Brain Advance Access originally published online on November 29, 2005
Brain 2006 129(1):168-181; doi:10.1093/brain/awh690
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Goal-driven selective attention in patients with right hemisphere lesions: how intact is the ipsilesional field?
Cognitive Neuroscience Laboratory, School of Behavioural Science, University of Melbourne, Victoria, Melbourne, Australia
Correspondence to: Jason B. Mattingley, Cognitive Neuroscience Laboratory, School of Behavioural Science, University of Melbourne, Victoria 3010, Australia E-mail: j.mattingley{at}psych.unimelb.edu.au
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
Patients with right hemisphere (RH) lesions often display a spatial bias in attention towards the ipsilesional hemifield. The behavioural manifestations of this spatial bias are typically interpreted as reflecting increased or enhanced attention for stimuli within the ‘intact’ ipsilesional field, and impaired attentional functioning within the contralesional field. In the healthy brain, goal-driven and stimulus-driven attentional processes interact to determine which stimuli should be prioritized for selection. Although unilateral brain damage increases the relative attentional salience of stimuli within the ipsilesional field, it might also cause problems in filtering or attenuating task-irrelevant information. We examined whether goal-driven attention modulates the processing of ipsilesional and contralesional information in 6 patients with unilateral brain damage following RH stroke (5 male, 1 female; mean age 60.8 years) and a group of age and sex-matched controls. We used a flanker task in which participants made speeded judgements on a central target item (a coloured letter). On each trial the target was flanked by a coloured letter in the left and right hemifields. In separate blocks, participants were instructed to judge either the identity or the colour of the central target and to ignore the flankers. The flanker on one side could be congruent, incongruent or neutral with respect to the target, on either the letter or the colour dimension, whereas the flanker on the other side was always neutral on both dimensions. Healthy controls showed significant interference from incongruent flankers on either side. Crucially, however, this effect only occurred for the task-relevant dimension [F(2,10) = 24.60; P < 0.001]. For patients, however, both the task-relevant and task-irrelevant dimensions of ipsilesional flankers interfered with response times [task-relevant: F(2,10) = 7.50, P < 0.05; task-irrelevant: F(1,5) = 6.20, P < 0.05]. Conversely, contralesional flankers influenced response times only when the target and distractor were incongruent on the task-relevant dimension [F(2,10) = 4.85; P < 0.05]. Our findings demonstrate that following RH damage, goal-driven biases cannot constrain the processing of task-irrelevant features of ipsilesional stimuli. We speculate that a lateralized bias in spatial attention leads to unselective prioritization of all feature-based attributes of stimuli appearing within the ipsilesional hemifield, whether or not they are relevant to performance. Attentional selection for ipsilesional stimuli in disorders such as spatial neglect and extinction may not therefore be entirely normal, as previously assumed.
Key Words: spatial extinction; selectivity; goal-driven; stimulus-driven; feature-based selection
Abbreviations: RT = reaction time; RH = right hemisphere
Received June 14, 2005. Revised October 20, 2005. Accepted October 20, 2005.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
Whether a feature or location is selected for attention is determined by the relative conspicuity, or salience, of elements in a visual scene. In the healthy brain, behavioural goals bias the extent to which particular stimulus attributes are prioritized for selection. Patients with brain damage provide a unique opportunity to examine the relative influence of stimulus-based conspicuity and goal-driven biases in determining salience and the consequences of salience for selection (Driver and Vuilleumier, 2001b
; Behrmann et al., 2004; Rorden and Karnath, 2004). Lateralized biases in spatial selection arise after damage to one of a number of cortical or subcortical regions, but are most commonly associated with parietal lobe damage (Vallar et al., 1994; Karnath et al., 2003; Mort et al., 2003). In severe cases, patients display visual extinction or unilateral neglect, in which information on the ipsilesional side is attended, whereas information on the contralesional side is ignored (Brain, 1941; Bender and Teuber, 1946; Heilman and Valenstein, 1979; Kinsbourne, 1993; Driver and Vuilleumier, 2001a). Although patients who have recovered from neglect and extinction are able to direct their attention contralesionally when required, a lateralized bias typically persists at a subclinical level (Karnath, 1988; Mattingley et al., 1994). In the current study we measured the extent to which goal-driven selection processes can modulate the processing of stimuli within the ipsilesional and contralesional hemifields in right hemisphere (RH)-lesioned patients, using a modified version of the standard flanker paradigm (Eriksen and Hoffman, 1973; Eriksen and Eriksen, 1974). In separate blocks of trials using visually identical coloured letter stimuli, a single featural attribute of a central target was made relevant for the patient's responses (e.g. colour); the other target attribute (e.g. letter identity) had to be ignored. Our results suggest a paradoxical impairment in selectivity for stimuli in the ‘good’ ipsilesional field, despite intact goal-driven selection processes in the ‘impaired’ contralesional field.
In humans, unilateral brain damage leads to a spatial imbalance in the salience ascribed to stimuli across the visual field such that information falling within the ipsilesional field is over-prioritized relative to stimuli located further towards the contralesional side (Pouget et al., 1999; Pouget and Driver, 2000; Driver and Vuilleumier, 2001a). Evidence from single-cell recordings in monkeys suggests that this gradient in spatial attention after unilateral lesions occurs as a result of loss of neurons in areas such as the parietal lobe that represent specific locations in space (Andersen et al., 1985; Gottlieb et al., 1998). The parietal lobe has been implicated as the site of a ‘priority map’ that continuously represents the relative importance of information across the visual scene, ultimately determining where attention is allocated (Gottlieb et al., 1998; Colby and Goldberg, 1999; Kusunoki et al., 2000; Bisley and Goldberg, 2003). Physiological evidence indicates that the parietal lobe is densely interconnected with subcortical forebrain regions including the basal ganglia (putamen, caudate nucleus and globus pallidus), thalamus (pulvinar) and superior colliculus (Selemon and Goldman-Rakic, 1988; Robinson and Petersen, 1992; Leichnetz, 2001), several of which are argued to form a network involved in encoding stimulus salience (Vallar et al., 1994; Downar et al., 2000; Itti and Koch, 2001; Corbetta and Shulman, 2002).
