Gestalt: What Replicated and What Did Not

The paper everyone cites runs from page 301 to page 350 of volume 4 of Psychologische Forschung, published in 1923. Its title is Untersuchungen zur Lehre von der Gestalt II. Most designers have never read it, and neither had I until a few years ago, because the version we all learn comes through a chain of intermediaries: Willis Ellis translated a condensed form of it for A Source Book of Gestalt Psychology in 1938 under the title Laws of Organization in Perceptual Forms, and almost every design book since has quoted the condensation rather than the paper.

This matters more than it sounds. The gap between what Max Wertheimer demonstrated and what a slide deck says he proved is wide enough to fall into, and I have fallen into it in front of clients. What follows is an attempt to be accurate about which parts of the Gestalt account survived a century of testing, which parts turned out to be vaguer than the design literature assumes, and how to keep using the useful parts without claiming they are laws of nature.

What is actually in the 1923 paper

Wertheimer's method was demonstration rather than measurement. He printed rows of dots, arrays of lines and simple figures, and asked the reader to notice what the eye does with them. The factors he names, in the Ellis translation, are the factor of proximity, the factor of similarity, the factor of uniform destiny (also given as common fate), the factor of objective set (Einstellung), the factor of direction, the factor of closure, the factor of the "good curve", and past experience or habit.

Two things in that list should stop a designer short. The first is that past experience is on it. The popular framing treats Gestalt grouping as pre-attentive, automatic, culture-free wiring, and then treats learned convention as a separate and lesser force. Wertheimer put habit in the same list as proximity. The second is that figure and ground is not on it. Figure and ground organisation is the other half of the field, developed principally by Edgar Rubin, and it gets bundled into "the Gestalt principles" in design writing as though Wertheimer had enumerated it alongside the rest.

On Prägnanz, the paper is suggestive rather than definitional. Wertheimer writes of Prägnanzstufen, levels or degrees of Prägnanz, "with their appropriate realms or regions", and says an observer recognises a resultant "good Gestalt" by its own "inner necessity". That is a phenomenological claim, not an operational one. It tells you what the experience is like. It does not tell you how to compute which of two layouts is the better Gestalt.

The origin story is wrong in a specific way

Almost every design-side account of Gestalt begins with a train. The version on Jon Yablonski's widely used Laws of UX site, for instance, credits "psychologist Max Wertheimer's 1910 observation of flashing lights at a railroad crossing, which resembled a movie theater marquee effect". The 1910 journey is real: the record has Wertheimer travelling from Vienna towards the Rhineland when the idea took hold. The flashing-lights apparatus belongs to the Frankfurt laboratory work that followed, not to the carriage window.

The more consequential error is downstream. What Wertheimer reported in 1912, in Experimentelle Studien über das Sehen von Bewegung (Zeitschrift für Psychologie 61, pages 161 to 265), was not simply that two flashing lamps look like one moving lamp. That effect, beta movement, was already known. His claim was stranger: at certain intervals observers see movement with no moving thing in it. Steinman, Pizlo and Pizlo set this out in Vision Research in 2000 under the title "Phi is not beta, and why Wertheimer's discovery launched the Gestalt revolution". They describe phi as "movement which connects the objects and has direction between them but seems not in itself to be an object", and they show that Edwin Boring's influential 1942 history "listed the ϕ-phenomenon in the wrong place in the switching sequence", with the result that "one will not see ϕ if one looks for it where Boring suggested". Phi occurs near simultaneity, not near succession.

So the textbook animation that illustrates "the phi phenomenon" is usually illustrating beta movement, and has been for eighty years. That is not a small pedantry. It is the difference between "the brain fills in a plausible object path" and "the brain can represent motion without an object", and the second is the observation that made the Gestalt programme feel urgent.

What the centenary review concluded

The serious assessment of where all this stands is the pair of reviews Johan Wagemans assembled for Psychological Bulletin in 2012. Part I, "A century of Gestalt psychology in visual perception: I. Perceptual grouping and figure-ground organization", with Elder, Kubovy, Palmer, Peterson, Singh and von der Heydt, runs from page 1172 to page 1217 of volume 138. Part II, "Conceptual and theoretical foundations", with Feldman, Gepshtein, Kimchi, Pomerantz, van der Helm and van Leeuwen, runs from 1218 to 1252.

