Brain Advance Access originally published online on April 27, 2005
Brain 2005 128(12):2843-2857; doi:10.1093/brain/awh522
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Attentional responses to unattended stimuli in human parietal cortex
1 Neurology Department, 2 Cognitive Neurology Laboratory, Experimental Neurology Division, KU Leuven, 3 Psychology Faculty, KU Leuven, 4 Radiology Department, University Hospital, Gasthuisberg and 5 Laboratorium voor Neuro- en Psychofysiologie, KU Leuven, Belgium
Correspondence to: Rik Vandenberghe, MD, PhD, Neurology Department, University Hospital Gasthuisberg, Herestraat 49, 3000 Leuven, Belgium E-mail: rik.vandenberghe{at}uz.kuleuven.ac.be
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
Right-sided parietal lesions lead to lateralized attentional deficits which are most prominent with bilateral stimulation. We determined how an irrelevant stimulus in the unattended hemifield alters attentional responses in parietal cortex during unilateral orienting. A trial consisted of a central spatial cue, a delay and a test phase during which a grating was presented at 9° eccentricity. Subjects had to discriminate the orientation of the grating. The unattended hemifield was either empty or contained a second, irrelevant grating. We carried out a series of functional MRI (fMRI) studies in 35 healthy volunteers (13 men and 22 women, aged between 19 and 30 years) as well as a behavioural and structural lesion mapping study in 17 right-hemispheric lesion patients, 11 of whom had neglect. In the patients with but not in those without neglect, the addition of a distractor in the unattended hemifield significantly impaired performance if attention was directed contralesionally but not if it was directed ipsilesionally. In the healthy volunteers, we discerned two functionally distinct areas along the posterior–anterior axis of the intraparietal sulcus (IPS). The posterior, descending IPS segment in both hemispheres showed attentional enhancement of responses during contralateral attentional orienting and was unaffected by the presence of an irrelevant stimulus in the ignored hemifield. In contrast, the right-sided horizontal IPS segment showed a strong attentional response when subjects oriented to a stimulus in the relevant hemifield and an irrelevant stimulus was simultaneously present in the ignored hemifield, compared with unilateral stimulation. This effect was independent of the direction of attention. The symmetrical left-sided horizontal IPS segment showed the highest responses under the same circumstances, in combination with a contralateral bias during unilateral stimulation conditions. None of the six patients without neglect had a lesion of the horizontal IPS segment. In four of the 11 neglect patients, the lesion overlapped with the horizontal IPS activity cluster and lay in close proximity to it in another four. The remaining three patients had a lesion at a distance from the parietal cortex. Our findings reconcile the role of the IPS in endogenous attentional control with the clinically significant interaction between direction of attention and bilateral stimulation in right parietal lesion patients. Functional imaging in neglect patients will be necessary to assess IPS function in those cases where the structural lesion spares the middle IPS segment.
Key Words: spatial cognition; vision; lesion studies; fMRI; extinction
Abbreviations: EPI = echoplanar image; fMRI = functional MRI; IPL = inferior parietal lobule; IPS = intraparietal sulcus
Received October 11, 2004. Revised March 11, 2005. Accepted April 5, 2005.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Visual extinction and neglect are frequent clinical signs of focal brain injury, especially when the lesion encompasses the inferior parietal (Vallar et al., 1993
, 1994; Mort et al., 2003) or posterior temporal (Karnath et al., 2001, 2003, 2004) cortex, with, at least for neglect, a clear right-hemispheric dominance (Weintraub and Mesulam, 1987; Heilman et al., 1993; Mesulam, 2000). From a cognitive processing point of view, visual extinction and neglect are striking clinical examples of how stimuli compete for attentional selection (Desimone and Duncan, 1995; Duncan et al., 1999; Driver and Vuilleumier, 2001; Pessoa et al., 2003). Patients with visual extinction are able to detect a contralesional stimulus when it is presented on its own but not when a second ipsilesional stimulus is added (Bender, 1952; Mattingley, 2002), despite the fact that visual afferents and early visual processing areas are functionally intact (Rees et al., 2000; Driver and Vuilleumier, 2001; Vuilleumier et al., 2001). In standard clinical neglect tests, such as the target cancellation tasks (Wilson et al., 1987; Heilman et al., 1993; Mesulam, 2000), both contra- and ipsilesional stimuli are present simultaneously, but when ipsilesional stimuli are experimentally removed, this markedly improves performance (Mark et al., 1988; Kaplan et al., 1991). Neglect and visual extinction therefore highlight the critical importance of two major factors in the control of spatially distributed attention: the direction of attention, leftward or rightward, and the type of stimulation, unilateral or bilateral (Mark et al., 1988; Eglin et al., 1989).
In healthy volunteers, attention studies that compared uni- and bilateral stimulation point to superior parietal involvement. In a transcranial magnetic stimulation (TMS) study, subjects had to detect stimuli that could appear to the left, to the right or bilaterally (Pascual-Leone et al., 1994). Stimulation over the left and right intraparietal sulcus (IPS) led to omissions of contralateral stimuli under bilateral but not under single stimulation conditions (Pascual-Leone et al., 1994) as well as improvement of detection rates ipsilaterally (Hilgetag et al., 2001). Using PET, Fink et al. (2000) presented subjects with a row of three letters in the left or in right hemifield or bilaterally for 100 ms and asked subjects to report as many letters as possible. The contrast between bilateral and unilateral stimulation revealed bilateral superior parietal activation (Fink et al., 2000).
Neuroimaging studies until now have studied the effect of direction of attention (leftward or rightward) under either bilateral (Heinze et al., 1994; Vandenberghe et al., 1997, 2000; Hopfinger et al., 2000; Macaluso et al., 2002; Yantis et al., 2002) or unilateral stimulation conditions (Corbetta et al., 1993; Nobre et al., 1997; Woldorff et al., 2004). They have provided robust evidence for modulation of ventral occipital cortex by the direction of attention (e.g. Heinze et al., 1994; Tootell et al., 1998; Martinez et al., 1999). Surprisingly, direct comparisons between left-sided and right-sided attention yielded relatively weak and inconsistent parietal differences (Corbetta et al., 1993; Nobre et al., 1997; Vandenberghe et al., 2000, 2001a; Macaluso et al., 2002).
