Brain (1998), 121, 12811294

The functional anatomy of attention to visual motion

A functional MRI study
Christian Buchel, Oliver Josephs, Geraint Rees, Robert Turner, Chris D. Frith and Karl J. Friston
The Wellcome Department of Cognitive Neurology, Institute of Neurology, London, UK Correspondence to: Christian Buchel, The Wellcome Department of Cognitive Neurology, Institute of Neurology, 12 Queen Square, London WC1N 3BG, UK. E-mail: [email protected]

Summary
Attention can enhance or modulate neural responses to stimuli at early and late stages of sensory processing. We were interested in the modulatory effect of attention to visual motion on cortical responses as measured by functional MRI. Subjects were scanned during repeated presentations of identical stimuli (visual motion) while only the attentional component of the task was varied. Enhanced haemodynamic responses during attentive conditions dened an occipitoparietofrontal system, including sensory and association areas, as well as the medial thalamus and superior colliculus. Attentional modulation was not restricted to extrastriate areas (including V3a and the V5 complex) but was also evident, to a lesser degree, in early visual areas close to the calcarine ssure (V1/V2 border). Attention-related enhancement of cortical responsiveness is discussed in terms of data that implicate modulatory short-term changes in synaptic efcacy and reciprocal connections between striate, extrastriate, parietal and frontal areas. Given the similarity of our attentional network to that controlling eye movements, the results of this study are in accord with theories linking oculomotor control and attention.

Keywords: V5 complex; V3a; dorsolateral prefrontal cortex; frontal eye elds; modulation Abbreviations: BA Brodmann area; BOLD blood oxygenation level dependent contrast; ERP fMRI functional MRI; SPM statistical parametric map event related potentials;

Introduction
Motion perception is an important faculty. Driving at night reduces the amount of information available about the car ahead to two red rear lights. The ensuing radial motion is a two-dimensional representation of motion in depth. The velocity of the two radially moving dots allows one to estimate how fast one is approaching the other car. However, the depth of processing depends on whether the observer is the driver or a passenger. Both observe the same scene but the driver has to pay attention to the rear lights and to the optical ow engendered. This emphasizes the fact that visual motion can be attended to. Visual stimuli are multidimensional (e.g. colour, shape and motion) and a mechanism to weight their current contextual importance is necessary (Posner and Petersen, 1990; Posner, 1994). In the above example only the velocity of visually presented moving dots is salient, their colour (red) is less important. In this paper we have used functional MRI (fMRI) to identify the neural systems implicated in attending to visual motion. Oxford University Press 1998 Electrophysiological studies have identied a motion sensitive area in the middle temporal region in the Macaque (Dubner and Zeki, 1971). A homologous area (human V5) can be dened in man by functional imaging using PET (Zeki et al., 1991) or functional MRI (fMRI) (Tootell et al., 1995). Recent primate studies (Treue and Maunsell, 1996) and fMRI data (OCraven et al., 1997) demonstrate that activity in the V5 complex (or the middle temporal region in the Treue and Maunsell study) can be modulated by attention. This is in accord with earlier results by Corbetta et al. (1991), who studied selective attention to different aspects of visual stimuli using PET. We refer to the middle temporal region or human V5 complex simply as the V5 complex in the remainder of this paper. These results suggest that the responses of extrastriate areas which process visual signals are modulated according to task demands; this may provide a mechanism whereby individual stimulus attributes such as colour and motion can

1282

C. Buchel et al. lasted 32.2 s giving 10 volumes per condition. In all conditions the subjects looked at a xation point (size 0.3) in the middle of a screen. Images were back-projected onto the screen from an LCD video-projector. In conditions with visual motion, 250 white dots (size 0.1) moved radially from the xation point in random directions towards the border of the screen, at a constant speed of 4.7/s, where they vanished. The active screen area was circular with a diameter of 17. The screen refresh rate was set to 67 Hz. The high density of dots in the foveal region and the speed employed suppressed any tendency for the subjects to track individual dots with smooth pursuit eye-movements. As shown previously (OCraven et al., 1997), radial motion of the type we used, makes it easy for subjects to maintain xation. These authors, using EOG and an infrared tracking device, reported no signicant modication of eye movements, i.e. no differences between the two visual conditions, with and without attention. We also used EOG recordings to measure eye movements during our experiment. As EOG recording is not possible during echo-planar imaging, we performed these measurements during a fth experimental run without scanning. The EOG measurements were calibrated and sensitive to saccadic movements of ~1. No saccadic eye movements were seen during visual stimulation. We dened four conditions: xation, attention, no attention and stationary. During the attention and no attention conditions subjects xated centrally, while 250 white dots emerged from the xation point and moved to the edge of the screen. The difference between the conditions lay in the explicit instructions given to the subjects. In the attention condition the subjects were asked to detect changes and during the no attention condition the subjects were instructed to just look. The verbal commands were recorded digitally and played 10 s before the condition started. During the stationary conditions the stimulus consisted of the xation point and 250 stationary dots. During the xation condition, only the central xation point was visible. Before scanning, subjects were exposed to ve 30-s trials of the stimulus. The speed of the moving dots was changed ve times during each trial. Subjects were asked to indicate any change in speed. Changes in speed were gradually reduced over the ve trials, until a 1% change was presented on the last occasion. During scanning, the speed of the moving dots was kept constant for both the attention and no attention condition, so that identical visual stimuli were shown in all motion conditions. In the scanner, the subjects were asked to detect changes silently, without any response. The subjects were also instructed not to count changes in speed, and subjects were not asked about the frequency of changes during the scanning session. However, after completion of the experiment, all subjects were asked to estimate, post hoc, the frequency of speed changes. The subjects denied counting the speed changes. However, we were not able to verify the reliability of the subjects reports with an independent measure.

be selected by attention. The locus of this modulation appears to be early in the sensory processing pathways, compatible with early selection theories of selective attention (Treisman and Gelade, 1980). Such modulation presupposes the interaction of sensory input and modulatory top-down inuences from other cortical areas. In other words, the expression of an early selection process, operating in primary and secondary sensory cortex, cannot be isolated from the question of the locus of top-down inuences exerted by higher (i.e. parietal and frontal) association areas (Fink et al., 1996; Rees et al., 1997). The functional architecture of selective attention has been inferred on the basis of labelling (Morecraft et al., 1993) and lesion studies (Mesulam, 1981). This work points to a largescale neural network that includes the frontal eye elds, the cingulate, premotor, lateral prefrontal, orbitofrontal, opercular, posterior parietal, lateral and inferior temporal, parahippocampal and insular regions as well as the thalamus. The present work aimed to identify regions that are modulated by attention to visual motion, including areas outside the visual and parietal cortex (Bushnell et al., 1981; Mesulam, 1981; Mountcastle et al., 1981; Assad and Maunsell, 1995) (i.e. frontal cortex and subcortical structures). For instance, the frontal eye elds are thought to play an important role in visual attention, as exemplied in the premotor theory of attention (Rizzolatti et al., 1987). To address these issues we designed an fMRI experiment that repeatedly used the same visual stimuli presented during a series of tasks that involved different attentional sets. The idea of using identical visual stimuli in the context of different attentional sets goes back to the study by Corbetta et al. (1991).

