Compared to previous work, the present study is, to our knowledge, the first to simultaneously assess behavioural performance on an additional visual task (CT) that closely matches the EFT in all cognitive processes except the major local visual processing component itself. While our behavioural results do not support the notion that autistic patients have an absolute
advantage for local visual processing, they do suggest that autistic individuals have a relative
advantage for local processing. This was reflected by a task-by-group interaction (p
= 0.054) in terms of RTs. This interaction indicated that autistic patients were equally fast on the EFT as the control subjects (; note that error rates were also very similar), but showed slower responses on the CT where local processing requirements were minimal. How does this finding fit with previous behavioural studies of EFT performance by autistic and control subjects? In the classical study by Shah and Frith (1983)
as well as in the study by Jolliffe and Baron-Cohen (1997)
autistic children were found to respond both faster and more accurately than control children. This superior performance of the autistic group was consistent with the findings of Kagan and Kogan (1970)
who found an inverse relation between sensitivity to social cues and performance on the EFT in children. Similarly, Baron-Cohen and Hammer (1997)
reported that parents of children with Asperger syndrome were significantly faster on the EFT than a control group of parents with non-autistic children. Several other studies provided additional evidence for better performance of autistic patients in detecting embedded figures (de Jonge et al., 2006; Jarrold et al., 2005; Pellicano et al., 2006
). Two other studies, however, deviated from this pattern. Ropar and Mitchell (2001)
found that autistic and Asperger children performed equally well on the EFT as control children. A study by Brian and Bryson (1996)
also failed to find superior performance of autistic patients compared to controls. The latter study, however, is difficult to interpret because the groups differed in age.
The current study achieved a fairly good replication of our previous fMRI results of adult volunteers performing the EFT (Manjaly et al., 2003
). In that study contrasting EFT to CT revealed significant activations in the left posterior parietal cortex, including the IPS, and in the left ventral premotor cortex (posterior IFG). This finding is consistent with the well-established left-hemispheric dominance for local visual processing (see Introduction
). Using a probabilistic cytoarchitectonic atlas, the premotor activation was found to overlap with areas 44 and area 6. A subsequent connectivity analysis demonstrated that this activation functionally interacted with areas commonly involved in visuospatial processing (Manjaly et al., 2005
). In the present study, comparing EFT to CT in the control group activated the left IPS and the left dorsal premotor cortex. While the parietal activation reported here was very similar to the one found by Manjaly et al. (2003)
, the premotor activation was more dorsally located. This may be due to differences in local processing in adults and adolescents, or it may result from differences in the design of the two studies. For example, the stimuli used by the current study were less complex (eight instead of twelve composing lines) and were shown in horizontal orientation only.
Importantly, the patient group showed a different activation pattern when comparing EFT to CT. Here, we did not find a further increase in the left-lateralisation of activity in IPS and other “higher” areas as implied by the classical WCC theory (see Introduction
). Instead, activations were found in the cortex surrounding the right calcarine sulcus and in the extrastriate cortex bilaterally. These activations at the early stages of visual processing in autistic individuals are compatible with the hypotheses by Plaisted et al. and Mottron and Burack, respectively, that the advantage of autistic patients for local visual processing may be due to differences in basic perceptual processes. According to Plaisted et al., in the early sensory cortices of autistic patients the salience of individual stimulus features is boosted without compromising processes of global integration (Plaisted et al., 1998, 2003; O’Riordan and Plaisted, 2001
). Similarly, the EPF theory by Mottron and Burack holds that superior perceptual processing and enhanced “perceptual traces” in early visual cortex could induce better memory for local properties of visual stimuli (Mottron and Burack, 2001
). In a recent review of the evidence for this, Mottron et al. (2006)
suggested that enhanced V1 activation may be a general phenomenon for visual tasks in autism. Finally, we also found an activation in the right cerebellar hemisphere, extending into the vermis, which is neither predicted by the WCC theory nor by the other theories. Currently, we cannot offer a good explanation for this finding.
The direct group comparison between patients and controls for the EFT vs. CT contrast did not yield any brain regions with between-group differences in local processing that were significant at p < 0.05 corrected. However, at uncorrected levels of significance, some of the regions (e.g. the calcarine sulcus) which were significantly activated in the EFT vs. CT contrast in the autistic group, also showed a task × group interaction, i.e. higher EFT vs. CT differences in the autistic group compared to the control group (see ).
One limitation of our study is the restricted sample size. Even though we studied twice as many patients (12) than the only other EFT fMRI study in autism available so far (Ring et al., 1999
), this is still a limited number given recent evidence that there may be considerable variance with regard to EFT performance among patients with autistic spectrum disorder (Edgin and Pennington, 2005
). Ideally, one would subdivide the patients into subgroups with different performance levels on the EFT. With the current sample size, however, this was not feasible statistically. This approach is suggested for future studies with larger sample sizes, and the present findings should be treated as preliminary results until confirmed in larger samples.
Notwithstanding the above caveats, our present results, which point to differential information processing at early stages of visual perception in autism, are in line with evidence from a growing body of psychophysical, electrophysiological and fMRI studies (reviewed by Plaisted et al., 2003; Mottron et al., 2006
). For example, the psychophysical experiments by Bertone et al. (2005)
and Caron et al. (2006)
pointed to differences in processing orientation information and perceptual cohesiveness of visual stimuli, respectively, at the level of early visual areas in autistic patients. McPartland et al. (2004)
demonstrated an abnormal N170 event-related potential at early stages of face processing. Brown et al. (2005)
reported abnormal EEG gamma activity over visual cortex, possibly reflecting a diminished signal-to-noise ratio due to decreased inhibitory processing at the early stages of perception in autism. Koshino et al. (2005)
reported increased activity in extrastriate areas of autistic patients, compared to controls, during a visual N-back working memory task studied with fMRI. Belmonte and Yurgelun-Todd (2003)
applied fMRI to a simple visuospatial attention task and found activations in primary visual and ventral occipital cortex of autistic subjects, but not of controls.
Concerning local visual search in particular, so far only one previous study has examined the neural mechanisms underlying local visual processing in autistic patients (Ring et al., 1999
). This fMRI study investigated six autistic subjects and twelve controls, contrasting the EFT with an unspecific ‘fixation only’ condition. When comparing the EFT-related activations between groups, Ring et al. found that the only regions exhibiting higher activity in the autistic group were primary and secondary visual areas. The far more specific EFT vs. CT contrast in our study produced results for the autistic group that are compatible with the findings by Ring et al. (1999)
The present study included a carefully designed control condition (CT) for a task probing local visual processing (EFT). However, it should be noted that while this task controlled for all motor, cognitive, and perceptual processes (except local visual search) of the EFT, it did not directly probe the capacity for global integration. A next step therefore is to directly compare with fMRI local and global processing capacities in larger samples of autistic patients, e.g. using identical, hierarchically structured stimuli (Rinehart et al., 2000; Fink et al., 1997
). In addition, future studies should focus on the direct comparison between different childhood psychiatric disorders characterized by attentional problems in order to clarify how specific these cognitive profiles are for patients with autism.