Still ahead of their time: Gestalt theory's intriguing ideas and research in memory

Abstract

This article is a sequel to “Gestalt Theory: Its Past, its Stranding, and its Future.” The aim of this article is to bring to light the conceptual and empirical contributions of Gestalt theory within the field of memory. It is typically believed that Gestalt theory is a theory about perception only. This, however, is not true. The first part of the article discusses some critical thoughts about memory processes as presented by Kurt Koffka in his Principles of Gestalt (1936) book. These involve Koffka’s proposal about the involvement and effects of memory processes in the perception of successive Gestalts; a discussion of the similarities and differences between percepts and memory traces; Koffka’s reference to research suggesting that memory traces are dynamic such that, depending on their Prägnanz, they will or will not change during storage in a way that can even be predicted in some cases. The article then reviews one of the most powerful empirical studies on memory within a Gestalt framework, i.e., Hedwig von Restorff’s 1933 dissertation demonstrating figure-ground dynamics in memory tasks. In the final part of this article, I present the main ideas of an utterly ignored memory researcher, Erich Goldmeier, from his 1982 book The Memory Trace: Its Formation and Its Fate. It is dismaying that these very original and interesting studies went unnoticed by mainstream cognitive psychology.

Gestalt Theory’s Less Known Studies: Memory

Source: Erich Goldmeier (1982)

1. Introduction

In this article I will review and discuss Gestalt theory’s approach and work on memory. Gestalt theory was born a century ago and continues to exist till today, although no longer with its impressive theoretical grounding but rather as a simplified, almost caricaturized, and even distorted version of it (cf. Maniatis, 2021a, 2021b; Mungan, 2020). This theory has been known to contribute mostly to the area of perception, and even not all of perception but mostly visual perception, despite the fact that within their proposals on perception there were almost as many references to auditory perception as there were to visual perception. Surely, their proposals were not restricted to the visual and auditory domains either, since perception was regarded as a whole. While scholars working on problem solving have always known about the work of Gestalt theorists in their domain (cf. Mungan, 2021b), scholars working on memory seem to completely have missed or ignored Gestalt theory’s proposals on memory. The goal of this paper is to make visible again what has been turned invisible. The most conspicuous proposal of Gestalt theory within the domain of memory is its claim that the memory trace is a Gestalt, that it is a dynamic trace that is subject to change even during storage due to possibly predictable Prägnanz dynamics. In that sense, for Gestalt theory, whereas continual change is an exception rather than a norm in perception (e.g. ambiguous figures), it is a norm rather than an exception in memory.

The present article will start with the chapters on learning and memory in Koffka’s 1936 book Principles of Gestalt, a book that even today would be a very inspiring textbook. The aforementioned sections, in which Koffka covers, discusses and provides examples for the “dynamic memory trace” proposal, make up no less than a third of his 685 pages long book! This is yet one more clear example that memory was not something “left out” in Gestalt theory as today’s scholars like to believe. In his book, Koffka also makes multiple references to Hedwig von Restorff’s[1] PhD thesis which she did under the supervision of Wolfgang Köhler. Her important contribution was to apply one of Gestalt theory’s most important proposals, the figure-ground conceptualization, to a memory setting. The article will subsequently discuss Koffka’s emphases regarding the relationship between learning and memory and the attention he draws to the phenomenon of implicit memory.

In the second half of this article, I will discuss Erich Goldmeier’s 1982 book The Memory Trace: Its Formation and Its Fate which, despite its extremely interesting and intriguing studies, went unnoticed in mainstream memory research within cognitive psychology. Despite its having been written 40 years ago, this book, I believe, is still deeply inspiring for any cognitivist, and for that matter, for any psychologist. Because of the article’s wide range of headings and subheadings, Table 1 is intended to help the reader orient themselves.

Table 1

Outline of the Article

CONTENTS

1. Introduction 5

Table 1 6

2. Kurt Koffka’s Analysis and Proposal on Learning and Memory 7

2.1 Gestalt formations in successive sequences (“Successive Gestalts”) 8

2.2 Are perception and memory processed in similar parts of the brain? 10

2.3 The Gestalts of memory traces 10

2.4 Dynamic properties of memory Gestalts 11

2.5 The difficulty of learning the monotone 13

2.5.1 Hedwig von Restorff (1906-1962)‘s 1933 dissertation 14

2.5.2 The change dynamics of mnemonic and perceptual traces 18

2.6 Learning and remembering: Friedrich Wulf’s 1922 study and dynamic memory trace theory 21

2.7 Does incessant repeating lead to learning? 25

2.8 First implicit memory tests: Claparède 1911, Maccurdy 1928 31

3. Erich Goldmeier’s Studies 31

3.1 What can be called “same”? 36

3.2 What is singularity? 37

3.3 What is encoded? 40

3.3.1 The difference between looking and seeing 41

3.3.2 Encoding the non-existent, not encoding the existent… 42

3.3.3 Processes that stabilize memory traces… 44

3.3.4 Grouping, meaning-making, noticing and encoding a Gestalt 44

3.4 Phenomenal asymmetries in stimuli 45

3.5 The sine qua non of memory experiments 46

3.6 Main findings of Goldmeier’s empirical studies 47

4. Finishing remarks, in place of a conclusion 50

5. References 52

2.Kurt Koffka’s Analysis and Proposal on Learning and Memory

Gestalt theorists appear to focus on the concept of “memory traces” when talking about memory. Koffka also uses the term “engram”[2]. But unlike Semon and later scholars of memory who followed in his footsteps, for Gestaltists, a memory trace is not a static trace. This perspective is somewhat unusual even today but was so particularly back then, if one considers that the formation and retrieval of memories had been discussed for a long time using metaphors such as the gramophone or the camera (cf. Roediger, 1980). A gramophone, for instance, apart from some additional noise that occurs during replay, does no more than play back what is recorded on the grooves of a record. This idea of recording and faithful replay (i.e., in which ever way something is engraved into the brain network, that is how it will remain and be ready for future play back or replay) is still surprisingly prominent particularly in neuroscience (cf. Sutherland & McNaughton, 2000; Eichenbaum, 2013, but also see Olafsdottir et al., 2018 for some signs of change regarding this rather static position of the engram). In cognitive psychology, in turn, the most prominently discussed change is the weakening of a memory trace due to disuse or some changes in its accessibility due interference or inhibition of competing traces. Even the immense false memory literature of the past 30-40 years (e.g., Roediger and McDermott, 1995; McCloskey ve Zaragoza, 1985; Zaragoza, Belli, & Payment, 2007) mostly talks about changes and overwritings of the original memory encodings as caused via manipulations in-between encoding and retrieval (or simply during retrieval, as in the misinformation effect, cf. Loftus & Hoffman, 1989; Loftus, 2005). This is quite different from proposing (even warning?) that changes in a memory trace can occur as it “rests in the mind”. Such a proposal is not only frightening but also rather difficult to study scientifically, which might explain the reluctance to endorse this possibility. Other than the Gestaltists, the person who came closest to considering this possibility, daring to look into this with considerable creativity, was most likely Frederic C. Bartlett (cf. Bartlett, 1932). Unsurprisingly, Gestalt theorists mention Bartlett and his work a couple of times in their writings but they also had some important criticisms about his work, which I will discuss in the last section of this article.

2.1 Gestalt formations in successive sequences (“Successive Gestalts”)

Gestalt theory approaches each aspect of memory formation, storage and retrieval again from a Gestaltist perspective. Its greatest objection to the empiristic[3], mechanistic accounts is that, whether in perception or memory, stimuli are not researched in their wholeness but rather in fragmented, decomposed forms. Koffka gives a nice example to show why this approach is problematic. To begin with, he remarks that the processing of successive (i.e., temporally unfolding) sequences relies not only on perceptual but also on memory-related functions. For this, he gives a rhythmic example, as rhythmic processing was his dissertation topic. When someone hears a pulse sequence such as xxXxxXxxX (x=unaccented, X=accented), they will hear a beat[4]. A grouping of triples (xxX or Xxx) will emerge quite naturally (cf. “emergent properties”, Palmer, 1999). Gestaltists noted that compared to simultaneous Gestalt formations, successive (visual, auditory, tactile or other) Gestalt formations involve more intense use of memory processes (which we today would call working memory processes). Hence, the perception of an X in isolation is quite different compared to the perception of an X preceded by two unaccented x pulses. Likewise, in a sequence of xxX groups, the perception of the first X is not the same compared to the perception of the subsequent Xs because now we have a memory formation. Therefore, if we happen to drop one or even more than one of the Xs in the sequence (e.g., xxXxxXxx_xx_xxX), the holistic triple Gestalt formation (xxX) will not dissolve. In music, such occasional omissions are used to break monotony without breaking the overall beat which holds the entire piece together.

After giving the xxX sequencing example, Koffka then proceeds to another example where this time X is the unaccented pulse in the series XXXXXXXX pointing to the fact that now, the same X is an unaccented pulse in the context of its neighboring X. Hence, what gives a certain pulse the property of being ‘accented’ or ‘unaccented’ is not its physical decibel but the proportion of the two different decibels. On the other hand, what is common between the two sequences is their triple beat property, that is, their Gestalts. Now consider a situation where the listener simply forgets each preceding pulse. If that happened no Gestalt would emerge[5]. Koffka furthermore remarks, referring to both music and language, that the last note in a melody or the last word in a sentence does not only arise from the just preceding note or word but is already contained in the entire musical or linguistic sentence. Here he emphasizes yet another critical point. When hearing do re mi fa, we hear a holistic melody. However, this melody is not a simple result of sequentiality per se because if in the transition from mi to fa a honking sound interferes, it will not naturally become a part of the successive Gestalt. In other words, it will not turn do-re-mi-fa into do-re-mi-honk-fa because the honking sound is not part of the musical system, hence does not belong to the musical “whole”[6].

2.2 Are perception and memory processed in similar parts of the brain?

Koffka asks whether perception and memory are processed in the same brain regions. In one of his dissertation experiments, participants were presented with a rhythmic light pulse appearing in the center of a screen. While the brightness of the light pulse was kept constant, a rhythm was created by varying the inter-stimulus durations. At the end of the sequence, participants were asked to continue its rhythm in their minds for a while and thereafter, to reproduce it by tapping. He observed that participants were doing the task by grouping the pulse sequence and, most importantly, that they did not remain loyal to the exact time intervals between pulses. This finding made him conclude that perceptual[7] versus mnemonic time might be processed in different brain regions. He further remarks that when one remembers an event, say a walk along the shore, one is able to reinstate, i.e., phenomenally re-experience, the spatial flow though not exactly the temporal flow of it. Instead, the temporal flow seems conceptually “added” to the memory rather than being phenomenally replayed. On this basis, he concludes that during the perception of the temporal flow of successive Gestalts different brain regions might be involved compared to those active when re-instating the temporal flow later on[8].

2.3 The Gestalts of memory traces

Koffka asks whether the reason why we seem unable to retain durational information (e.g., if we use the same ‘taking a stroll along the shore’ example, remembering, for instance, how long one stopped at a given moment to watch the ocean) could be related to the figure-ground dynamics of Gestalt. Here Koffka makes reference to Hedwig von Restorff’s dissertation studies where she showed that participants’ memory was quite bad for successively presented items if they belonged to the same kind (e.g., a series of numbers or a series of more or less similar drawings), hence carried an inconspicuous ‘ground’ quality, but impressive if they belonged to a different kind compared to the other ones (e.g., a drawing among a series of various numbers, or a number among a series of drawings), hence carried a conspicuous, figural quality. This study was important in revealing that figure-ground differentiation was playing a role not only in perception but also memory. Based on this finding, Koffka asks whether the duration of an event might be experienced like a ground perception and that because of time’s more homogenous ground quality, we may have a hard time remembering durations accurately when recalling an event. If there is indeed no concrete region responsible for the recording of duration, this line of reasoning does, I believe, deserve attention even today.

