To understand the mechanism of cognitive control dysfunction in schizophrenia, it is critical to characterize brain function without the confounding effect of medication. It is also important to establish the extent to which antipsychotic medication restores brain function and whether those changes are related to psychosis improvement. Twenty-two patients with schizophrenia, initially unmedicated and after a 6-week course of risperidone, and 20 healthy controls (HC) studied twice, 6 weeks apart, performed an fMRI task. We examined group and longitudinal differences in anterior cingulate cortex (ACC), striatum, and midbrain functional activity during performance of a Stroop color task as well as activity patterns associated with improvement in psychosis symptoms. Unmedicated patients showed reduced functional activity in the ACC, striatum, and midbrain compared to HC. Post hoc contrasts from significant group-by-time interactions indicated that, in patients, drug administration was associated with both activity increases and decreases. In unmedicated patients, greater baseline functional activity in the striatum and midbrain predicted subsequent better treatment response. Greater changes in functional activity in ACC and ventral putamen over the course of 6 weeks positively correlated with better treatment response. Unmedicated patients show reduced activity in brain networks pivotal for cognitive control and medication is associated with functional changes in these regions. These results suggest a mechanism by which antipsychotic medication has a beneficial effect on cognition. Our results also support the notion that treatment response is determined by a combination of the baseline pattern of brain function and by the pharmacological modulation of these regions.