Fear of Holes

Abstract

Phobias are usually described as irrational and persistent fears of certain objects or situations, and causes of such fears are difficult to identify. We describe an unusual but common phobia (trypophobia), hitherto unreported in the scientific literature, in which sufferers are averse to images of holes. We performed a spectral analysis on a variety of images that induce trypophobia and found that the stimuli had a spectral composition typically associated with uncomfortable visual images, namely, high-contrast energy at midrange spatial frequencies. Critically, we found that a range of potentially dangerous animals also possess this spectral characteristic. We argue that although sufferers are not conscious of the association, the phobia arises in part because the inducing stimuli share basic visual characteristics with dangerous organisms, characteristics that are low level and easily computed, and therefore facilitate a rapid nonconscious response.

Keywords

phobia trypophobia holes evolution visual stress aposematism threat fear evolutionary psychology vision avoidance

According to the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2000), a phobia can be defined as a marked and persistent fear of a specific object or situation that invariably provokes anxiety. Furthermore, the individual may recognize that the fear is excessive and unreasonable. Although setting out criteria for such aversions is relatively easy, identifying the etiology of the fear is difficult. Indeed, isolating cause has been one of the main challenges of phobia research. For instance, there have been accounts of phobia based on evolutionary principles (Marks & Nesse, 1994), classical conditioning (e.g., Merckelbach & Muris, 1997), and the role of thoughts and beliefs about objects and situations (e.g., Hertel & Brozovich, 2010).

Such theories of phobia acquisition can have difficulty explaining many phobias. A case in point, and central to this article, is trypophobia—the fear of holes. Sufferers report aversion to visual stimuli comprising particular configurations of holes. The stimuli are usually clusters of holes of any variety that are almost always innocuous and seemingly pose no threat. Although no peer-reviewed articles currently exist concerning the phenomenon, its documentation on the Internet surpasses that of several more widely recognized phobias, and there are a number of Internet-based support groups, including a Facebook site (https://www.facebook.com/groups/3318322299/), where people provide testimonials. It is clear from these accounts that for many people, trypophobia affects their everyday lives and can be quite debilitating. For instance, one sufferer reported, “[I] can’t really face small, irregularly or asymmetrically placed holes, they make me like, throw up in my mouth, cry a little bit, and shake all over, deeply.” The image most often reported as inducing the phobia is the seed head of the lotus flower (see Fig. 1). Other examples include soap bubbles and the holes in aerated chocolate. To obtain an initial estimate of how common trypophobia is, we asked 286 adults (91 male and 195 female; age range = approximately 18–55 years) to view an image of the lotus seed head. The participants indicated whether the image was “uncomfortable or even repulsive to look at.” Ten males (11%) and 36 females (18%) reported aversion.

Fig. 1.

Lotus seed head. Images of lotus seed heads are often reported as inducing trypophobia.

Sufferers of trypophobia report that it is the visual percept that is particularly aversive. This aversion can be contrasted with, for instance, an aversion to cats, in which a person with ailurophobia will be uncomfortable in the presence of a cat even if it is not visible. Furthermore, trypophobia seems to increase if the holes occur on human skin. It is only in this respect that the phobia involves any reference to the semantics of the image. The visual nature of trypophobia provides a clue as to its cause.

For many years, researchers have been aware of aversion and discomfort caused by the viewing of certain geometric patterns (Wilkins et al., 1984). Motivated by sporadic media reports during the past four decades of public artworks inducing migraines, Fernandez and Wilkins (2008) examined the spectral characteristics of images that induce aversion. Any visual image can be analyzed with respect to its fundamental visual properties. For instance, chromatic and brightness (luminance) contrasts can be computed at any point in a scene. A major property of the visual world is luminance contrast, which can be derived at various spatial scales. An image can be constructed from Fourier components consisting of luminance varying sinusoidally at different spatial frequencies, phases, contrasts, and orientations. One of the fundamental properties of a visual scene is the relationship between luminance contrast and spatial frequency. In scenes from nature, the spatial frequency and contrast of the components are related such that contrast increases as spatial frequency decreases. When log contrast energy is plotted against log spatial frequency, a straight line with a slope close to −1 is typically found (Field & Brady, 1997). The image is then scale invariant: The complexity of the scene is independent of spatial scale. In other words, the natural visual world has a characteristic visual property revealed with a spectral analysis.