Our aim was 2-fold: first, to determine in the intact brain the neural consequences of the presence of unattended stimuli in the ignored hemifield during unilateral orienting. Secondly, through the combined study of direction of attention (leftward versus rightward) and unilateral versus bilateral stimulation in patients and in healthy controls, we wanted to resolve the discrepancy between patient lesion data, implicating mainly the inferior parietal lobule (IPL) in unilateral orienting, and neuroimaging data, which show the most reliable direction-sensitive effects occipitally.
We used orientation discrimination and peripheral gratings given the extensive neuroimaging data available on this task and stimuli (Vandenberghe et al., 1996, 1997, 2000). Trials consisted of a central spatial cue pointing leftwards or rightwards, a delay and a test phase (Fig. 1A). During the test phase, subjects had to discriminate the orientation of the peripheral grating in the cued hemifield. Subjects had to fixate the centre of the screen: the direction of attention, leftward or rightward, and the attended hemifield, left or right, coincided. The grating was presented either on its own or together with a second, irrelevant grating in the opposite, unattended hemifield.
|
We validated our paradigm in stroke patients living at a rehabilitation unit (Geeraerts et al., 2005
). Fifteen right-hemispheric lesion patients suffering from visual neglect (Heilman et al., 1993; Mesulam, 2000) 1–34 months after stroke participated. Neglect was defined as impairment on two out of three tests: star cancellation (Wilson et al., 1987), line bisection (Schenkenberg et al., 1980) or lateralized errors during spontaneous behaviour at the rehabilitation unit. Seven also had clinical extinction. The three control groups consisted of 10 right-hemispheric lesion patients without neglect, 14 left-hemispheric lesion patients without neglect and 14 age-matched healthy controls. All 15 neglect patients showed significantly higher orientation discrimination thresholds during contralesional orienting when an ipsilesional grating was also present compared with the conditions where the contralesional stimulus was presented on its own or where subjects oriented ipsilesionally. The magnitude of the ipsilesional interference effect did not differ between neglect subjects with or without extinction. This clearly demonstrates that our paradigm captures a spatial–attentional process that is impaired in neglect patients (Geeraerts et al., 2005). Once we had established the external validity of our paradigm, we conducted three functional MRI (fMRI) experiments in healthy individuals: The main experiment (Fig. 1A), conducted on a 1.5 T MRI scanner, involved 23 subjects. A sensory control experiment (Fig. 1B) involved 16 of these subjects. A replication study on a 3 T MRI scanner where we controlled for differences in task performance between uni- and bilateral stimulation conditions involved 12 additional subjects. In this manner, we dissociated sensory from attentional effects and examined how suppression of irrelevant events in the unattended hemifield affects responses during attentional orienting to the left or to the right.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Subjects
Thirteen healthy women and 10 men, between 19 and 30 years of age, participated in the main fMRI experiment. Sixteen of these also took part in the sensory control experiment and four also in a motion localizing experiment. Twelve additional subjects, three men and nine women, between 19 and 26 years of age, participated in the replication experiment. All participants were strictly right-handed, free of psychotropic or vasoactive medication, had no neurological or psychiatric history, and had a normal brain MRI.
All participants gave written informed consent in accordance with the Declaration of Helsinki. The experiment was approved by the Ethische Commissie, University Hospital Gasthuisberg, Leuven.
fMRI experiments: stimuli and tasks
All fMRI experiments were conducted using SuperLab for PC version 2.0 (Cedrus, Phoenix, AR). For the main experiment as well as the sensory control experiment and the motion localizing study, visual stimuli were projected from a Barco 6300 LCD projector (1280 x 1024 pixels) onto a screen 28 cm in front of the subjects' eyes. Eye movements were monitored on-line using an MRI-compatible eye movement tracking device (Ober2, Permobil Meditech, Timra, Sweden) and stored for subsequent semi-quantitative analysis.
The replication experiment was conducted on a different, 3 T MRI scanner. Visual stimuli were projected from a Barco 6400i LCD projector (1024 x 768 pixels) onto a screen 36 cm in front of the subjects' eyes. Eye movements were monitored on-line using the Applied Science Laboratory infrared system (Waltham, MA), stored and analysed in an automatic, quantitative manner after filtering for eye blinks.
Each experiment was conducted in different runs, a run being defined as a continuous series of volume acquisitions. Subjects were instructed to fixate a 0.27° fixation dot that was present throughout the entire run. Each run contained 40 trials of each of four types plus 40 null events. Event onset asynchrony was 2250 ms. The event sequence was optimized so as to induce maximal signal variation (Burock et al., 1998).
In the main experiment (Fig. 1A), each event consisted of a central spatial cue pointing to the left or to the right (100 ms), a delay (300 ms) during which only the fixation point was shown, and a test phase (100 ms) during which a 4.96° square-wave grating (spatial frequency 0.60 cycles/°; mean luminance 194 cd/m2) was presented in the attended hemifield on the horizontal meridian. The grating's centre was positioned at 9.4° eccentricity. Its phase was randomly shifted to prevent use of cues other than orientation, and a 50% noise background was superimposed upon the grating. The grating's orientation was pseudorandomized between two alternative orientations centered around a 45° angle. Subjects had to press a left-hand or a right-hand key depending on the orientation of the grating.
The unattended hemifield was either empty or contained a grating at the symmetrical position of the relevant grating. This resulted in four conditions: left-sided attention, bilateral gratings; right-sided attention, bilateral gratings; left-sided attention, unilateral grating to the left; right-sided attention, unilateral grating to the right. In one-third of the bilateral cases, the orientation of the unattended grating matched the relevant grating's orientation, in one-third it corresponded to the alternative orientation and in one-third it was orientated more outward with respect to the relevant grating so that subjects could not solve the task by comparing the orientations of the two stimuli.
