whole, or, say, a re-grouping for a shared function (hence a relating of pieces that at first sight did not belong together but were then sensed, intuited or fully understood as belonging together). Sometimes haphazard shuffling of things helps to change the “meaning” of the whole, e.g., by making earlier unnoticed “ground” objects figural and hence potentially relatable and by pushing distractingly salient yet useless “figure” objects back to the ground.

3. Today’s Gestalt

In this section we will take a short look at where Gestalt theory could be situated in today’s research across a variety of scientific fields.

3.1 Cognitive psychology

The spring of Gestalt theory is clearly perception. As much as we now understand some of color perception and some of single, simple shape perception (of edges, corners, lines of different orientations but see Maniatis, 2021) in terms of which brain areas seem to be involved, we still lack a full-fledged understanding of the perception of complex figures within figure-ground constellations and the dynamics thereof. While there is frantic scientific work on perception, almost all of it assumes that we can walk our way up to whole perception by understanding piecemeal perception of parts and a bit of their close surrounding dynamics. Even articles focusing on Gestalt concepts seem to be limited to models and brain mechanisms that account for only very local, small groupings and figure-ground segregation (e.g., most of the studies reported in Wagemans, Elder, Kubovy et al., 2012 and Wagemans, Felder, Gepshtein et al., 2012 are of that kind). Moreover, none of these “Gestalt” articles seem to ever care about the theory’s proposition about phenomenality which necessitates the collection and careful analysis of qualitative data[23]. Also, all current studies that touch upon Gestalt concepts are mostly limited to perception. As mentioned repeatedly, the mainstream psychology community is rather unaware that Gestalt theory came up with a multitude of ideas and understandings that would span far beyond perception and even cognition in general. I think we need to reflect on why such a comprehensive theory has been grasped in such a narrow and “myopic" way. Maybe this has to do with the over-specialization and mental narrowing that necessarily resulted from the immense amounts of publications that have become impossible to keep track of. Bregman’s seminal work on auditory stream segregation would count as one of those rare exceptions where we see a very detailed and very comprehensive analysis of Gestalt dynamics in auditory perception (Bregman, 1994). It is as if Bregman completed in the auditory domain what Wertheimer left incomplete in his landmark 1923 publication.

Yet, when we turn to the body of literature on memory and thinking we do not encounter any Gestalt conceptualizations. This is particularly so in memory where Hedwig von Restorff’s tour de force work is presented as a single experiment stripped off its Gestalt theoretical grounding. In thinking and problem-solving, we see more reference to Gestalt (mostly Duncker) but more of a lip service kind of reference devoid of the depth and accuracy it deserves[24].

3.2 Developmental psychology

Within developmental psychology we once more see that Gestalt theory is covered almost exclusively in relation to visual perception. Let’s take a look at Spelke, Breinlinger, Jacobson & Philips (1993), Quinn, Bhatt & Hayden (2008), and Bhatt & Quinn (2011). Common to all three studies is their understanding of Gestalt as a nativist theory. Unfortunately, in the American literature on Gestalt this mistaken understanding of their proposals as 100% nativist and universalist continues till today despite the brilliantly written texts by Köhler (1938; 1950) where he clearly outlines the non-dichotomous, multifaceted basis of the theory. Mainstream scientists like to believe that Gestalt theory proposes that grouping and figure-ground segregation are 100% bottom-up processes that exist in full-fledged form right from birth. Gestalt theory is framed as if experience plays absolutely no role and neither individual differences in the unique ways a stimulus might be experienced. It seems that more Gestaltist developmental psychologists are needed to correct these misunderstandings, to bring to foreground the many neglected developmental works of past and present Gestaltists and to contribute with novel research[25]. As early as a century ago, no one else but Max Wertheimer mentioned the influences of culture on even very basic acts of perceptual grouping, such as grouping by proximity (Wertheimer, 1912). Hence, the mainstream reading of Gestalt theory as “nativist” and “universalist” is wrong.

