This is the first study comparing brain function during reversal learning in SMD, BD, and HV. We examined diagnosis-by-phase-by-accuracy, diagnosis-by-phase, and diagnosis-by-accuracy interactions. In the ROI analyses, only the diagnosis-by-accuracy interaction resulted in significant clusters; subsequent analyses demonstrated that the most salient neural differences between SMD, BD, and HV were in response to errors, regardless of phase. To elucidate these findings, the difference in activation during incorrect vs. correct trials was calculated. In caudate, this value was smaller in SMD and BD than in HV. In IFG, however, this value was smaller in SMD than in both BD and HV. Post-hoc analyses indicated that comorbid ADHD might influence these findings, particularly in caudate. Whole-brain analysis confirmed the IFG and caudate findings.
Behaviorally, SMD made more errors across the entire task than did the other two groups; their deficit was not limited to the reversal phase. Some out-of-scanner work in larger samples demonstrates specific reversal-learning deficits in both SMD and BD9–11
; thus, our study resembles others that, across a variety of paradigms, find intact in-scanner performance despite in-clinic deficits56,61
. Taken together, current and prior data suggest that behavioral deficits observed in the scanner in SMD and in out-of-scanner testing in SMD and BD9,11
reflect deficient engagement of IFG in SMD and of caudate in both SMD and BD following errors. Neither SMD nor BD exhibited the normative increase in caudate activity to incorrect trials, suggesting both have difficulty learning from errors, an important caudate function46
that could reflect dopamine dysfunction. Dopamine signaling modulates reward-based38
learning; such learning deficits can result in perseveration63
Unlike caudate dysfunction, which manifested in both groups, SMD, but not BD, failed to show the expected, normal increase in IFG activity during incorrect trials. Thus, while caudate dysfunction characterized both SMD and BD, frontal dysfunction was unique to SMD. Frontal projections modulate striatal activity, so IFG and caudate are part of a circuit that adjusts behavior following an error28,64
. IFG also mediates functions necessary for response selection, including attention maintenance and representation of contingencies, context, and goals. Right IFG plays an important role in response inhibition44
, and a recent study indicates that IFG activity is modulated by the response control demands of a motor task65
Speculatively, chronic SMD symptomatology may result from persistent fronto-striatal dysfunction, while episodic impairment in BD may reflect intermittent prefrontal dysfunction. SMD exhibited IFG and caudate dysfunction, as well as fewer correct trials than the other groups. In BD, on the other hand, IFG response and task performance are more similar to HV; perhaps intact IFG function enables BD subjects to compensate for basal ganglia dysfunction. Most patients were euthymic when scanned; while IFG dysfunction is amongst the most consistently reported finding in BD, a recent metaanalysis suggested such IFG dysfunction may be present during mania but not euthymia or depression66
. Future longitudinal studies of BD patients in different mood states are required to test this possibility.