At the start of the training session of the main experiment, subjects had to discriminate an orientation difference of 16° (centered around a 45° oblique). This was lowered to 12° and, subsequently, to 8° when they performed above 75% correct on all conditions. The smallest orientation difference that allowed for a performance of 75% correct during training in all conditions was used during MRI scanning. During training, eye movements were monitored on-line using an infrared Viewpoint Eye Tracker (Arrington Research Inc., AR) and subjects were informed if saccades occurred during covert orienting.
In an additional behavioural study outside the fMRI scanner, we examined whether the spatial cue effectively directed the spatial expectancy of the subjects in the cued direction. The experimental conditions were identical to those used in the main experiment except for the inclusion of 20% invalid trials during the unilateral stimulation conditions.
In the sensory control experiment, we delineated the sensory fMRI responses to peripheral gratings during a central discrimination task (Fig. 1B). The peripheral stimuli were identical to the first experiment but attention was directed to the centre of the screen. The cue pointed upwards or downwards instead of left or right. Subjects were instructed to press left or right depending on the direction of the arrow, upward or downward. The cue duration was 70 ms instead of 100 ms and the delay was 200 ms instead of 300 ms. These changes in timing ensured that subjects were engaged in the central discrimination task when the peripheral stimuli came up. Otherwise, the first and second experiment were entirely identical.
In a replication experiment, we adapted the orientation difference so as to match performances between the unilateral and the bilateral stimulation conditions. This allowed us to rule out aspecific effects of task performance as an explanation. In the replication experiment, a trial consisted of a central spatial cue pointing to the left upper, left lower, right upper or right lower quadrant (duration: 200 ms) followed by a 200 ms interval. During the test phase (duration: 200 ms), a grating (4.29° diameter) appeared in the cued quadrant at 9.46° eccentricity and subjects had to press a left- or right-hand key depending on the grating's orientation. The grating's orientation was pseudorandomized between two alternative orientations, rotated a fixed angle either clockwise or counterclockwise with respect to the horizontal meridian. The intertrial interval was 3200 ms. The grating was presented alone or together with a second, irrelevant grating. This irrelevant grating appeared at 9.46° eccentricity in the symmetrical quadrant in the opposite hemifield. The distance between the relevant and the irrelevant grating was 13.28°. This resulted in four event types: left-sided attention, unilateral single stimulation; left-sided attention, bilateral stimulation; right-sided attention, unilateral single stimulation; and right-sided attention, bilateral stimulation. During the training session of the replication experiment, the starting values for the orientation differences were 8° (bilateral stimulation) and 6° (unilateral stimulation) and this was lowered to 6 and 4°, respectively, according to the same rules as those applied during training in the main experiment.
In a motion localizing test, subjects underwent two runs of five epochs of coherent motion alternating with epochs of a stationary stimulus under passive viewing conditions. Stimuli (7° diameter) consisted of circular random textured patterns consisting of 50% white dots on a black background. The random textured pattern remained stationary in one condition and moved coherently at 6°/s in the other condition. Stimulus details have been described elsewhere (Sunaert et al., 1999).
fMRI experiments: image acquisition
For the main experiment as well as the sensory control experiment and the motion localizing study, a 1.5 T Siemens Sonata system (Erlangen, Germany) equipped with a head volume coil provided T1 anatomical volume images [repetition time (TR) 1950 ms, echo time (TE) 3.93 ms, in-plane resolution 1 mm] and T2* echoplanar images (EPIs) with blood oxygenation level-dependent (BOLD) contrast. EPIs (TR 2 s, TE 47 ms) comprised 23 axial slices acquired continuously in ascending order (voxel size 3 x 3 x 5 mm3). A total of 228 volumes were acquired during each run, covering the cerebrum entirely. The first six volumes were discarded to allow the MRI signal to reach steady state. Eight subjects underwent six runs, five subjects four runs, seven subjects three runs and two subjects two runs. The difference in the number of runs is due to the fact that, within the same session, some subjects also participated in the sensory control or the motion localizing experiment, and also because data had to be discarded due to motion or scanner artefacts.
For the replication experiment, a 3 T Philips Intera system (Best, The Netherlands) equipped with a head volume coil provided T1 anatomical volume images (TR = 1975 ms, TE = 3.93 ms, in-plane resolution 1 mm) and T2* EPIs with BOLD contrast. EPIs (TR 2 s, TE 30 ms) comprised 36 axial slices acquired continuously in ascending order (voxel size 2.75 x 2.75 x 3.75 mm3). A total of 180 volumes were acquired during each run, covering the cerebrum entirely. The first four volumes were discarded to allow the MRI signal to reach steady state. All subjects underwent four runs.
fMRI experiments: image analysis
Analysis was performed using Statistical Parametric Mapping version 2002 (Wellcome Department of Cognitive Neurology, London, UK). Following correction for differences in timing of slice acquisition within a volume, EPI volumes were realigned and resliced using sinc interpolation. A mean EPI volume was obtained during realignment and the structural MRI was co-registered with that mean volume. The structural scan was normalized to the Montreal Neurological Institute (MNI) T1 template in Talairach space (Talairach and Tournoux, 1988; Friston et al., 1995) using non-linear basis functions. The same deformation parameters were applied to the EPI volumes. The EPI volumes were spatially smoothed using a 5 x 5 x 7 mm3 filter. Data from different runs were proportionally scaled to a grand mean of 100 arbitrary units to account for overall differences in the intensity of whole brain volumes across the time series. The time series for each voxel were high-pass filtered to (1/128) Hz. The event-related response, synchronized with the acquisition of the top slice, was modelled by a canonical haemodynamic response function (HRF), consisting of a mixture of two gamma functions (Friston et al., 1999). The temporal derivative of the HRF was also included in the model. Statistical inference was corrected for intrinsic autocorrelations. A statistical parametric map of the t statistic for the parameter estimates was generated and subsequently transformed to a Z map. Data were analysed using a random effects general linear model. For the main experiment and the replication experiment, one contrast image per individual was calculated for each of the following a priori comparisons: (i) (right-sided attention, bilateral stimulation + right-sided attention, unilateral stimulation) – (left-sided attention, bilateral stimulation + left-sided attention, unilateral stimulation) and the inverse; (ii) right-sided attention, bilateral stimulation – left-sided attention, bilateral stimulation, and the inverse; (iii) right-sided attention unilateral stimulation – left-sided attention unilateral stimulation, and the inverse; (iv) (right-sided attention, bilateral stimulation + left-sided attention, bilateral stimulation) – (right-sided attention, unilateral stimulation + left-sided attention, unilateral stimulation), and the inverse; (v) right-sided attention, bilateral stimulation – right-sided attention, unilateral stimulation, and the inverse; and (vi) left-sided attention, bilateral stimulation – left-sided attention, unilateral stimulation, and the inverse.