When we look at the conclusions of the three studies, Spelke et al. (1993) report that 3-month old infants do not yet show a Gestalt type of holistic perception and that 5- and 9-months olds still show only a slight tendency of holistic perception. They present these findings as contradicting the claims of Gestalt theory. While Quinn et al. (2008) state that 3- and 4-months old babies perceive layouts as predicted by the Gestalt principle of proximity, Bhatt and Quinn’s (2011) review paper once again conceptualizes perception as a bottom-to-top associationistic construction of larger wholes via the knitting of smaller units through attentional processes. In their paper, the term “gestalt” appears 25 times, yet just like Spelke et al., Gestalt theory is again presented as a “nativistic” theory and as nothing more than figure-ground segregation and a bunch of grouping principles. Moreover, Gestalt’s most important statement that it is the perception of the whole that occurs first and instantaneously and that it is this perception that determines how parts are perceived again goes unnoticed, let alone its discussion of phenomenal perception, Prägnanz dynamics, and pattern transposability (as the litmus test of a Gestalt pattern).

A very recent study that instantly connects with Gestalt theory is Dillon’s 2020 study titled “Rooms Without Walls: Young Children Draw Objects But Not Layouts”. Just reading this title immediately brings to mind Gestalt theory’s figure-ground conceptualization, yet, nowhere in the article do we even see the word “Gestalt” let alone a careful examination and discussion of their findings in Gestalt theoretic terms. Hence, in the absence of any theory to explain this finding, we end up with one more “interesting finding” to be listed in future developmental psychology textbooks. Yet, even a quick look at their stimuli and the drawings of the 4-year olds shows us a strong case of figure-ground segregation where the objects are salient figures against a homogenous, much larger-sized background doomed to remain “ground”, even if presented with attention-grabbing colors. The idea of figure-ground segregation goes back to Edgar Rubin’s studies in 1915 where he tried to understand which shape qualities make certain parts emerge as figures against grounds, using inventive black-and-white drawings. If Gestalt theory had remained a mainstream theory within psychology, we would expect this study to be run the latest by the 1950s or 1960s. Yet, because of this strange loss of knowledge we encounter an important study as this one as late as 2020, and then even only as a ‘puzzling empirical finding that still demands solid explanation’.


Figure 14 (Source: Dillon, 2020)

3.3 Neuroscience

On the neuroscientific end of visual perception, the proposal that comes closest to Gestalt theory might be Ahissar and Hochstein’s (2004) Reverse Hierarchy Theory. According to this theory, the brain’s visual network first perceives the larger and the more holistic in a given layout. Despite this critical, Gestaltian perspective, the term Gestalt is mentioned only in their conclusion where they mention, almost in passing, that their theory might be called “a neo-Gestalt view”. Probably the biggest misfortune for Gestalt theory that led to its full dismissal in the neuroscientific field (so much so that Ahissar and Hochstein might have felt the need to mention it with a “neo-“ prefix and without further explanation) was the debate between Lashley and Sperry on the one side and Köhler and Wallach on the other. In their studies they tried to show that damage to distant areas did not affect the normal functioning of a target cortical area (Lashley et al., 1951; Sperry et al., 1955). Hence, they proposed that Köhler and Wallach’s 1955 field theory about the functioning of the brain was disproven. Köhler wrote a comprehensive and detailed objection but “the damage was done”, a field theory conceptualization of the brain was seen as 100 % refuted and something to stay away from in neuroscience[26]. As mentioned, this might explain the rather shy use of the term Gestalt in the Ahissar and Hochstein paper.

If we look at the post-1950s neuroscientific proposals about visual perception, we see a complete opposite perspective to Gestalt theory (and this perspective is still the one and only perspective accepted by mainstream research). According to this perspective, more complex objects are grasped via a stepwise perception of their singular parts which are then integrated, somehow knitted, into a larger whole. For instance, a baby would first only see disconnected parts thanks to their specialized neurons responsible for single feature detection (cf. Hubel & Wiesel, 1964). But there is a major gap in this proposal, and that is the absence of a convincing mechanism, even today, to explain how all that integration with all its complexity is realized. Let’s take a look at Irvin Rock’s example of the triangle (see Figure 15) and elaborate on it somewhat more. How logical is it to think that a baby will reach the image of a triangle by repeated firing of the “corner detectors” in her brain which will then somehow knit them into a coherent whole. Will the baby always be looking at those corners in exactly the same order? If, on the other hand, her eye fixations and saccade directions will every time be different, how then can a meaningful, holistic perception of a triangle come about? Note that we have not yet included the line detectors. The empirist perspective (i.e., the perspective that sees the beginnings of knowledge acquisition as a linear, step-by-step, successive, from-part-to-whole association mechanism that applies to any layouts, be they random or not) I believe is insufficient to explain how a baby learns shapes, for instance, be it visually or through touch, or even with the benefit of all sensory modes together. Theories in which top-down processing is assigned a major role (e.g., template theories or prototype theories) still do not explain a baby’s acquisition process as those very templates or prototypes are again assumed to be constructed from part-to-whole.


