In macaque ventral premotor cortex, we recorded the activity of neurons that responded to both visual and tactile stimuli. For these bimodal cells, the visual receptive field extended from the tactile receptive field into the adjacent space. Their tactile receptive fields were organized topographically, with the arms represented medially, the face represented in the middle, and the inside of the mouth represented laterally. For many neurons, both the visual and tactile responses were directionally selective, although many neurons also responded to stationary stimuli. In the awake monkeys, for 70% of bimodal neurons with a tactile response on the arm, the visual receptive field moved when the arm was moved. In contrast, for 0% the visual receptive field moved when the eye or head moved. Thus the visual receptive fields of most "arm + visual" cells were anchored to the arm, not to the eye or head. In the anesthetized monkey, the effect of arm position was similar. For 95% of bimodal neurons with a tactile response on the face, the visual receptive field moved as the head was rotated. In contrast, for 15% the visual receptive field moved with the eye and for 0% it moved with the arm. Thus the visual receptive fields of most "face + visual" cells were anchored to the head, not to the eye or arm. To construct a visual receptive field anchored to the arm, it is necessary to integrate the position of the arm, head, and eye. For arm + visual cells, the spontaneous activity, the magnitude of the visual response, and sometimes both were modulated by the position of the arm (37%), the head (75%), and the eye (58%). In contrast, to construct a visual receptive field that is anchored to the head, it is necessary to use the position of the eye, but not of the head or the arm. For face + visual cells, the spontaneous activity and/or response magnitude was modulated by the position of the eyes (88%), but not of the head or the arm (0%). Visual receptive fields anchored to the arm can encode stimulus location in "arm-centered" coordinates, and would be useful for guiding arm movements. Visual receptive fields anchored to the head can likewise encode stimuli in "head-centered" coordinates, useful for guiding head movements. Sixty-three percent of face + visual neurons responded during voluntary movements of the head. We suggest that "body-part-centered" coordinates provide a general solution to a problem of sensory-motor integration: sensory stimuli are located in a coordinate system anchored to a particular body part.