Before proceeding to the second level of analysis, we explicitly tested homogeneity of the sample subjects' data sets by means of DISTANCE software, available at www.madic.org (Kherif et al., 2003): the data did not show any outlier or clustering. It was therefore statistically sound to represent the group results through the mean using a random effects analysis (Kherif et al., 2003). We weighted the individuals' contrast images for the total number of runs per subject. We examined for each of the a priori comparisons whether, on average, the contrast images revealed significant differences (one-sample t test).
In a separate analysis, we tested whether the differences between conditions differed between the main and the sensory control experiment. We calculated for each of the 16 individuals who participated in the two experiments the following contrast images, using data from the same number of runs in the two experiments: (vii) (right-sided attention, unilateral stimulation – left-sided attention, unilateral stimulation) – (passive viewing, right-sided unilateral stimulation – passive viewing, left-sided unilateral stimulation), and the inverse; and (viii) [(right-sided attention, bilateral stimulation + left-sided attention, bilateral stimulation) – (right-sided attention, unilateral stimulation + left-sided attention, unilateral stimulation)] – [(passive viewing, bilateral stimulation + passive viewing, bilateral stimulation) – (passive viewing, right-sided unilateral stimulation + passive viewing, left-sided unilateral stimulation)].
The significance map for the group random effects analysis was thresholded at a voxel level inference threshold of P < 0.001 uncorrected combined with a cluster-level inference threshold of P < 0.05 corrected for the entire brain search volume. Except when mentioned otherwise, P values reflect cluster level inference.
Structural lesion mapping
Clinical scans that could be used for lesion mapping were available in 11 out of the 15 right-hemispheric lesion patients with clinical neglect who participated in the study of Geeraerts et al. (2005) (two MRI and nine CT scans) and in six out of the 10 right-hemispheric patients without neglect (three MRI and three CT scans) (Table 1). Using MRIcro (Rorden and Brett, 2000) (www.mricro.com), the lesions were manually outlined onto an intact brain normalized to the MNI template. For the patients with (Fig. 6A and D) and for the patients without neglect (Fig. 6B), the colour density-coded lesion maps were overlaid with the functional images from the contrast between bilateral and unilateral stimulation (contrast iv) (Fig. 6A and B) and from the contrast between left-sided and right-sided attention (contrast i inverse) (Fig. 6D). We also mapped the difference of lesions between neglect and non-neglect patients (Fig. 6C and E) and superimposed onto this map the functional images from the contrast between bilateral and unilateral stimulation (contrast iv) (Fig. 6C) and from the contrast between left-sided and right-sided attention conditions (contrast i inverse) (Fig. 6E).
|
To analyse the congruence between the structural lesion maps and the activity maps in a statistical manner, we derived a total lesion volume from the sum of the lesions of all neglect patients and used this as a search volume for the contrasts iv and i inverse, correcting the significance levels for this volume of interest.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Behavioural data
In the main fMRI experiment (Fig. 1A), reaction times and accuracies were analysed using repeated-measures analysis of variance (ANOVA) with two within-subject factors: direction of attention (two levels: left-sided versus right-sided) and type of stimulation (two levels: bilateral versus unilateral). Subjects were significantly less accurate [F(1,22) 24.7, P < 0.001] and slower [F(1,22) 35.2, P < 0.001] with bilateral compared with unilateral stimulation (Table 2A). There was no main effect of direction of attention (P > 0.3). Analysis of accuracy revealed a significant interaction between direction of attention and type of stimulation [F(1,22) 8.65, P < 0.01]: when subjects attended to the right compared with when they attended to the left, the addition of a stimulus in the irrelevant hemifield elicited a larger cost compared with single stimulus presentation (Table 2A). Subjects fixated well during MRI scanning: across the entire duration of a run (7.5 min), they made on average 2.48 (SD 1.51) saccades (varying between zero and maximally six saccades per run).
|
In the additional behavioural study, reaction times were significantly slower following an invalid cue [valid cue to the left, 678 ms (SE 60 ms); invalid cue to the left, 889 ms (SE 39 ms), valid cue to the right, 698 ms (SE 50 ms), invalid cue to the right, 821 ms (SE 83 ms), F(1,6) 50, P < 0.001]. This demonstrates that subjects developed a cue-related spatial expectancy prior to presentation of the test stimuli.
In the sensory control fMRI experiment (Fig. 1B), accuracies or reaction times did not differ between conditions (Table 2B).
In the replication study, reaction times or accuracies did not differ significantly between conditions (Table 2C). Globally, accuracies were higher than in the main experiment. This is related to differences in timing parameters as well as in the reference orientation (45° in the main experiment, 90° in this experiment). The number of leftward or rightward saccades (amplitude >2°) did not differ between conditions: left-sided attention, bilateral stimulation events, 0.29 saccades per run (SE 0.26); right-sided attention, bilateral stimulation events, 0.32 (SE 0.22); left-sided attention, unilateral stimulation events, 0.31 (SE 0.32); right-sided attention, unilateral stimulation events, 0.23 (SE 0.38) [F(7,6) = 0.68, P > 0.5].
