Visual processing in anorexia nervosa and body dysmorphic disorder: similarities, differences, and future research directions

Abstract

Anorexia nervosa (AN) and body dysmorphic disorder (BDD) are psychiatric disorders that involve distortion of the experience of one’s physical appearance. In AN, individuals believe that they are overweight, perceive their body as “fat,” and are preoccupied with maintaining a low body weight. In BDD, individuals are preoccupied with misperceived defects in physical appearance, most often of the face. Distorted visual perception may contribute to these cardinal symptoms, and may be a common underlying phenotype. This review surveys the current literature on visual processing in AN and BDD, addressing lower- to higher-order stages of visual information processing and perception. We focus on peer-reviewed studies of AN and BDD that address ophthalmologic abnormalities, basic neural processing of visual input, integration of visual input with other systems, neuropsychological tests of visual processing, and representations of whole percepts (such as images of faces, bodies, and other objects). The literature suggests a pattern in both groups of over-attention to detail, reduced processing of global features, and a tendency to focus on symptom-specific details in their own images (body parts in AN, facial features in BDD), with cognitive strategy at least partially mediating the abnormalities. Visuospatial abnormalities were also evident when viewing images of others and for non-appearance related stimuli. Unfortunately no study has directly compared AN and BDD, and most studies were not designed to disentangle disease-related emotional responses from lower-order visual processing. We make recommendations for future studies to improve the understanding of visual processing abnormalities in AN and BDD.

Introduction

Anorexia nervosa (AN) and body dysmorphic disorder (BDD) are psychiatric disorders characterized by disturbances in the experience of one’s physical appearance. In AN, individuals are preoccupied with body weight and size, often resorting to caloric restriction to maintain a low body weight. They hold often-delusional convictions of being overweight, despite substantial evidence to the contrary. Additionally, they focus on specific body areas that they believe appear “fat,” such as the abdominal region, hips, and face. In BDD, individuals are preoccupied with misperceived defects in appearance (Phillips et al., 2010). As a result, they believe that they look deformed or ugly, even though the perceived abnormalities are not noticeable to others or appear minor. They are often concerned with specific details, typically of the face or head (e.g. skin blemishes, hair texture, shape of nose), although any body part may be of concern. As in AN, they also are highly convinced of their perceptions, and 27–60% are classified as currently delusional (Mancuso, Knoesen, & Castle, 2010; Phillips, Menard, Pagano, Fay, & Stout, 2006). Both disorders may manifest similar phenomenologic patterns involving hypervigilant attention to details of appearance, which are perceived as flawed, likely contributing to often-delusional distortions in perception.

AN and BDD are associated with substantial psychological distress and functional impairment. Underscoring the broad public health significance of these conditions, the lifetime risk of attempted suicide in BDD is 22–27.5% (Phillips et al., 2005a; Phillips & Diaz, 1997; Veale et al., 1996), and the risk of completed suicide is 30 times that of the general population (Phillips & Menard, 2006). AN is associated with a mortality rate of 5–7% per decade, and an overall standardized mortality higher than any other psychiatric illness (Sullivan, 1995).

In addition to similarities in phenomenology, AN and BDD share a peak onset during adolescence, high risk for chronicity, and have similar comorbidity patterns (although there are higher rates of generalized anxiety disorder in AN and higher rates of panic disorder in BDD) (American Psychiatric Association., 2000; Phillips & Kaye, 2007; Phillips, Menard, Fay, & Weisberg, 2005b; Swinbourne & Touyz, 2007). AN and BDD co-occur frequently; up to 32% of BDD patients also have a lifetime comorbid eating disorder (Ruffolo, Phillips, Menard, Fay, & Weisberg, 2006) and 25–39% of those with AN are diagnosed with comorbid BDD (Grant, Kim, & Eckert, 2002; Rabe-Jablonska Jolanta & Sobow Tomasz, 2000). There is also overlap in specific areas of appearance concerns, e.g. size of abdomen, hips, and thighs (Grant & Phillips, 2004). Approximately 30% of individuals with BDD report significant weight concerns, a characteristic linked to greater symptom severity and morbidity (Kittler, Menard, & Phillips, 2007). The few studies that have directly compared AN and BDD found similarities on clinical and psychological measures, with both groups exhibiting severe body image symptoms and low self-esteem compared to healthy controls (Hrabosky et al., 2009; Kollei, Brunhoeber, Rauh, de Zwaan, & Martin, 2012; Rosen & Ramirez, 1998). There are also important differences, most notably that the gender distribution is less skewed toward females in BDD (Buhlmann et al., 2010; Koran, Abujaoude, Large, & Serpe, 2008; Rief, Buhlmann, Wilhelm, Borkenhagen, & Brahler, 2006).

The similarities in clinical features suggest that AN and BDD may represent overlapping body image disorders (Cororve & Gleaves, 2001). However, BDD is currently categorized as a somatoform disorder in DSM-IV-TR and as a form of hypochondriasis in ICD-10, while AN is categorized as an eating disorder in both systems (American Psychiatric Association., 2000; World Health Organization., 1992). Moreover, BDD is often considered to be on the obsessive-compulsive disorder (OCD) spectrum, due to similar phenomenology, demographics, heredity, course of illness, and response to treatment (Hollander & Wong, 1995; Phillips et al., 2007). (Of note, AN also has some features suggestive of overlap with OCD, including obsessive thoughts and ritualized eating behaviors, high comorbidity of OCD, and a high proportion of first degree relatives with OCD (Phillips et al., 2007).)