Our results are consistent with prior work. Using the stop signal task in a partially overlapping sample, we found that, relative to controls, BD had hypoactivation in IFG and striatum during unsuccessful attempts to inhibit motor responses47
(i.e., in response to error). Thus, using first a motor inhibition task in BD47
and then a response reversal task in BD and SMD, we identified neural dysfunction that may compromise the ability of both groups to adapt their behavior in response to changing contingencies, causing an increased propensity to experience frustration and irritability67
Post-hoc analyses suggest that comorbid ADHD may contribute to our caudate findings: patients with ADHD differed from those without ADHD in response to incorrect vs. correct trials, and patients without ADHD more closely resembled controls. In IFG, patients with ADHD did not differ from those without ADHD in response to incorrect vs. correct trials, rendering it less likely that the between-group differences we observed were secondary to ADHD. Our caudate results here are consistent with our previous stop signal findings, where we could not rule out the role of comorbid ADHD in the aberrant neural responses of BD patients during unsuccessful inhibition47
. A study using the same response reversal paradigm as here and a different analytic strategy found no neural deficits in ADHD patients56
, but other data suggest abnormal activation in ADHD during related cognitive tasks (e.g., behavioral deficits and aberrant IFG and caudate activity in ADHD during response inhibition68–70
The role of ADHD in our findings is difficult to ascertain because ADHD is present in 77% of the SMD sample; SMD inclusion criteria require three “hyperarousal” symptoms common to ADHD and mania. However, it is unclear whether the pathophysiology of ADHD in the context of SMD or BD is the same as that of ADHD without irritability. Studies report different neural activity5
and neurological symptoms71
in SMD or BD (including those with comorbid ADHD) vs. ADHD alone, again suggesting the importance of using behavioral and neuroimaging data, in addition to symptoms, to differentiate syndromes. Importantly, the current study was not designed to ascertain explicitly the impact of ADHD on the pathophysiology of SMD or BD, since to do so one would need to compare patients with “pure” ADHD without comorbid irritability to patients with SMD, BD, and HV5
A major strength of this study is the comparison of two clinical populations to each other and a healthy group. Most psychiatric imaging studies compare one clinical population to a healthy group, and thus cannot differentiate disease-unique and disease-common abnormalities in diagnoses with overlapping symptoms. The extent to which SMD and BD children differ from controls but resemble each other indicates overlapping disease substrates. Here, both disorders exhibit striatal dysfunction. On the other hand, differences between clinical groups may indicate disorder-specific abnormalities with important diagnostic and treatment implications; here, we found differences between patient groups in prefrontal cortex function. However, there may be a dimensional component to IFG dysfunction in this task: like SMD, BD may not increase IFG response to incorrect trials as much as HV; the comparison between BD and HV is at a trend level and therefore equivocal. Nonetheless, direct comparison between SMD and BD indicates a categorical between-group difference in IFG activity.
Exploratory whole-brain analyses confirmed the IFG and caudate ROI findings and identified dysfunction in other regions. Like the ROI results, whole-brain results suggest dysfunction in SMD in regions that mediate detecting and learning from errors and executing an alternative response (e.g. superior/medial frontal gyrus72,73
, right IFG, caudate). Also similar to the ROI analyses, the whole-brain analysis found that dysfunction was more consistent and pervasive in SMD than BD. Comparing BD and HV, our results are similar to Dickstein et al., who used a partially overlapping sample12
. Specifically, in BD vs. HV, both studies found parietal hyperactivation during incorrect reversal trials and hyperactivation to all incorrect trials in superior frontal gyrus.
The study has limitations. First, though larger than those in most pediatric fMRI studies, the sample sizes used here are small. Second, we did not obtain frustration ratings in the scanner, and therefore cannot correlate activation and affective response. Adding such measures would have limited comparability with other studies and altered the psychological processes engaged, perhaps activating top-down regulatory regions. Studies suggest that caudate and IFG mediate switching responses after negative feedback28,64
, but we cannot rule out the possibility that the between-group differences we observed are associated with psychological processes not measured by the task, such as increased frustration in response to negative feedback or decreased motivation in patients vs. controls. Also, because we lack dimensional symptom measures for comorbid disorders such as ADHD and ODD, we cannot correlate such symptoms with activation. Finally, many patients were medicated; ethical concerns preclude withdrawing ill children from medication for research. However, data suggest that medication may increase noise, rather than bias towards false positive errors74
. While not all subjects were euthymic, most were, and post-hoc analyses suggest that mood state does not account for our findings (see supplement 1, avaliable online
). Although post-hoc analyses suggest that the higher ADHD comorbidity, higher error rates, and slightly younger age of the SMD group are not driving between-group differences, these three factors may have interacted to influence the results.
This study is the first to examine the neural underpinnings of response reversal in children with SMD and BD compared to controls, finding deficits in both groups in response to errors. Hypoactivation during incorrect responses occurs in caudate in both SMD and BD and in IFG in SMD. Such hypoactivation to errors may reflect deficits in response inhibition signaling or new response selection, which may result in increased frustration and irritability. Future work should elucidate the role of comorbid ADHD, specifically in the caudate, in error-processing deficits in SMD and BD.