In summary, Gestalt theory proposes that memory traces are organized systems and that apart from the memory trace itself, its pattern is also a carrier of memory. To make the latter more explicit, Koffka gives a few examples. In one example he draws attention to how we can hum back a melody starting from a different note after first hearing (hence without ever having heard it starting from that note). And this is the case even though we could as well have hummed it along in its original register. Hence, as we hum back the tune later on not from its, say, original “do” but from “mi”, we are not literally recalling the original notes themselves but their pattern. Likewise, someone who has learned to read and write can read completely unfamiliar handwritings because what they learned was not the literal shapes of the letters but their overall, holistic configurations, says Koffka.

2.4 Dynamic properties of memory Gestalts

Koffka talks about an interesting finding in studies that measure perceptual thresholds (e.g., the smallest noticeable pitch or brightness difference) that psychophysical laws have a hard time explaining. When presenting stimulus pairs not simultaneously but sequentially, some asymmetric findings emerge. For instance, difference thresholds in comparisons of object weights were smaller when the first object was lighter than the second. In other words, a {light  heavy} line-up rendered a smaller threshold than a {heavy  light} line-up with exactly the same weights in reverse order. The critical difference, according to Koffka, is that a simultaneous presentation creates a perceptual task whereas a sequential presentation involves a memory component[9]. This finding is in line with Gestalt theory which claims that the x’s and X’s in xX and Xx are perceived differently due to the difference in the configurations they are in. A related finding comes from one of Köhler’s studies. When presenting a consecutive, equal-accented (X-X) pulse pair with an ISI of less than 4 seconds, the second pulse was heard as softer. In contrast, when the ISI was larger than 4 seconds, the second pulse was heard as louder. Again, from a Gestalt perspective, we are likely seeing the effect of grouping. If the ISI is below 4 seconds, one might conjecture that X-X is perceived as one group (hence the second pulse being “shadowed” by the first pulse, the pulse which starts the group), whereas at an ISI that is above 4 seconds, the two consecutive Xs are segregated (and hence the second X is a salient first element of a new group, hence it is perceived as if louder than the just preceding X).

At this point, it is worth mentioning a comment by Köhler (1918, cited in Koffka, 1936). Köhler remarks that in a task of judging two consecutive stimuli (be it based on their weight, their brightness or any other property), what seems to be happening is not so much an analytical comparison of the retained image of the first stimulus (hence its memory trace) and the currently perceived second one, but rather a very quick, spontaneous, almost automatic “same/different” judgment. Köhler points out that the moment the participant is perceiving the second stimulus they seem to know whether it is heavier, brighter or whatever the property of interest. The idea here is that if the two stimuli are perceived as one group, hence as one Gestalt, the awareness of the sameness or difference between the two is like an emergent property. In this example we once more see how careful and detail-oriented Gestaltist were in the way they inspected observed phenomena. Instead of a rote collection of psychophysical data and calculation of thresholds, Gestaltists made careful observations about participant behavior as well as their phenomenal experience. This very sensitivity allowed them to understand more complex dynamics that would otherwise go unnoticed. Max Wertheimer’s 1912 phi phenomenon article is yet another powerful example of how critical it is to not just collect simple reaction time or accuracy data but also place value on the experiences of participants to discover what otherwise would remain undiscovered.

2.5 The difficulty of learning the monotone

Why is it so difficult or even impossible to learn monotonous events? Koffka starts this section with a 1913 study by empirist Georg Elias Müller. Müller notices that when Ebbinghaus’ nonsense syllable lists were read in a monotonous way, without accenting, they were not learnt at all. This is an important point because in his 1885 book, Ebbinghaus reports that he read the lists to himself with a certain rhythmic grouping (stressing every four or five syllables). This information is mentioned almost in passing, as a single sentence under the section where he describes his experimental setup. To my knowledge, this critical information has gone completely unnoticed in mainstream cognitive psychology. One almost suspects that its English translation was not read with the care it deserved. Sadly, when referencing such classical works (and Ebbinghaus’s 1885 work appears to be cited 2264 times which is likely to be an underestimate), the works themselves no longer seem to be read that much. Anyone who works on memory should have immediately noticed that little sentence since we all know how critical rhythm is for memory if one only remembers the role of rhythmic narratives such as poems and legends in pre-literate human cultures. The empirist contemporaries of the Gestaltists were aware of the potential importance of the rhythmic component in Ebbinghaus’ experiments, leading them to notice that without rhythm those nonsense syllable lists were not learned. Yet, this finding constitutes a problem for the associationistic, empiristic viewpoint which sees repetitive, consecutive learning as sufficient for the formation of memories. Things like rhythm, meter, grouping, factors that clearly play a role both in perception and memory, are not easy to deal with from an empirist’s perspective. For example, in an ab/cde stimulus, that is, a stimulus where “ab” is grouped against “cde” (say via a slight accent on “a” and “c”), the ability of “a” to trigger “b” is more than the ability of “b” to trigger the post-grouping element “c” even if the durational gaps between all 5 elements are kept exactly the same (cf. Bower ve Winzenz, 1969; Restle, 1970). This phenomenon is rather difficult to explain within the empiristic tradition. The most it can come up with is something like some prior knowledge of the “ab” and “cde” groups leading to this result. Yet, what then to expect in an immediate regrouping of ab/cde to abc/de. For Gestalt theory the answer is simple and in line with the findings: now the ability of “b” to trigger “c” would be stronger than for “c” to trigger “d”. What could an empirist now say? Claiming that the slight accent on “d” creates some attentional interference which decreases the triggering power of “c” to “d” is probably not a very convincing explanation. In other words, the short-cut, often imaginary “past experience” joker card may not work anymore, nor any other explanation that ignores the grouping phenomenon. What is consistent in both cases is that the triggering power of consecutive elements within a group is always stronger than that for consecutive elements across groups.

From a Gestalt perspective rhythmic structure is important because it creates groupings. Once there is grouping, there is a figure that can segregate itself against a ground and as such make itself more memorable. After stating this, Koffka goes on to discuss in detail von Restorff’s findings of her series of studies looking at long-term memory performance in a variety of stimulus sets, various durational manipulations, tested on a range of age groups, including children. In his 2020 article on Hedwig von Restorff, MacLeod rightfully calls her dissertation a “tour de force”.

2.5.1 Hedwig von Restorff (1906-1962)‘s 1933 dissertation

Hedwig von Restorff’s thesis focuses on the process of assuming figure or ground characteristics as memory traces interrelate with each other. In her thesis, she reports a total of eleven studies, the first four of which might be seen as pilot studies. Instead of presenting each of her studies one by one, I will focus on two studies which nicely reflect the main idea of her work. The general structure of her experiments was like this: Participants received a list of 10 successive items which were presented at a regular 1.5-second per item rate; after this, all participants received a 10-min distractor task where they were asked to read a text for future rote recall; after this they were given a free recall task where they were to write all the items they remembered from the previous list, regardless of their order of appearance[10]. The critical variable was the configuration of the lists. Each list’s second (or third) item was of a different type compared to the remaining nine items. In some lists, that item was a 2-digit number amongst nine nonsense syllables, in some it was just the opposite, a nonsense syllable amongst nine 2-digit numbers (Figure 1a and 1b, respectively). In addition, to serve as a control, there was a fully heterogeneous list with ten items which were all of a different type (Figure 1c). In each session, participants would receive a single list and between sessions there would be a time interval of at least one day. The first session would always start with the control list of heterogeneous items. The experiment consisted of three sessions, each session with a different list configuration.

Figure 1a-c (1a and 1b items were constructed by the author for representation purposes; the control list in 1c, on the other hand, includes the different types of images as they were described in von Restorff, 1933)

Findings were clear. In the experimental lists (Fig.1a and Fig.1b), each of the “ground” items, that is the items that were all of the same kind, were remembered on average 22% of the time[11], whereas the recall rate for the critical “figure” item (i.e., of “sül” in Fig. 1a and “19” in Fig. 1b) was around 70%. The average recall rate of any given item in the control list, on the other hand, was 40%. The most critical part of this simple and elegant experiment was that by the time participants encountered the distinct item (e.g., “sül” in Fig.1a), they could not know yet that it was different from all the other items because of the fully mixed control list which each participant had received as their Day 1 list. This was also the reason why the distinct item was presented in the second (or third) position. If it had been presented towards the end of the list it would have drawn extra attention, something that von Restorff wanted to prevent. Anyone who would read “67 – sül – ” the day after receiving the control list would not likely pay any extra attention to “sül”. More important, “sül” in a list such as in Figure 1a will gain its “singularity” only as the list goes on. In other words, “sül” assumes its singular status only post hoc, which in turn is only possible through a memory process. Von Restorff furthermore explains the 40% recall rate in the control list as still some kind of a ground memory performance since in a list where every single item is different from the other, there is again a certain type of homogeneity emerging from full irregularity. In a list where each item is of a different type (where none of the types stand out from the other types), none of the ten items can assume the quality of singularity after all.

In another experiment, the setup was more or less the same, except that this time instead of a free recall test, a recognition test was used where participants had to identify which items were presented from among a list of presented and novel items. Moreover, this time participants were high-school students. Findings were rather similar: the average rate of recognition for a ground versus a figure item was 22% vs. 100%, respectively. On the other hand, the average rate of recognition of the corresponding second item in the control list was 56%, i.e., again a value in-between the recognition rates of the two contrasting (figure vs. ground type) target items. Hence, when looking at these results with today’s knowledge, we can say that the lower performance of ground items in the free recall setup was likely not an artifact of output interference (cf. Roediger, 1978) as it was also observed in a recognition test setting[12]. Von Restorff uses the term “monotone” to emphasize that items of such quality have a lower chance of forming memorable memory traces compared to items that carry a certain singularity within a whole. Critically, as also emphasized by Koffka, it is not the item per se that carries a quality of singularity. Hence, an item type that was monotone in one list could turn into a distinctive one in another list, depending on the items that “surround” it. In other words, whether an item has figure status or not does not have to depend on the specific features of the item per se but may as well arise from its relative status within a given environment[13].

Both von Restorff and Koffka touch upon comparable instances of monotony in daily life. Allusively, von Restorff remarks that the nonsensical, disconnected, monotonous stimuli in the experiments of researchers like Ebbinghaus and Georg Elias Müller were not necessarily removed from the reality of daily life where within its boring and monotonous rhythm people are indeed doomed to forget most of the things they experienced through the very forgetting mechanisms proposed by those empirists. Koffka likewise gives an example of a person whose daily routine was to wind their wristwatch before bed (those were the times when watches needed to be winded to ensure that they would not stop). They would remember their act of winding their watch that evening, but later on they would not keep any memory of any of the separate acts of winding anymore. In other words, such routine, monotonous acts do not turn into memories. This process of the monotonous to turn into a “ground”, hence becoming less and less memorable, can be seen in many other real life examples. As one gets older with a more and more settled life, every day becomes more routine, more cliché and the events within a given, ordinary day, turns into von Restorff’s nine homogenous items which, with each next week, build on top of each other and thus become less and less memorable as they melt into each other into a single “ground” memory. A similar phenomenon can be observed during long stays in prisons or hospitals[14]. Each day is just like the other and like this, with every next day time flows, or rather, steals itself away from one’s precious life.