Since distorted perception of appearance is a key feature of both AN and BDD, examining visual processing as a phenotype may provide a level of understanding about the relationship between these two disorders, and about the neurobiology behind this phenomenon, which is less likely to be captured by examining individual categorical diagnoses (Insel & Cuthbert, 2009). This has important clinical relevance, as persistent perceptual disturbance is a strong predictor of relapse in AN (Keel, Dorer, Franko, Jackson, & Herzog, 2005). There is a considerable need for understanding the neurobiology of perception in AN and BDD, including any similarities and differences, to help guide the development of rational treatments. To maintain focus on the phenotype of abnormal visual perception of appearance, we did not include other disorders in this review such as OCD or social anxiety disorder; these disorders may also be related to AN and BDD, although perhaps via different overlapping phenotypes (heightened self-consciousness, tendencies for obsessive thoughts and compulsive behaviors, etc.). We have not included other eating disorders, such as bulimia nervosa (BN), for several reasons. Among BN, AN, and BDD there is overlap of certain common clinical features (perceptual distortions, high trait perfectionism, and high comorbid anxiety) (American Psychiatric Association., 2000; Phillips et al., 2005b; Phillips et al., 2010; Sutandar-Pinnock, Blake Woodside, Carter, Olmsted, & Kaplan, 2003; Swinbourne et al., 2007). However, BN has additional characteristics that set it apart from AN and BDD with respect to perception and visual processing. For one, distorted body image perception is required for a diagnosis of AN or BDD, but not for BN (American Psychiatric Association., 2000). While many individuals with BN do have body image disturbances (Jansen, Nederkoorn, & Mulkens, 2005; Schneider, Frieler, Pfeiffer, Lehmkuhl, & Salbach-Andrae, 2009), this disorder is characterized by a preoccupation with shape and weight, along with body dissatisfaction (even if shape and weight are accurately perceived) (Stice & Agras, 1998). Thus, BN is more heterogeneous when it comes to perceptual distortions, with less consistency than in AN and BDD. Another characteristic that sets BN apart from AN and BDD, in general, is that individuals with BN have higher rates of impulsivity (Claes, Mitchell, & Vandereycken, 2012a; Claes et al., 2012b) and comorbidity with impulse control-related disorders (Fernandez-Aranda et al., 2008). Finally, there is support for the conceptualization of AN and BDD, unlike BN, as including individuals with low insight or delusional beliefs (Hartmann, Greenberg, & Wilhelm, 2013; Konstantakopoulos et al., 2012; Mancuso et al., 2010).

Conscious perception is a complex phenomenon that relies on multiple visual processing systems in the brain, along with tightly linked cognitive and emotional processes that contribute to the subjective perceptual experience (Moutoussis, 2009; Zeki & Bartels, 1999). Visual information is exchanged through functional connections between lower- and higher-order visual areas (occipital, temporal, and parietal), and centers for emotion, cognition, and memory (Lamme & Roelfsema, 2000). This facilitates both bottom-up, perceptually driven visual inputs to emotion and cognitive systems, and top-down modulation of visual input based on conscious interpretation (Hanson, Hanson, Halchenko, Matsuka, & Zaimi, 2007; Iaria, Fox, Chen, Petrides, & Barton, 2008). An individual’s current psychological state and past experiences with emotionally charged visual stimuli (e.g. images of bodies and faces for AN and BDD) are ever-present confounds in studies assessing visual processing (Rossignol et al., 2012; Schettino, Loeys, Bossi, & Pourtois, 2012). Pre-existing or symptom-dependent abnormalities in the function of lower-order visual systems, higher-order cognitive and emotional systems, or both, could be involved in abnormal perception. The majority of studies performed in AN and BDD thus far, unfortunately, have not been designed to discern top-down from bottom-up phenomena. This review focuses on studies that have addressed visual processing in individuals with AN or BDD. We define visual processing as phenomena involved in any of the following steps: acquisition of visual input in the peripheral sensory system (ophthalmologic), relay of this information to the central nervous system, neural processing of visual information in occipital and occipito-parietal regions (from basic feature characteristics to more complex aspects), and further elaboration and integration into representations of whole percepts (e.g. face or body images) in (primarily) temporal brain areas.

Our goal was to examine evidence for abnormalities of different aspects of visual processing in AN and BDD, from the function and structure of the eye to higher order processing of human face and body images. To focus the review, we excluded studies that lacked information on visual processing itself but may have otherwise investigated consciously or unconsciously held beliefs about appearance, emotional reactions to visual stimuli (including food stimuli), facial emotional recognition, and visual memory or attention. Another goal was to compare visual processing abnormalities between these related disorders; definitive conclusions, however, were limited because no study directly compared these two groups.

We organized these studies into: a) ophthalmologic findings; b) perceptual organization as assessed through neuropsychological tests of visuospatial, global/local processing, or multi-sensory integration that included visual stimuli; c) visual processing of naturalistic images (face or body images); and d) functional and structural brain imaging studies of visual processing. The latter category could provide information about visual processing at any of the aforementioned steps. In addition, we included studies that examined evidence for any abnormalities as either state (secondary to symptoms of the illness) or trait (pre-existing) characteristics in AN or BDD.