Figure 15 (Rock, 1982)

Let’s go back to Irvin Rock and look at the shape in Figure 16. Rock remarks that if we perceive our surroundings under strong guidance of top-down processes, i.e., prior-knowledge-guided processes (I intentionally added “prior” to “knowledge” to emphasize its ‘empirical acquisition’), we would expect a literate person, who has been showered with letters throughout their life, to first see a “B”. Also template theories would predict this as the shape contains all features in right constellation of the letter “B”. Moreover, the B template is expected to have fast accessibility due to frequent use. So we would expect it to be activated instantaneously, but it is not. Hence, it seems, a simplistic, top-down processing explanation, that cognitivists overly use to “fill in the blanks” also falls short of explaining our perception of Figure 16. A lot of time will still have to pass, and be even wasted, if we continue to fail taking into account (1) the intricate inner dynamics of stimuli caused by their unique holistic configurations, (2) their unique phenomenal effects (that may change from person to person, from situation to situation, just as we observed in Köhler’s chimpanzees and Duncker’s participants, where one notices the similarity between things while the other fails to see it), and (3) the neuroscientific underpinnings of (1) and (2). Hence, as mentioned in an earlier footnote, from the proposals of attractor neural networks that acknowledge the work of non-linear, dynamic forces as depicted in Figure 4c in Yuste’s 2015 article one wants to grow hopes for a major paradigm shift in the decades-long empiristic mindset of neuroscience.


Figure 16 (Rock, 1982)

3.4 Computer science

We see that particularly with the help of immense data sets processed by deep learning algorithms, artificial intelligence studies almost made a “quantum leap” with respect to object recognition, visual scene analysis, handwriting reading, natural language processing, gaming, and medical diagnostics (cf. Fayek, Lech, Cavedon, 2017; for a nice popular science account of AI and its exciting history, see It could be the case that the models that are more successful across tasks are those that include at least some non-linear, hierarchical structure, in which distant elements (say color processors and shape processors) are interconnected through hierarchical “cluster heads[27]”. We see, for instance, that Fleming and Storrs use the term “distal causal factors” in their 2020 article where they report the successful generation of realistic, novel human faces by way of unsupervised deep learning. Yet, despite all the excitement around deep learning, Heaven, in a 2019 article with the telling title Why Deep-Learning AIs Are So Easy to Fool[28] , presents the example of a self-driving car that would speed up to a stop sign, just because there were four tiny stickers on it, which made it misread the sign as a speed limit sign. Reporting similar problems across various other domains of stimulus identification, Heaven finishes his article by quoting someone from the field, who admits that no one seems to know how to better the system.

A “deep learning and gestalt” google scholar search delivers 61,600 results. Notable articles include Hörhan and Eidenberger's 2020 "Gestalt Descriptions for Deep Image Understanding" and Amanatiadis, Kaburlasos and Kosmatopoulos' 2018 "Understanding Deep Convolutional Networks Through Gestalt Theory". Yet, let us not forget that Gestalt theory does not deal with perception, memory and thinking as a purely externally driven process. The theory also cares about the phenomenal, and hence, examines the complex world of idiosyncratic meanings that can change from individual to individual, from infant to adult, from culture to culture. The equivalent of this in deep learning systems, for now, would probably be in the form of different solutions and outputs produced by systems fed with different data sets. However, it is questionable whether this could be called phenomenal differences.

