A new illusion of height and width: taller people are perceived as thinner
It is commonly said that tall people look thinner. Here, we asked whether an illusion exists such that the taller of two equally wide stimuli looks thinner, and conversely whether the thinner of two equally tall stimuli looks taller. In five experiments, participants judged the horizontal or vertical extents of two identical bodies, rectangles, or cylinders that differed only in their vertical or horizontal extents. Our results confirmed the folk wisdom that being tall makes you look thinner. We similarly found that being thin makes you look taller, although this effect was less pronounced. The same illusion was present for filled rectangles and cylinders, but it was consistently stronger for both photographs and silhouettes of the human body, raising the question of why the human form should be more prone to this illusion.
KeywordsVisual illusion Body image Integral dimensions Body shape
Being tall and thin is valued in the fashion world, and a number of fashion trends have purported to perceptually elongate the wearer. The fact that both dimensions (as opposed to just one) are valued suggests that the two may be related. Indeed, it is commonly said that tall people look thinner. Here we tested whether judgments about a person’s width are influenced by his or her height. Specifically, we asked participants to make judgments about the width (horizontal extent) of two bodies that also differed in their heights. If an illusion exists in which tall people are seen as being thinner than their shorter counterparts, then participants should tend to choose the taller body as thinner more often than the shorter one, despite the stimuli being identical in width. Although it is less commonly said that thin people look tall, in Experiment 1B we reversed the manipulation and asked whether judgments of height are influenced by a body’s width.
In addition to the folk wisdom that tall people look thinner, there is another reason to suspect that such an illusion might exist. Although it was apparently overlooked by most English-speaking researchers, both Lipps (1897) and Müller-Lyer (1889) reported a similar illusion with rectangles. For example, Lipps reported that as the width of a rectangle is shortened, its perceived height elongates, and similarly, as the height is shortened, the perceived width elongates (as cited in Vicario, 2011). Thus, in Experiments 2A and 2B, we extended our paradigm to simple rectangles to verify that the same illusion exists for rectangles. In Experiment 3, we asked whether the illusion extends to both silhouettes of the human body and shaded cylinders.
Experiments 1A and 1B: Human bodies
Participants were asked to compare the horizontal and vertical extents of two human bodies in Experiments 1A and 1B, respectively. On two-thirds of the trials, the stimuli actually differed on the judged dimension, but on a critical one-third of the trials, the judged dimension was identical, and the two images differed only in their extents on the orthogonal dimension. The presence of an illusion was assessed on the critical trials by a systematic preference (deviation from 50 %) for one image over the other, indicating that the nonjudged dimension influenced participants’ judgments.
A group of 20 right-handed healthy participants (ten females, ten males; five apiece in each experiment) with normal or corrected-to-normal visual acuity participated in Experiments 1A (n = 10) and 1B (n = 10). Their ages ranged between 18 and 36 years (mean 24.75 years). Participants were naïve to the purpose of the experiments, which were carried out according to the principles laid down in the 1964 Declaration of Helsinki, and the participants gave their informed consent prior to taking part.
Stimuli and procedure
Results and discussion
Results for critical trials in Experiments 1A and 1B are shown in Fig. 1A and B, respectively. In Experiment 1A, participants judged the taller body to be narrower, or the shorter body to be wider, on 82.66 % of the trials. A one-sample t test showed that this perceptual bias was significantly different [t(9) = 9.596, p < .001, ηp2 = .91] from chance (50 %). Although this effect represents the combined contributions of participants choosing the narrower or wider body, the perceptual biases computed separately for the narrower (81.41 %) and wider (83.91 %) judgments were both significantly different from chance [t(9) = 10.712, p < .001, ηp2 = .93, and t(9) = 8.403, p < .001, ηp2 = .89, respectively], indicating that the illusion was present regardless of the direction of the horizontal judgment. The mean accuracy on trials that actually differed in width was 95.01 %. Moreover, participants chose the physically wider (or narrower) stimulus significantly more often than the illusory wider (or narrower) stimulus for both the 1-deg [t(9) = 2.971, p < .05, ηp2 = .50] and 2-deg [t(9) = 3.907, p < .01, ηp2 = .63] differences, indicating that the illusion resulted in a perceptual difference of less than a degree.