Behavioural evidence in patients with lateralized attentional bias following unilateral brain damage suggests that the ‘intact’ ipsilesional field benefits from increased or enhanced attention (Heilman et al., 1985; Heilman and Valenstein, 1993; Kinsbourne, 1993; Vallar, 1998). For example, ipsilesional hyperattention (Ladavas, 1993; Vallar, 1998) is reflected behaviourally in ipsilesional deviation of eye and head movements (Behrmann et al., 1997; Karnath et al., 1998) and early orienting of attention towards the ipsilesional side of search arrays (Gainotti et al., 1991), with an increased number of fixations and duration of inspection for ipsilesional targets (Behrmann et al., 1997). Unilateral RH-damaged patients also show facilitated reaction times (RTs), rather than the normal pattern of inhibition of return, for ipsilesional targets appearing at previously explored locations (Bartolomeo et al., 1999, 2001; Vivas et al., 2003). Similarly, these patients show a paradoxical reduction in RTs to stimuli located further towards the ipsilesional periphery from a central fixation point (De Renzi et al., 1989; Smania et al., 1998). Consequently the ipsilesional manifestation of attentional bias has often been viewed as being productive or ‘positive’ in its effect on perception and action, whereas that of the contralesional side has been characterized as defective or ‘negative’ (Vallar, 1998; Rusconi et al., 2002).
Goal-driven selectivity in patients with unilateral attention bias
Salience, and therefore the likelihood of selection for attention, is determined by two distinct component processes (Driver, 2001; Pessoa et al., 2003). Stimulus-driven competitive mechanisms draw attention rapidly and involuntarily towards stimuli that have distinctive physical characteristics within a given context, such as their colour, orientation and brightness (Desimone, 1999; Kastner and Ungerleider, 2000; Itti and Koch, 2001; Bisley and Goldberg, 2003; Treue, 2003; Connor et al., 2004; Ogawa and Komatsu, 2004). In this context, stimulus-driven processes include both feed-forward and feedback effects on visual processing (Desimone and Duncan, 1995). Conversely, goal-driven (feedback) priorities allow attention to be biased voluntarily towards stimuli and/or responses that are meaningful for the control of action (Egeth and Yantis, 1997; Desimone, 1999; Kastner and Ungerleider, 2000, 2001). Importantly, the salience of representations across the visual scene is determined by an interaction between goal-driven priorities and stimulus-driven competitive interactions (Desimone and Duncan, 1995; Kastner and Ungerleider, 2000; Itti and Koch, 2001 Corbetta and Shulman, 2002; Graboi and Lisman, 2003).
Although many studies have examined stimulus-driven processes in patients with unilateral attentional biases (Driver et al., 1992; Ward et al., 1994; Mattingley et al., 1997; Pavlovskaya et al., 2000; Vuilleumier and Sagiv, 2001; Driver and Vuilleumier, 2001a; Ptak et al., 2002), relatively few have investigated the nature and extent of goal-driven biases for selection in these patients. Several studies have found that visual extinction is exacerbated when ipsilesional and contralesional targets are similar on a dimension that is relevant to the patient's task (Baylis et al., 1993; Vuilleumier and Rafal, 2000). Likewise, a number of studies suggest that detection of contralesional stimulus features can be improved via task-based strategies, either with external instructions or via endogenous cueing paradigms that increase patients' expectancy of contralesional events (Baylis et al., 1993; Luo et al., 1998; Smania et al., 1998; Duncan et al., 1999; Vuilleumier and Rafal, 2000; Bartolomeo et al., 2001; Ptak et al., 2002).
Prioritizing feature-based stimulus information
Although attention can boost salience at some locations relative to others (i.e. ‘location-based selection’) (Posner, 1980; Treisman and Gelade, 1980), mechanisms of attention may also enhance some feature-based stimulus properties over others, independently of their location in the visual field (i.e. ‘feature-based selection’) (Duncan, 1984; Baylis and Driver, 1992; Triesman, 1993; Vecera and Farah, 1994; Desimone and Duncan, 1995; Olson, 2001; Saenz et al., 2002; Martinez-Trujillo and Treue, 2004). For example, the magnitude of neuronal responses within early feature-based visual areas, such as V1, V2, V3, V4 and MT, in response to an unattended stimulus has been shown to be dependent upon the featural properties of an attended stimulus (Treue and Martinez Trujillo, 1999; Saenz et al., 2002; Martinez-Trujillo and Treue, 2004). Neural responses in early visual areas coding for feature-based stimulus properties, as described above, are determined by both stimulus-driven (i.e. feed-forward and feedback) and goal-driven processes (Reynolds et al., 2000; Martinez-Trujillo and Treue, 2004). For example, Reynolds et al. (2000) found that the response of V4 neurons to an orientation-grating stimulus was comparable when the grating was attended versus when it was unattended but its luminance contrast was increased. Feed-forward and feedback influences can therefore have equivalent effects on neural responses. Considering that the activity of early striate and extrastriate populations serves as input in subsequent stages of visual processing (e.g. object recognition further along ventral stream temporal areas), a change in any one of these influences will ultimately determine representational prioritization in ventral stream areas.