4. In Conclusion

The main purpose of this series of three article is to show that Gestalt theory is something far beyond “a charming school of interesting visual phenomena claiming that the whole is different from the sum of its parts”. The first article focused on the coming about of the theory, its theoretical proposals and its studies on perception which were likely an important trigger for the development of the theory (Mungan, 2020). The goal of the second article and this one was to demonstrate that, contrary to general belief, Gestalt theory was not just a theory about (visual) perception but that it also made important contributions to two other main fields within cognition, memory and thinking. My goal was to show not only how Gestalt theory’s conceptual and empirical contributions were mostly ignored, misrepresented or insufficiently represented (more so in memory, less so in thinking), but also how expansive and fertile a theory it is with respect to generating hypotheses even today.

Overall, we see that with their elegant and cleverly designed experiments and without any need for complicated and expensive technology, the Gestaltists were able to make sense of many things about the perceiving, remembering, thinking, interacting being. In his preface to Goldmeier’s 1972 monograph, Irvin Rock remarks that empirical psychology is more and more moving away from psychology. Five decades later, I think, this diagnosis is even more valid. As early as back then, Rock remarks that psychology is moving more and more towards “prestigious” areas such as computer science as a result of which it has become more sterile and oblivious to its big questions. He then goes on to ask a very simple question: How come that when tilting our heads clockwise by 45 degrees, the tilted square in Figure 17b (symmetrical across multiple axes) still looks like a tilted square (hence a diamond) whereas the tilted triangle in Figure 18b (symmetrical only across a vertical axis) tends to revert to a non-tilted triangle? I think that Gestalt theory is just the kind of theory that brings out such wonderful simple yet unexplained questions[29]. So maybe rather than being enslaved by neuroscience or computer science, a theoretically grounded, coherent psychology needs to emerge to show those sciences the way and force them to revise their current models rather than the other way around.


Figure 17a ve 17b


Figure 18

To use sciences famous “elephant in the dark room” analogy, it seems as if psychology is stuck. It seems as if we produce a plethora of findings about those parts of the elephant that receive the highest funding (say its tail or its trunk) and then go about to quickly generate models and coin effect names, even worse, propose novel theories that might be nothing more but theory surrogates (cf. Gigerenzer, 2010). Thus we seem to end up with today’s psychology as a discipline with disconnected tail, trunk, leg, hoof etc. models which do not coherently integrate and even contradict each other. Or, even worse, all those “models” may not even speak to each other so that we have tail theorists who are ignorant about what the trunk theorists have accumulated about the trunk of the elephant, and trunk theorists, in turn, who are ignorant about what the ear theorists have accumulated about the ear. And then again, how can they even be knowledgeable about each other’s work in this publication inflation we are experiencing and the tendency for each area to generate their own esoteric language. To me, Gestalt theory is a theory about the entire elephant. And this elephant is not just perception, memory and thinking but it is the living being as a whole.

As mentioned by Wolfgang Köhler in his wonderful 1959 address as president to the American Psychological Association, the theory started out with perception because that was the easiest starting point to examine. All that was needed was a stimulus and someone to perceive that stimulus. According to Köhler, Gestalt theorists derived general principles and dynamics from their discoveries in perception and applied those to learning, memory, thinking, and motivation[30]. With respect to motivation and group dynamics, he points to three important Gestaltists, Kurt Lewin, Solomon Asch and Fritz Heider. I believe that the works of these three also deserve to be brought to light again with a meticulous "archaeological excavation" with informed knowledge about the theory[31]. But this theory seems even more expansive than that. This impression grew even stronger in me when I read Max Wertheimer’s challenging yet brilliant articles on truth (1934), ethics (1935), democracy (1937), and freedom (1940). In these articles, he does a meticulous concept analysis grounded in Gestalt theory of each of these huge concepts. All four articles deserve to be re-read in today’s miserable world of what has come to be called “post-truth[32]”.