Materials and Methods

We searched for articles in ISI Web of Knowledge, PubMed, and PsychINFO databases. We used keywords of either “body dysmorphic disorder” or “anorexia nervosa” along with the following: “vision,” “eye tracking”, “eye movements”, “visual processing,” “visual perception,” “body processing,” “face processing,” “central coherence,” “global local processing,” “Navon,” and “Rey-Osterrieth Complex Figure Test.” We also used the keyword “visual” along with “body dysmorphic disorder,” but not with “anorexia nervosa” because the latter generated excessive unrelated results. We excluded articles that were not peer reviewed (n=15), not written in English (n=4), or did not provide data or clinical descriptions of visual processing of human bodies or faces in AN or BDD (n=200). We did not include articles describing visual memory or attention alone because we felt it was impossible to disentangle elements of visual processing from top-down and bottom-up modulation from other cognitive domains. Some versions of the neuropsychological tests included in our search terms also involve a memory component, although modifications helped separate this confound in some studies. We also included relevant manuscripts that were cited by articles found through the literature search, but were not otherwise retrieved using our search terms. In addition, we performed a search on Google Scholar (www.scholar.google.com) using the same search terms to locate any relevant articles that the other search methods may have missed. We did not impose a limitation on publication date of the articles.

Forty-four journal articles for AN and 15 for BDD were included in this review, ranging in publication date from 1973 to 2012. Of these, two articles for AN and one for BDD were literature reviews, one article for AN was a meta-analysis, and three articles for AN were case reports. All studies of BDD included adults only. Most studies of AN included adults (26 total), with ten studies also including adolescents and five studies including only adolescents and children. Most BDD studies did not list illness duration; the three studies that included this information reported mean illness durations of approximately 10–20 years (Kiri, 2012; Stangier, Adam-Schwebe, Muller, & Wolter, 2008; Yaryura-Tobias et al., 2002). Twenty-two of the AN studies listed illness duration, with means ranging from less than a year to over 20 years across studies and substantial variability within studies. Most AN studies included only females, with the exception of three studies that included one or two males (Andres-Perpina et al., 2011; Castro-Fornieles et al., 2009; Slade & Russell, 1973) and one large study that included 7 men with AN (Stedal, Rose, Frampton, Landro, & Lask, 2012). BDD studies were generally of mixed gender, with the exception of three studies that included only women (Clerkin & Teachman, 2008; Deckersbach et al., 2000; Stangier et al., 2008). A minority of studies included populations that were medication-free (six BDD studies, seven AN studies), with most studies not listing medication status, or else including individuals taking antidepressants, anxiolytics, or antipsychotics. Most studies excluded individuals with neurologic conditions or substance abuse, yet allowed other psychiatric comorbidities – with depression, anxiety, and OCD being the most common.

Results

II. Anorexia Nervosa

A. Ophthalmologic Findings

Two published studies have investigated vision at the ophthalmologic level in AN. Both found decreased retinal nerve fiber layer thickness in patients compared to healthy controls (Caire-Estevez, Pons-Vazquez, Gallego-Pinazo, Sanz-Solana, & Pinazo-Duran, 2012; Moschos et al., 2011). One study specifically found lower mean foveal thickness in AN (Moschos et al., 2011). In the two studies, there were inconsistent differences in visual acuity and visual fields; Moschos et al. found no differences between AN patients and controls in visual acuity or visual field, whereas Caire-Estevez et al. found worse visual acuity and visual field sensitivity in the AN group. It is unclear whether these differences are secondary to malnourishment and weight loss, as each tested underweight patients. Further research in recovered patients would help understand whether these differences persist after nutrition is improved.

B. Neuropsychological tests of visuospatial and global/local processing

Studies evaluating visuospatial abilities in AN have focused primarily on the integration of sensory information and on cognitive style, the latter suggesting that “weak central coherence” is present in AN (Lopez, Tchanturia, Stahl, & Treasure, 2008). Weak central coherence implies a lack of global and integrated processing and enhanced focus on detail (Frith, 2003). Studies investigating central coherence often use the Rey-Osterrieth Complex Figures Task (RCFT) (Shin, Park, Park, Seol, & Kwon, 2006) or a variation of the Embedded Figures Task (Witkin, 1950).

The RCFT requires participants to draw a complex figure and can be scored on different aspects, including performance and strategy on copy, although the interpretation of visual processing is confounded by memory in both immediate and delayed recall conditions. Studies measuring the accuracy of copy have shown either equivalent (Castro-Fornieles et al., 2009; Danner et al., 2012; Lopez et al., 2008; Sherman et al., 2006; Stedal et al., 2012) or poorer performance in AN relative to control groups (Kim, Lim, & Treasure, 2011; Lopez, Tchanturia, Stahl, & Treasure, 2009). One study found that underweight participants with AN performed worse than controls on copy, but improved after gaining 10% of body weight (Kingston, Szmukler, Andrewes, Tress, & Desmond, 1996), suggesting that the deficit may be weight-dependent. The majority of studies, which included adolescents and children along with adults, found significantly worse delayed recall in AN (Andres-Perpina et al., 2011; Camacho, 2008; Favaro et al., 2012; Lopez et al., 2008; 2009; Mathias & Kent, 1998; Pendleton-Jones, 1991; Sherman et al., 2006; Stedal et al., 2012; Tenconi et al., 2010; Thompson, 1993), or a trend for worse delayed recall (Castro-Fornieles et al., 2009), but some found equivalent performance (Danner et al., 2012; Kim, 2011; Kingston et al., 1996; Murphy, Nutzinger, Paul, & Leplow, 2002).