Neuroimaging data
Right-sided versus left-sided attention
In the main experiment (Fig. 1A), right-sided attention activated the posterior, descending segment of the left IPS compared with left-sided attention (–21,–87,30, Z = 4.55, extent 87, corrected P < 0.001) (Fig. 2A and B) (Duvernoy, 1999) (contrast i). The inverse contrast, left-sided minus right-sided attention, activated the symmetrical right-sided region (15,–87,33, Z = 3.96, ext. 54, corrected P < 0.001) (Fig. 2E and F) (contrast i inverse). Both clusters also contained more inferior (left minus right, 27,–90,27, Z = 4.01; right minus left, –18,–93,24, Z = 4.52) and more superior peaks of activation (left minus right, 30,–84,36, Z = 3.94; right minus left, –21,–87,39, Z = 3.97). Figure 2C and G shows the evoked response in this left and right posterior IPS segment, respectively, averaged over all subjects and all voxels belonging to these clusters and located at a z-level 
27 mm (45 voxels to the right, 60 voxels to the left). The anatomical criterion ensured that the analysis was restricted to the IPS component, excluding the more inferior component (Duvernoy, 1999). There are three important findings: first, the response during contralateral orienting with a single contralateral stimulus was almost identical to that obtained during contralateral orienting with bilateral stimulation (Fig. 2C, magenta dash-dotted line versus red dashed line; Fig. 2G, blue solid versus green dotted line) (contrasts v and vi). In other words, responses were unaffected by the presence of a second stimulus in the ipsilateral, unattended hemifield. As will be clear from the following section, this contrasted with the middle, horizontal IPS segment where the presence of a stimulus in the unattended hemifield made a clear difference (Fig. 5).
|
Secondly, under bilateral stimulation conditions, the response was higher during contralateral compared with ipsilateral orienting (left minus right, 18,–84,33, Z = 4.42, ext. 31, corrected P < 0.005, Fig. 2G, blue solid versus red dashed; right minus left, –21,–87,27, Z = 3.02, uncorrected P < 0.005 voxel level inference, Fig. 2C, red dashed versus blue solid) (contrasts ii and ii inverse). This differs from the profiles obtained during the sensory control experiment (Fig. 2D and H) where the two bilateral stimulation conditions (blue solid and red dashed) showed a narrow overlap.
Thirdly, under unilateral stimulation conditions, the difference between left-sided and right-sided attention conditions (green versus magenta, Fig. 2C and G) is larger than the difference between left-sided and right-sided stimulation in the sensory control experiment (green versus magenta, Fig. 2D and H). The attentional enhancement of the sensory response during the main experiment was confirmed by the presence of a bilateral interaction between stimuli (left-sided versus right-sided stimulation) and task (peripheral versus central discrimination) (interaction at –21,–87,30, Z = 2.94; at 15,–87,33, Z = 2.38) (contrasts vii and vii inverse). According to a motion localizing experiment, this region corresponded to what has been previously labelled VIPS (Orban et al., 1999) [Fig. 3A (right minus left), overlap marked in yellow].
|
The response profile in the posterior tip of the descending segment of the IPS was very similar to that seen in ventral occipital areas. Figure 4A and C shows the right and left ventral occipital activations in the main experiment during left-sided minus right-sided attention (30,–69,–12, Z = 5.39, ext. 397, corrected P < 0.001) (contrast i inverse) and during right-sided minus left-sided attention conditions (–27,–60,–9, Z = 5.73, ext. 654, corrected P < 0.001), respectively (contrast i). Responses during the sensory control experiment are shown in Fig. 4B and D, respectively. In the main experiment, responses were enhanced when attention was directed contralaterally (Fig. 4A, red dashed versus blue solid; Fig. 4C, blue solid versus red dashed) (contrasts ii and ii inverse). Importantly, responses were not affected by the presence of an ipsilateral irrelevant stimulus (Fig. 4A, red dashed versus magenta dash-dotted; Fig. 4C, blue solid versus green dotted) (contrasts v and vi).
|
Bilateral versus unilateral stimulation
In clear contrast to the posterior IPS segment (Fig. 2), the horizontal, middle IPS segment to the right responded strongly to the simultaneous presence of an irrelevant stimulus in the unattended hemifield compared with single stimulation conditions, regardless of the direction of attention (30,–72,36, Z = 4.28, 30,–72,48, Z = 3.62, ext. 21, corrected P < 0.05) (Fig. 5E and F, blue solid and red dashed versus green dotted and magenta dash-dotted) (contrast iv). A second activity cluster in the horizontal IPS segment (39,–54,51, Z = 3.79, ext. 11) remained below the significance threshold.
|
The horizontal IPS region did not show activity increases in the sensory control experiment (Fig. 5G), even at a voxel level threshold of uncorrected P < 0.05. The presence of an interaction between stimuli (bilateral versus unilateral) and task (peripheral versus central discrimination) (interaction at 30,–72,36, Z = 3.32; at 30,–72,48, Z = 3.27) (contrast viii) confirmed the task-dependent attentional nature of this activation.
The simultaneous presence of an irrelevant stimulus in the ignored hemifeld also caused a significant main effect in the symmetrical left-hemispheric region (–30,–60,48, Z = 4.28, ext. 50, corr. P < 0.001) (Fig. 5B and C) (contrast iv). Again, no activity increases were obtained in the sensory control experiment in this region (Fig. 5D) and there was an interaction between stimuli (bilateral versus unilateral) and task (peripheral versus central discrimination) (interaction at –30,–60,48, Z = 3.67) (contrast viii). This region lay laterally to the motion-responsive region DIPSM (Orban et al., 1999) (Fig. 3B).
