In a comparison of periods of sustained visual spatial attention to left and right hemifields, we have found enhanced activation of ventral occipital cortex contralateral to the attended hemifield, and enhanced activation of posterior ventral intraparietal sulcus ipsilateral to the attended hemifield, while activation of superior parietal cortex showed no effect of attended hemifield. The finding of ventral occipital activation agrees with electrophysiological studies which have demonstrated a small but significant modulation by visual spatial attention of steady-state visual evoked potential amplitude over the occipitoparietal scalp that overlies this area [Morgan & al. 1996; Belmonte 1998; Müller & al. 1998]. Confirming the attentional nature of this modulation is the observation that the hemispheric pattern of modulation shifts as attention shifts across the midline of the visual field [Belmonte 1998; Müller & al. 1998]. Three-dimensional dipole source localisation and fMRI have placed this attentional effect in the areas of posterior fusiform and lateral occipitotemporal gyri [Hillyard & al. 1997]. This localisation is corroborated by two studies that combined PET with source localisation of the P100 evoked potential [Heinze & al. 1994; Mangun & al. 1997]. This convergent evidence suggests that the fMRI and PET effects identified in ventral occipital cortex and the EEG modulation measured at the overlying occipitoparietal scalp sites are manifestations of one and the same neurophysiological process.
Our analysis also revealed effects in parietal cortex. In superior parietal lobe, although there was strong activation both during attention to the left hemifield and during attention to the right, the laterality of attention had no effect on the lateralisation of brain activation. Instead, superior parietal lobe seemed to respond equally or nearly equally to stimuli in both visual fields, attended and unattended. This absence of lateralised activation in superior parietal lobe during lateralised attention replicates the results of other studies that have presented attended and ignored stimuli simultaneously at fixed, single, peripheral locations in each hemifield [Heinze & al. 1994; Vandenberghe & al. 1997], but is at odds with a previous finding of lateralised activation during shifts between multiple locations within a hemifield [Corbetta & al. 1993]. A possible explanation of this set of findings is that the onset of a stimulus, whether attended or unattended, primes superior parietal cortex to attend to the location of that stimulus. Only in cases where a shift of attention is actually implemented would an additional, shift-related activity be superimposed on this exogenous, stimulus-related activity.
Our finding of intraparietal activation ipsilateral to the attended hemifield - or, equivalently, contralateral to the unattended hemifield - can be explained both in terms of the hypothesis of suppression of irrelevant stimuli and in terms of an alternative hypothesis of spatial orienting. The spatial orienting hypothesis is supported by a convergence of evidence from studies in which a frontoparietal network including intraparietal sulcus is activated by both exogenously and endogenously cued spatial shifting of attention [Nobre & al. 1997; Corbetta 1998; Gitelman & al. 1999; Hopfinger & al. 2000; Nobre & al. 2000; Beauchamp & al. 2001], by visual conjunction search [Donner & al. 2000], and by covert attentive tracking of moving targets [Culham & al. 1998]. This same network also seems active during tasks that demand sustained spatial attention during discrimination of orientation [Vandenberghe & al. 1996, 1997], form [Martínez & al. 1999], internal spatial relationships [Vandenberghe & al. 1997], motion coherence [Shulman & al. 1999], or visual flow [Büchel & al. 1998]. Remarkably, though, even when the factor of sustained spatial attention is controlled, intraparietal sulcus is activated by shifting [Le & al. 1998] and conjunction [Wojciulik & Kanwisher 1999] of purely feature-based criteria.
One theory that may explain this generality of intraparietal involvement in attention tasks is the idea that intraparietal sulcus subserves suppression of irrelevant distractors [Wojciulik & Kanwisher 1999]. Behaviourally, the existence of some suppressive process is suggested by the phenomena of negative priming [Tipper 1985] and inhibition of return [Posner & al. 1985]. In terms of functional anatomy, suppression within intraparietal sulcus would be consistent with intraparietal activation in spatial and non-spatial tasks that include distractors as discussed above, and with the finding that, unlike frontal activity, parietal activity in Stroop tasks is more sensitive to variations in the task-irrelevant dimension (colour-word versus colour-object) than to variations in the task-relevant dimension (colour-word versus spatial-word) [Banich & al. 2000]. In addition, the suppression hypothesis garners some support from studies of selective attention that find an absence of intraparietal activation in the absence of distractors: no activation of intraparietal cortex was detected in a comparison of selective attention to shape, colour, and speed with divided attention to all three attributes [Corbetta & al. 1991], nor in a comparison of feature conjunction to simple categorisation [Rees & al. 1997], nor during sustained attention to a single, instantaneously moving target [Vandenberghe & al. 2001].