The results of studies on goal-driven influences on selection raise the important question of whether patients with spatial attentional biases are able to selectively prioritize task-relevant feature-based information across the visual field. If feed-forward (i.e. stimulus-driven) and feedback (i.e. both stimulus-driven and goal-driven) influences can have equivalent effects on neural activity in early visual areas (Reynolds et al., 2000; Martinez-Trujillo and Treue, 2004), a spatial imbalance in the representation of salience might be expected to result in pathological prioritization of all feature-based attributes of ipsilesional items. Consequently, such patients may be impaired in their ability to suppress task-irrelevant feature-based information at ipsilesional locations. If, however, as is often assumed, attention in patients with unilateral brain damage is ‘enhanced’ for stimuli located in the ipsilesional hemispace (Heilman et al., 1985; Heilman and Valenstein, 1993; Kinsbourne, 1993; Vallar, 1998), then the reverse prediction follows: selectivity for task-relevant stimulus features at ipsilesional locations should remain intact (i.e. task-irrelevant information at ipsilesional locations should be suppressed).
The flanker-compatibility effect as an indirect measure of attentional prioritization
One experimental technique that has been used to index the influence of task-irrelevant stimuli on behaviour is the ‘flanker-compatibility effect’. In a typical flanker task (Eriksen and Hoffman, 1973; Eriksen and Eriksen, 1974) participants are instructed to make a speeded response to a central target, while ignoring flanker stimuli positioned on either side. The nature and extent of processing of lateralized flanking stimuli is indexed by their influence on RTs and errors to the central target. RTs are appreciably slower when the flankers are incongruent with the target, as compared with conditions in which the flankers are neutral [the flanker-compatibility effect (Miller, 1991)]. Because attention is directed centrally and efficient performance is achieved by ignoring flankers, the paradigm serves as an index of selectivity: the ability to inhibit irrelevant or distracting information (Tipper et al., 1994; Milliken and Tipper, 1998).
In the current study we explored whether the processing of feature-based characteristics of flankers can be modulated by goal-driven priorities at the target location. In the standard flanker paradigm the target is defined in visuospatial terms (e.g. it is positioned centrally with respect to the flankers) and responses are determined on the basis of featural properties of the target along a prespecified dimension (e.g. an ‘A’ or ‘B’ on the dimension of letter identity). To examine the influence of goal-driven or task-based effects on flanker processing, a second dimension can be incorporated so that stimuli can be defined by their congruence along two orthogonal dimensions (e.g. the letter ‘A’ or ‘B’, presented in red or green) (Cohen and Shoup, 1997; Mordkoff, 1998; Danckert et al., 1999; Maruff et al., 1999). Only one dimension of the target stimulus is relevant to an observer's behavioural response on each trial: the ‘task-relevant’ dimension (e.g. letter). The properties of the target on the other ‘task-irrelevant’ dimension (e.g. colour) are ignored.
In healthy observers, multidimensional flankers influence RTs to a central target, but only as a function of their congruence with the target on the task-relevant dimension; task-irrelevant flankers do not affect target RTs (Cohen and Shoup, 1997). For example, Maruff et al. (1999) presented healthy observers with a target and one flanker stimulus that varied simultaneously on the physical dimensions of letter and colour, or shape and colour. In separate tasks, observers identified one dimension of the central target (e.g. letter) and ignored the other (e.g. colour). Maruff et al. (1999) found that when observers reported, for example, the colour of the target only flankers that were incongruent with the target on the (task-relevant) colour dimension interfered with RTs, whereas the congruence between the target and flanker on the letter (task-irrelevant) dimension did not. This ‘top–down’ modulation of flanker processing has been observed for coloured lines (Cohen and Shoup, 1997), coloured letters and shapes (Mordkoff, 1998; Maruff et al., 1999), and holds under conditions in which observers are provided with a task rule a priori in order to choose the appropriate dimension on which to respond (Maruff et al., 1999).
Several studies have employed the flanker task (or variations thereof) to study the extent to which ipsilesional and contralesional stimuli are processed in patients with spatial attention biases (Audet et al., 1991; Cohen et al., 1995; Fuentes and Humphreys, 1996; Ro et al., 1998; Lavie and Robertson, 2001). In general the results from these studies have suggested that contralesional flankers exert a small but measurable influence on performance, whereas ipsilesional distractors have a sizeable influence on performance at an attended location. Of these studies, several have employed target/flanker stimuli that vary along two featural dimensions [e.g. blue target letter versus green flanker letters (Fuentes and Humphreys, 1996); small, coloured ‘O’ target versus large, coloured ‘O’ flankers (Cohen et al., 1995)], primarily to help patients differentiate the target from the flankers.
Only one study to date has incorporated multidimensional target and flanker stimuli specifically for the purpose of examining goal-driven selectivity following brain damage. Danckert et al. (1999) tested a single RH-damaged patient with left neglect in a study that aimed to determine whether changing the patient's behavioural goal would have a significant influence on the magnitude of contralesional flanker effects. Target and flanker stimuli were E's and O's (i.e. the letter dimension) that were red or green (i.e. the colour dimension). In half of the trials the patient identified the target letter and ignored its colour; in the other half he identified the colour and ignored the letter identity. As with several previous studies (Cohen and Shoup, 1997; Mordkoff, 1998; Maruff et al., 1999), target/flanker compatibility was defined according to goal-relevant and goal-irrelevant dimensions. Danckert et al. (1999) found that the similarity between target and flankers on the task-relevant dimension had a marked influence on their patient's RTs, whereas similarity on the task-irrelevant dimension did not. Crucially, this pattern of results was similar in the left and right visual fields, leading the authors to conclude that their patient was able to use top–down attentional goals to bias the processing of otherwise neglected stimuli.