For a long time, von Restorff’s important finding was referred to as the “isolation effect”. But from MacLeod’s 2020 article we understand that in the 1950s, a scientist from the UK referred to it as the “von Restorff effect”, which later on became its official name. As also stated by MacLeod, this effect turned out to be one of those few robust effects that passed the test of time. And because of this, it deserves to be understood and covered as a Gestalt effect as its entire experimental and conceptual framing is theory-driven. Once more, it is a big disappointment that in the mainstream Anglophone world of psychology this effect goes often unmentioned or when it is mentioned, it is referred to as “some curious finding” and misrepresented as “people will remember well a strange word in a middle of a list”, which completely misses the main point.

2.5.2 The change dynamics of perceptual Gestalts and memory Gestalts

When discussing her findings, von Restorff compares perceptual Gestalts and Gestalts emerging from memory traces. She remarks that dynamic changes in perception are more of the kind seen, say, in visual illusions or ambiguous figures. What she means here are figures like the Müller-Lyer illusion, the duck-rabbit illusion (Figure 2) or Penrose’s impossible figures (Figure 3), in which the emerging holistic percept is “distorted” (e.g., Müller-Lyer illusion) or unstable (e.g., ambiguous figures, Penrose impossible figures) due to the way specific parts are configured. She emphasizes that such cases are an exception rather than a rule in perception.

Figure 2 (Wiki commons)

Figure 3 (Hochberg, 2003)

In contrast, such changes and “distortions”, she remarks, are a rule rather than an exception when it comes to memory. She draws attention to the fact that forgetting mechanisms such as retroactive and proactive interference/inhibition effects are common for “ground” elements, but not for “figure” elements, where such effects do not seem to occur at all. Even today, almost all mainstream memory books and chapters refer to interference/inhibition effects as general forgetting mechanisms that apply (more or less) to all kinds of items[15]. Yet, in a different set of experiments in her dissertation, von Restorff had shown that this was not the case. This important distinction between kinds of items is still completely overlooked by mainstream psychology which for decades (and still) prefers to use simple, isolated, random materials. Hence, by constantly using the same type of items that lack any inner coherence among themselves, findings of interference/inhibition between items have been replicated again and again, leading to an illusion of having found an overarching principle. Certainly, exceptions were carefully marked, such as “release from PI” (e.g., the case where proactive interference builds up as long as items are from the same category, say, animals, and is “released” when there is a switch to a different category of items, say, vegetables). For a Gestaltist such “releases” would be nothing else but an isolation effect occurring right at the transition point as that item would assume figure quality (and later on lose it as the next set of disconnected same-type items within the new category would continue building up with no inner structure other than being disconnected members of a given category, e.g., vegetables).

Von Restorff also stresses that we should not a simply look at how much has been remembered but also what has been remembered and what has been falsely remembered. Because only then can we understand the change dynamics that memory traces undergo in between encoding and retrieval. Hence, I would like to finish this section with a report of an interesting false memory I experienced lately and a possible Gestalt-based analysis of it. In spring 2020, Dr. Rabia Ergin’s COGS 579 seminar course was hosting a series of cognitive science talks as part of our Cognitive Science Master’s Program, here at Boğaziçi University. In one of the weeks (Feb. 26), there was a talk on contour mechanisms in the early visual cortex and their role in shape perception, and a week later (March 6), a talk on symmetry perception and percept completions and how these can be understood via an analysis of frequency tagging EEG signals. Many months later, when I wanted to recall the talks (without looking at the seminar program list) I was perplexed to see that what I remembered as a single talk were actually two distinct talks by two different persons. One of the presenters (Dr. Funda Yıldırım from Yeditepe University) was someone I knew from the field. And what my mind did here, was to merge the two events into a single event and pick her as the single speaker, i.e., the “figure”. When I googled the second presenter (Dr. Nihan Alp from Sabancı University) I first felt a very low sense of familiarity but when continuing to look at her picture, I slowly started to remember her, and even voice and style of talking. I believe the reason why these two distinct events were merged into a single event were the overlaps that both events shared (thus, a consequence of the Gestalt principle of sameness, often known as the principle of similarity). Firstly, physically speaking both speakers looked somewhat similar to me and secondly, both talks had a relation to Gestalt theory, hence there was a certain similarity also in content. As a result, it seems that after some while, (1) my mind merged the two events into a single event instead of retaining them as two separate episodes, and (2) once the two episodes were reduced to one episode, the two speakers had to reduce to one speaker, and my mind, naturally, “chose” the speaker with who I had some earlier acquaintance. If the two speakers had been very different from each other physically (say one, a young female researcher and the other an older, bearded male researcher) or if the two talks had been as exciting yet very distinct talks (say one on visual perception, the other on children’s sign language), it is likely that my mind would have retained the February 26 and the March 6, 2020 talks as two separate episodes, that is as two unique memory Gestalts. I believe that memory research should show more effort in addressing such more complicated yet more “real life” events in order to understand memory in all its complexities. Even if such an endeavor may appear frightening at first sight, Gestalt theory would, I believe, be a good guide as to how to address these phenomena experimentally. This is exactly what Hedwig von Restorff did way back in 1933 as a young PhD student, and moreover, as a young woman in a men’s world of science…

2.6 Learning and remembering: Friedrich Wulf’s 1922 study and dynamic memory trace theory

Koffka presents a study by Wulf to exemplify the dynamic character of memory traces.Empiristic theories suggest that memory traces, that is, “engrams” can only change in the form of becoming weaker either due to the mere passage of time or due to interference mechanisms[16]. Gestalt theory objects to this unidirectional, linear-minded perspective and proposes instead that in some cases, with the mere passage of time, the engram may become even more singular, more “prägnant”, in other words, stronger rather than weaker, and that this cannot be simply explained away with a simple “they paid more attention” kind of explanation. In one of Wulf’s studies, participants were presented a set of drawings (e.g. left stimuli in Figure 4) each of which they had to study for a couple of seconds. After each stimulus they were asked to re-draw them after a time between 30 seconds to one week[17]. Two types of cases were observed: The general Gestalt of a given stimulus was either recognizable albeit with a few changes or it completely vanished and was replaced by a different shape (and here it was noted that participants would say that they were not sure about the shape they drew). In the cases where the general shape was retained during recall, an additional analysis was performed on the kind of transformations that occurred. Here, two main tendencies were observed. In one, a given feature of the drawing was drawn in a more pronounced way, this he called ‘sharpening’ (in German “Präzisierung”, e.g., right drawing of Stimulus 1 in Figure 4, with a more edgy, narrow, hence pronounced zigzagging). In the other, just the opposite happened, which he called ‘leveling’, i.e., a given feature became less pronounced at test (in German “Nivellierung”, e.g., right drawing of Stimulus 2 in Figure 4, where the concavity was slightly subdued at recall).

Figure 4 (Stimulus 1 and 2 from Wulf, 1922; the two on the left are the stimuli that were presented, those on the right are examples of their reproductions)

Wulf remarks that the sharpening of a memory trace cannot be sufficiently explained by a heightened attention during perception and accuses empirists of using attention as a fill-in concept whenever needed. Instead, he proposes that Prägnanz dynamics are at play rather than some laws of sensation, association or attention. Likewise, he sees the leveling of a stimulus at recall again as a Gestalt dynamic rather than the “wearing off” as observed in cobble stones as they “grow older”. Wulf also scrutinizes the dynamics of different repetitive instances of recall at different time periods. First and foremost, he notices that immediately at the first instance of recall there are already some changes, and that exact replication is rare. He then goes on to look at what happens at later stages when at first a given drawing is recalled with a sharpening versus a leveling. He remarks that if leveling had been something like a fading away of a memory trace, as suggested by empirists, one would expect that once leveling is seen, the image would stage by stage become less and less clear. This, however, was not the case. Instead, the leveled images continued to be recalled throughout long retention intervals. Wulf notes that be it in the direction of sharpening or leveling, what is happening is a Prägnanz dynamic where less stable forms resolve into more stable, singular forms. While in perceptual processes the degrees of freedom for Prägnanz dynamics is restricted due to the fact that the object is “out there” in physical space, this is not the case for memory, which enjoys considerably more degrees of freedom for the transformation of the engram. Wulf furthermore emphasizes that such changes do not have to be related to language-mediated or other top-down processes but that they may as well occur through a bottom-up process of perception and memory formation, something that is utterly ignored by empirists.

Koffka allocates a considerable number of pages to Wulf’s 1922 study as it reveals the trademark features of Gestalt theory. In his work, Wulf tries to understand which possible mechanisms could be at work as well as what kind of meaning impositions seem to occur for each participant. This is why he not only collects and categorizes the recalled versions of the drawings but asks every participant about their experience, their phenomenal perception and their thoughts and ideas about the process. Hence, with this qualitative data, instead of reverting to a cold, mechanistic perspective, he instead handles the material with all its richness and complexity as it varies from person to person. The resulting picture, when taking all findings into consideration, does not support empiristic expectations of “that which is more familiar will be recalled the way it is more familiar to the participant” but instead hints to the recalled image being determined by a resolution into a more stable version of the original drawing. In Figure 5a, for instance, given that each participant would have been very familiar with a similar shape of a cradle[18] it was not recalled as such but instead could turn into something as depicted in Figure 5b. The participant who drew Figure 5b reported that he had coded 5a as “columns with arc”. What is critical here is not to get stuck between an either-or ultimatum of whether a falsely recalled drawing occurred strictly due to past experience kind of top-down processes or strictly due to the inner dynamics of the stimulus. Instead, what Gestaltists propose is that all possible factors need to be considered, not just one of them.

Figure 5 (Koffka, 1936; “a” is what is presented, “b” is its recalled version)

Koffka emphasizes that Wulf’s repeated reproduction findings show that memory traces are dynamic not static. As mentioned earlier, the proposal that engrams are dynamic forms was received rather harshly by Donald Hebb and Karl Lashley. Today, however, many findings in cognitive psychology (e.g., the “post-event misinformation effect”) similarly point to the fact that memories can change. The main difference here, though, is that according to Gestalt theory, such changes do not necessarily require external manipulations (such as providing misinformation before testing) but that change is a natural property of memory traces which can happen as a result of figure-ground dynamics, Gestalt principles of grouping and Prägnanz dynamics. This is an important difference and it seems that the American empirist school’s resistance is directly against the “radicalness” of this proposition[19].

2.7 Does incessant repeating lead to learning?

After posing the question as to whether intense repetition is sufficient for learning, Koffka refers to Kurt Lewin’s interesting 1922 study. In this study, Lewin asked participants to read a list of nonsense syllables out aloud 300 times. In one condition, participants were told right from start about a subsequent memory test. In the other condition, they were not informed about a memory test to come. Compared to the first groups the second group performed miserably. This study was to show that, unlike the claims of empirists, mere repetition was not enough to ensure learning. Interestingly, this study is reminiscent of Nickerson and Adams’ 1979 study, so we are talking of yet another study that sadly has gone unnoticed only to be “re-invented” five decades later. In the Nickerson and Adams study, participants were asked to draw or recognize from among multiple options the American 1 penny. Almost none of the participants could correctly draw or recognize the penny. This study showed that mere repetition, in this case, frequent exposure, was not enough for learning to occur. All articles and textbooks present this finding as if it was first discovered by Nickerson and Adams whereas Lewin tested the same thing in a much more sophisticated, controlled way as attention was also ensured in the incidental setup by having participants read the material aloud.