Methods Experimental design and image acquisition

Data were acquired with a 2-T Magnetom VISION (Siemens, Erlangen, Germany) whole body MRI system equipped with a head volume coil. Contiguous multislice T2-weighted fMRI images (TE 40 ms; 90 ms per image; 64 64 pixels: area 19.2 19.2 cm) were obtained with echo-planar imaging using an axial slice orientation. A T2-weighted sequence was chosen to enhance the blood oxygenation level dependent (BOLD) contrast. The volume acquired covered the whole brain except for the lower half of the cerebellum and the inferior most part of the temporal lobes (32 slices; slice thickness 3 mm, giving a 9.6-cm vertical eld of view). The effective repetition time was 3.22 s per volume. Seven young, right-handed normal, healthy subjects (two male and ve female) gave informed consent to participate in this study, which was approved by the Joint Ethics Committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery, London. They were scanned during four runs, each lasting 5 min 22 s. In each run, 100 image volumes were acquired. Each condition

Attention to visual motion in fMRI Table 1 Experimental design

Run 1 Condition Volume images Time (s) M 10 32 F 10 32 A 10 32 F 10 32 N 10 32 F 10 32 A 10 32 F 10 32 N 10 32 S 10 32 M 10 32 Run 2 F 10 32 A 10 32 F 10 32 N 10 32 F 10 32 A 10 32 F 10 32 N 10 32

1283

S 10 32

Run 3 (counterbalanced) Condition Volume images Time (s) M 10 32 F 10 32 N 10 32 F 10 32 A 10 32 F 10 32 N 10 32 F 10 32 A 10 32 S 10 32 M 10 32

Run 4 (counterbalanced) F 10 32 N 10 32 F 10 32 A 10 32 F 10 32 N 10 32 F 10 32 A 10 32 S 10 32

Each subject was scanned during four runs. M denotes the rst 10 (discarded) scans of each run that contained magnetic saturation effects. F xation condition; A attention condition; N no attention condition; S stationary condition. Each condition lasted 32 s, corresponding to 10 volume images (see Methods for a description of the conditions).

Each of four runs comprised four xation periods, two attention, two no attention conditions and one stationary condition (Table 1). Each run was preceded by 10 scans of a blank screen, which were discarded to eliminate magnetic saturation effects and then continued with a xation task, followed by attention, xation, no attention and so on. To avoid habituation effects the third and fourth run were counterbalanced, starting with a no attention condition. There was a gap of 5 min between each run, while the scanner image processor reconstructed the images.

Image analysis and categorical comparisons

Image processing and statistical analysis were carried out using SPM96 (Friston et al., 1995c; Worsley and Friston, 1995). All volumes were realigned to the rst volume (Friston et al., 1995a). A mean image was created using the realigned volumes. A structural MRI, acquired using a standard 3D T1-weighted sequence (each voxel, 1 1 3 mm), was coregistered to this mean (T2*) image. This ensured that the functional and structural images were in the same space. Finally, the structural image was spatially normalized (Friston et al., 1995a) to a standard template (Talairach and Tournoux, 1988; Evans et al., 1993), using non-linear basis functions. The non-linear transformation mapping the structural T1MRI scan onto the template, was applied to the fMRI data. The data were smoothed using a 6-mm (full width at half maximum) isotropic Gaussian kernel for individual analyses and an 8-mm kernel for the group analysis. The use of a wider smoothing kernel for the group comparison was chosen to compensate for residual variability in gyral anatomy after spatial normalization. Data analysis was performed by modelling the different conditions (attention, no attention, xation and stationary) as reference waveforms (i.e. box car functions convolved with a haemodynamic response function) (Friston et al., 1995b) in the context of the general linear model as employed by SPM96 (Friston et al., 1996). Specic effects were tested by applying appropriate linear contrasts to the parameter estimates for each condition, resulting in a t-statistic for each and every voxel. These t-statistics (transformed to Z-statistics) constitute a statistical

parametric map (SPM). These SPMs were then interpreted by referring to the probabilistic behaviour of Gaussian random elds. Data were analysed for each subject individually and by including all subjects in a group analysis. The threshold adopted was P 0.05 (corrected for multiple comparisons). Less signicant activations are reported only if an a priori hypothesis existed for the location in question. Two comparisons were performed: (i) attention and no attention (i.e. visual motion conditions) against stationary dots (stationary) to identify motion sensitive cortical areas and (ii) the attention condition against the no attention condition to identify regions that show differential responses due to attentional set. Given our question, the second comparison between attention and no attention was the most important, whereas the rst localized regions on the basis of motion processing per se. The xation dot condition was used as a low-level baseline. Comparisons including this low-level baseline are not reported here.

Psychophysical validation of tasks

We performed an additional psychophysical experiment to validate the attention and no attention conditions employed in the functional imaging experiment. This experiment involved four of the original seven subjects who were involved in the fMRI study and a further ve subjects. We presented eight attention and eight no attention trials in the same order as in the imaging experiment. The only difference was that the moving dots stopped abruptly after each visual motion condition, resulting in a pronounced motion after-effect. Subjects were instructed to maintain xation after the dots stopped moving and to press a button when the percept of the motion after-effect ceased. It has been shown that the motion after-effect is signicantly shorter when the subjects are involved in a distracting task (i.e. letter discrimination) (Chaudhuri, 1990). Therefore, to endorse the attentional manipulation used in our experiments empirically, we hoped to demonstrate a signicantly longer motion aftereffect after the attention than after the no attention condition.