In Experiment 1B, participants judged the narrower body to be taller, or the wider body to be shorter, on 70.93 % of trials. This perceptual bias was again reliably different [t(9) = 5.771, p < .001, ηp2 = .79] from chance (50 %) , as were the perceptual biases computed separately for the taller [70.45 %; t(9) = 5.208, p < .001, ηp2 = .75] and shorter [71.41 %; t(9) = 5.236, p < .001, ηp2 = .75] judgments, indicating that the illusion was present regardless of the direction of the vertical judgment. The mean accuracy on trials that actually differed in width was 95.17 %. Again, participants chose the physically taller (or shorter) stimulus significantly more often than the illusory taller (or shorter) stimulus for both the 1-deg [t(9) = 6.807, p < .001, ηp2 = .84] and 2-deg [t(9) = 6.772, p < .001, ηp2 = .84] differences, indicating again that the perceptual illusion was smaller than a degree.
In short, Experiments 1A and 1B both indicated that a body’s height or width can influence judgments about the other dimension. In particular, we saw that taller people are perceived as thinner, and thinner people are perceived as taller.
Experiments 2A and 2B: Filled rectangles
The illusion observed in Experiment 1 is similar to one reported by Lipps (1897) and Müller-Lyer (1889) for filled rectangles. In Experiments 2A and 2B, we asked whether the same illusion observed with human bodies occurred when participants were asked to compare the horizontal and vertical extent of two filled rectangles.
A group of 20 right-handed healthy participants (ten females, ten males; five apiece in each experiment) with normal or corrected-to-normal visual acuity participated in Experiments 2A (n = 10) and 2B (n = 10). Their ages ranged from 18 to 33 years (mean 25 years). Participants were naïve to the purpose of the experiments, which were carried out according to the principles laid down in the 1964 Declaration of Helsinki, and the participants gave their informed consent prior to taking part. None of the participants run here had taken part in Experiments 1A and 1B.
Results and discussion
The results for the critical trials in Experiments 2A and 2B are shown in Fig. 2A and B, respectively. In Experiment 2A, the participants judged the taller rectangle to be narrower, or the shorter rectangle to be wider, on 66.25 % of trials. This perceptual bias was reliably different [t(9) = 7.736, p < .001, ηp2 = .87] from chance (50 %), as was the perceptual bias computed separately for the narrower [64.09 %; t(9) = 4.153, p < .01, ηp2 = .66] and wider [68.42 %; t(9) = 8.072, p < .001, ηp2 = .88] judgments. The mean accuracy on trials that actually differed in width was 98.87 %. Participants chose the illusory wider (or narrower) rectangle less often than they did for either the 1-deg [t(9) = 13.839, p < .001, ηp2 = .96] or 2-deg [t(9) = 13.437, p < .001, ηp2 = .95] physically wider (or narrower) rectangle, indicating again that the perceptual illusion was smaller than a degree.
In Experiment 2B, participants judged the narrower rectangle to be taller, or the wider rectangle to be shorter, on 58.06 % of trials. This perceptual bias was reliably different [t(9) = 11.838, p < .001, ηp2 = .94] from chance (50 %), as were the perceptual biases computed separately for the taller [61.25 %; t(9) = 7.754, p < .001, ηp2 = .87] and shorter [56.88 %; t(9) = 3.869, p < .01, ηp2 = .62] judgments. The mean accuracy on trials that actually differed in height was 95.04 %. Finally, once again participants chose the illusory taller (or shorter) rectangle less often than they did for either the 1-deg [t(9) = 13.839, p < .001, ηp2 = .96] or 2-deg [t(9) = 13.437, p < .001, ηp2 = .95] physically taller (or shorter) rectangle, indicating again that the perceptual illusion was smaller than a degree.
