Although we cannot reverse the flow of time, motion pictures technology allows us to render temporal events in reverse. Often, such reversals are immediately detectable: natural laws, from gravitational forces to the physiology of locomotion, imbue dynamic scenes with an easily detectable directionality. Other reversals, such as the order of events in a developing plot, may require analysis at longer time scales. What are the brain mechanisms involved in the analysis of temporal relationships, and do they differ for these different cases? This research is based on previous findings of voxel-wise correlations to repeated presentations of movies, which was effective for revealing brain areas that exhibit consistency in their response to external stimuli (Hasson et al. Science 2004). Here we presented observers with two repeated presentations of original (‘Forward’, F) and time-reversed (‘Backward’, B) movie clips while collecting whole-brain fMRI activity. There were three movie categories: ‘inanimate’ (e.g. collapsing buildings), ‘animate’ (e.g. people moving) and ‘plot’ clips taken from silent classic movies. Time-courses obtained from each Backward clip were reversed (rB), corrected for hemodynamic delay, and then correlated (‘C’) with those obtained from the Forward clip (CrB,F). Results were compared with those obtained by correlating two forward presentations (CF1,F2). Posterior temporal and occipital cortices exhibited comparable CrB,F and CF1,F2 maps, suggesting that activity in these areas depends primarily on the content of individual frames. Sensitivity to the arrow of time in the ‘inanimate’ and ‘animate’ clips was revealed in the intraparietal sulcus (IPS), where CrB,F values were significantly lower than CF1,F2 values. For the ‘plot’ clips, time-sensitivity was also revealed in the left planum temporale (Wernicke’s area). Our results reveal distinct cortical areas sensitive to different temporal grains, ranging from individual frames to discrete motion clips to longer plot-related events.