Part I is explicit that the strongest progress since Wertheimer has been in two areas: the discovery of new grouping principles, and "the experimental measurement of the strength of grouping factors and the development of quantitative laws". Both halves of that sentence should interest designers, because both cut against the way the principles are usually presented.

New principles, from the 1990s

Three of the principles in current design curricula are not Wertheimer's. Common region, the tendency for elements inside the same bounded area to group, was proposed by Stephen Palmer in 1992. Element connectedness, the tendency for distinct elements sharing a common border to group, comes from Palmer and Rock in 1994. Synchrony, the tendency for elements that change at the same moment to group, is later still.

Palmer and Rock's paper, "Rethinking perceptual organization: The role of uniform connectedness" in Psychonomic Bulletin & Review 1(1), pages 29 to 55, makes a claim that design writing almost never repeats. They argue that uniform connectedness "cannot be reduced to such grouping principles, because it is not a form of grouping at all". It is the process that carves the field into units in the first place, before the classical principles assemble those units into larger ones. If that is right, then drawing a card around a set of items is not the same kind of move as putting the items close together. It operates earlier.

They also report that uniform connectedness in the retinal image is necessary but not sufficient: elements seen as lying in different depth planes fail to group even when they touch. Grouping runs on perceived relations, not raw image relations. Anyone who has watched a shadow or an overlap destroy a card's apparent unity has met this.

Which principles have real quantitative support

Proximity is the success story. It is the principle that has actually been turned into a function. Part I reports Oyama's 1961 finding that grouping strength follows a power function of distance, and describes data in which "all data points... could be fitted well by a single straight line", a result usually called the pure distance law: what matters for grouping in a dot lattice is the ratio of the competing distances, not their absolute size. Michael Kubovy, Alex Holcombe and Wagemans set this out in "On the Lawfulness of Grouping by Proximity" in Cognitive Psychology 35, pages 71 to 98, in 1998. That is a principle with a curve attached to it, which is more than most of the others can say.

Common fate is the other strong one. Part I calls it "one of the most powerful of the classic grouping principles", the tendency for elements that "move together" to be perceived as a unitary entity. This is the principle that interface work under-uses, because static comps cannot show it. When four things animate on the same easing curve at the same moment, they become one thing, and no amount of spacing will undo that.

Similarity is real but underspecified. Similarity in what, and how much? The review's own complaint about the classical tradition is relevant here: Gestalt work was criticised for "formulating laws with little precision, and adding new 'laws' for every factor shown to have an influence on perceptual organization".

Prägnanz is the weak link, and design books lean on it hardest

Dejan Todorović's Scholarpedia article on Gestalt principles states the problem better than anything I have read in a design context. Collecting the principles under Prägnanz, he writes, buys "explanatory economy and unity" but does so "at the cost of clarity and operationalizability: whereas it may be relatively simple to point out the presence of continuity, closure, etc, it is more difficult to establish what exactly makes a pattern visually good, simple, unified etc."

The principles operate as tendencies, not rules. In Todorović's words, "the perceptual groupings are in some cases strong and unambiguous, but in other cases they are better described as tendencies."

Compare that with the design gloss. Laws of UX renders the Law of Prägnanz as: "People will perceive and interpret ambiguous or complex images as the simplest form possible, because it is the interpretation that requires the least cognitive effort of us." The first clause is defensible. The causal clause after "because" is an addition. The Gestaltists were not proposing a cognitive-effort account; they were proposing that the perceptual field settles into a minimum the way a physical system settles into one, without anybody doing any work. Part II of the centenary review describes the modern reconstruction as "the brain's tendency of being attracted towards states corresponding to the simplest possible organization, given the available stimulation", and then spends its length showing how much theoretical machinery it takes to make "simplest" mean anything: self-organisation in a complex adaptive system, neural resource economy, Bayesian formulations, structural information theory. Four incompatible-looking programmes, none of which reduces to effort.