These two alternative explanations are difficult to differentiate on the basis of purely correlational methods such as functional imaging, since both predict intraparietal activation contralateral to unattended stimuli. Under the suppression hypothesis this activation reflects active inhibition of distractors, whereas under the spatial orienting hypothesis it reflects the preparation of a reflexive shift of attention evoked by detection of these distractors - a shift that need not actually be executed. Both explanations seem particularly relevant to tasks such as the current one, in which periods of sustained attention are punctuated by shifts. In the task that we have used, subjects are asked to respond selectively to stimuli that satisfy a combination of spatial and chromatic criteria. Whereas the chromatic criterion is constant, the spatial criterion shifts from time to time: the stimuli to which subjects are responding are always red, but not always in the same hemifield. Under such circumstances, in which one is always anticipating an upcoming shift in the spatial focus of attention, spatial attention may not settle completely on one location. Instead, a somewhat weaker engagement of the spatial filter may yield the best compromise between speed of response to stimuli at the currently attended location and speed of disengagement and shifting to the next attended location. Under the spatial orienting hypothesis, this weak engagement at the current focus may augment the reflexive shifting response to stimuli outside the current focus. Under the suppression hypothesis, weak spatial engagement may demand more vigorous suppression of unattended stimuli that have passed the spatial filter. In this scenario, ventral occipital enhancement and intraparietal suppression would play complementary roles, with the intraparietal suppressive filter plugging the leaks in the ventral occipital selective filter.
This complementary activity between ventral occipital and intraparietal regions may be illustrated by contrasts between certain populations. Although with our small numbers of female and male subjects we did not find a significant sex difference in any of our measures, we did note what appeared to be complementary trends in ventral occipital and intraparietal regions in the two sexes: in the ventral occipital region men's z-scores tended to be higher in magnitude than women's, while in the intraparietal region women's z-scores tended to be higher than men's. In a study of subjects with autism performing this same task [Belmonte & Yurgelun-Todd 2001bc], we find that people with autism show no measurable attentional activation in the ventral occipital region, and significantly supranormal activation in the intraparietal region. Both the sex contrast and the autism contrast suggest that as differential activation in occipital cortex diminishes, differential activation in intraparietal cortex increases, as if to compensate.
Unlike the current study, the work of Wojciulik and Kanwisher [1999] in a series of selective attention tasks found no lateralisation of the intraparietal effect as a function of attended hemifield. This absence of lateralisation is not unexpected, though, since their peripheral spatial attention task included distractors in both hemifields, their peripheral object matching task used distractors that were split across hemifields, and their conjunction task involved only foveal stimulation. The current study, in contrast, used only one irrelevant location which was always in the hemifield opposite to the attended location.
Our intraparietal region of interest is also included in a posterior region identified by Martínez et al. [1999] as active in attention to peripheral stimuli. In that study, 3x3 arrays of stimuli were presented bilaterally. Subjects attended to one hemifield at a time, and had to press a button if the stimulus at the centre of the array in the attended hemifield were a rare target. This task demanded selective attention to the central stimulus within the attended hemifield, and suppression of the eight distractors in the attended hemifield as well as the nine stimuli in the unattended hemifield. In this paradigm, the distractors within the attended hemifield likely would have been more salient than those in the hemifield opposite, and thus the demand for suppression would have been greatest in the attended hemifield, and greater than in our study in which distractors were confined to the unattended hemifield. Accordingly, the intraparietal activation detected by Martínez et al. was localised contralateral to the attended hemifield, and the effect size exceeded that observed in the current study, reaching significance within a majority of individual subjects.
