Error detection in the stop signal task

Abstract

Previous error detection research has focused on error processing functions in the anterior cingulate cortex or on putative reinforcement learning roles of midbrain dopamine pathways. We studied error detection in 14 healthy adult volunteers using a novel fMRI design in the stop signal task (SST), a task which invokes numerous errors in performance and frequent instances of post-error slowing. The imaging design accommodated variability immediately before errors (handedness of response) and immediately after (degree of post-error slowing) using distinct within-trial regressors. This approach revealed a whole-brain view of error detection in a reinforcement-learning pathway. Error detection deactivated the midbrain in the vicinity of dorsal substantia nigra where dopamine neurons originate, and the primary targets of dopamine neurons: dorsal striatum and ventral anterior cingulate. Error detection also deactivated posterior hippocampus, which is highly sensitive to long-term synaptic plasticity effects of dopamine. Errors that led to slowed responses deactivated structures in the reciprocal pathway that are known to exert control over dopamine output, and which have been shown to encode error magnitude: ventral midbrain, ventral striatum, and caudal orbitofrontal cortex. Consistent with the role of these structures in modulating dopamine output, post-error slowing also increased activities in the same structures that deactivated on error detection. These results are consistent with the view that errors deactivate structures that receive input from dopamine neurons, followed by deactivations related to requisite behavioral adjustments in structures that exert control over dopamine output.