A few of the studies that found poorer recall on the RCFT in individuals with AN compared to healthy controls also investigated potential mechanisms for this deficit. Three studies, one of which included children and adolescents (Stedal et al., 2012), evaluated the order of construction and found that individuals with AN draw the detailed aspects of the figure first and show less continuity in their drawing (Lopez et al., 2008; Sherman et al., 2006; Stedal et al., 2012). In two of these studies, copy organization significantly mediated the relationship between diagnostic group and recall accuracy, suggesting that individuals with AN may not encode information efficiently for accurate retrieval (Lopez et al., 2008; Sherman et al., 2006).

Most of these studies involved underweight participants, which may explain poorer cognitive ability. However, one study tested a sample of females at risk for developing an eating disorder based on subclinical symptoms (Alvarado-Sanchez, Silva-Gutierrez, & Salvador-Cruz, 2009). This group evidenced more fragmented completion of the figure, although overall accuracy was equivalent to a non-risk comparison group. Two studies of weight-restored participants with AN showed a lack of significant differences on accuracy compared to healthy controls on copy and recall (Kingston et al., 1996; Pendleton-Jones, 1991), but did not evaluate strategy or style. A recent study found worse performance on the RCFT and lower central coherence in a cohort of underweight AN participants, but no significant difference between a separate cohort of recovered AN participants and healthy controls (Favaro et al., 2012).

Other studies have examined central coherence with the Embedded Figures Task (EFT), in which participants locate a shape embedded in a complex figure, with shorter response times attributed to bias towards detailed processing. Studies have also used the similar Matching Familiar Figures (MFF) test, which asks participants to identify which one of eight figures matches one previously viewed. Both of these tasks require detailed searching of the test images and visual working memory to recall the previously viewed figure.

Studies utilizing the EFT have shown inconsistent results. Three studies, one including adolescents, found that individuals with AN identified the embedded figures more quickly and with higher accuracy than healthy controls when the embedded figure was available for reference during the task (Lopez et al., 2008; 2009; Tokley & Kemps, 2007). However, Pendleton-Jones et al. (1991) found that longer time was required in both underweight and weight-restored AN adults relative to healthy controls, using the original version of the test, which required holding the figures in working memory, creating a confound between visual processing and memory. When adults and adolescents were administered a time-constrained EFT task, the AN group correctly located fewer shapes than controls (Kim et al., 2011).

Two studies have tested individuals with AN using the MFF. One study in adults (Toner, Garfinkel, & Garner, 1987) found superior accuracy in AN, suggesting a bias toward detail level processing. They also found faster response time; while this is generally interpreted as suggesting less bias toward detail processing, which requires greater time to perform, it could alternatively be suggestive of bias toward attention to detail along with abnormally high speed of detail processing. The other study in adolescents and adults found no difference in performance or response time compared to controls (Southgate, Tchanturia, & Treasure, 2008).

Other studies have utilized novel methods to investigate visuospatial processing, as well as sensory integration, in individuals with AN. In a task requiring participants to compare two visual stimuli, AN participants performed as quickly and accurately as healthy controls, although they were slower when the comparison was lexical (Eviatar, Latzer, & Vicksman, 2008). In order to evaluate body perception while minimizing the confounding emotional impact of body images, Nico et al. (2010) had participants follow a stimulus on a trajectory and estimate whether it would hit their body (Nico et al., 2010). Participants with AN were worse at detecting their left body boundary, showing a tendency to underestimate it. Although this was incongruent with feelings of “fatness,” which is expected to expand the body boundary, it was similar to the performance of stroke participants with right parietal damage, who were also evaluated. Another study of visuospatial processing used a task of manually matching the angle of a moveable bar. Adolescents with AN performed worse than healthy controls when using their right hand (Grunwald et al., 2002). Taken together, findings from these two studies (Grunwald et al., 2002; Nico et al., 2010) suggest deficits both on the left and right side of the body. This could be due to problems with hemispheric integration, since integration of multisensory information occurs in the right posterior parietal cortex (Grunwald et al., 2002). Individuals must perceive the visuospatial field before they can act on it, possibly explaining why one study found perceptual differences on the left side of the body (Nico et al., 2010) and another found performance differences on the right (Grunwald et al., 2002).

Guardia et al. (2012) also found evidence of visuospatial deficits in adolescents and adults. Participants with AN overestimated their body size (but not the body size of others) compared to healthy controls when asked to estimate if the body would be able to pass through a doorway (Guardia et al., 2012).

Integration between sensory modalities is important for adjusting visual perception and correcting errors in perception. This integration can be tested with a size-weight illusion, where one must integrate tactile with visual information to estimate weight in two differently sized objects of the same weight. One such study showed that individuals with AN perform better than controls, suggesting a reduced reliance on visual information in judgment of weight (Case, Wilson, & Ramachandran, 2012). A possible explanation is greater reliance on proprioceptive information, although the mechanism of enhanced performance is unclear.