The aim of the present study was to examine goal-driven selectivity for feature-based information within the ‘intact’ ipsilesional hemifield of patients with unilateral brain damage. If goal-driven biases for selection are intact in the ipsilesional field, then only the task-relevant features of ipsilesional stimuli should influence response times to central targets. Conversely, if an ipsilesional bias in attention leads to a relative increase in the salience of all stimulus-based properties of ipsilesional stimuli, then significant flanker-compatibility effects should be evident for both the task-relevant and task-irrelevant dimensions of ipsilesional flankers. To the extent that contralesional flankers are able to influence RTs to central targets in unilateral brain damaged patients (Audet et al., 1991; Cohen et al., 1995; Fuentes and Humphreys, 1996; Lavie and Robertson, 2001), intact feature-based selectivity should lead to significant compatibility effects associated with the task-relevant features of a contralesional flanker, with no such effect for the task-irrelevant dimension. In our study, stimulus displays were balanced, with flankers appearing briefly on both sides of a central target to eliminate the possibility of eye movements. If goal-driven biases influence the processing of flanker stimuli, it is also important to determine whether the processing of all task-irrelevant information is affected (e.g. flankers that are congruent or incongruent), or whether goal-driven modulatory effects only extend to stimuli that potentially interfere with performance (e.g. incongruent flankers) (Neill and Westberry, 1987; Tipper et al., 1994; Fuentes and Humphreys, 1996). Congruent and incongruent flanker conditions were therefore compared with a neutral condition to determine the direction of goal-driven bias.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
Participants
Six patients who had suffered a first-episode, unilateral RH stroke, participated in the study (5 male, 1 female; mean age = 60.8 years, SD = 15.1). All patients showed signs of rightward attentional bias, and consequent left inattention, on at least one of the following standard clinical measures: line cancellation (Albert, 1973
); star cancellation (Wilson et al., 1987); the Balloons test (feature and conjunction search tasks; Edgeworth et al., 1998); line bisection; and grey-scales judgement task (Mattingley et al., 2004). All patients were right-handed by self-report, and all had full visual fields (confirmed by confrontation during neurological examination). Clinical data for each patient are presented in Table 1. Lesion reconstructions from MRI for each patient are presented in Fig. 1. Six right-handed, age- and sex-matched healthy controls also participated (mean age = 60.8 years, SD = 17.0). All participants were English speaking and were alert and oriented at the time of testing. For all participants, informed consent was obtained according to the Declaration of Helsinki, and the Ethical Committees from each major hospital involved approved all research.
|
|
Apparatus and stimuli
Stimuli were presented on a 14 in. LCD monitor (60 Hz refresh rate) controlled by a Pentium III 847 MHz laptop computer. DMDX software (Forster and Forster, 2003
) was used to present stimuli and record manual response times. Responses were made using a mouse interfaced with the serial port of the computer. Experimental trials consisted of a central letter target flanked on both sides by a task-irrelevant letter flanker. Stimuli were the uppercase letters A, B and X in size 22 Arial font, and were red, green or yellow in colour. From a viewing distance of 
57 cm, each letter subtended 1.21° visual angle vertically and 0.97° horizontally. The target letter was centred 0.5° above the vertical midline. A white placeholder, subtending 0.95° of horizontal visual angle, appeared underneath the central target letter along the vertical midline, so that the target could be distinguished without spatial reference to either flanker. Each flanker was centred 1.5° from the horizontal midline. Each participant completed two main tasks in separate blocks of trials that were alternated within each testing session. In the Target Response task participants were required to categorize a central target letter as quickly and as accurately as possible, while ignoring bilateral flankers. In the Flanker Detection task participants had to report the presence of unilateral or bilateral letter flankers, while ignoring the central target stimulus.
At the commencement of each block of the Target Response task participants were instructed to respond to either the letter or the colour of the central target and to indicate their response by pressing one of two buttons with the right (ipsilesional) hand. The bilateral arrangement of the flankers was designed to create a balanced, competitive display. As outlined below, one of the two flankers was always neutral with respect to the target, whereas the other flanker (termed the ‘manipulated’ flanker) could be congruent, incongruent or neutral on either the letter or the colour dimension (see Fig. 2A).
|
Each target was defined with respect to its identity (A, B) and colour (red or green). Flankers were an A, B or X, in red, green or yellow; the letter X and the colour yellow served as neutral flankers (i.e. they did not map onto a response). On each trial, one of the two flankers was always a yellow X (the ‘neutral’ flanker). Thus, for example, on the letter dimension, the target and flanker could be congruent (e.g. target ‘A’ and flanker ‘A’), incongruent (e.g. target ‘A’ and flanker ‘B’) or neutral (e.g. target ‘A’ and flanker ‘X’). Similarly, on the colour dimension the target and flanker could be congruent (e.g. target red and flanker red), incongruent (target red and flanker green) or neutral (target red and flanker yellow). The manipulated flanker varied simultaneously on both of these dimensions, thus yielding nine different conditions (see Fig. 2B). The manipulated flanker appeared equally often on the left and right sides of the target. Nine congruence conditions and two manipulated flanker sides (left/right) yielded a total of 18 conditions. Presenting each of the four central targets (A and B in red or green) through the 18 stimulus conditions resulted in 72 different display types. Each of the displays was presented 6 times, yielding 24 items in each of the stimulus conditions (4 target types x 6 repetitions), for a total of 432 items. Each participant completed a total of 864 trials (except for patients C.P. and I.S., and their corresponding age-matched controls, who completed 1728 trials).
In half of the trials participants identified the central target letter (i.e. the ‘Letter Instruction’) and ignored its colour. In the remaining trials (i.e. the ‘Colour Instruction’) participants identified the central target's colour and ignored its identity. Thirty-six practice trials were constructed for both the letter and colour instructions. In order to minimize any task-switching effects, an additional 10 trials were added to the start of each task and removed from the subsequent analyses.
