The aim of this study was to assess the cerebral blood flow (CBF) changes induced by the semantic incoherence of sentences whose words were to be monitored by the subjects. According to Marslen-Wilson and Tyler, word processing would be parallel (global) in semantically coherent sentences, and serial (local) in semantically incoherent ones. We hypothesized that, in comparison with the parallel mode, the serial mode of word processing would result in a preponderant activation of the frontal lobes and/or of the left hemisphere, hence in a flow increase in these regions. In addition, one could assume that the increase in task complexity resulting from the breaking of semantic coherence would reinforce the functional links between hemispheres. Furthermore, as interactions between the processing of semantically coherent vs incoherent of verbal materials and the processing of imageable vs non-imageable stimuli was recently suspected, we designed a study combining coherent vs incoherent sentences and imageable vs non-imageable words. The results we obtained in 12 normal volunteers by measuring regional cerebral blood flow by SPECT and IV injection of Xenon 133 during the four experimental conditions, did not allow us to firmly establish our hypotheses. Indeed, we failed to demonstrate any significant CBF changes across conditions and there was no interaction between coherent vs incoherent and imageable vs non-imageable conditions. However, the analysis of the inter-regional correlations pointed to an increase in the functional links between the hemispheres in the incoherent conditions, whatever the imageability.