C. Visual processing of naturalistic images

Studies in AN in adults and adolescents have found abnormalities in attention, and overestimation of body size, specific to images of their own bodies (Garner, Garfinkel, Stancer, & Moldofsky, 1976; Slade et al., 1973; Smeets, Smit, Panhuysen, & Ingleby, 1997; Urgesi et al., 2012). An eye tracking study showed that AN participants focus visual attention on body parts they are dissatisfied with, whereas controls tend to scan the whole body image (Freeman, 1991). Another study showed that AN participants saccade more quickly to their own picture compared to other pictures, unlike controls (Blechert, Ansorge, & Tuschen-Caffier, 2010). These differences are consistent with symptoms of increased attention to body areas that evoke strong feelings of dissatisfaction. Without further investigations into visual processing, however, one cannot conclude if these findings are due to abnormalities in sensory-level visual information processing, cognitive and evaluative processes related to AN, or both.

Several studies indicate that adults and adolescents with AN overestimate their overall body size (Garner et al., 1976; Slade et al., 1973; Smeets et al., 1997; Urgesi et al., 2012), even though processing of their own body parts (Garner et al., 1976; Slade et al., 1973; Smeets et al., 1997; Urgesi et al., 2012) and of the body size of other women (Garner et al., 1976; Slade et al., 1973; Smeets et al., 1997; Urgesi et al., 2012) does not differ or is more accurate than controls. These studies also found that AN participants correctly judge height, body movements (Garner et al., 1976; Slade et al., 1973; Smeets et al., 1997; Urgesi et al., 2012), and objects (Garner et al., 1976; Slade et al., 1973; Smeets et al., 1997; Urgesi et al., 2012). These findings suggest a specific disturbance of own body image. However, other studies have found perceptual abnormalities when viewing images of others’ bodies. AN adults are better at detecting “thinner than” differences in others’ bodies (Smeets, Ingleby, Hoek, & Panhuysen, 1999) and are more accurate than controls in a delayed matching-to-sample task of pictures of male bodies, with no performance differences when matching pictures of body movements in adolescence (Urgesi et al., 2012). In summary, there appears to be evidence in AN for disturbances in visual processing of both own and others’ bodies, although there may be slightly different patterns in each.

D. Brain imaging studies of visual perception

Several brain imaging studies presenting naturalistic images of bodies to participants with AN found abnormal brain activity. In a review, Pietrini et al. (2011) report relatively consistent findings of abnormal activity in frontal (anterior cingulate, and frontal visual system (right superior frontal (Beato-Fernandez et al., 2009) and right dorsolateral prefrontal (Wagner, Ruf, Braus, & Schmidt, 2003), parietal (inferior parietal lobule), and striatal (caudate) regions (Pietrini et al., 2011). In one functional magnetic resonance imaging (fMRI) study, AN participants showed higher ventral striatal activity when viewing underweight images compared to normal weight bodies, and also preferred underweight images, unlike controls (Fladung et al., 2010). When asked to compare their own body to an image of another body, AN participants showed less activation in the insula and premotor areas and more activation in the anterior cingulate compared to controls (Friederich et al., 2010). This comparison of own vs. other body was associated with greater anxiety in AN participants, who, unsurprisingly, were less satisfied with their own body image. Another study contrasted own body images that had been altered to appear overweight vs. unaltered images. Left medial prefrontal cortex activation was reduced for the restrictive subtype of AN compared to healthy controls, while amygdala activation was normal for the combined AN group (Miyake et al., 2010). A structural MRI study found reduced gray matter density in the left extrastriate body area, which is involved in processing images of human body parts, in individuals with AN compared to controls (Suchan et al., 2010).

Taken together, the functional and structural brain imaging evidence suggests that AN participants demonstrate functional and structural abnormalities in brain areas that are involved in processing visual images of human bodies, as well as, systems involved in anxiety and emotion. Preexisting abnormalities in brain function/structure could predispose individuals to developing AN; however this is difficult to separate from the effects of low weight and poor nutrition, as these were predominantly studies of underweight individuals.

E. State vs. Trait

Some aspects of abnormal visual processing in AN may represent state characteristics (secondary to weight, nutrition, or other symptoms, and modifiable by treatment and recovery) while others represent traits (pre-existing and usually stable across time). Weight gain in underweight AN participants was associated with improved copy scores on tests of global vs. local processing (Kingston et al., 1996) and reduced overestimation of own body width (Slade et al., 1973).

Treatment and recovery from AN have also been associated with changes in functional brain activity. After recovery, AN participants showed normalized activation in the amygdala and fusiform gyrus for happy and fearful faces (Cowdrey, Harmer, Park, & McCabe, 2012). After treatment with cognitive behavioral therapy that reduced negative body-related thoughts compared to a waitlist group, participants showed increased brain activation when viewing pictures of their own bodies compared to pre-treatment, whereas the non-treatment group actually showed a decrease in activity (Vocks et al., 2011). The increase in activation after treatment was seen in areas that process human body images (extrastriate body area (Downing, Jiang, Shuman, & Kanwisher, 2001), left middle temporal gyrus (Weiner & Grill-Spector, 2011)) and self-awareness (bilateral middle frontal gyrus (Platek, Wathne, Tierney, & Thomson, 2008)).