This astonishing reach of Gestalt theory to an immense number of fields within psychology, and possibly even beyond psychology[33] is probably one of its biggest strengths. And given the current replication crisis in the empirical, social and behavioral sciences and the lack of coherent findings that do replicate (see Zwaan et al., 2018; also see Witte & Zenker comment in Zwaan et al., 2018) something has to change probably in ways that are more than palliative solutions. And most important, empirical psychology has to value the phenomenal, experiential component in its search for truth. My biggest hopes are that these three articles will inspire the younger generations[34], and that they will help in opening new possibilities against current conceptual and methodological paradigms in which psychology is, in my opinion, almost imprisoned. I would like to end this article with a quotation by chemist Cyril S. Smith as cited in Arnheim’s brilliant 1971 Entropy and Art: An Essay on Disorder and Order monograph, where Smith calls for more attention to aggregates in the physical and chemical sciences by remarking that “the chemical explanation of matter is analogous to using an identification of individual brick types as an explanation of Hagia Sophia” (p. 25; cited in Arnheim, 1971).


[1] The book’s German title is “Intelligenzprüfungen an Menschenaffen”. For a strange reason, this has been translated into English in 1925 as “The Mentality of Apes” instead of “Intelligence Assessments in Great Apes”. One suspects an effort to prevent any “dangerous” ideas that there might indeed be other beings with intelligence.

[2] In all his English writings, Köhler prefers the terms “empirist” and “empiristic” over “empiricist” and “empiricistic” to mark the difference between the philosophical school of empiricism and the ‘aberration’ it underwent in empirical psychological research (cf. corresponding footnotes in Mungan, 2020, 2021a).

[3] Köhler’s main experiments were run with seven chimpanzees, whose unique characteristics such as their backgrounds, where and at what age they were captivated, how long they have been living at the station etc., he describes in the introduction section. But because there were also dogs and chickens around, and moreover the little children of the employees, he would also describe some of the little demonstrations he did with them.

[4] Köhler notes that a 15-months old toddler who had just started to walk since a few weeks, behaved very similarly.

[5] Just as the nature nurture-debate lost its meaning particularly with the latest developments in epigenetics (i.e., it is no longer meaningful to even draw the line between ‘nature’ and ‘nurture’ as they seem intertwined), debates about body versus mind, objectivity versus subjectivity, or for that matter, learning by trial and error versus insight will likely, I believe, lose their meanings as an either-or debate.

[6] None of the chimpanzees behaved in exactly the same ways, just the opposite, each had their own unique style of solving the problems and reacting in different ways when they failed to do so. After all, each had a different history, which should not be disregarded, according to Köhler, when trying to understand those differences.

[7] The term “singularity” or “Prägnanz” (e.g., Goldmeier prefers the former, Wertheimer the latter term) refers to anything that stands out as a figure against ground and has a resolved figural stability in terms of perceptual organization.

[8] Such behavior is also seen in infants.

[9] One can see this also in student exams. I suspect that few teachers care to make such a distinction and instead take all or close to all points off when the end result is wrong. Multiple choice exams (probably, and sadly so, the most widespread type of exams in today’s college mass education), by their very nature have an “all-or-nothing” scoring only.

[10] Köhler though was not at all ignored during the times he worked on animal intelligence and insightful behavior. It was later on with the domination of Behaviorism that his work became ever less known in mainstream psychology.

[11] For a good review of his life and the importance of his work see Schnall (1999) and Simon (1999). Duncker himself refers to one more person who devoted most of his work on thinking, i.e., N.R.F. Maier.

[12] Interestingly, the term “affordance” is widely known to have been coined by J. J. Gibson in 1966. However, sources indicate that Kurt Lewin used the term already (Affordanz in German) and Gibson also makes note of this (cf. Scarantino, 2003). Conceptually and earlier than that, we come across this idea in Köhler’s 1938 The Place of Value in a World of Facts book, where he defines the concept of “requirement” as he discusses a tip-of-a-tongue state in a name search (p. 276) and the case that one immediately detects a “sour” note in a melody or a wrong word in a sentence (p. 275). These, I believe, are rather close to Gibson’s concept of affordances and unlike Gibson’s belief that Gestalt talked about affordances in purely phenomenal rather than “direct perceptual” terms, this was not how Köhler discussed it. On page 276, Köhler writes: “Now, on phenomenal grounds I cannot decide whether these names are right or wrong, because here I have not yet the standard with which to compare. "Outside," however, the standard is present. And as I try name after name, the standard replies from beyond: "wrong" "a trifle better" "not yet quite" "there, perfect". Once more we have an amphibian context; requiredness is its most striking trait”. Since Lewin worked on social psychological group dynamics, it is understandable why his conceptualization was restricted to phenomenal affordances. However, Gestalt theory’s proposals on perception deal with both the phenomenal and the physical constellation at hand with their unique “requirements” upon their parts. Only a look at Wertheimer’s 1923 article is enough to see how much Gestalt theory acknowledged the importance of the affordances of the physical layout as is.