1284

C. Buchel et al. at the occipitotemporal junction bilaterally. These regions were very close to the co-ordinates of V5 as reported by Watson et al. (1993). This comparison also revealed motion sensitivity in the middle occipital gyrus. This location is close to area V3/V3a as shown by DeYoe et al. (1996) and Tootell et al. (1997). The motion sensitivity of this area however, suggests area V3a rather than V3 (Tootell et al., 1997). We will therefore refer to this area of activation as V3a in the remainder of this paper (Table 3).

Table 2 The psychophysical control experiment: motion after-effects

Subject Mean duration of motion after-effect (s) Attention condition A* B* C* D* E F G H I 11.4 16.8 22.3 10.2 22.0 17.6 22.1 10.2 9.3 No attention condition 7.9 11.0 16.8 8.2 18.7 10.8 12.1 6.7 4.9 1.6 2.5 2.3 2.0 2.9 2.2 4.9 1.9 3.3 0.1 0.05 0.05 0.06 0.05 0.05 0.05 0.08 0.05 t-value P-value

Attention versus no attention

The comparison between attention and no attention, in which the stimulus was identical but the instructions differed, showed signicant differential responses in occipital, parietal and frontal cortices. Figure 1 shows the network rendered on a canonical brain (the gyral and sulcal anatomy does not precisely reect the anatomy of the individual subjects studied) from the Montreal Neurological Institute (Evans et al., 1993). Activations above a threshold of P 0.001 (uncorrected), that lie within 10 mm of the cortical surface are shown.

*These subjects also participated in the imaging experiment. The t-values are for a t test with 14 degrees of freedom.

Results Psychophysical experiment

The motion after-effect varied considerably over subjects (between 5 and 22 s). The analysis of the whole group data showed a highly signicant (P 0.01) difference between the attention and no attention conditions (t(8) 6.2; paired t test). Individual analyses (t tests) showed a signicant (P 0.05) prolongation of the motion after-effect after the attention conditions compared with the no attention conditions in six out of nine subjects. The remaining three subjects showed a trend (0.1 P 0.06) (Table 2).

Occipital cortex
Signicant attentional modulation in the occipital cortex was found at the occipitotemporal junction bilaterally in, and close to, area V5 as identied by the previous comparison. However, the peak activation in this comparison was ~10 mm below the peak of activation in V5 dened using the comparison of motion with no motion (Table 3). Figure 2 shows the time-courses of activity in two voxels (from a single subject) in the inferior and superior parts of V5 complex showing different degrees of attentional modulation. Two nearly symmetrical foci of attentional effects were identied in the middle occipital gyrus, ~30 mm from the midline (Fig. 3B). The location is congruent with the Talairach isocontour plot of V3a given in an fMRI study by DeYoe et al. (1996) and, more recently, by Tootell et al. (1997). Another focus was detected in the lingual gyrus on the medial surface of the occipital cortex, extending dorsally into the calcarine ssure. Neuroanatomically this region includes the V1/V2 border (Shipp and Zeki, 1995; Tootell et al., 1995) (Figs 3A and 4B).

Imaging experiment
Functional MRI is sensitive enough to allow for inferences about the functional anatomy of single subjects. On the other hand a group analysis allows a population oriented view of the common activations associated with the tasks. Our strategy here is to describe the results of each comparison for the group and then show time-courses of activation in a single subject to exemplify the individual results that comprise the groups response.

Behavioural responses
Although subjects were not asked to count changes in speed, we asked them, after the experiment, to estimate the frequency of changes in speed during the visual motion conditions, i.e. attention and no attention. All subjects reported roughly the same number of changes (i.e. ve). The subjects also reported that there were no changes in speed during the no attention condition.

Parietal cortex
The most signicant attentional modulations were in the posterior parietal cortex. These involved the superior parietal gyrus bilaterally [Brodmann area (BA) 7], extending ventrolaterally along the intraparietal sulcus into BA 40. The activations were therefore superior to the location of lesions seen in patients with visual hemi-inattention (De Renzi, 1982). Figure 3C shows the activation within the intraparietal sulcus. Although the mean MRI reects the anatomy of all

Motion conditions versus stationary dot pattern

This comparison was used to identify motion sensitive extrastriate areas. Highly signicant activations were found

Attention to visual motion in fMRI Table 3 Occipital activations: attention versus no attentionand stationary versus moving dots
Region Activation maxima Attention versus no attention x, y, z V1/V2 V3a (R) V3a (L) V5 (R) V5 (L) 9, 78, 9 36, 84, 18 30, 90, 9 42, 78, 12 48, 72, 9 Z-score 4.3* 6.8 6.6 5.3 5.8 Stationary versus moving dots x, y, z 42, 84, 9 33, 87, 3 48, 69, 6 48, 75, 3 Z-score 9.3 11.7 9.2 12.9 10.8 7.3 10.8 13.7 Distance between maxima (mm)

1285

R right; L left. Coordinates (x, y, z) are in millimetres from the anterior commissure as dened by Talairach and Tournoux (1988). *Not signicant (P 0.1, corrected); the other activations are signicant at P 0.05 (corrected).

seven subjects, the intraparietal sulcus is clearly visible. Activations in the right hemisphere were more widely distributed than in the left, as reected by the larger number of local maxima (Table 4).

and discuss these effects although they only reached signicance at the level of P 0.001 (uncorrected).

Frontal cortex
In the frontal lobe we found bilateral activation in the precentral sulcus close to the junction with the superior frontal sulcus. The activation on the right was not signicant (P 0.16, corrected) but was located in the homologous region to that in the left hemisphere. This region corresponds to the location of the frontal eyeelds described by Paus (1996). Table 5 shows the coordinates for frontal activations and the distance between the location of the frontal eye eld in our study and that reported by Paus. We found signicant symmetrical foci, 20 mm inferior to the frontal eye eld, lying in the posterior part of the inferior frontal sulcus at the junction with the precentral sulcus (revealed by rendering the activations onto the individuals structural MRIs) (Fig. 3D). Cytoarchitectonically this focus would correspond to BA 6 close to the border with BA 8 (Vandenberghe et al., 1996). A distinct effect was found in the right dorsolateral prefrontal cortex located in the inferior frontal gyrus, 16 mm anterior to the focus above. The maximum activation was in the posterior part of BA 9/46 (Fig. 3E). Bilateral attentional effects were also seen ventrolaterally in the region of BA 47 (Fig. 3A). On the medial aspect of the frontal lobe two small effects were found ~15 mm rostral to the anterior commissure. The more caudal activation lay within the anterior cingulate, whereas the superior activation probably corresponds to the supplementary eye eld (Darby et al., 1996) (see Table 5 and Fig. 4B).

