Using visual half-field presentations of words to the right (RVF) and to the left visual field (LVF), this study investigated the time course of the hemispheric involvement in the processing of semantic category information. Multi-channel event related brain potentials (ERPs) were recorded from 15 healthy subjects during a categorisation task of sequentially presented word pairs. Subjects had to judge mentally after the appearance of the second word whether the words of a pair were semantically related (SR) or not (SU). ERPs were computed, from 100 ms before the onset of the second word to 600 ms, for SR and SU conditions in the LVF and in the RVF separately. The temporal segmentation of ERP map series into sequences of quasi-stable map configurations revealed a total of seven segments in each visual field of which only the first five (S1-S5, appearing between 70 and 400 ms) showed different map configurations as a function of visual field but presented a similar temporal sequence in both visual fields. By contrast, of the last two segments (S6 and S7) which appeared between approximately 400 and approximately 600 ms, only S7 differentiated SR and SU conditions in terms of its duration. Source localisation analysis of the segments showed that following the initial activation of posterior brain regions as a function of the visual field of presentation, a common neural network was activated in the left hemisphere (LH) although the dynamics of activation varied as a function of visual field. Concerning the role of the right hemisphere (RH) in lexico-semantic processing, the results presented here appear to be compatible with a 'callosal relay model' and suggest that, in healthy subjects, information is transferred rapidly ( approximately 150 ms) from the RH to the language dominant-LH.