We reviewed 107 blood flow activation studies carried out with positron emission tomography and published between January 1993 and November 1996. These studies had reported their findings as peaks of significant difference in cerebral blood-flow (CBF) between two scans/tasks and had located the peaks in standardized stereotaxic space. We coded each task along several dimensions, including the type and rate of input and output, the types of cognitive processes, and the relative difficulty of tasks within a study. Based on this coding, a difference score (A-B) was calculated for each subtraction. Subsequently, the frequency distributions of the difference scores for subtractions yielding a peak in the anterior cingulate region (cingulate peak) were compared with those distributions obtained from subtractions without a cingulate peak (no cingulate-peak). The cingulate peak subtractions (n = 158) differed from the no cingulate peak subtractions (n = 229) in terms of difficulty level (p = 0.001) and the presence of a remote memory component (p = 0.01). Regional differences in the frequency distribution of certain task parameters, such as difficulty level, recent memory and the use of the hand for responding, were also observed when peaks found in the anterior cingulate cortex (ACC) were further classified as located in the rostral vs caudal ACC, supracallosal vs subcallosal ACC, and limbic vs paralimbic parts of the supracallosal ACC. We conclude that task difficulty plays a major role in modulating blood-flow response in the ACC, possibly interacting with other parameters such as the nature of the response and memory demands.