However, this particular property is not found in images that are uncomfortable to look at. Fernandez and Wilkins (2008) asked participants to rate discomfort in response to a wide variety of images, including paintings, photographs, and meaningless images created from random noise. Images rated as being particularly uncomfortable to look at possessed Fourier spectra with an excess of contrast energy at midrange spatial frequencies relative to that expected elsewhere in the spectrum. Thus, uncomfortable images do not possess the characteristic visual property in which contrast amplitude decreases linearly with increasing spatial frequency; rather, they tend to have relatively large contrast at midrange spatial frequencies. The discomfort depends on amplitude rather than phase. These findings have since been confirmed by O’Hare and Hibbard (2011).

Given the knowledge that images associated with aversion have a characteristic spectral composition, we examined whether trypophobia arises partly because the inducing images possess this unusual feature, that is, relative excess of contrast energy at midrange spatial frequencies. We performed a spectral analysis on a range of trypophobia-inducing images and compared them with control images of holes that do not induce trypophobia.

Experiment 1: Analysis of Trypophobic Images

A total of 76 images were obtained from the trypophobia Web site (www.trypophobia.com), including images of a lotus seed head and a wide range of other images of clusters of holes. We took the first 76 images presented without prejudice; none were of the skin-lesion type. A Google search for “images of holes” provided a set of 76 control images of holes that were not exhibited on the trypophobia Web site as associated with trypophobia. Using MATLAB, we cropped the images to give the largest central square image, resized them to 512 × 512 pixels (using the nearest-neighbor algorithm), rendered them in gray level with the rgb2gray function (0–255), and normalized them so that the mean gray level was 125 (SD = 25). We applied a Hanning window that reduced the contrast to 0 at the periphery to remove edge effects. The fast-Fourier-transform algorithm provided an amplitude spectrum in two dimensions, and this matrix was sampled using a set of annular masks that summed the energy over all orientations. The internal and external dimensions of the annuli were such as to sum the energy in bins of equal size on a logarithmic scale of spatial frequency, with each bin differing from its neighbor by a factor of a square root of 2. The four lowest spatial-frequency bins were removed from analysis owing to the effects of the Hanning window.

Figure 2 shows the power spectrum of the control images and those images obtained from the trypophobia Web site. Overall, a power function (linear on log-log axes) accounted for more than 97% of the variance, a good fit to the prediction for natural images. The percentage variance explained by the linear fit to the average spectra for the trypophobic images of holes (95.7%) was significantly less than for the nontrypophobic images of holes (97.9%), t(121) = 3.31, p < .002.

Fig. 2.

Power spectra (Fourier power as a function of spatial frequency) of trypophobic (broken line) and control (solid line) images of holes analyzed in Experiment 1.

These findings are consistent with a greater energy at midrange and high spatial frequencies in the trypophobic images. In Figure 2, the spatial frequency has been expressed in cycles per image (cpi). Using a Bonferroni correction for 12 paired comparisons of a probability value of .0043, we found the difference in power between the two functions to be significant for spatial frequencies in the range of 45 to 181 cpi. Most photographic images subtend 10° to 30° of visual angle, so from the viewpoint of the camera, the excess in contrast energy ranged from a minimum of 45 cpi divided by 30° per image (i.e., 1.5 cycles per degree, or cpd) to a maximum of 181/10 (i.e., 18) cpd. Objects are usually photographed so that they fit most of the frame; consequently, small objects may be photographed from distances less than those from which they are typically viewed, and the reverse holds for large objects. Fernandez and Wilkins (2008) showed that the range of spatial frequencies for which an excess energy can be expected in uncomfortable images is from 1 to 8 cpd—a range of a factor of 8. Given this large range, it seems likely that, even allowing for the typical viewing distance of small and large objects, this critical spatial-frequency range expressed in cycles per degree is within the range for which the two curves are maximally and significantly divergent. In sum, Experiment 1 showed that trypophobic images have a visual property not usually possessed by natural images: They have relatively high contrast at midrange spatial frequencies.