I have used the effort version in a rationale. It is persuasive, it sounds like science, and it is not what the source says. That is the sort of thing worth correcting in public.

The second weakness is conflict. Todorović notes that "the significant theoretical problem of how to predict which principle will win in which circumstances remains to be worked out in much more detail." Real layouts put proximity against similarity against common region constantly. There is no settled account of how they combine. When a design book says "use the Gestalt principles", it is quietly assuming a resolution rule that the literature does not have.

A status table

The rule I use

The honest position is that Gestalt gives you a vocabulary for describing grouping effects that are real and reliably observed, plus one principle, proximity, with a measured function behind it. It does not give you a decision procedure, a conflict-resolution rule, or a justification that can carry weight on its own.

So: name the principle to describe what you observe, never to explain why you chose it. "These four items read as a set because they share a bounded region and the other groups do not" is a defensible sentence. "We used common region because of the Law of Prägnanz" is not, because the second half does nothing the first half has not already done, and it asserts a mechanism the literature has not settled.

Three practical consequences follow. First, if you are relying on proximity, state the ratio. A 24 pixel gap inside a group and a 48 pixel gap between groups is a claim you can test; "grouped by proximity" is not. Second, if you have drawn a container, accept that you have made a stronger move than spacing and have probably made the spacing decorative. Third, if you have motion available, remember that common fate outranks almost everything static, which means an animation that moves unrelated elements together will fuse them whatever your layout says.

And the unknown, stated plainly: there is no accepted model of how the grouping principles combine when they disagree. Anyone who tells you which one wins is telling you about their taste, which is fine, as long as they say so.

Sources

• Wertheimer, Max. "Laws of Organization in Perceptual Forms." Translation by Willis D. Ellis of Untersuchungen zur Lehre von der Gestalt II, Psychologische Forschung 4 (1923), 301–350, in A Source Book of Gestalt Psychology (London: Routledge & Kegan Paul, 1938), 71–88. psychclassics.yorku.ca/Wertheimer/Forms/forms.htm

• Wagemans, Johan, James H. Elder, Michael Kubovy, Stephen E. Palmer, Mary A. Peterson, Manish Singh and Rüdiger von der Heydt. "A century of Gestalt psychology in visual perception: I. Perceptual grouping and figure-ground organization." Psychological Bulletin 138, no. 6 (2012): 1172–1217. doi:10.1037/a0029333. mentalhealthce.com/courses/contentGE/GE-Gestalt-Psychology-Visual-Perception-19.pdf

• Wagemans, Johan, Jacob Feldman, Sergei Gepshtein, Ruth Kimchi, James R. Pomerantz, Peter A. van der Helm and Cees van Leeuwen. "A century of Gestalt psychology in visual perception: II. Conceptual and theoretical foundations." Psychological Bulletin 138, no. 6 (2012): 1218–1252. doi:10.1037/a0029334. scispace.com/papers/a-century-of-gestalt-psychology-in-visual-perception-ii-4pqm3sn198

• Palmer, Stephen, and Irvin Rock. "Rethinking perceptual organization: The role of uniform connectedness." Psychonomic Bulletin & Review 1, no. 1 (1994): 29–55. doi:10.3758/BF03200760. link.springer.com/article/10.3758/BF03200760

• Todorović, Dejan. "Gestalt principles." Scholarpedia 3, no. 12 (2008): 5345.

• Steinman, Robert M., Zygmunt Pizlo and Filip J. Pizlo. "Phi is not beta, and why Wertheimer's discovery launched the Gestalt revolution." Vision Research 40, no. 17 (2000): 2257–2264. sciencedirect.com/science/article/pii/S0042698900000869

• "Max Wertheimer." Wikipedia. Biographical dates, 1910 journey, and citations for the 1912 and 1923 papers. en.wikipedia.org/wiki/Max_Wertheimer

• Yablonski, Jon. "Law of Prägnanz." Laws of UX. Cited as an example of the design-side formulation. lawsofux.com/law-of-prägnanz/

Related reading

• Hangul and the Square

• Two Laws Designers Quote and Misuse

• The Caption Did the Work

• Who Are You Looking At?

More from the Moc Mallen Studio journal