We performed an fMRI one-back recognition study aimed at distinguishing the semantic versus perceptual aspects of how objects and their written forms are processed. There were three types of visually presented items: pictures (schematic drawings of objects); words identifying these objects; and a mixed condition in which pictures were interleaved with words. A semantic decision about object identity was required when pictures were interleaved with words. This condition, contrasted with the other two, invoked a larger signal in multiple areas, including frontal cortex, bilateral occipitotemporal cortex, and the right middle temporal gyrus. We propose that the left occipitotemporal and right temporal activations are indicative of the neural substrate mediating picture-word conversions, whereas the frontal activations reflect the coordinating functions of the central executive.