On the other hand, if we take learning as a process we do observe the beneficial effects of repetition, says Koffka. Here he points to images that though chaotic at first sight, become organized over time, sometimes with repeated or extended exposure (cf. emerging images, e.g., Figure 6 and Figure 7; Koffka presents Figure 8 as an example, yet, that figure would no longer be perceived as chaotic, not even at first sight[20]). The question he asks is whether we might start seeing learning effects as we present these in different orderings. This is a question that can be tested, for instance, by comparing detection times of the first half of stimuli to the last half per list and per participant and see whether across participants, performance consistently improved in the second halves. Again many control variables could be thought of. Figures 6 and 7 are, for instance, very different types of emerging images. Whereas in Figure 6, there is a plethora of different-sized black-and-white spots, Figure 7 is less fragmented with fewer and larger-size black-and-white spots. Moreover, Figure 7 is strongly implying information about light falling from top whereas in Figure 6, no such suggestive information seems to exist. So one could have two lists, one that has images of the kind seen in Figure 6 and the other with images of the kind seen in Figure 7, and compare whether learning rates were comparable. In Gestalt terms, this learning would be a learning of some kind of abstracted patterns which would likely be different across the two lists. And would it not be interesting to start understanding what that generalized, holistic pattern perception could be, if that is the case, of course.

Figure 6 (‘The Dalmatian’ by R. C. James)

Figure 7 (Rock, 1984, ‘Person Sitting on Bench”)

Figure 8 (Koffka, 1936; ‘Man with Pince-Nez”)

Koffka goes on to report about an experience that appears to be widespread but hard to explain, i.e., the experience of no longer being able to recall a building that was there for ages but was then torn down and replaced by a new building. Here he asks why we become utterly incapable of retrieving the earlier building which we looked at and even spent time in. This is something we experience quite frequently in the Beyoğlu district of Istanbul. There have been buildings we had looked at zillions of times, have read their name plates, inspected their showcases. One day, the shop is vacated and a new shop is opened in its place and very shortly thereafter we are perplexed to notice that we can no longer remember what was in its place before. Koffka remarks that the replacement of something by a full-fledged other thing, a full Gestalt might be responsible for this almost shocking erasure. Such an interesting idea could be tested using a retroactive inhibition (RI) setup. In this setup, where a List 2 would serve to inhibit an earlier studied List 1, one could have two versions of a List 2. In one version, it would be a full-fledged list comparable to the first list. The other version would have exactly the same items as the first version but a less complete, less stable rendering of them (e.g., something like the gaps we might have in low-quality photocopies where the identity of each word is nonetheless preserved; or, using letters instead of words as in Figure 9 so as to further control for top-down “fill-in” effects). One could then compare whether the amount of RI in the second condition would be less than in the normal condition. One strength of Gestalt theory, I believe, is the power it has in creating experimental ideas thanks to the guidance it provides as a full-fledged theory.

Figure 9 (From: Warrington & Weiskrantz, 1968)

Koffka remarks that when one talks of repetition, it is of critical importance whether one refers to the repetition of an outcome for a given behavior or a repetition of a process. He states that the empiristic view always and only focuses on the outcomes while ignoring the much more important process component of learning. As an example, he refers to the famous Tolman and Honzik (1930) study.

Tolman and Honzik’s latent learning study used a maze learning setup, as was common for the Behaviorist school of the time. Rats were randomly divided into three groups. The first group of rats was rewarded with food every time they reached the exit of the maze (“constantly rewarded group”- HR/hungry reward). The second group was never rewarded (“constantly unrewarded group”- HNR/hungry nonreward). The critical group was the third group of rats who were not rewarded till Day 11 when they unexpectedly were rewarded once they reached the exit of the maze (“group rewarded starting from Day 11” – HNR-R/hungry nonreward-reward). The original findings are presented in Figure 10.

Figure 10 (From: Tolman & Honzik, 1930)

The first group of rats made fewer and fewer error turns as they were rewarded every single time they reached the exit. This was a finding that was expected by the empiristically minded Behaviorist school. In contrast, rats in the second group, who never received an extra reward when reaching the exit, showed some though minimal decrease in errors (likely due to the fact that finding the exit was a certain kind of reward as the trial would come to an end and the animal would be free to go). The real surprise came with the third group. These rats behaved just as the rats in the second group. However, when rewarded on Day 11, their behavior on Day 12 was astonishing. Rather than showing a learning slope that was similar to that of rats in the first group, their decrease in errors was vastly accelerated, showing that they had learned the maze layout during their eleven days of strolling around. Tolman and Honzik were furthermore surprised to see that rats in the third group made even fewer errors than rats in the first group despite the fact that the latter had constantly been rewarded every single time over those eleven days. This study has come to be one of the most important studies to disprove the mechanistic account of the Behaviorist school which claimed that any learning was dependent on rewards, i.e., that without rewards no learning would ever take place. While discussing Tolman and Honzik’s latent learning study, Koffka remarks that the eleven days of free exploration allowed the rats to construct a map of the layout. Hence we would also expect rats in the second group to have learned a lot more even if they were not showing it because there was no incentive to do so. Koffka used indeed the term “map” in quotation marks in his 1936 book, which was a big taboo in the times of Behaviorism. This very term would reappear years later in Tolman’s 1948 article called Cognitive Maps in Rats and Men[21].

There was yet another critical finding, which Tolman and Honzik note in the discussion section. Given that the third group seemed to have learned the maze even better than the first group, they conclude that latent learning might indeed have been even more effective than overt learning. This sentence seemed to have gone unnoticed throughout all of mainstream psychology. Instead, the study is often portrayed to show that rats in the third group simply “catch up” with the first group, ignoring the fact that they actually outperform the all-time rewarded rats. Sometimes there is reference to this surprising “strange” finding only to be dismissed as to have been caused by the third group of rats “being more motivated”. How strange that no one cared to run follow up studies to see whether this was simply a difference in motivation or whether, as courageously alluded by Tolman and Honzik, the rats in the third group did indeed form –to use Koffka’s term- a much better map of the layout. Could it be that the third group of rats had the freedom of exploring the various alleys with no urge to reach the exit as fast as possible? On the other hand, could it be that when constantly rewarded, the animal’s learning became narrowed to the “safest, least risky exit route” which might have missed much better short-cuts available? Hence, as a result, while rats in the first group simply memorized a certain path without exploring the different alleys and corners of the maze, could it be that the unrewarded rats explored every single corner which then allowed them to reach the exit with much fewer errors? In other words, could it be that while rats in the first group were made to learn simply by rote, S-R association, the ones in the second and third group were latently learning a holistic map? With today’s knowledge of place and grid cells (O’Keefe & Dostrovsky, 1971, Hafting, Fyhn, Molden, Moser, & Moser, 2005) and possibilities of much more sophisticated behavioral and neuroimaging recording this should not be too impossible to investigate.

Tolman and Honzik’s study should have made a breakthrough, yet it did no more than come in handy for the so-called “cognitive revolution”, which I think, sadly missed what it could have achieved if it had better understood the conceptualizations of Gestalt theory (cf. Mungan, 2020/2021). The cognitive revolution, in that sense, might not really be a revolution because it did not bring about a radical change in perspective (despite the efforts of a very few, e.g., Neisser, 1976, also cf. Bruner, 1990). After all, the empirist’s mindset of believing that an understanding of the whole can be reached by summing up isolated “parts information” was tenaciously retained[22].

2.8 First demonstrations of implicit memory: Claparède 1911, Maccurdy 1928

Unlike in the time of old school empirists who excluded meaning and only focused on mechanisms, and preferably the simplest possible mechanism, in the 1950s with the “cognitive revolution”, concepts like meaning and context came back. Yet, it was as late as the 1980s that implicit memory became part of mainstream scientific research (e.g. Kunst-Wilson & Zajonc, 1980) and a subject of careful investigation (e.g., Hashtroudi, Ferguson, Rappold, & Chrosniak, 1989). Meanwhile, way back in the 1910s and 20s there were already very intriguing observations and analyses of implicit memory. In his book, Koffka draws attention to Claparède’s 1911 monograph where he reports about an advanced Korsakoff syndrome patient who would perform miserably on various kinds of memory tests and would never recognize the nurse who took care of him for the past 6 months but who would have no problems finding his way within the hospital. In the same memory chapter (and there are a total of four memory chapters in the book!) we see a little study by Maccurdy who, inspired by Claparède’s work, presents one of his severely amnesic patients his own name and address and asks the patient to repeat it back. The patient is unable to do so. He would thereupon present the patient a list of various names and last names, as well as street names and numbers and have the patient guess what Maccurdy’s name and address were. The patient would do so almost perfectly. Yet, Maccurdy notes that this memory was a memory without an owner, without feelings, without a “me-ness”, as Claparède would call it[23].

3. Erich Goldmeier’s Studies

In this section I will discuss Erich Goldmeier’s The Memory Trace: Its Formation and Its Fate book from 1982, published seven years before his death. Even the book’s title, reminiscent of a Borges story, is rather daring for any mainstream memory researcher. Yet, in this book, Goldmeier attempts a very meticulous analysis of changes in memory traces during their storage period.

The book starts with the criticism that almost all research on memory was overly focused on verbal memory. Yet, the human brain obviously did not evolve to store endless, disconnected words and syllables, and certainly their nonsensical versions neither. He remarks that we should therefore not be surprised by Shepard’s 1967 findings, which showed that participants were able to correctly recognize 600 studied pictures from among 1200 pictures with 98% accuracy[24], as they were stimuli with unified, inherent, rich meaning. He then goes on to talk about the studies by memory researchers such as Tulving, Craik and Lockhart, and remarks that their focus always seemed to be the stimulus list, the processing tasks participants were asked to do during encoding, and finally their memory performances at test. Yet, this, to Goldmeier, is insufficient as it ignores the most critical phase, i.e., the dynamics of what happens to the encoded material during the time between study and test. Those dynamics, Goldmeier states, are driven not only by how the person interacted with the material during encoding but also by the inherent, structural, “whole” characteristics of the stimuli. And it is this very aspect that he examines throughout his book, looking at visual stimuli but also texts and symbols as they are studied and retrieved across various experimental setups.

To facilitate the reading process, before moving on to the different sections of the book, I will first summarize the main tenets of Goldmeier’s Stress Theory, which might be called an elaboration of Wulf’s Dynamic Trace Theory.

1. The changeability of memory traces depends on their singularity/Prägnanz, hence on their stability. Patterns that are regular, that is, patterns that have a stable organization, form “strong wholes” that carry singularity (Goldmeier, 1982, pp. 63-64). Memory traces of such patterns will have close to zero chance of changing (e.g., Figure 11a). The second type of patterns are those that are almost singular (near-singularity). Those he describes as a “stable organization except for minor perturbations” (pp. 64). For those patterns we might expect a change towards singularity. In case of Figure 11b, it could be a change where the lower right dot might vanish and hence be no longer recalled at test. And finally, there are patterns that are nonsingular, i.e., patterns where parts do not have meaningful interrelations but instead are randomly configured, just as envisioned by the empirists. Figure 11c would be an example for that. Such patterns are expected to fade during storage in random, hard-to-predict ways. At test, Goldmeier, remarks, it is possible that a person’s mind might construct something of the remaining meaningless pieces just to produce an output, or, earlier schemata may “come to help” to reconstruct something more meaningful from the randomly faded residues. It is the latter that the empirists have been showing again and again, believing that such a fading and reconstruction-via-schemata process is the general process to explain memory failures.

Figure 11a-c (Goldmeier, 1982)

2. The singularity of the memory trace is dependent on the phenomenal perception of the physical properties of a given stimulus. For instance, Figure 11a may not carry singularity for people from a non-Western culture, hence, it may instead be a near-singular pattern that, say, would turn into a 4-dot square with the dot in the middle removed, as the square might be a singular, stable form in that culture but not its 5-dot “dice” version. Whereas in Gestalt theory, properties like symmetry and regularity are proposed to enhance the singular character of physical stimuli, it also emphasizes that their counterparts in phenomenal perception (as surely influenced by culture) are as important[25].