Identical stimuli were used in the Flanker Detection task (A, B, and X in red, green or yellow). Flankers were presented bilaterally on 72 trials and on the left or right side only on 36 trials each. Eighteen catch trials in which there were no flanking stimuli were also included, yielding a total of 162 trials in total. The displays in the bilateral flanker condition were identical to those in the Target Response task and were comprised of one of each of the 72 different display types. For the unilateral left and unilateral right flanker conditions, the neutral yellow X from either the right or left side (respectively) of the target was omitted from each of the 72 display types. For catch trial displays (with no flankers), two of the four target types (red A, red B, green A, green B) appeared 4 times, and two appeared 5 times. The 162 trials were broken down into 8 blocks, each with an even distribution of target-flanker trial types.
Procedure
All stimuli were displayed on a uniform black background. Each trial in the Target Response task began with a white underscore at fixation for 500 ms. This was followed by the simultaneous onset of the central target and the left- and right-sided flanker stimuli. The flankers were presented for 100 ms only, whereas the central target remained on screen until response. The inter-trial interval was 1000 ms. The 24 trials from each of the 18 congruency conditions were divided equally into 6 blocks of 72 trials each, for the separate letter and colour instructions. Items were fully randomized within each block. Participants completed 3 blocks of one task (either letter or colour) per testing session. An ABBA design was implemented, with the order of administration of the letter and colour instructions counterbalanced both within and between participants.
All participants made responses with the right hand. In the Letter Task, ‘A’ and ‘B’ responses were mapped to left and right mouse buttons, respectively. In the Colour Task, ‘red’ and ‘green’ responses were mapped to left and right mouse buttons, respectively. Response times were later collapsed across both possible responses within each task (letter, colour) to avoid differences between fingers.
All participants completed a block of Flanker Detection trials prior to and following each of the four sessions of the Target Response Task. The number of correct detections in bilateral, left only, right only and catch trials for each participant was averaged across all sessions. Participants were instructed to indicate verbally whether they saw a flanker appear on the left or right side, on both sides or on neither side.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
Flanker detection trials
Our initial analyses focused on the patients' ability to detect flanker stimuli in the Flanker Detection task, as an index of the extent of their ipsilesional attentional bias. The healthy control group performed at ceiling for both unilateral and bilateral trials, and made no false alarms on catch trials. The RH patients correctly detected a single flanker on the left or right side on 69.8 and 100% of trials, respectively; they correctly reported seeing no flankers on 99.3% of catch trials. On bilateral trials, patients correctly reported 99.8% of right-sided flankers but only 49.5% of left-sided flankers, indicating a significant impairment of left detections in bilateral versus unilateral displays (
2 = 32.19; P < 0.000). These findings confirm the presence of a contralesional impairment in the RH group that was exacerbated by the presence of an ipsilesional competitor, as in conventional measures of extinction (Mattingley, 2002).
Target Response task
Our hypotheses for the Target Response task concerned potential differences in responses to central targets as a function of task-relevant versus task-irrelevant dimensions of left- and right-sided flankers. We therefore pooled the responses from the letter and colour tasks within each group to maximize statistical power. For all analyses, significant effects were explored with follow-up analysis of variance (ANOVA) and either planned comparisons (t-tests) or post hoc pairwise comparisons with Bonferroni adjustment. The Huynh–Feldt correction for violations of sphericity was applied where appropriate.
Errors
The proportion of errors made in categorizing the central target by both groups was very small. For the patient group, the mean proportion of errors in the left- and right-sided flanker conditions was 0.04 (SD = 0.04) and 0.05 (SD = 0.04), respectively. For the control group, the mean proportion of errors was 0.02 (SD = 0.01) in each of the left- and right-sided flanker conditions. Due to the small number of errors made by each group these data are not considered further.
Reaction times
RTs >5000 ms and <150 ms (0.4% of trials) were eliminated from the analyses. Median RTs for each participant were calculated for each task (letter/colour) and congruence condition (congruent/incongruent/neutral) on the task-relevant and task-irrelevant dimensions.
The influence of task-relevant flanker features on RT
Median RTs were initially examined with a three-way repeated measures ANOVA with the within-subjects factors of Task-Relevant Congruence (congruent/neutral/incongruent) and flanker Side (left/right) and the between-subjects factor of Group (patients/controls). There was a significant main effect of Task-Relevance [F(2,20) = 28.79; P < 0.001], and a three-way interaction between Task-Relevance, Side and Group [F(2,20) = 4.47; P < 0.05]. To examine this interaction further RTs for control and patient groups were examined separately.
Control group
Mean RT for the control group is shown in Fig. 3B (left panel), plotted as a function of congruency for left and right flankers separately. A two-way repeated measures ANOVA was conducted on the median RTs, with factors of Task-Relevant Congruence and Flanker Side. There was a significant main effect of Congruence [F(2,10) = 24.60; P < 0.001]. Pairwise comparisons revealed that RTs to the central target were significantly slower when the flanker was incongruent (RT = 590 ms) than when it was neutral (RT = 563 ms; P < 0.05) or congruent (560 ms; P < 0.05). There was no significant difference between RTs for left- versus right-sided flankers and no interaction between Congruence and Side.
|
Patient group
The pattern of performance of the patient group across Congruence conditions for each flanker Side manipulation was very similar to that of the control group (see Fig. 3A left panel). A two-way repeated measures ANOVA was performed on the median RTs for the within-subjects factors of Congruence (congruent/incongruent/neutral) and Side (left/right). The analysis revealed a significant main effect of target-flanker Congruence [F(2,10) = 10.73; P < 0.01]. Pairwise comparisons revealed that RTs were significantly slower when the flanker was incongruent with the target (RT = 768 ms) compared with the neutral condition (RT = 740 ms) (P < 0.05) on the task-relevant dimension. There was no main effect of flanker Side, and no Congruence x Side interaction, suggesting that the pattern of interference effects on the task-relevant dimension was similar for both the left and right sides in the patient group. To verify that the observed congruency effect was significant for both left- and right-sided flankers, further one-way ANOVAs were conducted.