[13] Karalyn Patterson notes an interesting situation in her patients with semantic dementia: All are almost addicted to doing puzzles and Sudoku. I wonder whether this could be explained by the Gestalt principle of closure. Could it be that these patients who one by one start to lose their world of concepts and meanings re-attain a sense of wholeness, hence meaning via these closures when they do the puzzles and Sudoku? To watch Karalyn Patterson describe the characteristics of these patients in rich and graceful detail, see Beyond the Scores.

[14] This reminds one of the story of Socrates in Plato’s Menon dialogues, where he has a slave find the solution to a geometry problem via questioning. Socrates would first draw a square and then ask the slave how much longer each side had to be to make the square twice as large. The slave would first answer “twice as long” whereupon Socrates would draw how the square would end up and thus make the slave notice his mistake. Using this method, Socrates, step by step, would allow the slave to find the correct answer without teaching any abstract geometry knowledge. However, there is an important difference. According to Plato, this knowledge exists innately, hence he sees this act of discovery as an act of remembering. According to Gestalt theorists, on the other hand, this knowledge is already directly available in space and can be discovered by 'direct perception', if, for example, the presented geometric problem is reconstructed in just the right way. It is interesting in itself that these two perspectives end up in the same place (since what Socrates actually does is to visually present the slave’s answers, hence reconstructions, till he can see the solution out there, rather than remember it from a long-forgotten past knowledge).

[15] When, after many years, people change the location of the things in their rooms, they often slam into them because they are still moving with the implicit guidance of the old mind map, i.e., the old synthetic insight.

[16] Alzheimer’s, Inc.: When a Hypothesis Becomes Too Big to Fail (2019)

[17] The other four problems are not presented here for lack of space but also, because the box problem is rich enough to deliver the main points of Duncker’s rationale. Yet, their results are presented so as to show the variance across the tasks.

[18] This can happen, for instance, if the boxes end up in a central position so as to catch the eye. Once this happens and as they may now lay closer to the candle and the tacks, the person may start forming a connection between the box and the latter.

[19] This also shows us how unfortunate it is that in the past 100 years, toys have become more and more perfect and similar to –sometimes even indistinguishable from—what they are to represent e.g., toy cars the child can sit in and command as well as the many virtual games with ever more ‘exact’ representations). This relentless tendency to turn toys more and more real might ever more decrease children’s ability to envision them as what they should be or even as something completely new. Likewise, the production of gadgets specifically and narrowly targeted for a given, single purpose might also kill one’s ability to break away from functional fixedness. In poorer countries we easily see how imaginatively people use certain things for very different, sometimes unusual purposes.

[20] She was the first woman to win that award.

[21] How exciting it would be if there were a Gestaltist who was also a mathematician to run such an analysis of both form and content.

[22] In a wonderful paper, Richters (2021) discusses how mainstream (American) psychology has been built on a completely unwarranted assumption of “psychological homogeneity”. I would like to thank my student Zehra Erkoç who brought this paper to my attention in our current PSY 411 Theories & Issues in Psychology class as we were discussing the problem of hidden assumptions in mainstream psychological research.

[23] Yet, interesting proposals relating to Gestalt theory (even if leaving out the phenomenal, hence psychological component) can be found in Sayım, Westheimer, & Herzog’da (2010) and von Gioi, Delon, & Morel’de (2012). More than that, I was recently told that nowadays even machine learning studies started to incorporate “explanations” where they ask their models why they did what they did (Aykut Erdem, personal communication, 2021).

[24] Gestalt theory is actually strongly valued in more applied disciplines such as the educational sciences and design. What I refer to here is the lack of empirical and conceptual work that derives from a full-fledged understanding of the theory. There are incredible scholars who have been doing just this, e.g. Gaetano Kanizsa, Riccardo Luccio, Gerhard Stemberger, Kurt Guss, Anna Arfelli Galli, Tiziano Agostini, Yoshie Kiritani, who not to my surprise are mostly of non-Anglo-American background (see ). These scholars mostly gathered around the scientific journal Gestalt Theory, which was founded in 1979. But, it seems to me, that not much of it was incorporated into Anglo-American mainstream cognitive psychology.