Discussion A distributed attentional system

This paper has presented a characterization of the functional architecture underlying attention to visual motion. We used manipulations of attentional set while subjects viewed the same visual motion stimulus in all conditions. The attentional system identied by comparing attention and no attention conditions included the frontal eye elds, posterior parietal, cingulate, premotor, ventral and lateral prefrontal cortices and the thalamus. This system corresponds remarkably well with a number of proposed neuroanatomical models of attention (Petrides and Pandya, 1984; Mesulam, 1990; Morecraft et al., 1993).

Experimental design
We used radially moving dots as the stimulus for a number of reasons. First, radial motion denes a natural xation point in the centre of the screen. Secondly, the high density of moving dots in the foveal region is a potent stimulus for motion responsive cortical areas (Tootell et al., 1995). With respect to attention, radial motion has another advantage: the subjects task was to detect near threshold changes in the speed of stimuli. Given a stimulus with a lower density of moving elements, a possible strategy might have been to select a single stimulus and then follow it, to evaluate changes in speed. This would confound attention to motion with object and spatial selection (Haxby et al., 1994). The density of the stimulus in the centre, the high speed and the fact that the dots were indistinguishable from each other, makes this explanation unlikely. The behavioural data were obtained by asking subjects to estimate the number of changes in speed after the experiment. The alternative, to ask subjects after each run, was dismissed

Subcortical activations
Two nearly symmetrical activations were found in the medial thalamus and the right superior colliculus (Table 6 and Fig. 4) As these areas are linked to spatial attention, we report

1286

C. Buchel et al.

Fig. 1 Activations revealed by comparing the attention and no attention conditions rendered on a standard anatomical template; activations with a Z-score 3.5 are rendered onto a standard 3D anatomical template. Activation clusters of 20 voxels which were within 10 mm of the surface are rendered. The occipitoparietofrontal system, including V1/V2, V3a, V5, posterior parietal complex (PP), frontal eye elds (FEF), Brodmann area 6 (BA6), dorsolateral prefrontal cortex (DLPFC) and ventrolateral prefrontal cortex (VLPFC), and supplementary eyeelds (SEF), is shown.

because of the ensuing demands that would have been placed on working memory. Indicating changes online by a button press was also dismissed because those movements and associated preparation would not have allowed any sensible interpretation of premotor responses known to be involved in attentional processing (Rizzolatti et al., 1987). We also established the attentional differences between our experimental conditions using a psychophysical

experiment. The signicance and robustness of the ndings, using the same experimental paradigm, both in subjects involved in the study and additional subjects, highlights the potent effect of the differences in attentional set used during scanning. The previous study by Chaudhuri (1990) demonstrated that attention affects the duration of the motion after-effect, which is consistent with our assumption that detecting changes in speed increases attention to motion.

Attention to visual motion in fMRI

1287

Fig. 2 Differential behaviour of areas in the V5 complex and their time-courses. The location (right) and time-course (left) of activation for V5 and related areas. The upper left panel shows the modulation of the responses by attention, and the inferior location of the maximum. The unbroken line indicates the best t of the reference waveform (delayed boxcar), whereas the dots show the actual signal. The rst positive half cycle denotes the attention condition, the second no attention, always interleaved with xation. The lower left panel shows the location and time-course of the responses in the superior V5 complex. There is no difference between activations during attention and no attention conditions. Note that from scan 180 onwards (third run), the scan order is counterbalanced. The vertical axis denotes the BOLD signal. Data are scaled to a whole-brain mean of 100.

The functional anatomy of attention to visual motion Striate and extrastriate visual areas
Extrastriate areas showing signicant attentional modulation included V5 complex and V3a bilaterally. The activation of the V5 complex when comparing the attention and no attention conditions was ~10 mm lower than the corresponding peak from a comparison of visual motion and stationary conditions. Possible explanations for this nding are that V5 complex or the middle temporal region in the monkey are functionally heterogeneous (Desimone and Ungerleider, 1986; Tanaka et al., 1986). Area V3a has motion selective neurons as demonstrated in primates (Galletti et al., 1990) and humans (Tootell et al., 1997). This was also demonstrated by our data, on comparing moving and

stationary dots, where V3a shows a signicant differential response. However, to our knowledge, this is the rst report of attentional modulation of V3a activity either in the primate or the human. Precise localization of V3a is only possible in single subjects through demonstrating the retinotopy of this area (Tootell et al., 1995, 1997). However, the Talairach isocontour plot for different visual areas given by DeYoe et al. (1996) indicates that the peaks of our activations, as given in Table 3, are close to area V3a. We have also demonstrated differential activation at the V1/V2 border when comparing attention and no attention conditions. However, this modulation was variable over subjects and there was only a non-signicant trend in the grouped data (P 0.1, corrected). A modulation of V1/V2 might be mediated by reciprocal connections between V1/ V2 and V5. This hypothesis is supported by reversible cooling

1288

C. Buchel et al.

Fig. 3 Cortical activations revealed by comparing the attention and no attention conditions, shown in sections, overlaid on the mean MRI of all seven subjects. The threshold for display is set to P 0.001 (uncorrected). (A) Bilateral activation in V5, activation in the lingual gyrus at the V1/V2 border and bilateral activation of the ventrolateral prefrontal cortex. (B) Bilateral visual area V3a. (C) Bilateral activation following the intraparietal sulcus is demonstrated in this coronal section. (D) Activation in the inferior frontal sulcus close to the junction with the prefrontal sulcus (area 6), and (E) activation further anterior in the inferior frontal sulcus in area 9 or 46.

experiments involving the middle temporal region (Girard and Bullier, 1989). Inspired by reversible cooling experiments concerning the interactions between V1 and V2 (Sandell and Schiller, 1982), Friston et al. (1995d) have shown modulation of V1 by V2 in humans using fMRI, by assessing activitydependent changes in effective connectivity between those areas. In a previous study (OCraven et al., 1997) modulation of lower visual areas by attention was not demonstrated.

