Event-related potentials (ERPs) were recorded from 40 locations, covering most of the scalp, during repeated tasks in which the observer (O) had to judge either the tense of a printed verb (V) or the symmetry of a spatial pattern (S). Stimuli were drawn at random from large ensembles. A simplified method of Laplacean analysis (MacKay 1983, 1984) allowed the corresponding source densities to be mapped at up to 28 locations, relatively free of artefacts due to eye movements or tongue movements. O signalled his judgement in each case by pressing one of two buttons on a given cue. The decision time allowed was kept short (about 1 s) but long enough for the task to be handled successfully. When stimuli 'V' and 'S' were drawn from geometrically different ensembles, the source-density distributions for the two tasks differed significantly at a number of locations. When 'V' and 'S' were drawn from a common ensemble, however, and O was instructed on each trial (in random order) to assess each stimulus as a word or as a geometrical pattern, the similarities in the source-density maps were more striking than the differences. It would seem that during sufficiently rapid verbal and spatial judgments, little sign of hemispheric specialization or task-specific differences may appear in the spatiotemporal profile of ERP source densities. More salient differences, some lateralized, appeared during the preparation interval prior to verbal and spatial tasks; but their pattern varied widely from subject to subject.