These results confirm modulation of the SSVEP by spatial attention. This modulation was strongest and most consistent at lateral occipital sites, near the occipitotemporal junction. The initial SSVEP response was an increase contralateral to the target, lasting about 300ms. Complementary to this phase-locked SSVEP increase was a decrease in background, non-phase-locked EEG amplitude. The latency of these responses decreased with increasing inter-target interval, confirming that this initial increase reflects target detection, a process that becomes more efficient the longer a subject has to orient attention to the location of the target. Behavioural and ERP studies confirm that there is a reflexive orienting of attention toward the location of a cue before attention can be voluntarily re-oriented toward the cued location. This reflexive orienting occurs rapidly, with peak facilitation of behavioural responses occurring 175ms after presentation of the cue, and resists being overridden by voluntary attention until about 400ms after presentation of the cue [Müller and Rabbitt 1989]. These values are consistent with the latencies of the early, contralteral modulations reported herein. Warner & al. [1990] confirmed this result for naïve subjects, but also demonstrated that normal subjects can learn to override involuntary orienting within 50ms to 100ms following the presentation of a cue. Folk & al. [1992] extended these results by showing that changes in luminance are not unique or invariant in their ability to cause involuntary attentional shifts. Rather, the type of stimulus change to which involuntary orienting is most responsive is a function of the relevance of the stimulus in question. If a task involves detecting luminance changes in target stimuli, then distractor stimuli that incorporate luminance changes will command involuntary orienting. If, on the other hand, the task involves detecting a particular colour, then colour-based distractors will be more effective in engaging involuntary orienting. In the present experiment, then, one might expect the momentary colour change that constitutes a cue initially and transiently to heighten attention to the location of the cue, and then to allow a slower, voluntary reorientation of attention to the opposite hemifield which is indicated by the cue.
Of greater interest for the present study is the later, ipsilateral increase in SSVEP amplitude which began to manifest by about 600ms after presentation of the target. This reflects heightened sensitivity contralateral to the newly attended, opposite hemifield. As was the case with the initial, contralateral augmentation of the SSVEP, background activity is modulated in the opposite direction; the non-phase-locked signal decreases at the same time and in the same hemisphere as this late SSVEP increase. These effects jibe with the behavioural data, which show a steep increase in target detection rate between 300ms and 600ms inter-target intervals. Like the early effect of target detection, this late effect of endogenous orienting showed a trend toward earlier latencies with longer inter-target intervals. The time course of these long-latency effects is similar to the latency of the positive shift-difference (Sd, or P700) wave observed by Akshoomoff and Courchesne [1994], and to the Late Directing Attention Positivity (LDAP) described by Harter & al. [1989] and by Harter and Anllo-Vento [1991]. The inversion of the phase-locked amplitude difference between about 300ms and about 600ms in the present study also is reminiscent of Harter's boundary between the Early Directing Attention Negativity (EDAN) and the LDAP. Although Harter's paradigm used a central cueing location instead of an oppositely lateralised one, both Harter's paradigm and the present one require processing of a cueing stimulus that appears in a location distinct from the location being cued. The similarities in time course suggest that the SSVEP amplitude modulations reported here may arise from the same processes as the slow potentials detected by these previous studies.
Several effects unique to the rapid-presentation format of the current experiment may have contributed to this result. Posner and Cohen [1984] and Maylor [1985] described another process affecting visual orienting, which biases attention in an opposite manner and has a longer time course. This `inhibition of return' is manifested in slowed reaction times to stimuli in previously attended locations in which stimuli have recently occurred than to stimuli in constantly unattended locations. Based on lesion studies, inhibition of return seems mediated by the midbrain, specifically the superior colliculus, and seems strongly linked to the saccadic eye movement system [Posner & al. 1985]. In the repetitive shifts of spatial attention demanded by this experiment, a steady-state form of inhibition of return may have affected behavioural reaction times and the time courses of EEG changes.