[25] In this respect, Anna Arfelli Galli’s 2013 Gestaltpsychologie und Kinderforschung book is important in that it presents the many Gestalt-theory-grounded developmental studies from 1921 to 1975. To my knowledge, this work is not yet translated into English.

[26] Interestingly, today the pendulum may have a chance to swing back. Just a look at Figure 4c in Yuste’s important 2015 paper with the telling title From the Neuron Doctrine to Neural Networks depicting an activity map of an attractor neural network, suffices I believe as a hint in that direction. Before that, we see neurosurgeon, psychiatrist and psychologist Karl Pribram’s (1919-2015) interesting Holonomic Brain Model. As much as mainstream neuroscience had problems with some aspects of that proposal (could it be that it sounded “too freaky”?), its holistic look at the brain’s electrophysiology deserves attention. Yet another place where I see this more holistic understanding is physicist and neurophysiologist Erol Başar’s work (e.g., Başar, 2010).

[27] We can liken this to syntax in language (cf. Lewis, Vasishth, Dyke, 2006)

[28] This article was brought to my attention by Lydia Maniatis.

[29] Nowadays we observe a second revival in AI caused by deep learning. And moreover, after the emergence of neuroimaging in the 1990s there appears to be an additional “enslavement” of psychology by the neurosciences.

[30] In fact, Gestalt theory used a lot of examples from music to exemplify their proposals. Hence, one would expect similar frequent references to language and language processing which share a lot of properties with music, to say the least, syntax. This, however, was not the case at all. Indeed, it is almost as if Gestalt theorists avoided any discussion of language as a psychological process. One possible reason for this could be the inner complexity of language. After all, all Gestaltists, Wertheimer, Köhler and Koffka, were students of musicologist Carl Stumpf and were themselves musicians. This explains why they were at ease with presenting musical rather than linguistic examples. Another reason could be that they did not live long enough to brood on language as well. Wertheimer and Koffka died in the early 1940s, that is, before Chomsky’s seminal work on syntax, and though Köhler lived till 1967 and hence was witness to the “cognitive revolution” after 1959, he seemed to have devoted his remaining energy to the isomorphism proposition of Gestalt theory.

[31] The founders of Gestalt therapy, for instance, declared with surprising ease that they developed their therapy without ever reading the founding works of the Gestalt theorists (cf. Henle, 1975). I now know that there also exists a theoretically grounded Gestalt therapy that is called Gestalt Theoretical Psychotherapy (cf. Stemberger, 2008).

[32] We are currently translating these into Turkish and hoping to publish them within this year of 2022.

[33] I strongly believe that Wertheimer’s article on truth should become a critical reading in logic; his articles on ethics and on freedom should be critical readings in philosophy, possibly also anthropology and sociology; and the one on democracy a critical reading in political science.

[34] I must admit that I am pessimistic about the older generations as there has been too much of investment into the old framework almost to a point of no return. What is shocking is that this empirist framework with its part-to-whole, linear, almost mechanistic and static understanding is never presented openly to students of psychology. What should have been done over the decades is to openly state in every single PSY 101 class the empirist assumption, on which all is built. Because when it remains unacknowledged (1) most researchers stop being aware that they are seeing everything from that single viewpoint, and (2) students are implicitly indoctrinated into believing that there is only one way of seeing things. I, myself, have been a victim of this, which might explain the fervor and eagerness with which I wrote these articles.

Esra Mungan

Dr. Esra Mungan is affiliated with the Department of Psychology, Boğaziçi University. Her main fields of study include verbal memory, musical memory, and musical cognition. After studying these issues for a long time, Dr. Mungan now continues her studies mainly on the Gestalt theory and the meta-theory of the psychological sciences.



Extending Cognition

The Cognizer is a publishing platform initiated by CogIST, a cognitive science community from Turkey. On this platform, articles and essays on different topics from different fields of cognitive science are published in a way that would bridge the gap between public audience and experts.

Copyright © 2023