Parietal cortex
Our results showed that the posterior parietal cortex was subject to clear attentional modulation; evidence for

attentional modulation of posterior parietal cortex activity comes from several other studies (Bushnell et al., 1981; Mesulam, 1981; Mountcastle et al., 1981; Assad and Maunsell, 1995). Goldberg and Segraves (1987) concluded that the attention-related enhancement of visual responses in posterior parietal cortex is a possible neural substrate for visuospatial attention. BA 7a contains light-sensitive neurons coding centrifugal or centripetal motion (Motter and Mountcastle, 1981) that implicate this area in the perception of optical ow, a phenomenon important for estimating velocity in forward locomotion. These cells are also important in terms of attraction of gaze and attention to objects and events in the peripheral visual eld. Optical ow, as used in

Attention to visual motion in fMRI our study, has been shown to activate occipitoparietal regions in a PET activation study by de Jong et al. (1994). The PET study of Corbetta et al. (1991), with attention to one component (motion) of a multi-component visual stimulus in a divided attention task, revealed activation in the intraparietal sulcus. Finally, bilateral lesions of the area around the intraparietal sulcus lead to peripheral visual inattention in patients (Pierrot Deseilligny et al., 1986).

1289

Neural mechanisms underlying attentional modulation

Attention can be expressed as the enhancement of one cortical pathway mediating the attended modality (e.g. motion) (La Berge, 1995). The question that arises is which part of the brain is capable of modulating such pathways? Selemon and Goldman-Rakic (1988) have shown that prefrontal cortex is connected to many cortical areas ascribed to spatially related behaviour and attention. They also note that prefrontal projections throughout the cortex terminate in layer I. This pattern is typical of feedback connections (Rockland and Pandya, 1979; van Essen and Maunsell, 1983). Therefore, the prefrontal projections to the posterior parietal cortex have both the macro- and microanatomical characteristics of a modulatory projection system. However, this does not explain the modulation of activity in extrastriate areas that we observed, because direct projections between prefrontal cortex and extrastriate areas have not been demonstrated (Selemon and Goldman-Rakic, 1988). However, the posterior parietal cortex, as the main target of modulation, has rich reciprocal connections with V5, the middle superior temporal region and V3a in the primate (Cavada and Goldman-Rakic, 1989), which themselves project back to striate cortex. Therefore, we suggest that activation of prestriate and striate cortical areas is modulated reciprocally in a chained cortical hierarchy. This idea would also explain the gradual decay of modulation from the posterior parietal cortex to V5 or V3a, and from there to V1/V2, as demonstrated for a single subject in Fig. 5. Modulation is an extremely exible mechanism, especially on a small time scale, a factor critical for attention. In a biologically compelling mathematical model, Olshausen et al. (1993) have used the concept of modulation to model involuntary shifts of attention between objects in a given visual scene. As a possible mechanism for modulation or gating at the neuronal level, they proposed voltage-gated calcium channels (Llinas, 1988) that would enable non-linear coupling between inputs. This has been shown for synaptic inputs into layer I of neocortex (Cauller and Connors, 1992),

Prefrontal cortex
The activation of right prefrontal cortex in the comparison of the attention and no attention conditions was localized to a medial part of the inferior frontal sulcus. As prefrontal anatomy varies between individuals (Rajkowska and Goldman-Rakic, 1995) we consider this region to lie in BA 9 or BA 46. The involvement of the prefrontal cortex in an attentional network is in accord with previous imaging studies (Pardo et al., 1991; Coull et al., 1996) and neuroanatomical data on connections between the prefrontal cortex and the posterior parietal cortex (Petrides and Pandya, 1984). Table 4 Attention versus no attention: parietal activations
Region Superior parietal (R) x, y, z 33, 63, 57 24, 60, 66 24, 69, 54 39, 42, 66 33, 48, 57 39, 42, 54 21, 66, 54 30, 54, 54 36, 45, 60 39, 36, 48 Z-score 8.4 9.0 7.7 6.5 6.0 5.6 10.8 6.6 6.5 4.7

Intraparietal sulcus (R)

Superior parietal (L) Intraparietal sulcus (L)

Activations reported are signicant at P

0.05 (corrected).

Table 5 Attention versus no attention: frontal activations

Region Ventrolateral prefrontal cortex (R) Ventrolateral prefrontal cortex (L) Inferior area 6 (R) Inferior area 6 (L) DLPFC (R) (area 9/46) Supplementary eye eld Anterior cingulate Frontal eye eld (R) Frontal eye eld (L) x, y, z 45, 24, 3 30, 27, 3 45, 6, 27 45, 6, 30 54, 21, 27 0, 12, 60 0, 15, 45 45, 3, 51 39, 3, 54 Z-score 4.5* 4.8 6.4 4.8 4.9 4.6 3.6* 4.2* 4.6 Remarks Corrected P 0.06

Corrected P 0.7 Corrected P 0.16, 11 mm from Pauss location 15 mm from Pauss location

DLPFC dorsolateral prefrontal cortex. The location of the frontal eye eld is compared with the mean location given by Paus (1996). *Not signicant; the other activations are signicant at P 0.05 (corrected).

1290

C. Buchel et al. in the right superior colliculus. This is in accord with previous studies on attention (Corbetta et al., 1991) and the above primate studies. Like the frontal eye elds, the superior colliculus is functionally implicated in the generation of eye movements. However, the supercial layers of the superior colliculus have visual receptive elds and receive information from the retina and visual cortex (Meredith and Stein, 1985). These cells could register the displacement of dots in our stimulus by eliciting a saccadic program in the deep layers (to follow the stimulus) while execution is blocked. Although radial motion has no unique direction to direct single saccades, it is possible that clusters of dots could engender a perception of transient directionality. This inhibition of saccade execution has been attributed to modulatory inuences from the frontal eye elds (Leichnetz et al., 1981), prefrontal cortex (Goldman and Nauta, 1976) and the posterior parietal cortex (Lynch et al., 1985). This could explain activation of the superior colliculus in the absence of eye movements but does not explain why the superior colliculus activity is greater during attention. Reciprocal inuences from V5 (Graham et al., 1979) and the posterior parietal cortex (Fries, 1984) to the supercial layers could, however, account for the observed modulation. There is considerable evidence to suggest that the system controlling eye movements and the attentional system cannot be separated. This has been shown by behavioural (Sheliga et al., 1994) and electrophysiological (Kustov and Robinson, 1996) experiments. As mentioned above, the moving dots of Table 6 Attention versus no attention: subcortical activations
Region Medial thalamus (R) Medial thalamus (L) Superior colliculus (R) x, y, z 9, 15, 9 12, 15, 9 6, 33, 3 Z-score 3.8 3.6 3.0

