THE functional anatomy of the interactions between spoken language and visual mental imagery was investigated with PET in eight normal volunteers during a series of three conditions: listening to concrete word definitions and generating their mental images (CONC), listening to abstract word definitions (ABST) and silent REST. The CONC task specifically elicited activations of the bilateral inferior temporal gyri, of the left premotor and left prefrontal regions, while activations in the bilateral superior temporal gyri were smaller than during the ABST task, during which an additional activation of the anterior part of the right middle temporal gyrus was observed. No activation of the occipital areas was observed during the CONC task when compared either to the REST or to the ABST task. The present study demonstrates that a network including part of the bilateral ventral stream and the frontal working memory areas is recruited when mental imagery of concrete words is performed on the basis of continuous spoken language.