Too often the conclusions of functional imaging studies are overly specific to the perceptual and cognitive problems presented by a particular experimental paradigm, or by a particular explanatory framework. For this reason we wish to stress that a role for intraparietal sulcus in suppression of irrelevant stimuli would not be incompatible with involvement in the preparation or implementation of attentional shifts, nor with other roles such as attentional integration of object features. One can imagine a neural network architecture in which inputs from various feature detectors, spatial maps, and object representations [Treisman 1993] impinge on interneurons, along with top-down attentional inputs that instruct each interneuron to select or to suppress its particular feature or set of features. This type of architecture would produce increased metabolic demand during conditions of feature integration or selection, feature or object suppression, and shifts of attention between objects, features, or locations. Gross techniques such as fMRI would be unable to detect differences in intraparietal functional anatomy between these conditions. Given the resolution limits typical of current functional imaging systems, one cannot look to fMRI for information on the microarchitecture of neural systems - for the same reason that one does not mend a watch with a sledgehammer.
We can, however, examine single-cell studies for evidence of separate selective and suppressive signals in primate parietal cortex. In macaque monkeys, signals consistent with active processing associated with irrelevant stimuli have been recorded both in area LIP and in nearby visual area 7a. In a memory-guided saccade task, a distractor stimulus presented within the receptive field of an LIP neuron during the delay period evokes a greater response when the target location of the saccade is outside the cell's receptive field than when the target location is inside the receptive field [Powell & Goldberg 2000]: in other words, the cell responds more vigorously when the stimulus is unattended than when the stimulus is attended. In an exogenously cued attention task, a subpopulation (23%) of LIP cells responded less to the onset of validly cued targets in their receptive fields than to the onset of invalidly cued targets in their receptive fields [Robinson & al. 1995]. A smaller subpopulation (13%) showed enhanced response to validly cued targets. Similar results were found with a foveal attention task and peripheral probes, and with a probabilistic attention task with probes in unexpected locations. In recordings from area 7a during a task of delayed matching to location, 55% of neurons responsive to visual stimuli were suppressed (versus 5% enhanced) by attended stimuli as compared to unattended stimuli within their receptive fields [Steinmetz & al. 1994]. These findings of greater single-unit response to unattended stimuli have been interpreted in the context of the spatial orienting hypothesis, as indications that parietal tissue is more involved in preparing shifts of attention to previously unattended stimuli than in enhancing responses to currently attended stimuli. However, they are also compatible with the suppression hypothesis. Electrode penetrations in each of these studies have been tangential to the cortical surface, which in these regions lies on the banks of sulci. Given our proposed model of interneuron-mediated suppressive and integrative effects, studies involving radial penetrations of these areas would be of interest. Such studies could categorise neuronal response properties as a function of laminar location and therefore make inferences as to response properties as a function of a cell's position within the local circuit.
In light of the high degree of normal variation in parietal anatomy, several authors [Nobre & al. 1997; Corbetta & al. 1998; Wojciulik & Kanwisher 1999] have noted the necessity of examining functional anatomy within individuals rather than in a spatial average of individual brains. In a full 30% of the normal population, the intraparietal sulcus follows a distinctly zigzag course through the parietal lobe, sending off varying numbers of small rami as it descends towards the transverse occipital sulcus [Naidich & al. 1995]. Given this degree of variation, spatial averaging based on gross cerebral landmarks would render subtle attentional effects either poorly localised or statistically undetectable. Even aside from the problem of spatial blurring, the attention effects that we have observed are small modulations superimposed on the larger effect of the task: despite the fact that the data when pooled across all subjects yielded a significant result, none of the attention effects were significant within individual subjects. Our alternative to these problems of spatial averaging across subjects and weak attention signals within subjects is to draw regions of interest within individuals, on the basis of individual functional mapping. Basing one's regions of interest on a prior functional analysis may seem at first glance to beg the question of whether an area is activated. The key to resolving this seeming confound is to note that our two functional analyses are completely independent of one another: since the analysis of left attention versus right attention is confined to the task periods, it cannot be affected by the prior comparison of task periods to fixation.
We have demonstrated that spatial enhancement of relevant stimuli in ventral occipital cortex is complemented by an intraparietal response associated with suppression of, or preparation of a reflexive shift of attention towards, irrelevant stimuli during bilateral rapid serial visual presentation. These findings are consistent with those of other functional imaging studies, as well as with those of single-unit recording. Furthermore, the possible suppressive function of intraparietal cortex is compatible with its involvement in other processes such as feature integration and attentional shifting.
ACKNOWLEDGEMENTS
This work was supported by a grant from the National Alliance for Autism Research. The authors thank two anonymous reviewers for helpful feedback.