Both in the left and in the right hemisphere, the horizontal IPS segment showed the highest responses during bilateral stimulation conditions (Fig. 5A–C, E and F). There was, however, a significant inter-hemispheric difference. To the left, the horizontal IPS segment showed a strong contralateral attentional bias during unilateral stimulation conditions (Fig. 5C, magenta versus green). This bias was much weaker or absent in the right hemispheric middle IPS segment (Fig. 5F, green versus magenta). To examine this inter-hemispheric difference statistically, we carried out a repeated-measures ANOVA on the peak response amplitudes in the horizontal IPS cluster with two factors: hemispheric side (left-hemispheric versus right-hemispheric) and side of stimulation (contralateral single stimulation versus ipsilateral single stimulation). The interaction effect was significant [F(1,22) 19.57, P < 0.001]: The difference between contra- and ipsilateral orienting was significantly larger in the left hemisphere than in the right hemisphere [F(1,22) 19.6, P < 0.0005].
In a third, replication study on a 3 T MRI scanner, we adapted orientation differences per event type so that accuracies were matched between the unilateral and the bilateral stimulation conditions (Table 2C). Again, the contrast of bilateral minus unilateral stimulation yielded significant activation of the horizontal segment of the IPS bilaterally (36,–54,54, Z = 4.23, ext. 24, corrected P < 0.001; 30,–63,36, Z = 3.75, ext. 16, corrected P <0.01; –27,–63,45, Z = 3.92, ext. 24, corrected P < 0.001), replicating the results of the main experiment. Differences in task performance therefore cannot account for the IPS effects.
Structural lesion maps versus activity maps
The maximal structural lesion overlap in neglect was localized to the IPL (Fig. 6A and C), in agreement with previous studies (Vallar et al., 1993; Mort et al., 2003). The activity cluster in the right horizontal IPS segment (contrast iv) overlapped with the lesion of four out of 11 patients (cases 1 and 8–10) and lay in close proximity to the lesion in four other cases (cases 2, 6, 7 and 11) (Fig. 6A and C). The remaining three patients had a lesion of the basal ganglia in isolation (case 4) or combined with an inferior frontal (case 3) or a posterior temporal lesion (case 5). Cases 3 and 5 experienced the lowest interference by an ipsilesional distractor of the entire neglect group (Table 1). None of the patients without neglect had a lesion that overlapped with the middle IPS activity cluster (Fig. 6B and C).
|
In addition to the visual overlay, we analysed the activations within the total lesion volume of the neglect patients statistically (Fig. 6A). The activation of the right middle IPS segment during bilateral versus unilateral stimulation conditions (contrast iv) was significant (30,–72,36, ext. 21, corrected P < 0.01). During left-sided minus right-sided attention, the activation of the posterior IPS segment was also significant (contrast i inverse) (27,–90,27, Z = 4.01, ext. 18, corrected P < 0.01) together with activation of the horizontal posterior segment of the superior temporal sulcus (45,–66,15, Z = 3.85, ext. 31, corrected P < 0.005) (Fig. 6D and E).
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
The most striking finding of this study is the strong attentional effect that an irrelevant stimulus in the ignored hemifield exerts upon activity of the horizontal segment of the IPS during unilateral orienting (Fig. 5B, C, E and F). This activity focus lay within a volume of interest defined on the basis of lesions of patients with neglect in whom an irrelevant ipsilesional stimulus caused pathological interference during left-sided attention (Fig. 6A and C). As our second main finding, direction of attention (to the left or to the right hemifield) and type of stimulation exerted dissociable effects along the posterior–anterior axis of the IPS (Fig. 2 versus Fig. 5).
Aspecific effects of task performance can be excluded as an explanation. In a replication experiment in 12 novel subjects, we matched performances by adapting the orientation differences that subjects had to discriminate during unilateral or bilateral stimulation conditions (Table 2C). Strong activation of the horizontal IPS segment was present under these circumstances as well. Neither can sensory effects account for the difference. We ran the same stimuli in a sensory control experiment where subjects had to carry out a central discrimination task. In this sensory control experiment, we used a relatively easy central discrimination task so as to maximize the chance of finding sensory differences (Rees et al., 1999). The horizontal IPS segment did not show a measurable sensory response (Fig. 5D and G). The posterior IPS segment showed a sensory response to contralateral stimulation. In the posterior IPS segment, therefore, the attention-related effect consisted of attentional enhancement (Fig. 2C and G) of a sensory response to contralateral stimulation (Fig. 2D and H).
Human IPS contains a number of motion-sensitive areas (Orban et al., 1999; Sunaert et al., 1999; Bremmer et al., 2001). In four subjects, we localized the IPS foci with respect to these motion-responsive areas (Fig. 3). The posterior descending IPS segment was motion sensitive (Fig. 3A) and probably corresponds to VIPS, which lies superior to V3A (Orban et al., 1999). In comparison with previous activations obtained in studies comparing left-sided and right-sided attention, the coordinates of activation in the left posterior descending IPS segment lay slightly superior to the superior occipital activation obtained in comparisons between left-sided and right-sided attention using bilateral stimulation (Hopfinger et al., 2000; Macaluso et al., 2002) and slightly posterior to the posterior IPS activation reported by Yantis et al. (2002).
The activated region in the horizontal IPS segment was not motion responsive and lay anteriorly and laterally with respect to DIPSM (Orban et al., 1999) (Fig. 3B). It lay inferior to a parietal region activated during visual search when the number (Nobre et al., 2003) or heterogeneity of distractors (Wilkinson et al., 2002) is increased. It lay lateral to the medial superior parietal area implicated by Pollman et al. (2003) in rejection of old stimuli in favour of novel ones. Our region lay lateral to the region of activation reported by Koyama et al. (–20,–63,49; 19,–63,49) during visually guided saccades compared with fixation (Koyama et al., 2004). It overlapped with an area (32,–68,46) that is activated during a delayed saccade paradigm in a retinotopic fashion, a putative homologue of area LIP in monkeys (Sereno et al., 2001).
Area LIP in monkeys is activated during the cue phase, the delay and the execution phase of visually guided saccades in a retinotopically specific manner (Colby et al., 1996; Gottlieb and Goldberg, 1999). To our knowledge, no data are available on the effect of competing stimuli within or outside a neuronal receptive field in LIP during covert peripheral discrimination tasks. We propose that the response of this region to bilateral stimulation reflects the selection of an attentional target among different potential targets. This interpretation is in line with neurophysiological evidence of a role for LIP in the representation of behavioural and visual salience (Gottlieb et al., 1998; Gottlieb and Goldberg, 1999).
