We examined, with event-related fMRI, two hypotheses about the organization of human working memory function in frontal cortex: (1) that a region immediately anterior to the frontal eye fields (FEF) (superior frontal cortex, SFC) is specialized for spatial working memory (Courtney, et al., 1998); and (2) that dorsolateral prefrontal cortex (PFC) plays a privileged role in the manipulation of spatial stimuli held in working memory (Owen, et al., 1996; Petrides 1994). Our delayed-response task featured 2-D arrays of irregularly arranged squares that were highlighted serially in a random sequence. The Forward Memory condition required maintenance of the spatio-temporal sequence, the Manipulate Memory condition required reordering this sequence into a new spatially defined order, the Guided Saccade condition required saccades to highlighted squares in the array, but no memory, and the Free Saccade condition required self-paced, horizontal saccades. The comparison of fMRI signal intensity associated with 2-D saccade generation (Guided Saccades) versus fMRI signal intensity associated with the delay period of the working memorials condition revealed no evidence for greater working memory-related activity than saccade-related activity in SFC in any individual subject, nor at the level of the group, and greater 2-D saccade than delay-period activity in three of five subjects. These results fail to support the hypothesis that spatial working memory-related activity is represented preferentially in a region of SFC anterior to the FEF (Courtney, et al., 1998). The comparison of maintenance versus manipulation of spatio-temporal information in working memory revealed significantly greater activity associated with the latter in dorsolateral PFC, but not in ventrolateral PFC or in SFC. These results suggest that the delay-related function of SFC is limited to the maintenance of spatial information, and that this region does not support the nonmnemonic executive control functions supported by dorsolateral PFC. These results also indicate that the preferential recruitment of dorsolateral PFC for the manipulation of information held in working memory applies to tasks employing spatial stimuli, as well as to tasks employing verbal stimuli (D'Esposito, et al., 1999); Petrides et al., 1993; Postle et al., 1999).