extrapersonal space (13–15). This control circuitry in turn determines which stimuli entering the visual pathways are to be facilitated or suppressed by virtue of anatomical projections to the ventral stream of extrastriate visual areas that encode stimulus features and objects (16, 17) and perhaps to dorsal stream areas that encode stimulus motion as well. In other words, this attentional control circuitry provides “bias signals” that either enhance or suppress sensory representations in the extrastriate visual pathways according to their momentary relevance (18, 19). In an alternative formulation, Crick (20) has proposed that the control circuitry for spatial attention may act via the reticular/perigeniculate nuclei of the thalamus to selectively modulate thalamic input to the cortex. If such a mechanism exerted control over the transmission of visual signals in the geniculo-striate projection, it would be expected that short-latency activity in primary visual cortex would be modulated during appropriate attentional tasks.
Mechanisms of Spatial Attention. ERP data have been very informative about the time course of visual processing in humans and its modulation by spatial attention. The visual ERP consists of several characteristic voltage deflections beginning about 50 ms after stimulus onset that have been labeled the C1 (50–90 ms), P1 (80–130 ms), and N1 (140–200 ms) components (Fig. 1). Directing attention to the location of a stimulus typically results in an amplitude enhancement of the P1 and multiple N1 components evoked by that stimulus with little or no change in component latencies or scalp distributions (reviewed in refs. 8, 21, and 22). This suggests that spatial attention exerts a gain control or selective amplification of sensory information flow in the visual pathways between 80 and 200 ms after stimulus onset (14, 23). Such an amplification mechanism would presumably give inputs from attended locations an improved signal/noise ratio so that more information can be extracted from relevant portions of the visual field.
This pattern of P1/N1 amplitude enhancement seems to be a general characteristic of the spatial focusing of attention
across a variety of task situations. Stimuli at attended locations elicit larger P1/N1 components than at unattended locations whether the stimuli are presented continuously in randomized sequences, as in Fig. 1, or cued on each trial as to the most probable location of the subsequent target stimulus. In such trial-by-trial cueing tasks, enhanced P1/N1 amplitudes to target stimuli at valid (precued) locations have been associated with speeded reaction times and improved detectability of target signals (22, 24–26). which lends support to the hypothesis that these ERP amplitude modulations reflect sensory information that is used for perceptual judgements. Similar P1/N1 modulations have been found in visual search tasks in which subjects had to deploy focal attention to identify the shape of a target defined by its color in an array (27, 28). In contrast, the earlier C1 component has been found to remain invariant as a function of spatial attention (29–32).
To investigate the anatomical level(s) of the visual pathways at which spatial attention affects processing, several studies have attempted to localize the respective neural generators of the C1 and P1 components. The C1 has a midline occipitoparietal scalp distribution that is well accounted for by a dipolar source in primary visual (striate) cortex (32). Moreover, the C1 inverts in polarity as a function of stimulus elevation in the visual field in a way that is consistent with the mapping of the retina onto the upper and lower banks of the calcarine cortex (29, 33). In contrast, the P1 component reportedly does not invert in polarity with stimulus position, and dipole modeling of its neural generators has indicated sources in the ventral-lateral extrastriate cortex of the occipital lobe (30, 32). Thus, spatial attention does not seem to influence visual processing in the striate cortex itself (indexed by C1) but rather acts to produce an amplification of stimulus-evoked activity in extrastriate cortex beginning at about 80 ms poststimulus (indexed by P1).
To verify the extrastriate localization of this early spatial attention effect, several recent studies have combined ERP recordings with PET obtained during performance of the same task and with the same subjects. In the first such experiment by Heinze and colleagues (7), subjects were required to direct their attention to the right or left half of bilateral symbol arrays that were flashed in rapid sequence, while maintaining central fixation (Fig. 2). Separate runs of attend-left, attend-right, and passive trials were carried out in two separate sessions, the first with ERP recordings from 30 scalp channels and the second with PET following intravenous injection of a positron emitting tracer (H2O15). In the ERP recording session, an enlarged P1 component was observed over the hemisphere contralateral to the attended visual field (Fig. 2A), in line with previous studies (34). Dipole modeling of the P1 attention effect was carried out on the subtracted attend-left minus attend-right scalp distributions to eliminate any nonspecific arousal or motivational effects. A pair of dipolar sources in the ventral extrastriate cortex of the fusiform gyrus provided an excellent fit to the P1 attention effect, accounting for 98% of the variance in the scalp voltage distribution within the P1 latency range (Fig. 2B).
In the PET session, significant blood-flow changes were observed in the posterior fusiform gyrus at a location that corresponded closely to that of the calculated P1 dipole (Fig. 2C). To further support the proposal that the enlarged contralateral P1 and the enhanced contralateral blood flow reflected the same pattern of attention-related neural activity, it was shown that dipolar sources calculated across the ERP waveform were situated closest to the locus of PET activation precisely at the peak of the P1 wave (110–130 ms). In addition, when dipoles were placed (seeded) at the fusiform site of PET activation, they accounted for over 96% of P1’s scalp voltage distribution. Thus, the combined ERP and PET evidence was consistent with the hypothesis that neural activity in the