Interestingly, the illusion that we observed for human bodies appears to be much greater than that for rectangles. To assess the relative sizes of the illusion in the four experiments, the data from all four experiments were submitted to a two-way ANOVA with Type of Stimulus (body, rectangle) and Judgment (horizontal, vertical) as between-subjects factors. We found a significant main effect of type of stimulus [F(1, 36) = 26.873, p < .001, ηp2 = .43], with bodies (76.80 %) yielding a stronger perceptual bias than did rectangles (62.66 %), and a significant main effect of judgment [F(1, 36) = 12.038, p < .001, ηp2 = .25], with horizontal judgments (74.46 %) yielding a stronger perceptual bias than did vertical judgments (64.99 %). No significant interaction was found between these two factors [F(1, 36) = 0.693, p = .411, ηp2 = .019]. The fact that the illusion is considerably stronger for human bodies than for rectangles raises some interesting questions. Is the illusion stronger for bodies than rectangles for perceptual reasons, or might it reflect more social influences? In Experiment 3, we asked whether a perceptual difference between the stimuli might account for the illusion. Specifically, we asked whether the implied depth and internal detail of the body as compared to a rectangle might contribute to the illusion.
Experiment 3: Silhouettes and cylinders
A group of 20 right-handed healthy participants with normal or corrected-to-normal visual acuity participated in experiment; ten viewed the silhouettes (five females, five males) and ten viewed the cylinders (five females, five males). Their ages ranged from 19 to 31 years (mean 24 years). Participants were naïve to the purpose of the experiments, which were carried out according to the principles laid down in the 1964 Declaration of Helsinki, and the participants gave their informed consent prior to taking part. None of the participants run here had taken part in Experiment 1 or 2.
Stimuli and procedure
The stimuli were black silhouettes (Fig. 3A) of the male and female bodies used in Experiment 1A, and shaded cylinders (Fig. 3B) with dimensions identical to those of the rectangles in Experiment 2A. All other aspects of the experiment were identical to those of Experiments 1A and 2A, including the task to choose the narrower or wider of the stimuli.
Results and discussion
As in Experiment 1A, participants judged the taller silhouette to be narrower, or the shorter silhouette to be wider, on 80.23 % of trials. This perceptual bias was reliably different [t(9) = 13.875, p < .001, ηp2 = .96] from chance (50 %), as was the perceptual bias computed separately for the narrower [84.22 %; t(9) = 16.802, p < .001, ηp2 = .97] and wider [76.25 %; t(9) = 8.428, p < .001, ηp2 = .89] judgments. As in Experiment 2A, participants judged the taller cylinder to be narrower, or the shorter cylinder to be wider, on 63.83 % of trials. This perceptual bias was reliably different [t(9) = 6.140, p < .001, ηp2 = .81] from chance (50 %), as was the perceptual bias computed separately for the narrower [66.09 %; t(9) = 4.986, p < .001, ηp2 = .73] and wider [61.56 %; t(9) = 6.294, p < .001, ηp2 = .81] judgments. Thus, the patterns of results for the silhouettes and cylinders did not differ from those for the bodies [t(18) = 0.599, p = .556, ηp2 = .02] and rectangles [t(18) = 0.876, p = .393, ηp2 = .04], respectively. Finally, a between-subjects t test revealed that, like the bodies and rectangles, the silhouettes produced a significantly greater illusion than did the cylinders [t(18) = 5.23, p < .001, ηp2 = .60].
In short, the pattern of results did not reverse, as would be predicted if implied depth and internal detail were responsible for the illusion. Instead, the body-like stimuli (silhouettes) produced the same illusion as had the color photographs of bodies, and the nonbody stimuli (cylinders) produced the same illusion as had the flat filled rectangles.