which supports the conjecture that attentional modulation could be exerted by prefrontal areas whose connections terminate predominantly in layer I. On the basis of our data alone it is not possible to prove that the prefrontal cortex modulates motion-dependent responses in posterior cortical regions. However, together with lesion data assigning an executive role to the prefrontal cortex this hypothesis may be more tenable. The role of a central executive with respect to attentional processing is demonstrated by patients with prefrontal lesions who show decits of voluntary attention (Brazzelli et al., 1994; Godefroy et al., 1996). Another source of evidence for the role of prefrontal cortex in modulating the posterior parietal cortex comes from human PET studies (Frith et al., 1991). However, further studies, assessing cortical interactions, e.g. effective connectivity (Friston et al., 1995d; Buchel and Friston, 1997), are necessary to investigate this hypothesis directly. Although not signicant, the almost symmetrical activation of medial thalamic nuclei is in accord with the hypothesis that the dorsolateral prefrontal cortex mediates part of its modulatory inuence via these thalamic nuclei (La Berge, 1995).

Oculomotor control and attention

Activations during attention as opposed to no attention involved regions associated with oculomotion and attention. These included the frontal eye elds, medial premotor areas (supplementary eye elds), inferior BA 6 and the superior colliculus. The role of the frontal eye elds and BA 6 in attention is suggested by lesion (Rizzolatti et al., 1983) and animal studies (Latto and Cowey, 1971; Collin et al., 1982). These results led to the premotor theory of attention (Rizzolatti et al., 1987; Sheliga et al., 1994) with the notion that the mechanisms responsible for spatial attention and the programming of saccades cannot be separated. The observed medial subcortical activation was most likely

Activations reported are signicant at P

0.001 (uncorrected).

Fig. 4 Subcortical activations revealed by comparing the attention and no attention conditions, overlaid on the mean group MRI. The threshold for display is set to P 0.01 (uncorrected). The coronal section (A) shows a signicant signal in the medial thalamic nuclei, and B shows activation in the right superior colliculus, VI/V2 border and medial thalamus.

Attention to visual motion in fMRI

1291

Fig. 5 Progressive reduction of modulation from anterior to posterior brain regions. Plots of the timecourse of activation in three different voxels of a single subject. Plots of the mean signal for the three conditions attention, no attention and xation are shown. The upper, middle and lower panels show the activation of single voxels in V1, V5 and the posterior parietal cortex, respectively. Z-scores are derived from the comparison of attention and no attention conditions. The associated P-values indicate the signicance of this difference. The vertical axis denotes the BOLD signal, which is arbitrarily scaled to a whole brain mean of 100. Note the progressive reduction of modulation (difference between attention and no attention) from posterior parietal cortex over V5 to V1; the difference between attention and no attention in V1 (top) did not reach signicance (corrected P 0.64). The horizontal time axis is calibrated in terms of scan number, each lasting 3.2 s. The broken line shows the modelled response.

1292

C. Buchel et al.

our paradigm are prepotent targets for eye movements, and therefore saccades might be programmed but not executed (Rizzolatti et al., 1987; Mesulam, 1990). Mesulam (1990) proposed that the frontal eye elds and area PG have a collective mechanism for specifying whether a location in space (and events within it) will become a target of enhanced neuronal impact, visual grasp, manipulation or orientation. Kustov and Robinson (1996) show that a shift in attention is followed by increased ring of neurons in the superior colliculus, which the authors attribute to the preparation of eye movements. These data support the suggestion that the activation of the superior colliculus (and the frontal eyeelds) in our paradigm is an expression of attentional processing and not related to executed eye movements.

Acknowledgements
We wish to thank A. Brennan, J. Galliers and G. Lewington for help with fMRI scanning. C.B., O.J., G.R., R.T., C.D.F. and K.J.F. are supported by the Wellcome Trust.

References Assad JA, Maunsell JH. Neuronal correlates of inferred motion in primate posterior parietal cortex. Nature 1995; 373: 51821. Baddeley A. Working memory or working attention? In: Baddeley A, Weiskrantz L, editors. Attention: selection, awareness, and control. Oxford: Clarendon Press; 1993. p. 15270. Brazzelli M, Colombo N, Della Sala S, Spinnler H. Spared and impaired cognitive abilities after bilateral frontal damage. Cortex 1994; 30: 2751. Bruce CJ, Goldberg ME. Primate frontal eye elds. I. Single neurons discharging before saccades. J Neurophysiol 1985; 53: 60335. Buchel C, Friston KJ. Modulation of connectivity in visual pathways by attention: cortical interactions evaluated with structural equation modelling and fMRI. Cereb Cortex 1997; 7: 76878. Bushnell MC, Goldberg ME, Robinson DL. Behavioral enhancement of visual responses in monkey cerebral cortex. I. Modulation in posterior parietal cortex related to selective visual attention. J Neurophysiol 1981; 46: 75572. Cauller LJ, Connors BW. Functions of very distal dendrites: experimental and computational studies of layer I synapses on neocortical pyramidal cells. In: McKenna T, Davis J, Zornetzer SF, editors. Single neuron computation. Boston: Academic press; 1992. p. 199229. Cavada C, Goldman-Rakic PS. Posterior parietal cortex in rhesus monkey. I. Parcellation of areas based on distinctive limbic and sensory corticocortical connections. J Comp Neurol 1989; 287: 393421. Chaudhuri A. Modulation of the motion aftereffect by selective attention [see comments]. Nature 1990; 344, 602. Comment in: Nature 1990; 344, 22. Collin NG, Cowey A, Latto R, Marzi C. The role of frontal eyeelds and superior colliculi in visual search and non-visual search in rhesus monkeys. Behav Brain Res 1982; 4: 17793.

Working memory and voluntary attention

The activation of dorsolateral and ventrolateral prefrontal cortex may also be related to the non-verbal working memory during the attention condition. Voluntary attention of the form used in our experiment requires some explicit information to be kept online. This possibility is in accord with the notion that the dorsolateral prefrontal cortex represents a system involved in monitoring information within working memory (Petrides, 1996). Furthermore, it has also been suggested that working memory and attentional processing are linked and should not be seen independently (Baddeley, 1993).

