Processing fluency

Evidence

Reber, Schwarz and Winkielman (2004) propose that aesthetic pleasure is a function of processing fluency: what is processed more easily triggers a positive affective reaction, and that reaction feeds into the judgment. They trace figure–ground contrast, symmetry, repetition, prototypicality and priming back to this single mechanism. The paper is a theoretical work with no new data of its own; it organizes other people’s experiments.

Its sharpest finding separates fluency from the stimulus feature itself. Reber and Schwarz (2001) showed circles of high or low figure–ground contrast for 0.3, 1, 3 and 10 seconds. Contrast influenced the aesthetic judgment only at the short durations — at 10 seconds it no longer did. If contrast itself were the beautiful thing, it should work all the more under long viewing. That it does not is the authors’ argument: what matters is the moment while processing is still work.

Two findings support the claim that the judgment runs through a feeling rather than through fluency itself. Winkielman and Cacioppo (2001) measured, by facial electromyography, stronger activity of the smiling muscle in the first three seconds after the stimulus, seconds before any judgment was given. And when participants attributed their affective reaction to background music, the effect on liking disappeared; when they attributed only the ease to the music, it remained.

Two things belong to the evidential picture. Several of the studies carrying the mechanism — including that very misattribution study — are listed in the paper as unpublished manuscripts. And the stimuli are dot patterns, circles, faces, pictures of animals and chord progressions. Prices, numbers and purchase decisions do not appear in the paper at all.

For one of the five mechanisms there is a large meta-analysis — and it is unambiguous. Bornstein (1989) summarised the research on the mere repetition effect from 1968 to 1987: 134 papers, 208 independent effect sizes. Across all of them the relationship between frequency and liking sits at r = .26 — the r measures how closely two quantities go together and runs from 0 to 1. The fail-safe N is 33,047 — a good thirty-three thousand unpublished null results would have to be sitting in drawers to overturn the finding. The direction is not in doubt.

What matters is what this paper does not measure. Bornstein records exposure frequency and liking, not processing fluency. That the one explains the other is Reber and colleagues’ interpretation, not Bornstein’s finding. His collection does contain individual studies on advertising, but he does not analyse them as a stimulus class of their own. Prices do not appear in his work either.

One moderator supports the mechanism directly. The shorter the exposure, the stronger the effect — the same curve as in Reber and Schwarz (2001), but now across 139 effect sizes rather than one.

The same holds for the number of repetitions: the relationship peaks early and falls away after that, from r = .21 at one to nine exposures to .06 at a hundred and more. Bornstein’s explanation is a two-factor model: habituation creates liking, satiation takes it away again.

The best-known offshoot of this theory has been independently tested, and it collapsed. In 2007 Alter and colleagues reported the mirror image of fluency: if ease makes people think superficially, a hard-to-read font should make them think carefully — and indeed their participants solved more reasoning problems. The finding appears in over 133 academic articles and in at least three bestselling books, among them Kahneman’s Thinking, Fast and Slow.

Meyer and colleagues (2015) pooled the original study with sixteen repetitions: 7,367 participants. The effect disappears entirely — 1.43 against 1.42 problems solved, p = .75. Of seventeen experiments, only the original found a significant difference. Sirota and colleagues (2021) added two pre-registered experiments; their updated meta-analysis across eighteen studies yields Hedges’ g = −.01 and a Bayes factor of 151.6 against the effect. That is: the data argue roughly 150 times more clearly that the effect does not exist than that it does.

What the original hung on is worth noting. Of its three problems only one showed a difference: 16 out of 20 against 4 out of 20. On the other two it was zero (16 to 15 and 18 to 18). The 20 percent solution rate in the control condition sits well below the 50 percent of a sample of 300 Princeton students. Meyer and colleagues do the arithmetic: had the control group performed like the other Princeton students, the original’s p value would have been .36 — and none of the sixteen repetitions would ever have been run.

Sirota’s control closes the last gap. One could object that the test does not measure careful thinking finely enough. So in a second, pre-registered experiment Sirota and colleagues simply asked participants to pay attention — and the solution rate rose substantially (d = .64, Bayes factor 575,000). The instrument does respond to more careful thinking. The hard-to-read font just does not trigger it.

What this means — and what it does not. These two papers test the reverse path, difficulty to deliberation, not the one described here, ease to liking. They do not refute the aesthetic finding. But they do show what happened to the most prominent prediction of the same framework once seventeen teams reproduced it independently. Anyone invoking processing fluency to justify a design decision should know that.

Limits

Simple does not automatically mean easier. The authors explicitly warn against inferring processing fluency from the simplicity of a stimulus. More complex shapes often carry more redundancy and are recognised faster than simple ones; a letter inside a word is recognised more easily than the same letter alone. Their conclusion: the actual ease of processing has to be examined empirically every time. For the example above that means: that €2,500 is easier to process than €2,499.99 is plausible, but it is not tested in this paper.

Where the source of the ease is obvious, the ease evaporates. On the discrepancy–attribution hypothesis the authors review, the feeling arises mainly when the ease comes as a surprise. An obvious repetition scheme or a very simple pattern creates a matching expectation — and so no feeling. In one reviewed study, liking even fell after 27 repetitions once the repetition became salient.

Practised viewers respond differently. Without art training people prefer simple and symmetrical forms, with training complex and asymmetrical ones; music novices like prototypical chord progressions, experts do not. Anyone comparing prices daily is, on this logic, closer to the expert than the novice.

Ease does not only make things more likeable, it makes them more believable. The same authors found that identical statements are more often judged true at high figure–ground contrast, and report that rhyming statements come across as truer than non-rhyming ones. For price communication that is the more delicate side of the mechanism: an easily read statement is not only liked more, it is believed more readily.

The stimulus class closest to a piece of design shows no effect in Bornstein. Across 38 effect sizes for paintings, drawings and matrices the relationship sits at r = −.03 — not merely small but slightly negative, with a fail-safe N of 3. Repetition works most strongly for meaningful words (r = .49) and photographs (.37); for real people and objects it sits at .20. A price tag is not a dot pattern, but it is not a drawing either.

Uniform repetition does not work at all. Where the same stimuli were shown in a block one after another, the relationship sat at r = −.02 and was not significant. Only mixed presentation, where a stimulus appears among others, produces the effect (r = .30). Bornstein cites this explicitly as the reason television advertising works: brief, varied, placed among other things. A price display that recurs identically on every page of a shop is the uniform case.

The more consciously the stimulus is perceived, the weaker the effect. Stimuli below the threshold of awareness produced a markedly stronger effect at r = .53 than briefly shown but recognised stimuli (.34). Bornstein’s conclusion is an awkward one: awareness of the stimulus inhibits the effect. A price that someone reads and compares is precisely the consciously perceived case.

That round prices work better because of processing fluency remains an interpretation. The field finding on round prices (Wieseke et al., 2016) speaks of perceived convenience, not of processing fluency, and rests on a single sales measurement of 306 drink purchases in three university cafeterias. A meta-analysis does exist, but it concerns repetition and liking, not prices. On the effect in pricing contexts there is still none.

The evidence falls into two halves. The repetition effect, one of the five mechanisms named above, is confirmed across 208 effect sizes. The disfluency offshoot of the same theory has failed across eighteen studies. One branch is therefore supported and the other refuted — and neither has been tested on prices.