Our data suggest that taller people are indeed perceived as thinner. Moreover, the reverse is also true, that thinner people are perceived as taller, although this illusion is not as great as the former one. The experiments with filled rectangles suggest that the illusion is not specific to bodies, but interestingly, this effect cannot fully account for the body illusion, as the illusion was significantly greater for bodies than for filled rectangles. The fact that the illusion was greater for bodies than for rectangles does not stem from the fact that bodies have greater implied depth or more internal details, as silhouettes produced the same illusion magnitude as had the photographs of bodies, and a greater illusion magnitude than the three-dimensional cylinders (comparable to the rectangles).
The tendency for judgments about width and height to be influenced by the orthogonal dimension is in accordance with work suggesting that height and width are integral dimensions (Dykes & Cooper, 1978; Felfoldy, 1974). Integral dimensions are defined as dimensions for which, when only one dimension is to be sorted (e.g., height), the other dimension (e.g., width) produces redundancy gains (faster sorting) when it is correlated with the judged dimension, and orthogonal interference (slower sorting) when it is varied orthogonally with the judged dimension (Garner, 1970; Garner & Felfoldy, 1970). Although these data suggest that it is difficult to process height and width separately, they do not, as our data indicate, suggest that the two dimensions result in an illusory perception of either the height or width. In other words, our data suggest that height and width are not only integral, but that variations in one can produce illusory perception of the other. Moreover, it would seem that whatever the mechanism is that links width and height in this particular direction, it is stronger for body-like stimuli than for rectangles or cylinders.
The illusion reported here does have some similarity with two other geometric illusions reported in the literature: the horizontal–vertical and Shepard illusions. We will discuss the relationship between our illusion and each of these illusions in turn. The horizontal–vertical illusion refers to the fact that a vertical line is perceived as being longer than an equivalent horizontal line (Fick, 1851), which is often illustrated as an inverted capital letter “T.” Although the horizontal–vertical illusion is also a misperception of height and width, it differs in a number of ways from our own. The horizontal–vertical illusion occurs when the observer is to judge the relative lengths of the horizontal and vertical segments of the object, whereas our illusion occurs when judging the horizontal (or vertical) aspects of two objects. In other words, the horizontal–vertical illusion requires the comparison of the vertical and horizontal dimensions, whereas our task explicitly focuses on just a single dimension, although it appears that participants are unable to discount the orthogonal dimension. Second, in the horizontal–vertical illusion, the two line segments are identical in length, whereas in our illusion the horizontal and vertical extents of the figure are very clearly different. In fact, one way to characterize our illusion is to say that the more different the vertical extent is from the horizontal (i.e., the taller that the object is), the more likely the horizontal extent is to be underestimated. This characterization raises the possibility that the two illusions draw on some of the same mechanisms. It could be, for instance, that rather than horizontal extents being generally underestimated relative to the vertical (or vertical being overestimated relative to horizontal), the horizontal–vertical illusion actually occurs due to an integration of the two extents that then, as in our illusion, serves to pull perceptions of the sizes of the two extents in opposite directions.
The Shepard illusion is apparent when comparing two three-dimensional parallelepipeds, in which the top surfaces (parallelograms) are identical but oriented differently; for one parallelepiped, the major axis of the parallelogram is oriented vertically, and for the other it is oriented horizontally. The resulting illusion is such that the vertically oriented parallelogram looks elongated relative to the horizontally oriented one. Thus, like our illusion, the Shepard illusion concerns height and width. Unlike our illusion, however, the Shepard illusion is thought to result from the fact that parallelepipeds are oriented in depth (Shepard, 1981). The vertically oriented parallelepiped appears to recede farther into the page (i.e., orthogonal to the line of sight) than does the horizontally oriented one. Due to projective geometry (from three dimensions to our two-dimensional retina), objects that extend in depth orthogonal to the line of sight are perceptually foreshortened. Shepard argued that our visual system compensates for this foreshortening and concludes that if the vertically oriented parallelogram subtends the same visual angle on the retina as the horizontally oriented one, then the vertically oriented parallelogram must in fact be elongated with respect to the horizontal one. In keeping with this explanation, the illusion is greater when the parallelograms are made to look like three-dimensional tables by adding table legs (Mitchell, Ropar, Ackroyd & Rajendran, 2005). Our stimuli, however, cannot be said to be oriented differentially in depth. They are identical in all respects except for a small difference in height or width. Thus, it cannot be said that one should suffer from more foreshortening than the other.