Experiment 2: Generality of the Aversion

In our second study, we examined whether aversion to trypophobic images extends across the general population. Fifty images obtained from the trypophobia Web site and 50 images of holes obtained from a Google search were presented in random order as a PowerPoint presentation to 20 undergraduate students at the University of Essex, none of whom reported being trypophobic. The students were asked to rate any discomfort in response to viewing the images, using scales from −5 (maximum discomfort) to 5 (maximum comfort). The mean ratings for the trypophobic and control images were −0.42 and 0.53, respectively, t(49) = 4.67, p < .0001. Evidently, trypophobic images are uncomfortable not simply for a minority of individuals who profess to a phobia but also for individuals in the general population.

Experiment 3: Analysis of Images of Poisonous Animals

In our third study, we attempted to identify the cause of trypophobia by assessing the spectral composition of poisonous animals. This procedure was motivated by an individual who reported a fear of holes and told us that certain animals also induced aversion (e.g., the blue-ringed octopus). The common aspect of the animals seemed to be that they were highly poisonous. We obtained images of animals that in a large number of Internet sources have been listed as “the 10 most poisonous animals.” These animals are extremely poisonous to humans and, consequently, are commonly considered dangerous. The 10 species were the blue-ringed octopus, the box jellyfish, the Brazilian wandering spider, the deathstalker scorpion, the inland taipan snake, the king cobra snake, the marbled cone snail, the poison dart frog, the puffer fish, and the stonefish. Ten different images of each species were obtained from the Internet. As with the trypophobic images, we analyzed the first 10 uncomfortable images that a Google image search generated without prejudice; our only constraint as to selection was that the size of the image had to exceed 300 pixels on its smaller dimension. The images were photographs of the individual animals on a variety of backgrounds, most of which were natural. In most of the photographs, the animal was close to the center of the image and, in its longer dimension, occupied more than 50% of the image. The images were in JPEG format, which involves lossy compression. To control for any artifacts such compression might introduce, we compared the images with images of otherwise similar but nonpoisonous species, sourced and prepared identically. These images were of various nonpoisonous octopus species, nonpoisonous jellyfish, spiders, crabs, nonvenomous snakes, nontoxic frogs, edible snails, and edible fish.

Figure 3 shows the log contrast energy as a function of log spatial frequency for the images of the poisonous and nonpoisonous animals. Overall, a power function (linear on log-log axes) accounted for more than 99% of the variance, which again provided a good fit to the prediction for natural images. The percentage of variance explained by the linear fit to the average log spectra for the 10 highly poisonous animals (99.1%) was less than for the 10 control animals (99.6%), t(9) = 2.58, p = .03. Notwithstanding the normalization of all images in terms of pixel mean and variance, there was 15% more contrast energy at midrange spatial frequencies in the images of the poisonous animals (p < .05; 16–32 cpi); the size of this difference is masked by the logarithmic scale in Figure 3. The excess was obtained not only in the gray-level images but also in the images formed from the R, G, and B pixels taken separately, suggesting that the excess was not dependent on a particular coloration or, indeed, a particular spectral sensitivity. In sum, these results show that the images of highly poisonous animals possess a spectral feature similar to that of the trypophobic images.

Fig. 3.

Power spectra (Fourier power as a function of spatial frequency) of images of poisonous (broken line) and nonpoisonous (solid line) animals analyzed in Experiment 3.

Experiment 4: Snakes and Spiders

Some of the more common phobias are those of snakes and spiders, and many individuals are unable to look at images of these animals without aversion (e.g., Ohman, Flykt, & Esteves, 2001). This is the case even in countries in which spiders are not venomous and present no threat. Furthermore, a number of standard behavioral measures have been used to assess the processing priority given to such objects. For instance, in a typical attention task, observers are required to detect the presence of a target item as quickly as possible. Relatively short response times are usually taken as a marker of cognitive biases toward particular stimuli (e.g., Crundall, Cole, & Galpin, 2007). Both adults and young children have repeatedly been found to detect snakes more rapidly than other kinds of stimuli. LoBue and DeLoache (2008) measured the time taken to detect images of snakes and frogs by young children. They found that the snakes were more rapidly detected and that it was the coiled body shape rather than the snakes’ sometimes-colorful markings that was largely responsible for the conspicuity. Such a spectral power distribution is likely to be conspicuous because it differs from the spectral energy most pervasive in nature.

