The neuropeptide oxytocin (OT) plays a critical role in modulating social behavior across a wide range of vertebrate species. In humans, intranasal oxytocin (INOT) has been shown to modulate various aspects of social behavior, such as empathy, trust, in-group preference, and memory of socially relevant cues. Most INOT studies employ cross-sectional designs despite the enhanced statistical power and reduction in error variance associated with individual differences characteristic of within-subject designs. Using the Prisoner Dilemma task, which models a real-life dyadic social interaction, our group has systematically explored the effect of INOT on social cooperation and non-cooperation using a cross-sectional design. In the current study, we investigated if the main findings from our cross-sectional study could be replicated in a within-subject design using the same paradigm and whether new findings would emerge. We found OT to attenuate the ventral tegmental area response to reciprocated cooperation in women, an effect that is also present in our cross-sectional sample. However, other cross-sectional findings, especially those found in men, were not observed in this within-subject study. We hypothesize that the discrepancy can be explained by differing OT effects based on the degree of stimulus novelty/familiarity. Our within-subject study also revealed new effects not found previously in our cross-sectional study. Most importantly, OT treatment on scan 2 blocked amygdala habituation to unreciprocated cooperation found in a group that received placebo on both scans among men. Our results suggest that exogenous OT reduces the salience of positive social interactions among women and prevents habituation to negative social interactions among men. These findings may have implications for the potential clinical utility of OT as a treatment for psychiatric disorders.