3. The trademark of phenomenal singularity is the instant detection of a small deviation that breaks it, i.e., its well-structured, stable, “good form”. For instance, a deviation is instantly detected in Figures 12b and 13b compared to the standard presented in Figures 12a and 13a (say in a recognition test) because it violates the stable parallel line structure of 12a, and straight line structure in 13a. On the other hand, the change in Figures 12c or 13c may go undetected because it only changes the angle without destroying the parallel line or straight line structure in 12a and 13a.

Figure 12a-c (Goldmeier, 1982)

Figure 13a-c (The lines were reconstructed by the author based on Goldmeier’s verbal descriptions in his 1942 book. In Fig. 13b, the “straight line” Gestalt of Fig. 13a is slightly deviated by an approximate 10-degree deviation on its upper part; on the other hand, in Fig. 13c, the straight line in 12a is rotated as a whole clockwise by approx. 10 degrees without breaking the “straight line” Gestalt[26].)

4. Stimuli that have singularity (e.g. Figure 14a) are encoded with the least effort and form memory traces that are the most resilient to time and interference. Near-singular stimuli (e.g., Figure 14b), in turn, lead to predictable memory trace changes (e.g., Fig. 14b turning into Fig. 14a), and once they resolve into a more stable, singular version, they become resilient to time and interference. In contrast, in the case of nonsingular stimuli (e.g., take Figure 14b and change it into a completely irregular, chaotic version of it with randomly long spikes that lack an overarching “good continuity” contour on top), we cannot predict the kind of change in the memory trace as that change will be a random change.

Figure 14a and 14b (Goldmeier, 1982)

According to Goldmeier, singularity means least complexity. Hence, if we take Figure 15a, its likely phenomenal perception will not be a grouping as depicted in Figure 15c but one as in Figure 15b since that one, phenomenally, is the simpler and, in terms of durability, the more “economic” one[27].

Figure 15a-c (Goldmeier, 1982; red circles added by the author)

Goldmeier further remarks that while the possible ways of grouping are rather restricted in singular patterns, just the opposite applies to nonsingular patterns. In the latter, there are multiple ways with none of them being the inevitable or even the dominant one. As an example, unlike Figure 16a, Figures 16b and 16c are nonsingular, hard to differentiate, and as such are ground-like rather the figure-like, are complex, and unsuggestive of any naturally emerging, stable kind of grouping.

Figure 16a-c (Goldmeier, 1982)

3.1 What is “same”: The mathematical vs. the phenomenal same

Goldmeier mentions that what is considered to be the same (or similar) in mathematical terms does not necessarily match with what is considered to be the same (or similar) in psychological terms. As an example he shows Figure 17a and remarks that while its mathematically enlarged (hence same) version (i.e., where all proportions are strictly retained) is Figure 17b, for the perceiving person it is rather 17c than 17b that most resembles 17a.

Figure 17a-c (Goldmeier, 1982)

Likewise, for the perceiving person it is Figure 18b that better mimics the pattern of Figure 18a rather than its mathematically exact stretched version in Figure 18c.

Figure 18a-c (Goldmeier, 1982)

Hence, in any memory study, the researcher has to consider phenomenal rather than mathematical similarity when planning which stimuli to match with which. This is particularly important if, aside from recall tests, the researcher plans to use recognition tests that require the generation of lure items. Goldmeier presents examples of similarity ratings provided by participants for different patterns and their variants, and notes that the ratings could not be predicted based on systematic, mathematical transformations of sameness. Hence, he warns that relying on mathematical assumptions regarding sameness judgments will be insufficient and even lead to misleading results.

3.2 What is singularity?

Goldmeier stresses that memory studies have to be as meticulous in their construction of stimuli as perception studies. After all, memory traces are often transformations of that which is perceived in physical space. Hence, it is of utmost importance to analyze the singularity qualities and the resulting stability of those traces, particularly so if it is indeed the case that they may go through transformations as they rest in memory.

Just like the three states of matter, Goldmeier proceeds to list the three states of memory traces:

(1) singular ones, (2) near-singular ones, and (3) nonsingular ones. Singular memory traces will show minimal change and display high resolution. Here he adds Bear’s 1973 definition of “a good shape is a shape where each of its parts strongly suggest each of its other parts”. Figures 11a, 12a, …, 16a, 17a are examples of singular images likely to stay unchanged as memory traces. Yet another example would be Figure 19a. In a memory experiment one would expect the recall of Figure 19a to be more stable and accurate compared to the recall of Figure 19b, despite the fact that Figure 19b is not at all that complex with its four straight lines and two instead of one angles.

Goldmeier also objects to the empirists’ “schema” proposal to explain the memory advantage for Figure 19a. Throughout the book, Goldmeier juxtaposes his stress theory with the empiristic fading-plus-reconstruction model. According to the empirists’ view, any fading memory trace will get activated via a resemblance to a schema and reconstructed accordingly. This change can happen through some manipulation during encoding (cf. Carmichael, Hogan, & Walter, 1932) or with or without a probe during retrieval. Yet, according to Goldmeier, this perspective completely ignores the change dynamics of the stimuli’s phenomenal and intrinsic characteristics. Moreover, it also cannot easily explain singularity effects with less familiar stimuli. The stimulus in Figure 20a, for instance, though less schematic compared to the one in Figure 19a, will nonetheless display much more stable reproductions compared to Figure 20b. If one claims that people might have used a schema for the recall of Figure 20a, then one would expect the same schema to be used in the recall of Figure 20b. Yet, while Figure 20a shows stable reproductions this is not the case for Figure 20b. In other words, as much as schemata may play a role, it is also the regularity or irregularity of a stimuli’s inner structure which will affect to what extent it will be remembered accurately and over long retention intervals.

Figure 19ab (Goldmeier, 1982)

Figure 20ab (Goldmeier, 1982)

In a different section, Goldmeier dwells on the possible sources, hence kinds, of singularity. One of them he defines as physiological singularity, which is a singularity that results from our “hardware”. He points to Eleanor Rosch’s 1971 research on category membership where she discovered, for instance, the existence of focal as opposed to nonfocal colors. Rosch found that within each color category of the typical rainbow colors, those that were focal (hence, of a given wavelength) were seen as clearer, rated as more prototypical, hence more memorable compared to their neighboring versions within the same color category. She found that this quality carried a universality that transcended culture, age and past learning experience, likely due to specialized retinal receptor cells that responded best for that given wavelength. Using Gestalt terminology, Goldmeier identifies this as an enhanced memorability caused by a singularity that is brought about by our unique physiological constitution.

As a second kind of singularity, Goldmeier proposes the one that results from imprinting, as observed in birds and mammals. A newly born gosling, for instance, will imprint on her mother as soon as she hatches, that is, she will instantaneously and for many months to come know her as her caretaker. Likewise, for a newly born baby, the face, voice and smell of her mother, in whose womb she grew and whose milk she sucked will assume singularity almost instantaneously and for years to come. These, too, are instances of very stable, insistent knowledge type memories.

The third kind of singularity he describes as the one that gradually evolves through learning. As an example, he mentions the singularities a radiologist perceives in tomography scans or the unique configurations a chess expert perceives in a chess display. Just as the other Gestaltists, Goldmeier, too, does not reject the phenomenal changes that occur in perception (and memory) via learning and experience. And just as the other Gestaltists, his objection is to reduce everything to mere learning as proposed by empiricism, as this disregards inherent structure of entities and instead takes the world of entities as equally random where all knowledge acquisition solely occurs through repetitive association. Such a view would likely not provide a satisfactory explanation for singularities, e.g., Rosch’s focal colors.

3.3 What is encoded?

So what is it that people encode as they process stimuli? Here, Goldmeier refers to Palmer’s 1975 study which showed that after studying holistic face drawings participants had a hard time recognizing pieces of it when presented in isolation, torn away from their intrinsic roles within the whole (see Figure 21a and its parts presented separately as in Figure 21b). Goldmeier remarks that this shows that at least in some cases we encode an instantaneous emergent whole (which makes its parts almost unrecognizable when presented in isolation) rather than single pieces from which a full image is constructed. Palmer likewise emphasizes that global processing inevitably weakens local processing. In other words, it seems that the two types of processing are opponent processes where one or the other dominates over the other.

Figure 21a and 21b (Palmer, 1975; cited in Goldmeier, 1982)

3.3.1 The difference between looking and seeing

Goldmeier refers to yet another study to address the question of what is encoded. It is a study by Wiseman and Neisser (1974) where participants received black and white images of faces of varying degrees of identifiability (Figures 22a-c). Their task was to try to detect a face in each of them. At the end, they were given a surprise recognition memory task where they had to mark all the images they had been presented. Findings were crystal clear: Images that could be identified as faces were recognized 1.5 times better than those that remained unidentified. Goldmeier interprets this as a difference in singularity. Once a participant was able to identify a face, that is, a whole, meaningful, structured image with inner coherence, it was now a singular pattern leading to a singular (hence stable) memory trace. In contrast, the same image would remain as a meaningless chaos for those who did not see the coherent face when looking at it. Hence the difference is one of looking versus seeing. Once the singular image emerges we deal with a stable and possibly very long lasting memory trace. Wiseman and Neisser, too, emphasize that what is important is not so much the stimulus itself but how it appears to the person, i.e., its phenomenal perception.

Figure 22a-c (Wiseman & Neisser, 1974; 21a is an easily identified face, i.e. identified by almost all participants, 21b is a moderately well identified face, i.e., identified by approximately half of the participants, 21cis a hardly identified face, i.e., identified by “relatively few”[28])

Goldmeier likewise discusses Blesser et al.’s (1973) handwriting stimuli (see Figure 23). These different “A”s do not have a common geometrical structure. What a person learns and remembers when learning to read, he says, is not so much geometrical structure but “the organization, the phenomenal parts and phenomenal features” (p.76). What he probably means is that what allows us to identify specific dynamics of the Gestalt principles of grouping such as good continuation, proximity, similarity etc.

Figure 23 (Blesser et al., 1973; cited in Goldmeier, 1982)

3.3.2 Encoding the non-existent, not encoding the existent…

In addition to analyzing visual images, Goldmeier also gives examples from narratives, texts, and symbols. For instance, when reading a text or listening to a story, things might get “added” just as the subjective contours in visual images (see Figure 24a and 24b), thus leading to the encoding of something that was not there.

Figure 24ab (Kanizsa, 1976)

Let us think of a short narrative such as this: “The woman wearily went to the supermarket. She chose 1 kilo of oranges from the fruit aisle and put it in the shopping cart. After she bent down and placed an orange on the floor back on the shelf, she went directly to the cash register, paid her bill and left the market[29].” It is quite likely that participants, when asked to recall the story thereafter, particularly a few hours later rather than immediately, will remember it as if they heard that the women herself dropped that orange. Yet, there is no such mention in the narrative as to who dropped the orange or whether the orange was laying on the floor all time through. But the almost reflexive inference that she dropped it makes the narrative more coherent, more stable, in other words, singular.

As an example of not coding the existent, Goldmeier refers to a study by Britton, Meyer, Simpson et al. (1979, as cited in Goldmeier, 1982). In this study, participants were given two different texts that had one paragraph in common. Whereas in one text, that paragraph was made to be a more central, figural component of the storyline, in the other text, it was made to be a more peripheral, ground-like component[30]. Findings showed that the very same paragraph was remembered well in the first case and almost not at all in the latter case. In other words, what turned out to be critical was not the inherent properties of the paragraph (say its own inner coherence) but its role in the larger, overarching whole. Goldmeier then cites additional studies which show that as long as the macro structure (i.e., the main storyline) is retained, microstructural features (i.e., parts, side narratives that do not affect the main meaning if left out) are typically lost in story recalls. Goldmeier remarks that to explain this away as being due to schematic, predictable, “ideal form” characteristics of the macro structure, as Mandler and Johnson suggest, is not fully convincing. He presents Alice in Wonderland as an example where the main storyline is absolutely far from being predictable and “ideal” but is still remembered very well. According to Goldmeier, what is critical is not so much experience-based acquired schemata but rather the singularity that emerges out of the inner coherence of a narrative.