Motivated by this prior work, we analyzed images of snakes and spiders. Twenty images of snakes and 20 images of spiders with smaller dimension of at least 300 pixels were sourced from Google—again in order of acquisition and without prejudice—and processed as in Experiment 1. Both spectra were curved downward, as reflected in Figure 3. A power function accounted for an average of 98.5% (SD = 1.1%) of the variance of the spectra of the snake images and 98.6% (SD = 1.0%) of the variance of the spectra of the spider images. These figures were substantially lower than those for the images of the control animals used in Experiment 3. Thus, as with the poisonous animals analyzed in Experiment 3, images of snakes and spiders did not show the usual linear relationship of log contrast energy to log spatial frequency.

General Discussion

We found that images responsible for a previously undescribed but relatively common form of visual phobia possess a property characteristic of images that are generally uncomfortable to view. Such images show comparatively high contrast energy at midrange spatial frequencies. This confirms the results of Fernandez and Wilkins (2008), who found a similar property in a variety of uncomfortable images. We also found that images of animals well known to be dangerous also possess this visual property. We therefore suggest that trypophobia arises because the inducing stimuli share a core spectral feature with such organisms—a feature that does not reach conscious awareness. In other words, if any stimulus, such as a configuration of holes, coincidently possesses this spectral feature, the stimulus may induce some form of aversion because of the survival value of such aversion.

This survival account is based on the notion that humans have been selected, via Darwinian principles, for their ability to notice poisonous organisms. The notion that phobias can be explained by an innate predisposition to fear potentially dangerous stimuli (e.g., Marks & Nesse, 1994) is often contrasted with the view that the etiology of phobia is due to a learning process. Aversion to dangerous objects is said to have resulted in modern humans’ possessing an innate predisposition to develop fears of certain objects, such as snakes, spiders, heights, and so forth. An alternative Darwinian explanation is that the ability to effectively process dangerous stimuli evolved before humans originated. Possessing patterns as a warning of unpalatability (i.e., aposematism) is a well-established method of defense (e.g., Santos, Coloma, & Cannatella, 2003). Such patterns tend to be characterized by their high-contrasting colors at midrange spatial frequencies. Furthermore, it is widely accepted that the visual system has been selected for its ability to orient attention to the location of a new object in the visual field (e.g., Abrams & Christ, 2003; Cole & Kuhn, 2009, 2010). However, conscious recognition of an object is a slow process taking up to 350 ms (Johnson & Olshausen, 2003). Responding to a potential threat, such as a snake, on the basis of relatively slow conscious perception could be costly to an organism.

An alternative, more effective detection-and-avoidance strategy might be to respond to the presence of an object via an early, fast-acting visual mechanism based on a simple feature that is common to most dangerous animals. In addition to the perception of motion (e.g., Cole, Heywood, Kentridge, Fairholm, & Cowey, 2003; Skarratt, Cole, & Gellatly, 2009), the computation of contrasts at various spatial scales provides just such a low-level mechanism. Support for this idea comes from other work that has examined threat-related stimuli with respect to low-level features. For instance, Bannerman, Hibbard, Chalmers, and Sahraie (2012) required observers to make a saccade to happy, fearful, or neutral faces that had been filtered so that they had predominantly low, high, or broad spatial frequencies. Among a number of effects, Bannerman et al. reported that at low spatial frequencies, fearful faces showed the fastest saccadic responses. In contrast, there were no differences in mean latency between any emotions for higher spatial frequencies. Similarly, Vuilleumier, Armony, Driver, and Dolan (2003) showed that amygdala activity was greater for the processing of fearful expressions of faces containing low spatial frequencies as opposed to high spatial frequencies.