On the other hand, certain abnormalities in global-local processing may be stable traits of individuals predisposed toward AN, as they have been found in studies of recovered AN participants, unaffected relatives, and at-risk populations with sub-clinical symptoms. (Of note, certain pathophysiological processes evident in recovered AN individuals may represent “scars,” or lasting effects of the underweight state or other aspects of the illness that persist after weight has been restored.) Superior attention to detail and poor central coherence compared to controls was observed in both active and recovered AN participants and their unaffected sisters, for adults and adolescents (Roberts, Tchanturia, & Treasure, 2012; Tenconi et al., 2010). A strong correlation between altered central coherence and deficits in set shifting also persisted in recovered AN participants (Danner et al., 2012). Females at risk for developing AN, with sub-clinical symptoms, also demonstrate worse organization (fragmented completion of figure) on the RCFT (Alvarado-Sanchez et al., 2009). Overall, there is some disagreement on whether and how global-local processing changes with treatment in AN. There is no available literature on the stability of other aspects of visual processing in AN.

II. Body Dysmorphic Disorder

A. Ophthalmologic Findings

There are no published studies yet on ophthalmologic abnormalities in BDD.

B. Neuropsychological tests of visuospatial and global/local processing

Several studies have investigated visuospatial processing in BDD. One study found that individuals with BDD, similar to those with OCD, performed normally on visuospatial construction and memory on the RCFT (Hanes, 1998). A subsequent neuropsychological study, on the other hand, found that the BDD group performed worse than controls on the RCFT (Deckersbach et al., 2000). In this study, group differences in free recall were mediated by deficits in organizational strategies, in which the BDD group selectively recalled details instead of larger organizational design features. The authors suggested that abnormalities in executive functioning might have explained these results. However, earlier perceptual abnormalities in global and local visual processing, or differences in selective attention, may have also contributed, since this task involved viewing and encoding a complex visual figure.

Dunai et al., (2010) administered a battery of executive functioning tests of planning, organization, working memory, and motor speed to participants with and without BDD (Dunai, Labuschagne, Castle, Kyrios, & Rossell, 2010). They found several domains of executive functioning were impaired in BDD, including difficulty manipulating visual information held in working memory on the Spatial Working Memory Task.

C. Visual processing of naturalistic images

Early experimental evidence that BDD may involve aberrant own-face perception comes from a study in which BDD participants and healthy controls viewed an image of their own face and indicated if any alterations had been made. A higher proportion of the BDD group perceived distortions of their faces, when in fact none were made (Yaryura-Tobias et al., 2002). Another study investigated asymmetry detection in individuals with BDD (Reese, McNally, & Wilhelm, 2010). Participants viewed others` faces that were unaltered or altered in symmetry, and also viewed arrays of dots that were symmetric or asymmetric. Individuals with BDD did not differ significantly from controls in accuracy of detecting asymmetry with faces or dot arrays, although they were slower in making decisions about symmetry. In another study, BDD participants were more accurate than controls at detecting changes made to facial features (e.g, distance between the eyes) of photos of others’ faces (Stangier et al., 2008).

Another investigation of face processing demonstrated that individuals with BDD were slower and less accurate than controls in matching the identity of an emotional face to the same face with a neutral expression (Feusner, Bystritsky, Hellemann, & Bookheimer, 2010a). This was evident regardless of the type of emotional expression. This suggests general abnormalities in visual processing of faces, which may be more pronounced when features are in a different configuration, such as occurs with emotional expressions.

The “face inversion effect” is a phenomenon in which recognition of inverted (upside down) faces is less accurate and slower relative to recognition of upright faces, due to the absence of a holistic template for inverted faces (Farah, Tanaka, & Drain, 1995). In a study using this task, the BDD group demonstrated a smaller inversion effect during the longer duration stimuli, but no differences were seen for shorter duration stimuli (Feusner et al., 2010b). This suggests that BDD individuals may have an imbalance in global vs. local processing, with a tendency to engage in highly detailed processing of faces, whether upright or inverted. This is in contrast to controls, who may primarily engage holistic processing for upright faces, yet rely on detailed processing for inverted faces (Freire, Lee, & Symons, 2000). This may explain the BDD group’s advantage in speed for inverted faces, yet only when stimuli were presented long enough to engage detail processing. Another study used inverted faces to test face recognition in individuals with BDD and healthy controls (Kiri, 2012). Although the study did not test the “inversion effect” per se, they found that individuals with BDD relative to controls had enhanced ability to recognize inverted famous faces, but did not demonstrate significant differences for upright famous faces.

Clerkin and Teachman (2008) tested visual processing of images of own faces morphed with those of highly attractive or unattractive others, in individuals with either high or low BDD symptoms (Clerkin et al., 2008). The low BDD symptom group demonstrated a normative self-enhancement bias (tendency to rate more attractive morphed image as representing themselves), which was not evident in the high BDD symptom group. This resulted in a non-significant trend for interaction between morphed photograph type and group.

A study using eye tracking investigated selective visual attention in BDD, social phobia, and healthy controls (Grocholewski, Kliem, & Heinrichs, 2012). Only the BDD group showed selective attention to their own areas of perceived defects of their faces, as measured by number of fixations per degree of visual angle.