It is possible, however, that this height/width illusion may also stem from our visual system’s compensating for a perceived difference in depth. Indeed, one of the more well-known explanations of a number of size illusions, including the Müller-Lyer (arrowhead) illusion and the Ponzo illusion, was put forth by Richard Gregory (Gregory, 1963). He argued that many geometric illusions stem from an erroneous application of size constancy on the basis of depth cues (although see Howe & Purves, 2005, for an alternative explanation). Our height/width illusion may, like the Shepard (described above) and the Müller-Lyer (arrowhead) illusions, stem from the visual system’s compensating for a perceived difference in depth. For example, in the case of two bodies (or rectangles) that differ only in their height, one interpretation of two images that are identical in all respects except height is that the shorter of the two is farther away. If the shorter of the two is farther away but still subtends the same visual angle horizontally, then (in keeping with size constancy) that object must in fact be wider, and this is what participants report. Such a mechanism may also explain why the illusion is greater for human bodies than for rectangles. The tendency to perceive the taller of two identical images as being farther away may be greater for human bodies than it is for filled rectangles, as it is more likely that two filled rectangles may be different instances of a shape (with different aspect ratios) than it is that two identical bodies (differing only in height) are two different people rather than the same person at varying depths. However, there are a couple of reasons to doubt this explanation. First, our effect does not interact with height in the picture plane; that is, the illusion is equivalent whether the shorter image is higher in the picture plane (consistent with its being seen as farther away) or lower in the picture plane [t(9) < 1, for both bodies and rectangles, ηp2 = .071], which should serve to reduce the perception of distance, and thus the illusion. Second, adding depth to the cylinders did not increase the illusion relative to rectangles, nor did making the bodies look more flat reduce the illusion relative to the photographs. Perceived distance in depth may still play a role in this illusion, but if it does it does not seem to be related in a simple way to known monocular cues to depth, such as height in the picture plane or implied depth from shading or other details.
Another possibility is that the illusion of height and width that we have demonstrated is actually a variant of a size contrast illusion, of which the Ebbinghaus illusion (Titchener circles) is probably the most famous (although see Roberts, Harris, & Yates, 2005, for an alternative explanation). Here, the width and height of the stimulus serve as contrasting elements, in the same way that the large and small circles do in the Ebbinghaus illusion, with the smaller width, for example, inducing the height to look larger by contrast. However, unlike the Ebbinghaus illusion, which requires sensitivity to global context (de Fockert, Davidoff, Fagot, Parron, & Goldstein, 2007; Doherty, Campbell, Tsuji, & Phillips, 2010), in this case the contrasting “elements” are integral dimensions of the same object.
Although the mechanisms underlying our illusion are currently unclear, the fact that it is considerably stronger for human bodies than for rectangles or cylinders may prove important. Further research will be needed to determine whether some other perceptual aspect of the human form makes it more susceptible to the illusion. There is another possibility, however, which is that our social experience with the human body may play a role. For example, perhaps the illusion taps into a culturally valued tall and slim silhouette. If so, we might expect exposure to such a cultural value to modulate the illusion. We might also ask whether individuals with body-image problems are more susceptible to this illusion. Regardless of the source of the illusion, the data suggest that a person’s width and height may be more interconnected and less veridical that we may have realized. However, further research will be necessary to determine whether this illusion also occurs for three-dimensional bodies, as opposed to photographs of three-dimensional bodies.
This study was supported in part by a Fulbright Research Scholar Grant to S.S. We thank Rossana Actis-Grosso and Giovanni Bruno Vicario for alerting us to the Lipps and Müller-Lyer versions of the illusion.