This experiment demonstrated that individuals with BDD show less of an inversion effect than healthy controls for long duration (5000 msec) but not short duration (500 msec) stimuli. This difference appeared to be explained by 8.5% faster response times in BDD subjects for the long duration inverted faces stimuli. These findings confirmed our hypothesis with respect to response time but not for accuracy rates, which were not significantly different between groups.
Reduced face inversion effect in individuals with BDD relative to healthy controls for the long duration stimuli may be due to a general propensity to engage in highly detailed and piecemeal processing. In healthy controls, processing of upright faces is believed to occur holistically while presentation of inverted faces engages slower processing of details and components, due to the lack of a holistic “template” for inverted faces (Freire et al., 2000
). If individuals with BDD encode details more rapidly than controls whether upright or inverted, this would result in less of a difference in response times for the different orientations.
The interpretation of greater detail-oriented processing in BDD is supported by a previous functional magnetic resonance imaging (fMRI) study that examined brain activation patterns in response to viewing others’ faces (Feusner, Townsend, Bystritsky, & Bookheimer, 2007
). The experiment utilized normal, high, and low spatial frequency photographs of other peoples’ faces in order to isolate specific neural pathways responsible for different spatial frequency domains. Individuals with BDD demonstrated abnormal left hemisphere hyperactivity in an extended face-processing network including temporal, parietal and inferior frontal gyrus regions. This occurred even for low spatial frequency images, which normally are processed by the right hemisphere (Hellige, 1996
). Predominant left hemisphere activity suggests greater detail encoding and analysis relative to configural processing for faces. Clinically, while individuals with BDD may perceive details accurately (and perhaps more accurately than healthy controls (Stangier, Adam-Schwebe, Muller, & Wolter, 2008
)), the perceptual distortions may come from a failure to integrate these details into a visual gestalt.
The reduced face inversion effect on response times for BDD subjects in the current study occurred only for long but not short duration stimuli. The fact that there were minimal and nonsignificant differences in inversion effect on response time for short duration stimuli suggest that BDD subjects and healthy controls may equally engage processing of holistic and configural information, conveyed by low spatial frequency information, when given a brief presentation. Long duration stimuli, on the other hand, may have allowed time for encoding of details, conveyed by high spatial frequency information. If the BDD group has a greater propensity for processing high spatial frequency information (and/or lesser low spatial frequency processing) this would confer an advantage for inverted faces, which require detail processing (Farah et al., 1995
). An alternative explanation is that BDD subjects may not have relied on normal holistic/configural processing during the short duration period but instead employed abnormally rapid detail processing, although it is unlikely given the 500 msec timeframe.
Another (possibly related) explanation for differences in inversion effect for long but not short durations is that longer times allow for multiple eye fixations. Only the longer duration stimuli would therefore have allowed sufficient time for visual scanning and subsequent detail encoding. If the BDD group has a greater propensity for visual scanning and subsequent detail encoding, this would confer an advantage for the inverted faces. Short durations stimuli did not allow sufficient time for multiple eye fixations, which may have resulted in normal holistic/configural processing in the BDD group and hence no significant differences from controls. Future studies with eye-tracking would be helpful to determine if this is occurring.
Perceived saliency of facial features may have driven enhanced detail processing in the BDD group and therefore differences in inversion effects. Barton et al. (2001)
found that salient features (especially the mouth, for healthy controls) were less susceptible to inversion effects than non-salient features (Barton et al., 2001
). Individuals with BDD may find multiple features and details of others’ faces salient (e.g. skin blemishes, lines, irregularities of the contour of the nose) because when viewing others they typically engage in comparisons to their own appearance, of which they typically have multiple concerns (Phillips, 2005
). Greater saliency of multiple facial features would then invoke greater part decomposition and detail encoding in the BDD group in this study, consistent with the previous explanations. Only longer duration stimuli may have allowed time for this, resulting in differences in inversion effects between groups.
To our knowledge no previous study has specifically tested the face inversion effect in BDD. However, several studies have investigated face identity recognition of upright faces. Using slightly different versions of face-matching tasks, these studies have demonstrated similar findings as the current study; individuals with BDD do not appear to differ greatly from healthy controls in terms of response time or accuracy with respect to behavioral performance on upright faces (Buhlmann et al., 2004
; Feusner, Bystritsky, Hellemann, & Bookheimer, in press
; Feusner et al., 2007
). However, because these studies all used stimulus durations of several seconds, it is possible that the BDD group may have utilized rapid local processing while controls may have relied primarily on global processing for the same task.
