| Functional magnetic resonance imaging (fMRI) approaches tend to focus on a particular process of interest, and model neural activity as a single phase of activity on a given trial. This thesis presents new methods for separating rapidly unfolding within-trial activities in a stop signal task using fMRI.;Current models of inhibitory control address processes that are engaged after the presentation of go- and stop-signals. The first study presented here separated warning- from response-phase activities in healthy adults (A. Chevrier, Cheyne, Graham, & Schachar, 2015). This work revealed activities in posterior networks that signal the need for top-down control followed by fronto-posterior activities involved in implementing top-down control. Both of these phases contained activities that influenced inhibitory control.;Reinforcement learning models treat errors as a unitary phase of activity. The second study presented here separated error detection from post-error slowing on failed stop trials (A. Chevrier & Schachar, 2010). Error detection deactivated structures that are richly innervated by midbrain dopamine neurons (dorsal striatum and ventral anterior cingulate cortex (ACC)), followed by deactivations during post-error adjustment in structures that modulate the ascending dopamine response (ventral striatum and caudal orbitofrontal cortex), which encode error magnitude.;The third study involved the separation of error detection and post-error slowing in adolescents with and without attention deficit hyperactivity disorder (ADHD) (Chevrier, Bhaijiwala, Cheyne, Graham & Schachar, manuscript in preparation). A recent study of these data from our laboratory (Bhaijiwala, Chevrier, & Schachar, 2014) showed that previous evidence of ADHD subjects under-activating inhibition networks, is instead the result of deactivating these networks during preparation, when healthy controls were pre-activating. This suggests that inhibitory control deficits in ADHD result from dysfunctional reinforcement, which tunes appropriate networks in an inappropriate way. Consistent with this hypothesis, we found that ADHD subjects failed to fully deactivate dopamine targets on error detection, and inappropriately activated the reciprocal pathway on post-error slowing. Dysfunctional reinforcement would explain the lack of potency of behavioral interventions that are based on the known properties of normal reinforcement. Further, the effects of dysfunction in moment-to-moment reinforcement would accumulate over experience, suggesting a mechanism for the delayed maturation of function and cortical thickness observed in ADHD. |