Given the large number of images that possess an excess of energy at midrange spatial frequencies (Fernandez & Wilkins, 2008), it is most unlikely that this spectral feature is a sufficient condition for phobia, even though it is associated with aversion. Nevertheless, it may prove possible to offer treatment by progressive spatial filtering of the offensive images. It is, of course, still unknown why some people develop an aversion to holes but others do not. This issue is common to all explanations of phobia; some people who have not suffered an animal bite become phobic to dogs, whereas others who have suffered such a bite do not become phobic. However, our results from Experiment 2 do suggest that nontrypophobic individuals are sensitive to the inducing stimuli in that they perceive trypophobic images of holes to be more aversive to look at than nontrypophobic images of holes. Perhaps the condition is a matter of degree, an exaggeration of a normal tendency. Finally, although the aversion has become known as the fear of holes, our data reveal that one essential characteristic that induces the aversion is a particular spectral property, a property often associated with relatively high-contrast material at midrange spatial frequencies and not necessarily involving holes.

Footnotes

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

References

Abrams

R. A.

Christ

S. E.

(2003). Motion onset captures attention. Psychological Science, 14, 427–432.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., rev.). Washington, DC: Author.

Bannerman

R. L.

Hibbard

P. B.

Chalmers

Sahraie

(2012). Saccadic latency is modulated by emotional content of spatially filtered face stimuli. Emotion, 12, 1384–1392.

Cole

G. G.

Heywood

C. A.

Kentridge

R. W.

Fairholm

Cowey

(2003). Attentional capture by colour and motion in cerebral achromatopsia. Neuropsychologia, 41, 1837–1846.

Cole

G. G.

Kuhn

(2009). Appearance matters: Attentional orienting by new objects in the precuing paradigm. Visual Cognition, 17, 755–776.

Cole

G. G.

Kuhn

(2010). Attentional capture by object appearance and disappearance. Quarterly Journal of Experimental Psychology, 63, 147–159.

Crundall

Cole

G. G.

Galpin

(2007). Object-based attention is mediated by collinearity of targets. Quarterly Journal of Experimental Psychology, 60, 137–153.

Fernandez

Wilkins

A. J.

(2008). Uncomfortable images in art and nature. Perception, 37, 1098–1113.

Field

D. J.

Brady

(1997). Visual sensitivity, blur and the sources of variability in the amplitude spectra of natural scenes. Vision Research, 37, 3367–3383.

10.

Hertel

P. T.

Brozovich

(2010). Cognitive habits and memory distortions in anxiety and depression. Current Directions in Psychological Science, 19, 155–160.

11.

Johnson

J. S.

Olshausen

B. A.

(2003). Timecourse of neural signatures of object recognition. Journal of Vision, 3, 499–512.

12.

LoBue

DeLoache

J. S.

(2008). Detecting the snake in the grass: Attention to fear-relevant stimuli by adults and young children. Psychological Science, 19, 284-289.

13.

Marks

I. M.

Nesse

R. M.

(1994). Fear and fitness: An evolutionary analysis of anxiety disorders. Ethology and Sociobiology, 15, 247–261.

14.

Merckelbach

Muris

(1997). The etiology of childhood spider phobia. Behaviour Research and Therapy, 35, 1031–1034.

15.

O’Hare

Hibbard

P. B.

(2011). Spatial frequency and visual discomfort. Vision Research, 51, 1767–1777.

16.

Ohman

Flykt

Esteves

(2001). Emotion drives attention: Detecting the snake in the grass. Journal of Experimental Psychology: General, 130, 466–478.

17.

Santos

J. C.

Coloma

L. A.

Cannatella

D. C.

(2003). Multiple, recurring origins of aposematism and diet specialization in poison frogs. Proceedings of the National Academy of Sciences, USA, 100, 12792–12797.

18.

Skarratt

Cole

G. G.

Gellatly

A. R. H.

(2009). Prioritisation of looming and receding objects: Equal slopes, different intercepts. Attention, Perception, & Psychophysics, 71, 964–970.

19.

Vuilleumier

Armony

J. L.

Driver

Dolan

R. J.

(2003). Distinct spatial frequency sensitivities for processing faces and emotional expressions. Nature Neuroscience, 6, 624–631.

20.

Wilkins

A. J.

Nimmo-Smith

M. I.

Tait

McManus

Della Sala

Tilley

. . . Scott

(1984). A neurological basis for visual discomfort. Brain, 107, 989–1017.