No other studies, to our knowledge, have investigated inverted face processing in disorders related to BDD such as obsessive-compulsive disorder, social phobia, or eating disorders. However, several studies have tested the face inversion effect in individuals with autistic spectrum disorders. Some of these (Bookheimer, Wang, Scott, Sigman, & Dapretto, 2008
; Hobson, Ouston, & Lee, 1988
; Langdell, 1978
; Rose et al., 2007
; Tantam, Monaghan, Nicholson, & Stirling, 1989
) but not all (Teunisse & de Gelder, 2003
) have demonstrated lesser inversion effect relative to healthy controls, attributed to enhanced featural processing and a local bias and/or lesser contextual processing.
Interestingly, children also show reduced face inversion effect, relative to adults, which similarly appears to be associated with a more piecemeal strategy (Aylward et al., 2005
; Joseph et al., 2006
). This raises the possibility that the abnormalities in visual processing of faces in BDD seen in this study may stem from aberrant neural development.
With respect to accuracy in the current study, both groups demonstrated an inversion effect. However, our hypothesis that there would be less of an inversion effect for the 5000 msec but not the 500 msec condition was not confirmed. A possible ceiling effect for the long duration stimuli may limit our ability to interpret this finding. There may have also been a tradeoff between accuracy and response time, as both were lower across groups for short duration stimuli.
Findings from this study may have clinical implications. BDD patients typically perceive defects of their own appearance, often a facial feature, which are not noticeable or appear minor to others. They also frequently check appearance features of others to compare to their own (Phillips, 2005
). The findings from this study add to converging (although preliminary) evidence suggesting that disordered visual information processing may not be limited to their own face, but may extend to others’ faces as well (Feusner et al., in press
; Feusner et al., 2010
; Feusner et al., 2007
; Yaryura-Tobias et al., 2002
). This suggests a more general visual processing diathesis, further supported by abnormal visuospatial performance for complex figures (Deckersbach et al., 2000
). However, it is still not clear if this represents an underlying trait that predisposes individuals to BDD, or a result of the illness itself. In these regards, if the reduced face inversion effect for long duration stimuli found in this study reflects faster detail processing than healthy controls, the question remains if this represents an underlying endophenotype that predisposes individuals to BDD, or if it is a result of habitually engaging in selective attention to details, driven by appearance concerns.
In this study, the random effects statistical models incorporate baseline heterogeneity in subjects’ tendencies to respond slower or faster than the average trend. The small sample size limited our ability to obtain precise estimates of both a more complex covariance structure and the fixed effects simultaneously. However, in exploratory models we did detect evidence of a random learning effect during the inverted 5000 msec period and heteroskedastic serial correlation over the four experimental conditions. This random learning effect occurred during what appeared to be the most difficult period for subjects, implying that there is additional variability in how quickly individuals adapt to the conditions beyond the average learning trend in this period. The serial correlation components suggest that an individual’s responses that are closer together in time are more correlated than those spaced further apart; this phenomenon is stronger in the two shorter-duration periods, when subjects have less time to process each face independently. The estimated fixed effects are virtually unchanged in this more complicated covariance model, but the standard errors are somewhat larger. We have chosen to present the parsimonious version in order to preserve power to test our hypotheses of interest, which we believe is more appropriate to the scope of the present experiment, but plan to pursue this issue in future, more highly-powered studies.
There are several limitations in this study that should be considered. The sample size was not sufficient to separately analyze individuals with comorbid depressive disorders, GAD, or both. Also, the sample size and the fact that there were a high number of covariates due to the four different experimental conditions and the learning effect limited our ability to investigate clinical variables such as anxiety and depression as covariates.
In sum, our data suggests that BDD subjects experience less of an inversion effect on response times compared to controls for long, but not short, stimulus durations. Although there are possible alternative explanations, this may provide evidence of a propensity to process faces in a piecemeal, detail-oriented manner, at least when viewed for longer durations. Clinically, abnormal face processing may explain perceptual distortions in the way that individuals with BDD view their own faces, as well as perhaps the faces of others. Further research involving testing of the inversion effect with other tools such as fMRI or event-related potentials may help to illuminate whether differences stem from encoding and/or recognition processes, or some other phenomenon. Additionally, eye-tracking techniques may provide useful analysis of what may behaviorally mediate abnormalities in facial recognition.