The role of IPS in processing bilateral stimuli fits with studies of TMS over the IPS in healthy volunteers (Pascual-Leone et al., 1994; Hilgetag et al., 2001). It confirms PET findings of superior parietal activation with bilateral stimulation (Fink et al., 2000) and, at the higher anatomical accuracy provided by fMRI, implicates the horizontal IPS segment rather than the superior parietal lobule.
According to our patient validation study, behavioural performance on this paradigm is closely linked to clinical neglect. Our paradigm may be closer to the classical neglect tasks than to clinical extinction in so far that it probes focal rather than divided attention (Mesulam, 2000). Clinical tests of neglect are sensitive to the simultaneous presence of stimuli in the ipsilesional hemispace (Mark et al., 1988; Kaplan et al., 1991).
All four patients in whom the lesion overlapped with the middle IPS activity cluster had neglect (Fig. 6A). Conversely, in seven out of 11 neglect patients, the lesion did not overlap with the middle IPS activity cluster. Several factors account for this lack of concordance. First, structurally intact cortex that borders a lesion does not necessarily function normally: white matter tracts leading up to the IPS may be interrupted and lesions of cortex that is connected to IPS may impact on IPS functionally. This factor applies to four patients who had a lesion in close proximity to the middle IPS activity cluster (Fig. 6A). Secondly, our statistical approach relies on a thresholding procedure. When we assessed the full extent of the middle IPS activation by lowering the threshold, a significantly larger proportion of neglect patients showed overlap with the activation of the horizontal IPS segment (eight out of 11 neglect patients) (Fig. 6A). In three patients (cases 3–5), the lesion was even not near the IPS or IPL. Two of these patients experienced the lowest interference by an ipsilesional distractor of the entire neglect group (Table 1). In neglect patients with lesions at a distance from the IPS, i.e. affecting inferior frontal cortex, Corbetta et al. (2003) reported decreased fMRI activity in the IPS during a spatial cueing paradigm that activates the same region in healthy volunteers (Nobre et al., 1997; Corbetta et al., 2003). Functional recovery was associated with reappearance of IPS activation (Corbetta et al., 2003). That study (Corbetta et al., 2003) demonstrates that dysfunctional regions in neglect patients extend beyond the regions that are structurally lesioned, even when these regions lie at relatively remote locations.
Lesions overlapped maximally in the IPL rather than the IPS. This replicates previous findings (Vallar et al., 1993; Mort et al., 2003) and fits with the prominence of the IPL in most theories of neglect (Heilman et al., 1993; Vallar et al., 1993). If the middle IPS segment mediates an attentional process that is impaired in neglect patients, why does IPL rather than IPS lie at the centre of lesion overlap? First, the attentional process we studied in neglect patients is mediated by the IPS like some of the other spatial–attentional deficits seen in neglect (Mannan et al., 2005), but other aspects of attentional impairment in neglect may be attributable to regions other than IPS. Neglect is a heterogeneous syndrome (Heilman et al., 1993; Bisiach and Vallar, 2000; Mesulam et al., 2000; Husain and Rorden, 2003; Mannan et al., 2005) and the exact spatial cognitive deficits may differ between neglect patients (Bisiach and Vallar, 2000), e.g. depending on the exact lesion site (Mannan et al., 2005). Functions that do not rely on IPS but on IPL and that are impaired in neglect are exogenous shifting (Posner et al., 1984; Corbetta et al., 2000; Rushworth et al., 2001) and spatial short-term memory (Vandenberghe et al., 2001b; Husain et al., 2003). Secondly, most of our neglect patients had suffered an ischaemic stroke. The outcome of our lesion mapping study therefore is influenced by the distribution of vascular territories and not only by regional function. When parietal branches of the right middle cerebral artery are occluded, a region at the centre of their vascular territory will be more systematically involved than one at the periphery given the between-subject variability of vascular territory boundaries (Duvernoy, 1999).
The absence of suppressive distractor-related effects in ventral occipital cortex (Fig. 4A and C) differs from what has been reported when multiple stimuli were presented within the same quadrant either simultaneously or sequentially (Kastner et al., 1998). This suggests that occipital competition effects arise within one quadrant or hemifield (Kastner et al., 2001) but not between hemifields. The absence of between-hemifield suppressive effects in occipital cortex (Fig. 4A and C) also differs from a previous PET experiment (Fink et al., 2000) that showed reduced striate and extrastriate activity to bilateral compared with unilateral stimulation. Eccentricities as well as stimulus duration were roughly similar between the two studies. We propose the following explanation: in our study, subjects had to attend exclusively to either the left or right hemifield, both in the bilateral and in the unilateral stimulation conditions. In the study by Fink et al. (2000), subjects attended exclusively to the stimuli presented in one hemifield during the unilateral stimulation conditions but, in the bilateral stimulation conditions, relevant stimuli were present in the two hemifields. Possibly, the difference in attentional resources spent on a given side in the study by Fink et al. (2000) during bilateral compared with unilateral stimulation explains the lower activity in the average of the bilateral stimulation compared with the left-sided or the right-sided unilateral stimulation conditions in that study. This effect would be reinforced further by the blocked trial presentation (Kastner et al., 1998; Driver and Frith, 2000) and the fact that the bilateral stimulation conditions were always analysed together regardless of side priority (Fink et al., 2000). According to this interpretation, the difference in results relates to the distinction between divided attention (Fink et al., 2000) and focal attention, as tested by our paradigm.
