We have investigated the effects of selective spatial attention on early and middle-latency SEPs. Baseline control responses to electrical stimulation of 2 digits of the hand were recorded first in conditions of mental relaxation, in the absence of any cognitive task, to obtain truly 'neutral' responses uncontaminated by cognitive components. Then, during a 'task condition,' identical stimuli were applied to the same two fingers, but the subject's attention was driven towards the stimulated territory by the bias of mechanical taps delivered to the same digits. The earliest effect of directing attention towards the territory stimulated was a positive shift on contralateral somatosensory responses, with onset at 27.4 +/- 4 msec post stimulus. This SEP modification: (a) did not entail any change in the scalp distribution of components, as assessed by topographic mapping, and (b) was not present when attention was directed towards the hand contralateral to that receiving electrical stimuli. A second effect was represented by a parieto-central negativity in the 60-80 msec latency range; this feature could also be observed during contralaterally driven attention and was associated with topographical changes in SEP scalp distribution. Finally, a late centro-frontal negativity beginning at 90-100 msec (N140) appeared during ipsilateral attention, while P100 was not enhanced. Subcortical P14 and primary cortical N20 were not significantly affected by the tasks. We conclude that the 'early positive shift' is linked to the spatial aspects of selective attention and represents in part modulation of obligatory components (P25 through P45) existing in control SEPs; it probably corresponds with the deflections with similar polarity and time-course that have been described by others in response to somatosensory target stimuli. Conversely, 60-80 msec negative enhancement is less spatially selective and may represent non-specific arousal effects. The late negative component (N140) shares several features with the 'processing negativity' described in auditory paradigms and could represent the equivalent of this effect in the somatosensory system.