3.3.3 Processes that stabilize memory traces…

In his book, Goldmeier also cites Craik and Tulving’s famous 1975 levels-of-processing study. In this study, participants were asked to engage in different processing tasks as they studied a list of words. For some words they were asked to mark whether or not it was written in upper case; for some, whether or not it rhymed with a sample item; for some, whether or not it belonged to a certain semantic category; and for some, whether or not it was a possible fill-in for an incomplete sentence provided next to it. After a while, participants were given a surprise memory test. Findings showed a very powerful levels-of-processing effect: Words processed semantically (i.e., in terms of their meaning, as in the latter two tasks) were remembered[31] five times as much as those processed with the nonsemantic tasks (i.e., the letter case and rhyme tasks). Yet there was another important finding, a curious finding that keeps replicating in various other studies but is still not too well understood: Across all tasks, words which received affirmative responses (regarding the processing questions) were significantly better remembered. Goldmeier states that for the “yes”-responded items, the words entered into a meaningful whole with the context and as a result assumed singularity, something that was not possible for the “no”-responded items. I believe that this a proposal worth considering and investigating.

3.3.4 Grouping, meaning-making, noticing and encoding a Gestalt

Goldmeier also scrutinizes our memory for symbols. Here he cites Katona’s 1940 study where participants are presented with various 12-digit numbers (e.g., 581215192226) in different ways. One group was asked to memorize each by reading it aloud in groupings of threes (“five hundred eighty-one, two hundred fifteen, …”). The second group always first received a sentence fragment “The federal expenditures in the last year amounted to $…….” and then the number which they again had to read out loud as a monetary value (“five billion eight hundred one million[32] ….”) . Hence, in this group, the number was embedded into a semantic context. The third group, on the other hand, was simply told to memorize it. Right after that, when asked to write down the number, all groups showed more or less equal performance. The only difference, Katona noted was that the third group was slower in retrieving compared to the other two groups. However, when given a surprise memory test a week later, neither the first group which had studied the sequence as triplets, nor the second group which studied it in a meaningful context were successful. On the other hand, some of the participants of the third group remembered the series flawlessly because it turned out that they managed to discover the rule behind the series (adding, in alternation, a 3 and a 4 to each number to obtain the succeeding one) and it was that rule which they remembered and hence applied during recall given they remembered the first number. We may liken this to what happened in the difficult-to-identify black and white face illustrations in the Wiseman and Neisser 1974 study. Those who detected the rule were like those who were able to discern a face in the black and white patches. Likewise, those who did not detect the rule saw the 12-digit number series as irregular, i.e. random, just as the participants who could not detect a face and hence only saw “chaos” instead of a face. Just as in the Wiseman and Neisser study, those who saw order outperformed those who did not. In Gestalt terms, the difference depended on whether participants were able to discern a figure or were lost in just seeing something with no singular character, i.e., a ground.

In summary, encoding is selective and economical according to Goldmeier. It tries to code the most important information with the least effort. Hence, says Goldmeier, you might forget what a person was wearing, or whether they had glasses or not but you would for sure know that they were not naked. After all, if they were naked, that would make them a “figure”, just as in the “von Restorff effect”, whereas their outfit (if it did not relate to the conversation or was not something of value in your phenomenal field of interest) would function as ground rather than figure and be hence encoded in low resolution, as a “blurred” memory trace.

3.4 Phenomenal asymmetries in stimuli

In the last section of his book, Goldmeier talks about how different stimuli (e.g., images, numbers, words, concepts) carry different degrees of stress (in the sense of field forces) depending on their singularity status. He remarks that it is these asymmetrical stress dynamics that will influence phenomenal perception and hence bring about different memory dynamics. Eleanor Rosch, for instance, found that when presenting participants focal and nonfocal colors, they would rate a purplish red as similar to a focal red but not vice versa, i.e., a focal red would not be rated as similar to a purplish red. Similarly, she found that the number “103” would often be coded as “close to 100” but the number “100” would never be coded as “close to 103”. In other words, while nonfocal category members would assimilate to the central, focal category member, a focal member would remain separate from its nearby nonfocal members. As such, this should not simply be taken as some isolated, “interesting phenomenon”. Instead, in a memory setup, we might expect that nonfocal category items would be more likely to be misremembered as focal items, whereas focal ones would not be subject to such memory trace changes. This is indeed what Rosch finds: a purplish red was often misremembered as red but a red was not misremembered as purplish red; the number “103” was often misremembered as “100” but “100” was not misremembered as “103” or even “101”. These asymmetries observed in both perception and memory are interpreted by Goldmeier from a Gestalt conceptualization of “field forces”. Hence, be it a number (“103”), a symbol (“~”), a concrete (“cupboard”) or an abstract (“agony”) word or a conceptual terminology (“semi-presidential system”), if each has a more prominent, a more focal neighboring number, symbol, word or conceptual terminology, they will likely turn into these in later recall (e.g., into “100”, “”, “cabinet”, “pain”, “presidential system[33]”..).

3.5 The sine qua non of memory experiments

Goldmeier raises two main criticisms to mainstream memory research, (1) that the stimuli are disconnected, random items, and (2) that the memory tests are mostly done after a short retention interval of ca. 15-20 minutes. Instead, according to Gestalt theory, if the goal is to do “ecologically valid” studies, to use Neisser’s terminology, that is, to run studies that relate to real life, the stimuli have to be meaningful (while a random item list should certainly be included but as a control list). In addition, memory tests should include more real-life retention intervals that expand over days even weeks rather than minutes. The basic experimental design of Goldmeier’s studies reported in his book is a 3 (stimuli: singular, near-singular, nonsingular)within-ss x 2 (type of memory test: free recall and recognition)within-ss x 4 (retention interval: immediate, 2 weeks, 4 weeks, or 6 weeks)between-ss. All participants were exposed to all three singularity types and their encoding periods were such as to allow for meaningful processing (e.g., if the to-be-learned stimuli were visual images, participants were granted 25 seconds per image during which they had to draw the figure twice on a piece of paper; that task would be presented as measuring participants’ copying ability, with no mention of a subsequent memory test). After a given time period, depending on their assigned retention interval, half of the participants were first given a free recall test, then a recognition test, whereas half were first given a recognition test and afterwards a free recall test. The first test always came as a surprise, hence, to keep conditions equal, only the data of the first tests were analyzed.

3.6 Main findings of Goldmeier’s empirical studies

Regardless of whether the items were visual images, verbal materials, or symbols, Goldmeier’s main findings were as follows.:

1. What was remembered was linked to the singularity characteristic of a given stimulus. Singular stimuli were remembered with almost no transformation across all retention intervals. Near-singular ones which were like singular ones except for an additional or missing feature were remembered by deleting the extra or adding the missing; this tendency was more pronounced at longer retention intervals.

2. In all stimulus materials (visual shapes, narratives, symbols) it was observed that there were some versions that would never appear as a possible transformation in free recall or be falsely chosen in a recognition setting. Figure 25a, for instance, was never reproduced (not by any of 23 participants) with the upper arc missing. Twenty participants reproduced the arc as it should be, i.e., with three vaults, one participant with four vaults, one with two vaults, and one with a single vault. In contrast, the lower straight line was kept in only 11 out of 23 participants (though not necessarily in horizontal orientation, see Fig. 25b). Ten out of 23 participants completely dropped that line. As can be seen, what is remembered or not in such nonrandom shapes is not at all arbitrary and also not necessarily predicted by its familiarity (a straight horizontal line is probably one of the most singular shapes around the world). It is the features that make a stimulus singular in its entirety that are resistant to forgetting. In contrast, those that do not are easily forgotten.

Figure 25a-d (Goldmeier, 1982; Fig. 25a, target stimulus; 25b-c-d different recalled versions)

Along the same lines, in Figure 26, if the studied picture is 26h, it can be misrecognized as 26c or 26d, but never as 26e, 26f, or 26g. Goldmeier explains this in terms of the differences in phenomenal stress caused by the gaps within (26h  26c and 26d) versus between (26h ≠ 26e, 26f, and 26g) continuous lines[34]. In contrast, if the studied image was the one in Figure 26e, no change was observed, even after 6 weeks for delay. Similarly, Figure 27a was never misrecognized as Figure 27b because of its unique inner coherence that nicely fits with the principle of good continuation, as proposed by Wertheimer in his seminal 1923 paper, introducing the Gestalt principles of grouping/organization.

In one or two places, Goldmeier refers to Bartlett’s and Wulf’s earlier studies emphasizing that these lacked a conceptual framework to explain their findings. Hence, even though both lines of research revealed interesting findings they were doomed to remain descriptive[35].

Figure 26 (Goldmeier, 1982)

Figure 27a and 27b (Goldmeier, 1982)

4. Finishing Remarks, In Place of a Conclusion

In this article, I presented the interesting yet little known conceptualizations of Gestalt theory on memory. As I mentioned in the introduction, Gestalt theory’s proposal that memory traces can undergo nonrandom changes as they rest in memory and without necessarily external manipulations (such as misinformation in between study and test) is probably one of its most central ideas. Another critical contribution is its ability to apply the figure-ground conceptualization of perception to memory in the most ingenious way. According to Gestalt theory, other than the intrinsic figural stability (which Goldmeier preferred to call “singularity”, and Wertheimer “Prägnanz”) of a stimulus, also its “figure” identity that emerges from the contextual environment it is embedded in (e.g. a single 2-digit number within a list of nonsense syllables or vice versa) will determine whether and what kind of forgetting mechanisms will be at work, as elegantly shown by Hedwig von Restorff. Gestalt theory criticizes mainstream memory research for having for decades studied mostly random, structureless items embedded in their ‘own copies’, fired at the participants at fast rates, one after the other. Hence all replicated effects regarding encoding, retrieval and forgetting mechanisms were mistakenly believed to apply across the board while they might so far have only revealed how memory works for nonsingular, random, ground-like items presented at hasty rates. Yet, we do not live in a world of ground, of randomness, devoid of structure and meaning. The mechanistic worldview of the empirist encounters serious challenges when trying to explain how we remember poetry, music, fairy tales, myths, even conspiracy theories as it shies away from trying to understand structured meaning that goes beyond the static idea of “schemata” (cf. Bruner, 1990). Hence, it is no wonder that we know very little about these as we are approaching 150+ years of psychological research. Instead, at least within memory research, we see a plethora of “findings”, “effects”, “phenomena[36]” , which can hardly be coherently explained even within themselves let alone across all domains of memory, and let alone, cognition in general.