Results from these psychophysical experiments suggest that imbalances in holistic vs. detailed processing may explain performance advantages in individuals with BDD relative for inverted faces (Feusner et al., 2010b), as well as for change detection in facial features of others’ faces (Stangier et al., 2008), and may be an inefficient or inaccurate strategy for identity recognition across facial expressions (Feusner et al., 2010a). In addition, heightened vigilance to details, particularly for areas of perceived defects (Grocholewski et al., 2012), may also increase susceptibility to errors of commission. This could result in “false positive” errors when scrutinizing own-face images (Yaryura-Tobias et al., 2002).

D. Brain imaging studies of visual perception

The first fMRI study to investigate the neural correlates of visual perception in BDD used others’ faces as stimuli (Feusner, Townsend, Bystritsky, & Bookheimer, 2007). BDD participants and healthy controls were scanned with fMRI while matching photographs of others’ faces with normal, high or low spatial frequencies (creating images that contained primarily high detail or configural/holistic information, respectively). The BDD group demonstrated left hemisphere hyperactivity relative to controls in an extended face-processing network for normal and low spatial frequency images. Within-groups results suggested that healthy controls only engaged the left hemisphere for high spatial frequency (high detail) images, whereas BDD participants engaged the left hemisphere for all image types.

An fMRI study using own-face stimuli in BDD participants and healthy controls found abnormal hypoactivity in the BDD group in striate and extrastriate visual cortex for low spatial frequency images, and hyperactivity in orbitofrontal cortex and caudate for normal images (Feusner et al., 2010c). BDD symptom severity correlated with orbitofrontal-striatal and extrastriate visual cortex activity. In a secondary data analysis of the same experiment, anxiety scores in BDD were regressed against fMRI signal changes in brain areas implicated in anxiety and visual processing of details (Bohon, Hembacher, Moller, Moody, & Feusner, 2012). Intermediate anxiety scores were associated with higher levels of brain activity than high or low scores in ventral visual processing areas. Interestingly, the relationship between anxiety and activity in ventral visual processing systems held regardless of BDD symptom severity.

Another fMRI experiment used inanimate object stimuli to investigate general abnormalities in visual processing in BDD (Feusner, Hembacher, Moller, & Moody, 2011). BDD participants and healthy controls matched photographs of houses that included normal, high or low spatial frequencies. The BDD group demonstrated abnormal hypoactivity in secondary visual processing systems for low spatial frequency images.

These functional neuroimaging studies provide evidence of abnormal visual processing in BDD, although they utilized relatively small sample sizes. The studies found abnormalities in primary and/or secondary visual cortical, temporal, and prefrontal systems and suggest imbalances in detailed vs. global/configural processing. Moreover, this overall pattern is evident not only for own and other’s appearance-related stimuli, but also for inanimate objects, suggesting more general aberrancies in visual processing.

E. State vs. Trait

There are currently no published studies on visual processing abnormalities in BDD as being either state or trait features.

Discussion

Summary of findings

Overall, the literature on AN and BDD suggests a pattern of abnormalities in visual processing and perceptual organization that includes over-attention to detail and reduced processing of larger global features. In both AN and BDD, cognitive strategy and attention may at least partially mediate abnormalities, as these groups tend to focus more on symptom-specific details (body parts in AN and facial features in BDD), and misperceive aspects of their own images. However, visuospatial abnormalities are also evident in both disorders for non-appearance related stimuli. In brain imaging studies, both disorders show abnormal brain activation in frontal, parietal, striatal, and visual systems. Since no study has yet to directly compare visual processing in AN and BDD, we consider the following conclusions separately for each group.

In AN, there is evidence of over-attention to detail and reduced processing of larger holistic features, which likely contribute to lower accuracy on visuospatial tasks. AN individuals tend to overestimate their own body size in images. There is also evidence of abnormal reward circuit and limbic system activity for specific, salient body images. Integration of information between the left and right hemispheres in the brain may also be impaired.

Several studies have found that patterns of visual processing in those with AN may depend on body weight. However, possible effects of weight on visual processing of bodies may be intertwined with the severity of the disorder or degree of recovery. For example, at lower body weight several domains of AN symptoms may be more severe, and ability to gain weight or successfully maintain a normal weight may be linked to improvement in different symptom (e.g. cognitive rigidity or anxiety related to weight gain). Thus, it can be difficult to determine whether differences in visual processing after weight gain are related to changes in weight or nutrition, or to improvement in other symptoms.

In BDD, we also see evidence for increased attention to detail, reduced global processing, and poorer performance in visuospatial tasks. In this group, spatial working memory has been found to be impaired, which may contribute to these effects. Individuals with BDD may employ brain systems normally reserved for detailed image processing and underutilize brain regions responsible for configural and holistic processing. They also identify non-existent distortions in their own face images and show abnormal sensitivity to detecting change in others’ faces, possibly due to heightened vigilance to details. Abnormal performance in tests of visual perception and brain activation patterns in BDD are present for own face images, other face images, and inanimate objects.

