Individuals with HFA failed to recruit neural circuitry typically engaged during the performance of an EF task designed to differentiate shifts in behavioral response from shifts in cognitive set. Specifically, we found that during response shift trials, control participants recruited a neural system comprising the DLPFC, ACC, premotor cortex, IPS, BG, thalamus, cerebellum, and VLPFC/anterior insula. Task-related activation in the autism group, however, was limited to the VLPFC (). Direct statistical comparison of BOLD activation between control and HFA participants showed greater activation to target trials for the control group in DLPFC, BG, and IPS (). This pattern of reduced activation in the autism group in frontal, striatal, and parietal regions was observed when target stimuli were subdivided into target-shift and target-maintain trials (), indicating that these functional deficits occurred for all trials requiring a response shift and not only during runs requiring a shift in cognitive set. Performance data indicated that HFA participants were less accurate than control participants in responding to target trials. Although performance on the task differed between groups, task performance was not correlated with brain activation in any a priori ROI. Therefore, these activation differences cannot be accounted for solely by performance differences; rather, the differences likely reflect neural dysfunction in HFA. We suggest that the reduced activation among HFA participants signifies a global neural processing deficit related to the disorder and might indicate an ineffective neural strategy for altering ongoing behavioral responses.
The current findings are important, given the empirical evidence suggesting that cognitive flexibility and other EF deficits are apparent in autism (5
) and that some EF deficits might be more severe in autism than in other childhood disorders (7
). These findings complement and extend recent imaging studies of the neural substrates of EF deficits in autism, such as reports of reduced PFC activation during motor sequence learning (34
), auditory target detection (44
), and response inhibition (36
). The task used in the current study isolated two components of EF often measured through the WCST (i.e., the shifting of behavioral responses and cognitive sets) (28
). Because the WCST is complicated by its vague instructions and complex design, we sought to simplify the behaviors associated with this task and make it easier for participants to understand the task instructions. Despite the simplification, our task measures at least three properties of EF and cognitive flexibility tested in the WCST, specifically the maintenance of a stimulus-response rule, the alteration of ongoing prepotent responses, and the periodic shifting of stimulus-response rules. Therefore, our findings are consistent with current views of executive dysfunction in autism as driven primarily by deficits in cognitive flexibility (17
) and the ability to inhibit prepotent ongoing responses (5
An intriguing finding in the current study was that activation to target stimuli within ACC and IPS declined as a function of RRB severity as measured by the ADI-R (). This negative relationship between RRB scores and BOLD activation in brain regions associated with executive and attentional processes (45
) might indicate that repetitive behaviors in autism are driven, at least in part, by more general deficits in attention, target detection, and response preparation (48
). Because the correlation results were part of an exploratory analysis, they should be considered preliminary; further investigation into the possible relations between neural activation during cognitive tasks and RRB severity in autism is necessary to draw more definitive conclusions. Despite their preliminary nature, these findings, together with the hypoactivation observed in the autism group, fit well within models suggesting that the restricted and repetitive behaviors commonly observed in autism might be related to deficits in certain domains of EF, such as cognitive flexibility and response inhibition (6
The results of the current study indicate that deficits in cognitive flexibility observed in the HFA group were limited to shifts of behavioral response, with cognitive set shifting ability unaffected by the disorder, at least when instructions to shift cognitive set are overt. We predicted that HFA participants would have the greatest difficulty with target-shift trials, because these trials require shifts in both behavioral response and cognitive set. To successfully respond to these trials, participants were required to alter the stimulus-response associations governing the previous run. For example, if the previous run required shifting responses for triangles but not for circles, the participant must now shift responses for circles and continue responding normally for triangles. Therefore, during shift runs, participants were required to generate shifts in responses for target stimuli in addition to creating shifts in cognitive set. Because there were no performance differences between the target-shift and target-maintain trials for HFA participants, we suggest that HFA individuals demonstrate an intact ability to shift cognitive set when the cognitive shift rule is overtly displayed.
Performance data for non-target novel trials provide corroborating evidence for an intact cognitive set shifting ability during our task in HFA individuals. These data indicated that accuracy for novel trials within the HFA group was not significantly different from the control group, even for novel trials during shift runs. If overt cognitive shifts were affected in autism, we would have observed poor performance in the HFA group during novel-shift trials in addition to target-shift trials. Although HFA individuals might be prone to fall into repetitive behavior patterns and thus would be expected to have the greatest difficulty with target-shift and novel-shift trials, this behavioral pattern was not observed in our data. Therefore, we conclude that the participants with autism did not have difficulty understanding that the shape previously identified as the target was now a non-target novel stimulus.
Our findings of reduced activation in multiple brain regions during set shifting are consistent with at least three other reports of hypoactivation in frontal regions during EF-related tasks (34
). However, Schmitz et al.
) reported significantly increased brain activation in multiple areas during a motor inhibition task, a cognitive interference task, and a set shifting task. Most relevant in the present context is the hyperactivation in the HFA group in response to set shifting reported by this research group. It should be noted that the set shifting task employed by this group required a shift of both behavioral response and cognitive set on all switch trials, whereas the current paradigm differentiated behavioral and cognitive set shifting. Nonetheless, even on trials requiring both behavioral and cognitive set shifts, we observed hypo- rather than hyper-activation in the HFA group. Moreover, our current finding of selective set shifting deficits in autism might explain why these behaviors seemed to be preserved in HFA individuals, as reported by the other research group (35
). Further imaging studies using paradigms that tap specific components of EF (e.g., by isolating the inhibition of prepotent responses from the execution of appropriate responses) are necessary to shed light on these contradictory findings.
In summary, our findings demonstrate neural hypoactivation during an EF task in HFA. Differences in brain activation between HFA and control participants were apparent both for between-task and for between-group comparisons. The most striking differences included an overall hypoactivation to target stimuli; lack of target-novel activation differences; and reduced activation in frontal, striatal, and parietal regions during target trials in the autism group. A limitation of the current study is the inclusion of only individuals with HFA, owing to the task demands and the confining nature of a functional imaging study. This restriction might limit the generalizability of our findings to individuals with more severe symptoms. In addition, the resulting restricted range of symptom severity might have limited our ability to detect correlations between BOLD activation, RRB scores, and task performance.
A second limitation pertains to the method for fMRI data analysis. We included all trials in the data analysis, regardless of whether those trials were correct or incorrect. Excluding incorrect trials from our analysis would have led to insufficient power to detect accurate BOLD signal changes, particularly for the HFA group. We acknowledge, however, that our results might have been influenced by including all trials in the analysis. Specifically, reduced neural activity both in frontal and in striatal regions in the HFA group could reflect reduced response inhibition related activity or reduced error related activity. Future studies should attempt to clarify this issue. Despite these limitations, our findings are consistent with current models of executive dysfunction in autism and provide novel insights into the putative neural mechanisms underlying EF deficits and stereotyped, repetitive behaviors often observed in the disorder.