Reeves and Sperling [1986] found that shifts in attention can affect the perceived order of rapidly occurring stimuli. Their task required the detection of a target in one location and the reporting of the several stimuli immediately following the target in time but occurring in a different location in space. In this paradigm, targets in the newly attended location may be erroneously perceived as having preceded the target in the originally attended location. This demonstrates that perception of temporal order at high rates of stimulus presentation depends on attentional salience as well as on the actual order of the stimuli. This effect no doubt contributes to the low accuracy figures for the shortest inter-target interval bin in the present experiment, and possibly to the time course of the change from contralateral to ipsilateral SSVEP augmentation.
The disappearance of the late, ipsilateral SSVEP increase in the right hemisphere at long inter-target intervals suggests that when given enough time to implement the attentional shift, the right hemisphere always becomes involved, regardless of which hemifield is being attended. (Keep in mind that what is being measured here is the difference between responses to left-field targets and responses to right-field targets, so if a hemisphere is equally involved in shifts to both hemifields, no effect will be measured, despite the presence of activity in that hemisphere.) This bilateral augmentation in the case of rightward shifts supports the widely held notion that the right hemisphere is somehow specialised for spatial attention. Heilman and Van Den Abell [1980] confirmed behavioural and lesion evidence of right-hemisphere dominance for attention [Mesulam 1981; Weintraub & Mesulam 1987] by measuring EEG desynchronisation in the alpha band in response to lateral visual stimuli. Whereas the left parietal lobe desynchronised most after right stimuli, the right parietal lobe desynchronised equally to stimuli on either side. Similarly, Harter & al. [1982] found a greater effect of selective attention on right-hemisphere potentials than on left-hemisphere potentials. Using functional mapping by means of positron emission tomography, Corbetta & al. [1993] confirmed that superior parietal cortex is activated bilaterally when spatial attention is shifted to the right visual field, but mostly contralaterally when spatial attention is shifted to the left visual field. This activation is specific to shifts of spatial attention, and is not present during conditions of statically maintained attention [Petersen & al. 1994].
However, the persistence of the non-phase-locked amplitude modulation under the same conditions in which the right-hemisphere phase-locked modulation disappears adds a curious postscript to this conclusion. If phase-locking can be interpreted as an indicator of the amount of neural tissue devoted to processing stimuli originating in the attended location, and absence of phase-locking as an indicator of the amount of background processing, then it seems that background processing is preferentially suppressed in the hemisphere contralateral to the attended location, even when attentional processing is distributed across both hemispheres.
The importance of controlling eye position in studies of covert orienting of attention cannot be emphasised enough. In addition to the retinotopic effects of eye position on the magnitudes of visual event-related potentials, there also seems to be a uniquely attentional effect: extremes (40° horizontal eccentricity in either direction) of eye position influence the magnitudes of auditory ERPs, too [Okita & Wei 1993]. In a design in which attention is static within each block of trials, effects of eye position can be eliminated by excluding subjects whose fixation is poor [Mangun & Hillyard 1991]. In paradigms that require rapid shifts of spatial attention within blocks, however, it may be almost impossible for any subject to suppress saccades completely.
This study included only twelve subjects, and, probably as a result of this small size, statistically significant intervals of left-field/right-field difference were widely spaced. Studies including more subjects will be able to close these gaps and thus to determine more precisely the time courses of these modulatory effects. Nevertheless, the current results demonstrate the usefulness of SSVEP measurements, and offer an opportunity for hypothesis generation as to the broadly defined time courses of the effects in question. These effects are small--a few tenths of a microvolt--and thus easily swamped by noise. Because of the small size of this study, it was thought best a priori to concentrate on subjects in whom any effects could be easily detected. Subjects manifesting large amounts of alpha activity were excluded because their behavioural performance was poorer than that of the more attentive subjects. In addition, the volume of alpha tended to swamp the attentional effects in single-subject averages. In a larger study in the future, it will be interesting to examine such noisy data in order to pick out small effects.
In summary, this study revealed that (1) the amplitude of the steady-state visual evoked potential is modulated by spatial attention during rapid, repetitive attentional shifting, (2) the time course of this modulation is consistent with behavioural data and single-trial ERP data, (3) the pattern of modulation reflects right-hemisphere specialisation for spatial attention and is consistent with recent functional imaging findings, (4) background EEG is in general modulated oppositely to the SSVEP, reflecting a specificity for task-relevant processing.