These analyses confirmed the significant effect of congruence on the task-relevant dimension held for both left- [F(2,10) = 4.85; P < 0.05] and right-sided [F(2,10) = 7.50; P < 0.05] flankers. For left-sided flankers, RTs were significantly slower for incongruent stimuli (744 ms) than for neutral stimuli [735 ms; t(5) = 2.930; P < 0.05]; RTs were not facilitated by congruent task-relevant features (744 ms). For right-sided flankers, RTs were significantly faster in the congruent condition (725 ms) than in the neutral condition [745 ms; t(5) = –3.417; P < 0.05], and marginally slower in the incongruent condition (775 ms) than in the neutral condition [t(5) = 2.37; P = 0.06].
The influence of task-irrelevant flanker features on RT
Median RTs were also examined according to the relationship between the target and flankers on the task-irrelevant dimension (i.e. collapsing across congruence on the task-relevant dimension). A three-way repeated measures ANOVA with the within-subject factors of task-irrelevant Congruence (congruent/neutral/incongruent) and flanker Side (left/right) and the between-subjects factor of Group (patients/controls) revealed a significant two-way interaction between Congruence and Side [F(2,20) = 4.27; P < 0.05]. This was further qualified by a three-way interaction between Congruence, Side and Group [F(2,20) = 3.99; P < 0.05]. To examine this interaction further RTs for the task-irrelevant dimension were analysed separately for control and patient groups.
Control group
The pattern of performance for the control group in each congruence condition and flanker side is illustrated in Fig. 3B (right panel). A two-way repeated measures ANOVA on the median RTs with the within-subject factors of Congruence and Side revealed no significant effects, indicating that the task-irrelevant dimension of left- and right-sided flankers had no effect on RTs to the central target in the control group.
Patient group
The pattern of performance of the patient group in each Congruence and flanker Side condition is presented in Fig. 3A (right panel). A two-way repeated measures ANOVA revealed a significant task-irrelevant Congruence x Side interaction [F(2,10) = 5.17; P < 0.05]. A subsequent one-way repeated measures ANOVA on the data for left- and right-sided flanker manipulations separately revealed no significant task-irrelevant congruency effects for left-sided flankers, but showed a significant effect of congruency for right-sided flankers [F(1,5) = 6.20; P < 0.05]. Planned comparisons for the right-sided flanker condition revealed that RTs were significantly slower for incongruent flankers (RT = 770 ms) than for neutral flankers [RT = 733 ms; t(5) = 3.43; P < 0.05], whereas RTs in the congruent condition (RT = 744 ms) were not significantly different from neutral.
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
The principal aim of the current study was to examine goal-driven selectivity within the ‘intact’ ipsilesional hemifield of patients with RH damage. We reasoned that if enhanced attention improves selectivity for ipsilesional events, RH patients with a spatial attention bias should show flanker-compatibility effects from only the task-relevant dimension of flanking distractors on the right side. Conversely, if the spatial bias leads to an unselective increase in the salience of ipsilesional stimuli, whether relevant to the current task or not, then all feature-based properties of ignored flanker stimuli on the right should affect performance at the central target location.
With respect to the task-relevant dimension, left- and right-sided flankers that were incongruent with the target slowed RTs significantly relative to neutral flankers in both the control and patient groups. The findings for patients suggest that features relevant for performance are prioritized in both the contralesional and ipsilesional fields. The significant contralesional flanker effect supports findings from a large number of studies that have documented significant processing of contralesional stimuli following unilateral RH damage (Karnath and Hartje, 1987; Berti et al., 1992; Driver et al., 1992; McGlinchey-Berroth et al., 1993; Mattingley et al., 1995; Ro and Rafal, 1996; Driver and Vuilleumier, 2001a; Rees et al., 2002). That task-relevant features of contralesional flankers influenced performance for the central target in the current study is particularly striking, given that the distractors were presented very briefly and were irrelevant to the central task. This finding is consistent with single-unit recording studies in primates (McAdams and Maunsell, 2000; Martinez-Trujillo and Treue, 2004) and ERP/event related magnetic field recordings in healthy humans (Hopf et al., 2004), which suggest that task-relevant stimulus features are prioritized for selection across the visual field in a location-independent manner.
In contrast, patients and controls differed with respect to the effect of the task-irrelevant dimension of ignored flankers. For the healthy controls, task-irrelevant features did not facilitate or interfere with response times to the central target. This finding replicates those of previous studies conducted in younger populations (Cohen and Shoup, 1997; Maruff et al., 1999), and suggests that in the healthy brain behavioural goals are used to bias selection in favour of task-relevant features by constraining the processing of task-irrelevant information at ignored locations. The results for the control group also confirm that the stimulus displays and task manipulation we used were sufficient to elicit goal-driven modulation of flanker effects. In contrast, patients with RH damage exhibited a significant interference effect for the task-irrelevant dimension of incongruent right-sided flankers, indicating a striking breakdown in selective processing of visual features in the ipsilesional hemifield. Conversely, for patients the task-irrelevant dimension of contralesional flankers did not interfere with response times to the central target.
With respect to the effects for contralesional stimuli, previous studies in patients with spatial attention biases have indicated that stimulus-driven competitive interactions operate for contralesional representations, despite the fact that patients often remain unaware of the contralesional event (Driver et al., 1992; Ward et al., 1994; Gilchrist et al., 1996; Mattingley et al., 1997; Pavlovskaya et al., 1997; Boutsen and Humphreys, 2000; Pavlovskaya et al., 2000; Vuilleumier and Sagiv, 2001; Ptak et al., 2002). The absence of flanker compatibility effects for the task-irrelevant dimension of contralesional flankers in our study cannot be explained by a failure to process stimuli in the contralesional field, because the same stimulus-based properties interfered reliably with performance in identical displays when those features were relevant for responses to the central target. Instead, our results suggest that goal-driven selection biases can also remain intact for contralesional stimulus representations (Danckert et al., 1999).
