Guilt is thought to maintain social harmony by motivating reparation. This study compared two methodologies commonly used to identify the neural correlates of guilt. The first, imagined guilt, requires participants to read hypothetical scenarios and then imagine themselves as the protagonist. The second, recollected guilt, requires participants to reflect on times they personally experienced guilt. In the fMRI scanner, participants were presented with guilt/neutral memories and guilt/neutral hypothetical scenarios. Contrasts confirmed a priori predictions that guilt memories, relative to guilt scenarios, were associated with significantly greater activity in regions associated with affect [anterior cingulate cortex (ACC), Caudate, Insula, orbital frontal cortex (OFC)] and social cognition [temporal pole (TP), precuneus). Similarly, results indicated that guilt memories, relative to neutral memories, were also associated with greater activity in affective (ACC, amygdala, Insula, OFC) and social cognition (mPFC, TP, precuneus, temporo-parietal junction) regions. There were no significant differences between guilt hypothetical scenarios and neutral hypothetical scenarios in either affective or social cognition regions. The importance of distinguishing between different guilt inductions inside the scanner is discussed. We offer explanations of our results and discuss ideas for future research.