Helmholtz Versus Haute Couture: How Horizontal Stripes and Dark Clothes Make You Look Thinner

Helmholtz Versus Haute Couture: How Horizontal Stripes and Dark Clothes Make You Look Thinner. Antonis Koutsoumpis, Elias Economou, Erik van der Burg. Perception, August 16, 2021. https://doi.org/10.1177/03010066211038158

Abstract: In Helmholtz’s illusion, a square with horizontal stripes appears taller than an identical square with vertical stripes. This effect has also been observed in experiments with human stimuli, where a human figure wearing a dress with horizontal stripes appears thinner than a drawing clad in vertical stripes. These findings do not agree with the common belief that clothes with horizontal stripes make someone appear wider, neither do they disentangle whether the horizontal or vertical stripes account for the thinning effect. In the present study, we focused on the effect of horizontal stripes in clothes comparing horizontal stripes against no-stripes (not against vertical; Experiments 1 and 2), using photos of a real-life female model, and controlling for the average luminance of the stripes (Experiment 2). Results showed that horizontal stripes and lower luminance have—independently—a small-to-moderate thinning effect on the perceived size of the body, and the effect is larger when the two variables are combined. In Experiment 3, we further show that the thinning effect due to the luminance of the dress is enhanced when the general background gets darker.

Keywords: size perception, horizontal stripes, luminance, Helmholtz illusion, clothes

Discussion

In the present study, we focused on the perceptual effect of horizontal stripes in clothes. Unlike most previous studies, we compared the effect of horizontal stripes against non-striped stimuli, that is, against a neutral condition. In Experiment 1, we found that participants perceived a striped dress (with light-green and black stripes) to be significantly thinner than an identical non-striped (light-green) version of the same dress, but not when compared to a non-striped black version. Since the thinning effect was significant only when the striped dress was compared to the light-green dress, we hypothesized that luminance might moderate the thinning effect. In Experiment 2, we dissociated the stripes and luminance effects by including two extra stimuli. In this way, it was possible to measure the effect of horizontal stripes and luminance on the perceived body width independently, as well as their aggregating effect. The results showed that, first, when the effect of luminance was controlled for, horizontal stripes had a thinning effect against non-striped stimuli. Second, when the effect of horizontal stripes was controlled for, stimuli with lower luminance had a thinning effect against stimuli with higher luminance, either when both stimuli were striped or non-striped. Third, when both stripes and luminance were taken together, the effect was even larger.

Although it seems that the stripes effect holds when the luminance of the stimuli and the background are held constant, the luminance of the background—against which the stimuli were placed—might have affected the thinning effect of darker stimuli. This is because, in the luminance effect, the darker (vs. lighter) stimuli were further away from the luminance of the background and this luminance difference might have affected size perception. In Experiment 3, we manipulated the background luminance (i.e., creating light, average, and dark backgrounds) to investigate whether the luminance effect of Experiment 2 was due to the increased contrast between the dark dresses and the light-grey background. The results showed that the thinning effect of dark dresses could not be attributed to stimuli-background contrast. Instead, we discovered that when the luminance of the general background is darker, the thinning effect becomes more pronounced.

The thinning effect of low luminance is in line with the irradiation illusion (Helmholtz, 1962), according to which a bright area seems larger than a darker area of similar size (Long & Murtagh, 1984; Oyama & Nanri, 1960), an effect attributable to the non-linear visual processing of dark and light both on retinal (Valeton & van Norren, 1983; Westheimer, 2008) and cortical level (Kremkow et al., 2014). On the cortical level for instance, Kremkow et al. (2014) showed that the ON and OFF neurons operate in dissimilar ways in our visual system, with the OFF neurons responding linearly to external stimuli, and the ON neurons responding in a curvilinear way (overresponding even to limited external stimulation). The practical implication is that dark stimuli are perceived as smaller than bright, not because the size of dark stimuli is underestimated, but rather because the size of bright stimuli is overestimated. As a consequence, the darker dresses in Experiments 2 and 3 were perceived as thinner presumably because participants might have overestimated the size of the brighter dresses.

So far only one study has tested the perceptual effect of horizontal stripes against non-striped stimuli. Thompson and Mikellidou (2011; Experiment 3) found that a cylinder with horizontal stripes was perceived as similar in size (neither wider nor thinner) to an identical non-striped cylinder. This finding is at odds with the results of the present experiments. Nevertheless, the present results suggest that the thinning effect of horizontal stripes seems to be consistent and strong (in terms of effect size and observed statistical power). They also suggest that the thinning effect is further decomposed into two separate effects, that is, both the orientation and the luminance of the stripes affected the perceived size of the dress, and the thinning effect of luminance was further amplified as the luminance of the background decreased.

Our results also extend and partially explain the Helmholtz illusion. In the original paradigm, a square with horizontal stripes is compared to a similar square with vertical stripes. In that context, the thinning effect of horizontal stripes cannot be attributed to luminance, since both stimuli have the same luminance regardless of stripe orientation. It can be attributed, however, to the thinning effect of horizontal stripes, as Experiment 2 showed. Whether vertical stripes also contribute to the overall effect or not, remains an open empirical question.

Future Research

The present results open two main venues for future research. The first is to further probe the boundary conditions that might account for the effect. In the present study we tested the effect of horizontal stripes, luminance, and contrast but the list of confounding variables may be longer. Some variables that might account for the effect include the orientation of the stripes (e.g., test the effect of vertical stripes against a neutral condition), their spatial frequency (the thickness of the stripes), or color. Moreover, previous studies reported that the body size moderates the effect, with bigger body types decreasing (Ashida et al., 2013) or even reversing the thinning effect of horizontal stripes (Imai, 1982). Future studies can test the effect in different body sizes and body shapes (e.g., hourglass, spoon, rectangle).

The second line of research calls for an interdisciplinary approach between the fields of social cognition (participants’ previous beliefs) and psychophysics (participants’ perception). It is striking that, although the results of the present experiments support the thinning effect of horizontal stripes, this finding clashes with participants’ beliefs and leads to a paradox: during the behavioral task in the lab participants perceived horizontal stripes in clothes as thinner, but in self-reports they stated the opposite.³ This inconsistency was also described by Ridgway et al. (2017) where they scanned the body of 15 women and their 3D avatars were presented on a computer. The avatars were dressed in horizontal stripes (and seven more optical illusions) and participants were asked to evaluate them. A thematic analysis (a qualitative in-depth interview method) revealed that a number of participants swayed away from their original negative attitude towards horizontal stripes once they saw themselves in the horizontally striped dress. Although the results of that study were based on qualitative analysis and did not provide any conclusive evidence, the discrepancy between previous beliefs and behavioral responses regarding the effect of horizontal stripes in clothes still remains.