The implication of this finding is that although brief contralesional events often go undetected by patients with unilateral lesions, intact goal-driven mechanisms nonetheless influence prioritization of such stimuli. This might seem surprising, given that patients typically perform so poorly with contralesional stimuli in attentional tasks. There is evidence, however, that substantial processing of contralesional stimuli may arise in the absence of focused attention and awareness (Driver and Vuilleumier, 2001b; Berti, 2002). For example, spatial neglect and extinction can be ameliorated by early stimulus-driven processes such as figure-ground segmentation (Driver, 1995; Mattingley et al., 1997) and grouping by symmetry in shape, colour or pattern (Driver et al., 1992; Baylis et al., 1993; Ward et al., 1994; Vuilleumier and Sagiv, 2001; Ptak et al., 2002), and by co-linearity of edges (Gilchrist et al., 1996; Pavlovskaya et al., 1997; Pavlovskaya et al., 2000). Even categorical and semantic information can be extracted from stimuli within the contralesional field (Berti et al., 1992; McGlinchey-Berroth et al., 1993), as reflected by significant neural responses within the preserved ventral processing stream of the affected hemisphere (Vuilleumier and Schwartz, 2001; Rees et al., 2002; Vuilleumier et al., 2002). The findings from our flanker task suggest that at least some aspects of goal-driven selectivity remain intact for stimuli that fall within the affected visual hemifield.
Our finding of significant interference from task-irrelevant features of right-sided flankers demonstrates for the first time that the spatial bias in RH-damaged patients does not improve selectivity of visual features. Rather, such patients have a specific impairment in the ability to selectively inhibit task-irrelevant information within the ipsilesional field. Our design enabled us to rule out possible alternative interpretations of the observed pattern of results. First, we used spatially balanced stimulus displays on every trial of the Target Response task to encourage an even distribution of attention across experimental conditions, rather than presenting left- and right-sided flankers individually (Audet et al., 1991; Cohen et al., 1995; Ro et al., 1998; Danckert et al., 1999; Lavie and Robertson, 2001). Secondly, we implemented a task in which flankers always varied on just two dimensions (letter and colour), to reduce any costs of task switching. If patients had been confused about which dimension of the target was currently relevant, we should have observed a large number of errors in target classification. On the contrary, the patient group made <5% classification errors overall (controls made 2% errors).
The question arises therefore as to why RH-damaged patients show reduced selectivity for stimuli appearing within the ipsilesional hemifield. Several authors have suggested that parietal structures are critical for generating control signals for goal-driven selection (Kastner et al., 1999; Culham and Kanwisher, 2001; Toth and Assad, 2002; Assad, 2003; Friedman-Hill et al., 2003; Yantis, 2003; Freedman, 2004; Stoet and Snyder, 2004). Conversely, studies using single-unit recordings in monkeys (Asaad et al., 2000; Wallis et al., 2001) and fMRI in humans (Kastner et al., 1999; Brass and von Cramon, 2004) have reported frontal lobe activation in either the period prior to or in the delay period following the presentation of a cue that indicates the task-relevance of an upcoming target. It is possible, in the current study, that brain areas thought to be involved in coding salience across the visual scene, such as parietal or interconnected subcortical networks (Robinson and Petersen, 1992; Desimone and Duncan, 1995; Gottlieb et al., 1998; Itti and Koch, 2001), received feature-based, goal-driven (feedback) signals, but this information was not effective in prioritizing stimulus properties. Interestingly, Peers et al. (2005) found a significant positive correlation between lesion volume and an index of top–down control in 25 patients with circumscribed frontal or parietal lesions. This was the case for both patient groups, suggesting that both frontal and parietal regions may have some involvement in goal-driven selectivity. For patients in our study, there was no significant relationship between lesion volume and feature-based selectivity (i.e. incongruence cost: neutral—incongruent flanker conditions); this was the case for both task-relevant [left flanker, r(4) = –0.685, P > 0.05; right flanker, r(4) = 0.418, P > 0.05) and task-irrelevant flanker features [left flanker, r(4) = –0.285, P > 0.05; right flanker, r(4) = 0.40, P > 0.05]. It is noteworthy that our task differed from that of Peers et al. (2005) in that we used an indirect measure of selectivity, and the spatial location of irrelevant distractors remained unchanged across all trials. Further, given the relatively small sample size our data prevent a definitive conclusion on this issue.
Although the neural correlate of salience is still the subject of debate [e.g. whether it corresponds to increased neuronal firing rates (Desimone and Duncan, 1995), relative timing of spike activity within a neural population (VanRullen, 2003) or synchronicity in neural firing patterns (Fries et al., 2001; Yantis and Serences, 2003)], it is generally agreed that in a ‘winner-take-all’ fashion the most salient representation is selected for further processing (Desimone and Duncan, 1995; Egeth and Yantis, 1997; Kastner and Ungerleider, 2000, 2001; Itti and Koch, 2001; Corbetta and Shulman, 2002; Graboi and Lisman, 2003; Treue, 2003; Yantis, 2003). Our results suggest that a relative imbalance in the salience afforded to stimuli across the visual field following unilateral damage (Pouget et al., 1999; Pouget and Driver, 2000; Driver and Vuilleumier, 2001a) may have consequences for the balance between stimulus-driven and goal-driven selection signals. Given the qualitatively similar feed-forward (stimulus-driven) and feedback (stimulus- or goal-driven) influences in ventral areas coding for feature-based information (Reynolds et al., 2000; Martinez-Trujillo and Treue, 2004), an increase in feedback signals could represent stimulus-based conspicuity or heightened behavioural relevance. If the net result of such feedback is an increase in the competitive strength of the corresponding representations then these will be prioritized for selection, whether task-relevant or not. The outcome of spatial hyperattention, therefore, is that goal-driven signals might be swamped as a consequence of a biased neural representation of space.