The inter-hemispheric differences in the horizontal IPS segment may help us understand the higher prevalence of neglect with right- compared with left-hemispheric lesions. In the horizontal IPS segment of the right hemisphere, activity levels reflect what happens in the two hemifields: if activity levels are high, between-hemifield competition is present (Fig. 5F). In combination with the lateralized activity in posterior IPS (Fig. 2G), this provides sufficient information to distribute attentional weights to stimuli in both hemifields depending on their behavioural pertinence (Duncan and Humphreys, 1989). In the left hemisphere, high activity levels in the horizontal IPS segment may imply bilateral stimulation or contralateral single stimulation (Fig. 5C), thus creating ambiguity. If the right hemisphere is lesioned, the contralateral orienting bias of the left descending (Fig. 2C) and horizontal IPS segments (Fig. 5C) leads to a lateralized attentional imbalance in favour of the right hemispace. If the left hemisphere is lesioned, the direction-independent response in the right-sided horizontal IPS segment (Fig. 5F) may prevent the imbalance between hemifields. This model relates to the right-hemispheric dominance model proposed by Weintraub et al. (1987) (Anderson, 1996; Mesulam, 2000). Our data are in line with this model and provide direct evidence for a direction-independent role of the right-sided horizontal IPS segment in processing bilateral compared with unilateral stimuli (Fig. 5F) in combination with a direction-dependent role of the symmetrical left-sided area during the same processing conditions (Fig. 5C).
In the framework of the theory of visual attention (Duncan et al., 1999) and in line with single-neuron electrode recording studies (Gottlieb et al., 1998; Gottlieb and Goldberg, 1999), we propose that the horizontal IPS segment has a role in dividing the attentional weights between targets and non-targets. This attentional weighting may be composed of facilitatory effects favouring the processing of targets and suppressive effects inhibiting distractors. Potential sources of attentional inhibition are the medial superior parietal lobule (Wojciulik and Kanwisher, 1999; Pollmann et al., 2003) or prefrontal cortex (Vandenberghe et al., 1997). According to our proposal, when multiple potential attentional targets are present, the activated region in the horizontal segment of the IPS tags a stimulus that will be chosen as the subsequent target for an attentional shift. In this sense, the current data reconcile the role of IPS in endogenous selection, revealed by numerous recent neuroimaging studies (Coull et al., 1996; Wojciulik and Kanwisher, 1999; Corbetta et al., 2000) with a role in the attentional processing of bilateral stimuli. Further experiments in patients and healthy controls will allow us to tease apart whether the IPS activation is related to the bilateral stimulus location, the number of locations occupied or the number of objects presented (Rapcsak et al., 1989; Humphreys et al., 1994).
To conclude, an irrelevant stimulus in the ignored hemifield exerts a strong effect upon the horizontal IPS segment, in accordance with a role for this structure in dividing attentional weights among stimuli (Duncan et al., 1999; Corbetta et al., 2000). This study reconciles the role of bilateral IPS in endogenous stimulus selection with the striking effect of bilateral stimulation following right-hemispheric parietal lesions. Functional imaging in neglect patients will be required to assess the impact of structural lesions of IPL upon IPS activity levels in those cases where the structural lesion does not include the horizontal IPS segment.
|
Acknowledgements |
|---|
Supported by FWO grant G.0076.02 (R.V. and G.O.) and by KU Leuven Research Fund OT/04/41 (R.V.). R.V. is a Clinical Investigator of the Fund for Scientic Research (FW0), Flanders.
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Anderson B. A mathematical model of line bisection behaviour in neglect. Brain 1996; 119: 841–50.
Bender M. Disorders in perception (with particular reference to the phenomena of extinction and displacement). Springfield (IL): C. C. Thomas; 1952.
Bisiach E, Vallar G. Unilateral neglect in humans. In: Rizzolatti G, editor. Handbook of neuropsychology. Vol. I. Amsterdam: Elsevier; 2000. p. 459–502.
Bremmer F, Schlack A, Shah N, Zaris O, Kubischik M, Homann K, et al. Polymodal motion processing in posterior parietal and premotor cortex: a human fMRI study strongly implies equivalencies between humans and monkeys. Neuron 2001; 29: 287–96.[CrossRef][ISI][Medline]
Burock M, Buckner R, Woldorff M, Rosen B, Dale A. Randomized event related experimental designs allow for extremely rapid presentation rates using functional MRI. Neuroreport 1998; 9: 3735–9.[ISI][Medline]
Colby C, Duhamel J, Goldberg M. Visual, presaccadic and cognitive activation of single neurons in monkey LIP. J Neurophysiol 1996; 76: 2841–2852.
Corbetta M, Miezin F, Shulman G, Petersen S. A PET study of visuospatial attention. J Neurosci 1993; 13: 1202–26.[Abstract]
Corbetta M, Kincade J, Ollinger J, McAvoy M, Shulman G. Voluntary orienting is dissociated from target detection in human posterior parietal cortex. Nat Neurosci 2000; 3: 292–7.[CrossRef][ISI][Medline]
Corbetta M, Kincade J, Hahn M, Stanley C, Kwentus A, Sapir A. Reactivation and rebalancing of attention networks during recovery from spatial neglect. Soc Neurosci Abstr 2003; 33: 228.11.
Coull J, Frith C, Frackowiak R, Grasby P. A fronto-parietal network for rapid visual information processing: a PET study of sustained attention and working memory. Neuropsychologia 1996; 34: 1085–95.[CrossRef][ISI][Medline]
Desimone R, Duncan J. Neural mechanisms of selective visual attention. Annu Rev Neurosci 1995; 18: 193–222.[CrossRef][ISI][Medline]
Driver J, Frith C. Shifting baselines in attention research. Nat Rev Neurosci 2000; 1: 147–8.[CrossRef][ISI][Medline]
Driver J, Vuilleumier P. Perceptual awareness and its loss in unilateral neglect and extinction. Cognition 2001; 79: 39–88.[CrossRef][ISI][Medline]
Duncan J, Bundesen C, Olson A, Humphreys G, Chavda S, Shibuya H. Systematic analysis of deficits in visual attention. J Exp Psychol Gen 1999; 128: 450–78.[CrossRef][ISI][Medline]
Duncan J, Humphreys G. Visual search and stimulus similarity. Psychol Rev 1989; 96: 433–58.[CrossRef][ISI][Medline]