Attention to motion
Motion is displacement over time and is a spatialtemporal attribute. Therefore, there is no clear-cut distinction between featural attention for motion and spatial, or even temporal, attention. In our paradigm, subjects were asked to detect changes in speed. They therefore had to pay attention to the displacement of the dots over time, calling upon both spatial and temporal attention.

Conclusion
Our results show that attention to motion produces a modulation of the activity evoked by moving stimuli in extrastriate cortical regions V3a and V5. A wider network of parietal and frontal cortex is engaged by attention to motion but not by motion per se, suggesting that these areas may be the source of top-down modulatory inuences on extrastriate cortex. Selection based on stimulus motion occurs early in visual processing, compatible with early selection attentional theories. However, the widespread engagement of higher order association cortex by attention to motion argues in favour of an integrated mechanism of attention (Mesulam, 1990) in which selection may take place at multiple loci in the processing pathways for visual signals.

Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE. Selective and divided attention during visual discriminations of shape, color, and speed: functional anatomy by positron emission tomography. J Neurosci 1991; 11: 2383402. Coull JT, Frith CD, Frackowiak RS, Grasby PM. A fronto-parietal network for rapid visual information processing: a PET study of sustained attention and working memory. Neuropsychologia 1996; 34: 108595. Darby DG, Nobre AC, Thangaraj V, Edelman R, Mesulam M-M, Warach S. Cortical activation in the human brain during lateral saccades using EPISTAR functional magnetic resonance imaging. Neuroimage 1996; 3: 5362. de Jong BM, Shipp S, Skidmore B, Frackowiak RS, Zeki S. The

Attention to visual motion in fMRI

cerebral activity related to the visual perception of forward motion in depth. Brain 1994; 117: 103954. De Renzi E. Disorders of space exploration and cognition. Chichester (UK): John Wiley; 1982. Desimone R, Ungerleider LG. Multiple visual areas in the caudal superior temporal sulcus of the macaque. J Comp Neurol 1986; 248: 16489. DeYoe EA, Carman GJ, Bandettini P, Glickman S, Wieser J, Cox R, et al. Mapping striate and extrastriate visual areas in human cerebral cortex. Proc Natl Acad Sci USA 1996; 93: 23826. Dubner R, Zeki SM. Response properties and receptive elds of cells in an anatomically dened region of the superior temporal sulcus in the monkey. Brain Res 1971; 35: 52832. Evans AC, Collins DL, Mills SR, Brown ED, Kelly RL, Peters TM. 3D statistical neuroanatomical models from 305 MRI volumes. Proc IEEE-Nucl Sci Symp Med Imaging 1993; 18137. Fink GR, Halligan PW, Marshall JC, Frith CD, Frackowiak RS, Dolan RJ. Where in the brain does visual attention select the forest and the trees ? Nature 1996; 382: 6268. Fox PT, Fox JM, Raichle ME, Burde RM. The role of cerebral cortex in the generation of voluntary saccades: a positron emission tomographic study. J Neurophysiol 1985; 54: 34869. Fries W. Cortical projections to the superior colliculus in the macaque monkey: a retrograde study using horseradish peroxidase. J Comp Neurol 1984; 230: 5576. Friston KJ, Ashburner J, Frith CD, Poline J-B, Heather JD, Frackowiak RSJ. Spatial registration and normalization of images. Hum Brain Mapp 1995a; 3: 16589. Friston KJ, Frith CD, Turner R, Frackowiak RSJ. Characterizing evoked hemodynamics with fMRI. Neuroimage 1995b; 2: 15765. Friston KJ, Holmes AP, Poline J-B, Grasby PJ, Williams SC, Frackowiak RS, et al. Analysis of fMRI time-series revisited. Neuroimage 1995c; 2: 4553. Friston KJ, Ungerleider LG, Jezzard P, Turner R. Characterizing modulatory interactions between V1 and V2 in human cortex: a new treatment of functional MRI data. Hum Brain Mapp 1995d; 2: 21124. Friston KJ, Holmes AP, Ashburner J, Poline J-B. SPM96. 1996 [accessed 1997 Feb 27]. Available from: WWW: http:// www.l.ion.ucl.ac.uk/spm. Frith CD, Friston K, Liddle PF, Frackowiak RS. Willed action and the prefrontal cortex in man: a study with PET. Proc R Soc Lond B Biol Sci 1991; 244: 2416. Galletti C, Battaglini PP, Fattori P. Real-motion cells in area V3A of macaque visual cortex. Exp Brain Res 1990; 82: 6776. Girard P, Bullier J. Visual activity in area V2 during reversible inactivation of area 17 in the macaque monkey. J Neurophysiol 1989; 62: 1287302. Godefroy O, Lhullier C, Rousseaux M. Non-spatial attention disorders in patients with frontal or posterior brain damage. Brain 1996; 119: 191202.