|
Implications and conclusions |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
The notion that selective attention in the ipsilesional hemifield might more accurately be viewed as ‘dysfunctional’, rather than ‘intact’ as previously thought in patients with unilateral brain damage (Heilman et al., 1985
; Heilman and Valenstein, 1993; Kinsbourne, 1993; Vallar, 1998; Rusconi et al., 2002), is supported by commonly observed behavioural patterns, such as perseverative drawing and revisiting of ipsilesional targets on cancellation tasks (Rusconi et al., 2002; Mannan et al., 2005), and the failure of these patients to prevent attention from revisiting ipsilesional stimuli (inhibition of return) (Bartolomeo et al., 1999, 2001; Vivas et al., 2003).
Our findings lend support to the notion that unilateral brain damage results in distractibility towards ipsilesional stimuli. That spatial ‘hyperattention’ may lead to selection of all stimulus-based features, whether task-relevant or not, is consistent with Duncan's ‘integrated competition hypothesis’ (Duncan et al., 1997). According to this model, object-based features entering the visual system are encoded across a distributed network of cortical and subcortical ‘modules’. Within each module, activations from different object features compete for selection. Crucially, competition is integrated across a distributed network; as the properties of an object become dominant within one module, other featural properties associated with the object acquire a competitive advantage in their respective modules. As a result, all features of the object are ‘primed’ across multiple systems. It follows from this model that if the spatial position of a stimulus (e.g. as an integral property of an object) becomes dominant, other properties of the stimulus at the dominant location should become more salient, and therefore have a greater influence on behaviour. In the case of patients with RH damage, a bias in the neural coding of space should lead to enhanced representation of all features of a stimulus at the ipsilesional location, just as we observed in our flanker task. One drawback of increased salience is reduced goal-driven selectivity, with the consequence of indiscriminate behavioural prioritization of ipsilesional events, possibly leading to a reduction in resources allocated towards the processing of any competing contralesional stimuli.
|
Acknowledgements |
|---|
The authors wish to thank Adam Morris and Robert Hester for helpful comments on an earlier version of the manuscript. The authors also thank Louise James and Jacqueline Anderson from Austin Health for help with patient recruitment.
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionImplications and conclusionsReferences |
|---|
Albert ML. A simple test of visual neglect. Neurology 1973; 23: 658–64.
Andersen RA, Essick GK, Siegel RM. Encoding of spatial location by posterior parietal neurons. Science 1985; 230: 456–8.
Assad JA. Neural coding of behavioral relevance in parietal cortex. Curr Opin Neurobiol 2003; 13: 194–7.[CrossRef][ISI][Medline]
Asaad WF, Rainer G, Miller EK. Task-specific neural activity in the primate prefrontal cortex. J Neurophysiol 2000; 84: 451–9.
Audet T, Bub D, Lecours AR. Visual neglect and left-sided context effects. Brain Cogn 1991; 16: 11–28.[CrossRef][ISI][Medline]
Bartolomeo P, Chokron S, Sieroff E. Facilitation instead of inhibition for repeated right-sided events in left neglect. Neuroreport 1999; 10: 3353–7.[ISI][Medline]
Bartolomeo P, Sieroff E, Decaix C, Chokron S. Modulating the attentional bias in unilateral neglect: the effects of the strategic set. Exp Brain Res 2001; 137: 432–44.[CrossRef][ISI][Medline]
Baylis GC, Driver J. Visual parsing and response competition: the effect of grouping factors. Percept Psychophys 1992; 51: 145–62.[ISI][Medline]
Baylis GC, Driver J, Rafal RD. Visual extinction and stimulus repetition. J Cogn Neurosci 1993; 5: 453–66.
Behrmann M, Watt S, Black SE, Barton JJ. Impaired visual search in patients with unilateral neglect: an oculographic analysis. Neuropsychologia 1997; 35: 1445–58.[CrossRef][ISI][Medline]
Behrmann M, Geng JJ, Shomstein S. Parietal cortex and attention. Curr Opin Neurobiol 2004; 14: 212–7.[CrossRef][ISI][Medline]
Bender MB, Teuber HL. Phenomena of fluctuation, extinction and completion in visual perception. Arch Neurol Psychiatry 1946; 55: 627–58.
Berti A. Unconscious processing in neglect. In: Karnath H-O, Milner D, Vallar G, editors. The cognitive and neural bases of spatial neglect. London: Oxford University Press; 2002. p. 313–26.
Berti A, Allport A, Driver J, Dienes Z, Oxbury J, Oxbury S. Levels of processing for visual stimuli in an ‘extinguished’ field. Neuropsychologia 1992; 30: 403–15.[CrossRef][ISI][Medline]
Bisley JW, Goldberg ME. Neuronal activity in the lateral intraparietal area and spatial attention. Science 2003; 299: 81–6.
Boutsen L, Humphreys GW. Axis-based grouping reduces visual extinction. Neuropsychologia 2000; 38: 896–905.[CrossRef][ISI][Medline]
Brain WR. Visual orientation with special reference to lesions of the right cerebral hemisphere. Brain 1941; 64: 244–72.
Brass M, von Cramon DY. Selection for cognitive control: a functional magnetic resonance imaging study on the selection of task-relevant information. J Neurosci 2004; 24: 8847–52.
Cohen A, Shoup R. Perceptual dimensional constraints in response selection processes. Cognit Psychol 1997; 32: 128–81.[CrossRef][ISI][Medline]
Cohen A, Ivry RB, Rafal RD, Kohn C. Activating response codes by stimuli in the neglected visual field. Neuropsychology 1995; 9: 165–73.[CrossRef]