However, findings in both disorders at this point should still be considered inconclusive, as there are some discrepancies in results across studies. Some of the discrepancies may be explained by differences in the study populations, which could have affected the measurements being analyzed. All BDD and some AN studies included only adults, with several AN studies also investigating adolescents. For example, age differences may explain discrepancies in studies using the MFF; studies of AN adults showed higher accuracy and detailed processing, while studies of adolescents found no difference between AN and controls. All BDD studies were of adults, so no clear conclusions about age in this group can be made. Differences in illness duration across studies may explain discrepancies in results. For example, studies including individuals with AN with longer durations of illness demonstrated abnormally slow performance. This may be due to an accumulation of damage to the brain as a result of longer standing malnutrition, anxiety, depression, etc. In general, duration of illness was either not listed or spanned such a broad range (months to decades in some cases), that no clear conclusions can be drawn. Comparisons between AN and BDD should also be made in light of the fact that BDD studies included both genders, while AN studies included only women (with one exception (Stedal et al., 2012)). Thus AN findings may be more specific to women, while BDD findings may be more generalizable to males and females. Two additional factors that could affect results across studies are comorbidities with other psychiatric disorders (some studies included individuals with AN or BDD who had comorbid anxiety, depression, and OCD, along with other disorders), and current use of psychoactive medications. It is also important to note that not all studies provided detailed descriptions of their study populations, making it difficult to assess these factors thoroughly.

Also limiting conclusions is the fact that in general there is an insufficient body of research on visual processing in AN and BDD. In particular, many aspects of visual processing have not been investigated in either disorder (e.g. sensory-level striate and extrastriate visual cortical functioning).

Recommendations for future studies

The following recommendations are meant to address limitations in the current literature, and to expand our understanding of pathological processes that cross diagnostic boundaries. It is difficult to make conclusions about visual processing abnormalities in AN and BDD because few studies have adequately disentangled abnormalities in visual processing from effects due to factors that influence visual perception, particularly emotionally salient stimuli. These modifying variables may include anxiety (both general, disease-related, and task-provoked anxiety), depression, personality traits (such as perfectionism and cognitive rigidity), specific symptom severity (obsessive thoughts, or fear of being fat), insight/delusionality, or weight in AN (current weight, or weight gain longitudinally, under-weight vs. weight-restored). While some of these factors may be modeled as covariates, others could be manipulated experimentally to provide more power and sensitivity to detect subtle effects on visual processing. For example, emotionally neutral figures and objects can be presented in various degrees of complexity, contrast, and spatial frequency. Emotional or physical experiences can also be manipulated in experiments to understand their effects on visual processing. Future studies should include stimuli that both do and do not elicit a disease-related emotional response to separate the influence of emotion on visual processing.

Abnormalities in visuospatial performance on neuropsychological tasks in AN and BDD could be mediated by abnormal cognitive strategies for encoding or retrieval of visual information in working memory. Some studies used modified versions of these tasks to reduce confounds with memory. It would also be worthwhile for future studies to specifically investigate visual memory and attention abnormalities in AN and BDD, using studies designed to disentangle these cognitive functions from basic visual processing abnormalities. This could be done, for example, with eye-tracking studies using basic visual as opposed to symptom-relevant stimuli, or studies assessing the difference in response of individuals asked to consciously shift their attention to different stimulus features. As an example of the former, Pallanti et al. (1998) found that the severity of abnormalities in smooth eye movements and saccadic performance in AN to moving, low level visual stimuli was correlated with OCD symptoms, perfectionism, drive for thinness, and interoceptive awareness (they did not, however, measure perceptual distortions in their participants) (Pallanti, Quercioli, Zaccara, Ramacciotti, & Arnetoli, 1998). Widely used and well-validated, low level, neutral visual stimuli such as sine-wave gratings (Campbell & Green, 1965) or the contour integration task (Kovacs, Polat, Pennefather, Chandna, & Norcia, 2000) could be employed to assess relatively early visual system functioning. Moreover, brain imaging techniques such as electroencephalography (EEG) or magnetoencephalography (MEG) have the temporal resolution to discriminate early vs. later visual processing abnormalities, which is not possible with fMRI.

Conclusions about similarities and differences between AN and BDD are limited by the fact that these groups have not yet been directly compared on the same experimental tasks. To more definitively compare and contrast visual processing in these related disorders of body image, future studies should include both AN and BDD subjects, along with matched controls, within the same experiment to ensure methodological consistency. In addition, studies in large, analog populations that are assessed for dimensionality of body image disturbances and visual perceptual distortions may uncover dimensional abnormalities in visual processing that map to dysfunctional brain networks. Another line of research yet to be performed is an investigation of if, and how, prior trauma may relate to the development of aberrant visual processing.

Whether visual abnormalities are a contributing cause or a consequence of AN and BDD is still unclear. We attempted to address this question in the State vs. Trait section of our Results; however no papers were available on this topic for BDD. For AN, several studies suggest that heightened attention to detail and poor central coherence are stable traits that could have contributed to the development of the disorder. However, this inference is based on individuals in recovery or at-risk for AN, rather than currently unaffected individuals who later went on to develop AN. This type of powerful longitudinal study has not been done in AN or in BDD due to practical barriers, although it would be extremely valuable.

Replication of the current studies with larger sample sizes is also needed to understand which measures (neuropsychological assessments, behavioral measures, MRI measures etc.) are most reliable and informative for assessing visual processing in these populations. Multi-site studies are likely necessary to obtain large enough samples, due to difficulty in recruitment of these populations. Finally, it will be important to perform longitudinal studies to determine whether these putative phenotypes predict course of illness and response to treatment.