1293

Goldberg ME, Segraves MA. Visuospatial and motor attention in the monkey. Neuropsychologia 1987; 25: 10718. Goldman PS, Nauta WJ. Autoradiographic demonstration of a projection from prefrontal association cortex to the superior colliculus in the rhesus monkey. Brain Res 1976; 116: 1459. Graham J, Lin CS, Kass JH. Subcortical projections of six visual cortical areas in the owl monkey, Aotus trivirgatus. J Comp Neurol 1979; 187: 55780. Haxby JV, Horwitz B, Ungerleider LG, Maisog JM, Pietrini P, Grady CL. The functional organization of human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations. J Neurosci 1994; 14: 633653. La Berge D. Attentional processing: the brains art of mindfulness. Cambridge (MA): Harvard University Press; 1995. Latto R, Cowey A. Visual eld defects after frontal eye-eld lesions in monkeys. Brain Res 1971; 30: 124. Leichnetz GR, Spencer RF, Hardy SG, Astruc J. The prefrontal corticotectal projection in the monkey: an anterograde and retrograde horseradish peroxidase study. Neuroscience 1981; 6: 102341. Llinas RR. The intrinsic electrophysiological properties of mammalian neurons: insights into central nervous system function. [Review]. Science 1988; 242: 165464. Lynch JC, Graybiel AM, Lobeck LJ. The differential projection of two cytoarchitectonic subregions of the inferior parietal lobule of macaque upon the deep layers of the superior colliculus. J Comp Neurol 1985; 235: 24154. Meredith MA, Stein BE. Descending efferents from the superior colliculus relay integrated multisensory information. Science 1985; 227: 6579. Mesulam MM. A cortical network for directed attention and unilateral neglect. [Review]. Ann Neurol 1981; 10: 30925. Mesulam MM. Large-scale neurocognitive networks and distributed processing for attention, language, and memory. [Review]. Ann Neurol 1990; 28: 597613. Morecraft RJ, Geula C, Mesulam MM. Architecture of connectivity within a cingulo-fronto-parietal neurocognitive network for directed attention. Arch Neurol 1993; 50: 27984. Motter BC, Mountcastle VB. The functional properties of the lightsensitive neurons of the posterior parietal cortex studied in waking monkeys: foveal sparing and opponent vector organization. J Neurosci 1981; 1: 326. Mountcastle VB, Andersen RA, Motter BC. The inuence of attentive xation upon the excitability of the light-sensitive neurons of the posterior parietal cortex. J Neurosci 1981; 1: 121825. OCraven KM, Rosen BR, Kwong KK, Treisman A, Savoy RL. Voluntary attention modulates fMRI activity in human MT-MST. Neuron 1997; 18: 5918. Olshausen BA, Anderson CH, Van Essen DC. A neurobiological model of visual attention and invariant pattern recognition based on dynamic routing of information. J Neurosci 1993; 13: 470019. Pardo JV, Fox PT, Raichle ME. Localization of a human system

1294

C. Buchel et al.
Sheliga BM, Riggio L, Rizzolatti G. Orienting of attention and eye movements. Exp Brain Res 1994; 98: 50722. Shipp S, Zeki S. Segregation and convergence of specialised pathways in macaque monkey visual cortex. J Anat 1995; 187: 54762. Talairach J, Tournoux P. Coplanar stereotaxic atlas of the human brain. Stuttgart: Thieme; 1988. Tanaka K, Hikosaka K, Saito H, Yukie M, Fukada Y, Iwai E. Analysis of local and wide-eld movements in the superior temporal visual areas of the macaque monkey. J Neurosci 1986; 6: 13444. Tootell RB, Reppas JB, Kwong KK, Malach R, Born RT, Brady TJ, et al. Functional analysis of human MT and related visual cortical areas using magnetic resonance imaging. J Neurosci 1995; 15: 321530. Tootell RBH, Mendola JD, Hadjikhani NK, Ledden PJ, Liu AK, Reppas JB, et al. Functional analysis of V3A and related areas in human visual cortex. J Neurosci 1997; 17: 706078. Treisman AM and Gelade G. A feature-integration theory of attention. Cognitive Psychology 1980; 12, 97136. Treue S, Maunsell JH. Attentional modulation of visual motion processing in cortical areas MT and MST. Nature 1996; 382: 53941. Vandenberghe R, Dupont P, De Bruyn B, Bormans G, Michiels J, Mortelmans L, et al. The inuence of stimulus location on the brain activation pattern in detection and orientation discrimination. A PET study of visual attention. Brain 1996; 119: 126376. Van Essen DC, Maunsell JHR. Hierarchical organization and functional streams in the visual cortex. Trends Neurosci 1983; 6: 3705. Watson JD, Myers R, Frackowiak RS, Hajnal JV, Woods RP, Mazziotta JC, et al. Area V5 of the human brain: evidence from a combined study using positron emission tomography and magnetic resonance imaging. Cereb Cortex 1993; 3: 7994. Worsley KJ, Friston KJ. Analysis of fMRI time-series revisited again. Neuroimage 1995; 2: 17381. Zeki S, Watson JD, Lueck CJ, Friston KJ, Kennard C, Frackowiak RS. A direct demonstration of functional specialization in human visual cortex. J Neurosci 1991; 11: 6419.

for sustained attention by positron emission tomography. Nature 1991; 349: 614. Paus T. Location and function of the human frontal eye-eld: a selective review. [Review]. Neuropsychologia 1996; 34: 47583. Petrides M. Specialized systems for the processing of mnemonic information within the primate frontal cortex. [Review]. Phil Trans R Soc Lond B Biol Sci 1996; 351: 145562. Petrides M, Pandya DN. Projections to the frontal cortex from the posterior parietal region in the rhesus monkey. J Comp Neurol 1984; 228: 10516. Pierrot-Deseilligny C, Gray F, Brunet P. Infarcts of both inferior parietal lobules with impairment of visually guided eye movements, peripheral visual inattention and optic ataxia. Brain 1986; 109: 8197. Posner MI. Attention: the mechanisms of consciousness. [Review]. Proc Natl Acad Sci USA 1994; 91: 7398403. Posner MI, Petersen SE. The attention system of the human brain. [Review]. Annu Rev Neurosci 1990; 13: 2542. Rajkowska G, Goldman-Rakic PS. Cytoarchitectonic denition of prefrontal areas in the normal human cortex: II. Variability in locations of area 9 and 46 and relationship to the Talairach Coordinate System. Cereb Cortex 1995; 5: 32237. Rees G, Frackowiak R, Frith C. Two modulatory effects of attention that mediate object categorization in human cortex. Science 1997; 275: 8358. Rizzolatti G, Matelli M, Pavesi G. Decits in attention and movement following the removal of postarcuate (area 6) and prearcuate (area 8) cortex in macaque monkeys. Brain 1983; 106: 65573. Rizzolatti G, Riggio L, Dascola I, Umilta C. Reorienting attention across the horizontal and vertical meridians: evidence in favor of a premotor theory of attention. Neuropsychologia 1987; 25: 3140. Rockland KS, Pandya DN. Laminar origins and terminations of cortical connections of the occipital lobe in the rhesus monkey. Brain Res 1979; 179: 320. Sandell JH, Schiller PH. Effect of cooling area 18 on striate cortex cells in the squirrel monkey. J Neurophysiol 1982; 48: 3848. Selemon LD, Goldman-Rakic PS. Common cortical and subcortical targets of the dorsolateral prefrontal and posterior parietal cortices in the rhesus monkey: evidence for a distributed neural network subserving spatially guided behavior. J Neurosci 1988; 8: 404968.

Received October 18, 1997. Revised January 16, 1998. Accepted February 9, 1998