This paper should be seen as a presentation of the main points and highlights of Gestalt theory’s conceptualization of memory processes. I believe that even this little glimpse should surprise the reader as to its richness in ideas and methods. Then, one cannot help ask why neither psychology intro books and memory textbooks nor memory lectures at universities cover at least one or two of their studies with the theoretical credit it deserves. We should leave this strange “neglect” to scholars working in historiography, yet, anyone within psychology, and particularly within cognitive psychology should be wondering about this. Could it be that the proposals of Gestalt theory, which it took care to put to empirical test (with the least costly yet most ingenious experiments), were ignored because of its audacious tenet that memory traces have “inner dynamics”? If so, what was so utterly unacceptable about this? Could it be that the Anglo-American mainstream simply did not like it as they believed it would replace the mechanistic with the “uncanny”? Who knows…

All things aside, it is impossible not to see the creative, cleverly crafted, productive, stimulating and refreshing side of Gestalt theory’s proposals regarding memory. Maybe one of its most exciting aspects is that a lot of intriguing memory phenomena that appear in literary works, while unaddressed or unaddressable by current mainstream memory models, find their possible explanations in Gestalt theory. Even the idea of thinking of time as a “ground” type memory is potentially a brand-new way of looking at things as it provides a framework of thinking. After all, there should be a way to explain why Hans Castorp in Thomas Mann’s The Magic Mountain lost his sense of time when he first arrived as a visitor at a sanatorium in Davos to see his tubercular cousin only to be interned himself first for a few days due to health problems, then for another few days which “in a glimpse of a moment” turned into seven years. This experience of losing one’s seven years should not be too far from what we have been experiencing over the two years of the Covid-19 pandemic. A vast majority of people report feeling as if those two years were stolen from them, as if they were “unlived” with no clear memories to refer back to. If memory research continues to shy away from these real life experiences or does no more than give rating scales and collect data with no real overarching theory, I am afraid we will be doomed to be doing no more than beating the air, or as we say in Turkish, beat the water in the mortar.

Footnote

[1] For a inspiring article on two brilliant women of Gestalt theory, Hedwig von Restorff and Bluma Vulfovna Zeigarnik, see MacLeod (2020).

[2] The first person to propose the term “engram” was the German zoologist and evolutionary biologist Richard W. Semon (1859-1918), who was working on a brain-based organic model of memory. As such, “engram” refers to a property of long-term memory.

[3] Köhler (1950) coins the term “empiristic” to refer to experimentalists who knowingly or unknowingly subscribe to a simple, mechanistic, associationistic view. As such, he separates those from scholars with an in-depth empiricist philosophy.

[4] These type of groupings are discussed in detail by Max Wertheimer (see Wertheimer’s still not fully translated 1923 article; for a detailed treatise on Wertheimer (1923), see Mungan, 2020)

[5] This issue (and, for that matter, the intriguing issue of chunking in working memory) is still not studied sufficiently. For instance, if listeners are presented with an auditory sequence which does not allow for a clear grouping, say a sequence of pulses of various, random decibels (e.g., xxxxxxxxxx….), they would have a hard time to retain each preceding pulse due to its unpredictability. Grouping is one of the most important factors predicting working memory capacity (cf. Ericsson, Chase, & Faloon, 1980). It is of no surprise that all music across the world has some kind of beat, i.e., regularity.

[6] Certainly, experimental music can make such external sounds part of a musical whole by using rhythmic integration or massed repetition.

[7] Since there is no sensory receptor for time processing, we have to call it a perception rather than a sensation.

[8] While we know today that there are different brain regions with different temporal resolutions, we still do not know how the brain processes and stores time in its entirety (cf. Buhusi, 2020; Ünal & Ayhan, 2020)

[9] We found a similar threshold asymmetry in some of our studies (e.g. Mungan ve Kaya, 2020). There is yet no explanation for these. We know that Barbara Tillmann’s lab also found such asymmetries (Barbara Tillmann, personal communication). Sadly, because of either an ambition or pressure to keep things “simple”, psychophysical studies almost without exception report difference thresholds not separately (i.e., thresholds obtained in ascending vs. descending conditions) but as averages (i.e., the two conditions collapsed). This way, this rather curious asymmetry managed to go unnoticed again for the decades to follow…

[10] After the recall test, participants were asked to write down the text they had to memorize in phase 2 to ensure they would not suspect that that phase was simply a distractor phase.

[11] I think that this figure-ground conceptualization of Gestalt theory, be it within the domain of perception, memory, or problem solving, allows for very interesting predictions and understandings of various disconnected “phenomena” listed in mainstream psychology books. For instance, we know that people have a tendency to believe that a head/tail sequence such as HHHHHHH is less likely than a sequence such as HTTHTTT (cf. Tversky & Kahneman, 1971). This could as well be understood from a Gestalt perspective, in that the former has the qualities of a singular, prägnant shape, i.e., is a figure (hence of low likelihood), whereas the latter appears like any of those random alterations that are hard to differentiate, hence is seen as a ground (as more likely because most of the sequences are of this kind). Hence, it could be for this reason why the latter is failed to be perceived in its unique, figural form. In other words, unlike Tversky and Kahneman’s proposal that we are dealing with a mere fallacy in thinking we might propose that what is happening is a perceptual, even mnemonic phenomenon where the distinct HHHHHHH is retained flawlessly whereas the alternating sequence cannot be retained in its singularity due to its ground character.

[12] In free recall, the recall of a given item will not only influence the next item to be recalled but with every next recall, the chances of the remaining items to be recalled will decrease as each recalled word creates interference with the to be recalled upcoming items. Hence, as output increases, recalling additional items becomes less and less likely.

[13] In a critical footnote, von Restorff states that if the items in a given category form a meaningful whole, they no longer are monotonic but instead are well-remembered in later tests. She then cites a few studies that showed this. Interestingly, this phenomenon was presented as a novel finding in Chan’s (2009) study on retrieval-induced forgetting. This does not come as a surprise, given that mainstream memory literature is absolutely unfamiliar with the details of the von Restorff’s studies.

[14] A beautiful narration of this loss of a sense of time can be found, for instance, in Thomas Mann’s Der Zauberberg (The Magic Mountain) novel. For a way delayed scientific treatise of this phenomenon, see Draaisma’s 2004 Why Life Speeds Up As You Get Older book.

[15] As much as from the 1970s on and particularly post-1980s, memory research started to talk about context effects (e.g., Johnson, Doll, Bransford, & Lapinski, 1974; Bower, Karlin, Dueck, 1975; Jonker, Seli, MacLeod, 2013), distinctiveness effects (cf. Hunt, 2006, also see Hunt, 1995, for some discussion of what was largely misinterpreted in von Restorff’s findings), focal/nonfocal cues in forgetting (Kliegel, Jäger, Phillips, 2008), what has been lacking is a single, coherent understanding behind those multiple lines of mainly disconnected research. Mostly, such effects are explained based on attention (the “joker card” to use Koffka’s words). Attention is something that occurs at a given moment due to some salience (or, say, interest, as in the case of endogenous attention). Yet, if we look at von Restorff’s findings, for instance, the saliency of an item emerges afterwards, in the larger context, not at that instant moment, as required by attentional processes. This, I believe, is an important difference.

[16] Donald Hebb and Karl Lashley referred to Gestalt theory’s proposal of dynamic memory traces as its most “unacceptable” proposal. Koffka makes reference to this harsh objection and implies at several places that this upfront rejection by Hebb and Lashley is doomed to succumb in the time to come.

[17] We see that Wulf’s study precedes Bartlett’ın 1932 studies by 10 years. Moreover, Wulf’s research displays relatively meticulous experimentation (e.g., he pilots a large set of shapes and also different retention intervals to see which ones are more appropriate for the purpose of the study). He also came up with rather creative methods of testing. He would, for instance, construct a cued recall version for his studies where participants would receive a cue to help them recall the shape with the instruction that they were also free to revise the cue itself once they thought they remembered the target drawing.

[18] We must not forget that these similarities are bound by their era. Today, the same image might rather transform into a more straightened up smiley.

[19] As much as particularly with Loftus (1975), we see a plethora of studies in eyewitness research, false memory (Roediger & McDermott, 1995) and autobiographical memory (e.g., Platt, Lacey, Iobst, & Finkelman, 1998; Bahrick, 2014) that look into memory distortions, most of these use clear manipulations to bring about these changes. In cases where changes are observed as is (e.g., in autobiographical memory studies), the proposed mechanism is again a kind of “joker card” of top-down processing. Gestalt theory, in contrast, proposes additional mechanisms of change that are brought about not necessarily via top-down processes but via figure-ground, grouping, and most important, Prägnanz dynamics. These mechanisms are explained in more detail under the Erich Goldmeier section of this article. One could even say that Gestalt theory “dislikes” to artificially separate bottom-up from top-down, and likewise, the environmental from the phenomenal, as all these exist jointly and create their effects jointly.

[20] This change in itself is rather interesting. During the times that Koffka wrote his book, comic books, animations, above all avant-garde versions of them were not that widespread. Possibly, quickly understanding the strangest drawings may be acultural learning (cf. Cohn & Magliano, 2020).

[21] Even though Tolman never considered himself a Gestaltist, he did meet Kurt Koffka in the 1910s when he visited Germany to improve his German while he was a PhD student amongst Behaviorists at Harvard. That was just about the time when the Gestalt School was slowly taking over the current empirist structuralist (mostly stationed in Germany) and behaviorist (mostly stationed in the US) climate. Years later in 1935, Tolman joined a meeting by Kurt Lewin, who had settled in the US in 1933 and founded the Topology Society. There even is a picture of him, Lewin and Koffka (cf. Goodwin, 2005)

[22] Therefore, it is of no wonder that both in my country and in the world, educational systems are still almost blindly following a system of reward and punishment, giving scores for every single thing a student does, rewarding GPAs and hence inflating them, unaware of, or, oblivious of (or indifferent to?) the fact that humans (even rats!) can learn without that and might even learn much more when not imprisoned in a “Skinner box”. Worse still, today in this “Skinner box” of life, we are made to race against each other in fierce and ruthless competition at the expense of enjoying the process of open, explorative learning.

[23] In 1968, Zajonc used a very similar method in his famous mere exposure effect study.

[24] The stimuli were colored pictures. A 2-choice recognition test after 2 hours of delay led to close to 100% correct responses, after 3 days to 92%, and after one week still to 87% correct responses.

[25] As an example, we might think of distinctiveness as a more phenomenal than physical property. In von Restorff’s studies, it is the context which determines whether a given stimulus has figure or ground quality.

[26] Note that the angle values are only roughly represented.

[27] Here Goldmeier refers to Hochberg & Attneave’s attempt to quantify Gestalt theory’s “singularity” concept from an information processing theory (cf. Shannon,1948) as indexed by predictability.

[28] This expression is used by Wiseman and Neisser (1974) with no further detail.

[29] This narrative was inspired by Hannigan and Reinitz’ (2001) study.

[30] Goldmeier compares this latter version to Gottschaldt’s “embedded figures”, where a distinct part can become invisible once embedded within a larger form.

[31] This was so for both recognition and recall tests. Presenting items twice (not massed but spaced out) instead of once during encoding even further enhanced the effect.

[32] To make that possible, the 12-digit number was written like monetary values with a decimal point marker at the end, 581215922.26. This study was run in 1939 and Katona admits that this minute change does add a flaw to the study.

[33] Naturally, we would expect that the more complex and layered the item is, the more the field dynamics would vary from person to person and even from one time to another in the same person. Hence, we would expect a casual number 103 in a report to be later on remembered as “close to 100” or even “100” in most cases. In contrast, we would expect a terminology such as “semi-presidential system” casually mentioned in some report, to be less predictable in terms of whether and how it would change at later recall since this would more heavily be influenced by each person’s unique phenomenal world of meanings.

[34] Throughout his book, Goldmeier would compare his terminology with those of the empirists of his time, e.g., with Tversky’s 1977 “saliency” term in his feature theory of similarity or likewise with Rosch’s prototype and schema concepts in her feature theory of categorization. At places he would acknowledge that the same phenomena could be explained through different concepts. Yet what I think is the most important is that none of the other paradigms can say as much on perception, memory, productive thinking and possibly other domains in psychology as Gestalt theory with its overarching postulates of prägnanz dynamics, transposibility, figure-ground segregation, and grouping (cf. Mungan, 2020/2021). Here we have a theory that can explain a wide variety of phenomena in visual as well as auditory perception, in memory, in productive thinking and possibly more. In contrast, most models, theories, phenomena within cognitive psychology do not even generalize across its own domains but are mostly only applicable to a narrow range within a given domain, say memory, or perception, or thinking (and even worse, sometimes not even across a given domain).

[35] Goldmeier notes that what Bartlett observes in his The War of Ghosts study series is not so much a simple ‘assimilation into cultural schemata’ mechanism but rather a tendency to transform things into a more coherent, singular narrative. After all, the participants came all from a very homogeneous cultural and socio-economic background yet there were considerable differences in their reproductions. Hence, a simple schematic assimilation explanation is insufficient.

[36] When browsing the most “in” scientific journals in psychology at the turn of the previous century (e.g. Ebbinghaus’ famous Zeitschrift für Psychology) one is shocked to see some names with an incredibly high publication record with almost one per issue, i.e., the “high-scorers”, whose names no one would recognize today. This makes one wonder whether a similar thing will happen a century from now when people will look back at the “all-rewarded”, “all-promoted” research of today in this strange climate of publish or perish. How reminiscent of the Tolman & Honzik (1930) study…

[37] Let us only think of the countless number of models, theories, theory-like proposals within every, carefully separated niche of memory research, where researchers turn into experts of their little domains and subdomains, unaware of the findings being produced minute-by-minute in surrounding areas, as they are all drowned in their own terminology to make cross-talk close to impossible.

5.References

Bahrick, H. P. (2014). Loss and distortion of autobiographical memory content. In Autobiographical Memory (pp. 69-78). Psychology Press.

Bartlett, F. C. (1932/1977). Remembering: A Study in Experimental and Social Psychology. Cambridge: Syndics of the Cambridge University Press.

Blesser, B., Shillman, R., Cox, C., Kuklinski, T., Ventura, J. ve Eden, M. (1973). Character recognition based on phenomenological attributes. Visible Language, 7(3), 209-223. DOI: 10.1016/S0020-7373(74)80033-7

Bower, G. H., Karlin, M. B., & Dueck, A. (1975). Comprehension and memory for pictures. Memory & Cognition, 3(2), 216–220.

Bregman, A. S. (1994). Auditory scene analysis: The perceptual organization of sound. MIT press.

Bruner, J. S. (1990). Acts of Meaning. Cambridge, MA: Harvard University Press.

Buhusi, C. V. (2020). Episodic time in the brain: A new world order. Learning & Behavior, 189-190. DOI: 10.3758/s13420-019-00379-4

Carmichael, L., Hogan, H. P. ve Walter, A. A. (1932). An experimental study of the effect of language on the reproduction of visually perceived form. Journal of Experimental Psychology, 15(1), 73-86.

Chan, J. C. (2009). When does retrieval induce forgetting and when does it induce facilitation? Implications for retrieval inhibition, testing effect, and text processing. Journal of Memory and Language, 61(2), 153-170. DOI: 10.1016/j.jml.2009.04.004

Cohn, N. ve Magliano, J. P. (2020). Editors’ Introduction and Review: Visual Narrative Research: An Emerging Field in Cognitive Science. Topics in Cognitive Science, 12(1), 197-223. DOI: 10.1111/tops.12473

Craik, F. I., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104(3), 268-294. DOI: 10.1037/0096-3445.104.3.268

Darwin, C. J., Turvey, M. T., & Crowder, R. G. (1972). An auditory analogue of the Sperling partial report procedure: Evidence for brief auditory storage. Cognitive Psychology, 3(2), 255-267.

Ebbinghaus, H. (1885/1964). Memory: A Contribution to Experimental Psychology. New York: Dover Publications.

Eichenbaum, H. (2013). Memory on time. Trends in Cognitive Sciences, 17(2), 81-88.

Ericsson, K. A., Chase, W. G., & Faloon, S. (1980). Acquisition of a memory skill. Science, 208(4448), 1181-1182.

Goldmeier, E. (1982). The Memory Trace: Its Formation and its Fate. Hillsdale, NJ: Lawrence Erlbaum Associates Publishers.

Goodwin, C. J. (2005). A History of Modern Psychology. Hoboken, NJ: John Wiley & Sons.

Hannigan, S. L., & Tippens Reinitz, M. (2001). A demonstration and comparison of two types of inference-based memory errors. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(4), 931–940. DOI: 10.1037/0278-7393.27.4.931

Hashtroudi, S., Ferguson, S. A., Rappold, V. A. ve Chrosniak, L. D. (1988). Data-driven and conceptually driven processes in partial-word identification and recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14(4), 749-757. DOI: 10.1037//0278-7393.14.4.749

Hochberg, J. (2003). Acts of perceptual inquiry: Problems for any stimulus-based simplicity theory. Acta Psychologica, 114(3), 215-228. DOI: 10.1016/j.actpsy.2003.07.002

Hunt, R. R. (1995). The subtlety of distinctiveness: What von Restorff really did. Psychonomic Bulletin & Review, 2(1), 105-112. DOI: 10.3758/BF03214414

Hunt, R. R. (2006). The concept of distinctiveness in memory research. In R. R. Hunt & J. B. Worthen (Eds.), Distinctiveness and memory (pp. 3–25). Oxford University Press. DOI: 10.1093/acprof:oso/9780195169669.003.0001

Johnson, M. K., Doll, T. J., Bransford, J. D., & Lapinski, R. H. (1974). Context effects in sentence memory. Journal of Experimental Psychology, 103(2), 358–360.

Jonker, T. R., Seli, P., & MacLeod, C. M. (2013). Putting retrieval-induced forgetting in context: An inhibition-free, context-based account. Psychological Review, 120(4), 852–872.

Kanizsa, G. (1976). Subjective contours. Scientific American, 234(4), 48-53.

Kliegel, M., Jäger, T., & Phillips, L. H. (2008). Adult age differences in event-based prospective memory: A meta-analysis on the role of focal versus nonfocal cues. Psychology and Aging, 23(1), 203–208.

Koffka, K. (1935/1936). Principles of Gestalt Psychology. London: Kegan Paul.

Köhler, W. (1938). The Place of Value in a World of Facts. New York: Liveright.

Köhler, W. (1950). Psychology and evolution. Acta Psychologica, 7, 288–297. DOI: 10.1016/0001-6918(50)90020-5

Kunst-Wilson, W. R. ve Zajonc, R. B. (1980). Affective discrimination of stimuli that cannot be recognized. Science, 207(4430), 557-558. DOI: 10.1126/science.7352271

Loftus, E. F. (1975). Leading questions and the eyewitness report. Cognitive Psychology, 7(4), 560–572.

Loftus, E. F., & Hoffman, H. G. (1989). Misinformation and memory: The creation of new memories. Journal of Experimental Psychology: General, 118(1), 100-104. DOI:10.1037/0096-3445.118.1.100

Loftus, E. F. (2005). Planting misinformation in the human mind: A 30-year investigation of the malleability of memory. Learning & memory, 12(4), 361-366.

MacLeod, C. M. (2020). Zeigarnik and von Restorff: The memory effects and the stories behind them. Memory & Cognition, 48, 1073–1088. DOI: 10.3758/s13421-020-01033-5

Maniatis, L. (2021). Why did Karlovich & Wallisch (2021) paste the" Gestalt" label on notions the Gestaltists rejected and disproved?

Maniatis, L. (2021b). Why do vision researchers model data based on assumptions they know to be false?

McCloskey, M. ve Zaragoza, M. (1985). Misleading postevent information and memory for events: Arguments and evidence against memory impairment hypotheses. Journal of Experimental Psychology: General, 114(1), 1-16. DOI: 10.1037//0096-3445.114.1.1

Mungan, E. (2020/2021). Geştalt Kuramı: Bir “Nazariye” nin mazisi, akameti ve akıbeti. Nesne, 8(18), 585-618. DOI: 10.7816/nesne-08-18-15 / FOR ENGLISH,

Mungan, E. ve Kaya, E. (2020). Some Time Violations Go Less Noticed: Gestalt Grouping? A Brief Report. Timing & Time Perception, 1(aop), 1-13. DOI: 10.1163/22134468-bja10010

Neisser, U. (1976). Cognition and Reality: Principles and Implications of Cognitive Psychology. New York: Freeman

Nickerson, R. S. ve Adams, M. J. (1979). Long-term memory for a common object. Cognitive Psychology, 11(3), 287-307. DOI: 10.1016/0010-0285(79)90013-6

Ólafsdóttir, H. F., Bush, D., & Barry, C. (2018). The role of hippocampal replay in memory and planning. Current Biology, 28(1), R37-R50.

Palmer, S. E. (1999). Vision science: Photons to Phenomenology. MIT press.

Platt, R. D., Lacey, S. C., Iobst, A. D., & Finkelman, D. (1998). Absorption, dissociation, and fantasy‐proneness as predictors of memory distortion in autobiographical and laboratory‐generated memories. Applied Cognitive Psychology: The Official Journal of the Society for Applied Research in Memory and Cognition, 12(7), S77-S89.

Restle, F. (1970). Theory of serial pattern learning: Structural trees. Psychological Review, 77(6), 481–495. DOI: 10.1037/h0029964

Rock, I. (1984). Perception. New York: W.H. Freeman.

Roediger, H. L. (1978). Recall as a self-limiting process. Memory & Cognition, 6(1), 54–63. DOI: 10.3758/BF03197428

Roediger, H. L. (1980). Memory metaphors in cognitive psychology. Memory & Cognition, 8(3), 231-246. DOI: 10.3758/BF03197611

Roediger, H. L. ve McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(4), 803-814. DOI: 10.1037/0278-7393.21.4.803

Shepard, R. N. (1967). Recognition memory for words, sentences, and pictures. Journal of Verbal Learning and Verbal Behavior, 6(1), 156-163. DOI: 10.1016/S0022-5371(67)80067-7

Sutherland, G. R., & McNaughton, B. (2000). Memory trace reactivation in hippocampal and neocortical neuronal ensembles. Current opinion in neurobiology, 10(2), 180-186.

Tolman, E. C. ve Honzik, C. H. (1930). Introduction and removal of reward, and maze performance in rats. University of California publications in psychology, 4(17), 257-275.

Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189-208. DOI: 10.1037/h0061626

Tversky, A. ve Kahneman, D. (1971). Belief in the law of small numbers. Psychological Bulletin, 76(2), 105–110. DOI: 10.1037/h0031322

Ünal, G. ve Ayhan, İ. (2020). İşlevsel Özelleşmeye yeni bir bakış: Nöronal saatler. Nesne, 8(17), 270-283. DOI: 10.7816/nesne-08-17-08

Von Restorff, H. (1933). Über die Wirkung von Bereichsbildungen im Spurenfeld. Psychologische Forschung, 18(1), 299-342.

Warrington, E. K. ve Weiskrantz, L. (1968). New method of testing long-term retention with special reference to amnesic patients. Nature, 217(5132), 972-974. DOI: 10.1038/217972a0

Wertheimer, M. (1923). Untersuchungen zur Lehre von der Gestalt. II. Psychologische Forschung, 4(1), 301-350.

Wertheimer, M. (1945). Productive thinking. New York: Harper.

Wiseman, S. ve Neisser, U. (1974). Perceptual organization as a determinant of visual recognition memory. The American Journal of Psychology, 87(4), 675-681. DOI: 10.2307/1421974

Zaragoza, M. S., Belli, R. F., & Payment, K. E. (2007). Misinformation effects and the suggestibility of eyewitness memory. Do justice and let the sky fall: Elizabeth Loftus and her contributions to science, law, and academic freedom, 35-63.

Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2), 37-40. DOI: 10.1037/h0025848