Neuropsychologia 40 (2002) 2199–2205

Expectation-based attentional modulation of
visual extinction in spatial neglect
Radek Ptak a,∗ , Nathalie Valenza b , Armin Schnider a
a Clinique de Rééducation, Hôpital Cantonal Universitaire, 26, Av. de Beau-Séjour, CH-1211 Geneva 14, Switzerland
b Clinique de Neurologie, HUG, CH-1211 Geneva 14, Switzerland

Received 26 March 2002; received in revised form 25 July 2002; accepted 29 July 2002

Abstract
Visual extinction, the failure of patients with unilateral focal brain damage to report the contralesional of two simultaneously presented
stimuli, may be modulated by characteristics of the display such as similarity, collinearity, or connectedness. Since these factors affect
the perceptual configuration of stimuli, the modulation of extinction is believed to reflect low-level perceptual grouping. In the present
study, patient AG did not show any modulation of contralesional detection when the ipsilesional and contralesional stimulus grouped by
colour, by form, or both (Experiment 1). In contrast, identification of the contralesional stimulus was facilitated when the stimuli grouped
(Experiment 2), suggesting a modulation of extinction by specific task demands. Experiment 3 used a cueing procedure to demonstrate
modulation of extinction by expectation biases. Prior to stimulus presentation, AG was cued to attend to a particular feature (e.g. colour).
After stimulus exposure he was prompted to identify the expected feature on valid trials and the unexpected feature on invalid trials. AG
showed a significant validity effect for contralesional stimuli i.e. he identified the expected feature (e.g. colour) significantly better than
the unexpected feature (e.g. form). These results suggest that competition for selection between visual stimuli may not only be influenced
by perceptual characteristics of the display, but also by high-level factors such as the response criterion or expectation biases.
© 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Extinction; Neglect; Attentional modulation; Top–down processing; Feature-based attention

1. Introduction
Visual extinction is a frequent consequence of unilateral focal brain damage. Patients fail to detect or identify
the more contralesional of two briefly presented stimuli
although they correctly perceive isolated contralesional
stimuli [8,21]. Current theories propose that extinction is
a consequence of a diminished sensory or attentional processing capacity of the damaged hemisphere for the contralateral space [8,16]. According to the sensory hypothesis,
this processing limitation is due to a weakened or delayed
afferent input to the damaged hemisphere [12,22]. In contrast, the attentional hypothesis proposes that extinction is
the result of a biased competition for attentional selection
[11,20,31]. The attentional hypothesis fits well to current
models of attentional selection in the healthy brain. Selective attention is limited in capacity, hence visual stimuli
compete for selection in the healthy brain [5,9]. A unilateral
brain lesion biases this competition in favour of stimuli pre∗ Corresponding author. Tel.: +41-22-382-35-24;
fax: +41-22-372-36-44.
E-mail address: radek.ptak@hcuge.ch (R. Ptak).

sented in ipsilesional space [11]. A strong support for this
attentional explanation of extinction is the finding that the
ipsilesional selection bias can be moderated by perceptual
factors. Ward et al. [31] found that a contralesional stimulus
was better detected when it was perceptually similar (i.e.
two brackets) to the ipsilesional stimulus than when it was
dissimilar (i.e. a dot and a bracket). The authors argued that
two similar stimuli more easily grouped to form a pair and
were therefore more likely to be selected as a group than
two dissimilar stimuli. Further studies found that other perceptual factors such as connectedness [7], polarity [14], or
collinearity [25] also affected grouping of the ipsilesional
and contralesional stimulus. However, most of these studies
used detection paradigms, which have a very limited response set and do not allow a qualification of the patient’s
perception. The results of perceptual grouping studies therefore suggest that processing capacity for stimulus selection is allocated on an all-or-none basis: the patient either
detects the presence of a contralesional stimulus or not.
Based on these results, recovery of extinction by perceptual grouping is thought to result from segmentation of the
visual scene into more easily perceivable elements [6,19].
However, an alternative explanation for grouping effects

0028-3932/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 8 - 3 9 3 2 ( 0 2 ) 0 0 1 3 4 - 3

2200

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205

is that grouping does not affect the bottom-up perceptual
processes per se, but—at least in some cases—rather modulates the decision criterion of the patient while leaving
bottom-up processing unchanged. This hypothesis relies on
the assumption that spatial extinction is not an all-or-none
phenomenon, but that processing of the contralesional stimulus reflects a continuum between zero and total processing.
If such a continuum of more or less degraded representations exists, it should be possible to modulate extinction
top–down, by changing the expectancies of the patient. Evidence for a degraded representation comes from recent studies showing that the degree of extinction varies substantially
with the task demands. Vuilleumier and Rafal [29,30] asked
extinction patients either to localise or to enumerate briefly
presented stimuli. In contrast to a severe extinction observed
in the localisation task, contralesional stimuli were correctly
enumerated, suggesting that they had been processed at a
level allowing enumeration but not allowing the determination of their position. These results suggest that location
information was essential for awareness of the stimuli as
individual elements in space, but not for enumeration.
The present study was designed to investigate whether visual extinction can only be modulated by perceptual factors
or whether strategic factors and expectation biases equally
facilitate the access to perceptual information of competing
stimuli. Some evidence for top–down influences on extinction has been presented by Baylis et al. [1], who found that
the selection bias in stimulus identification and awareness
was strongly affected by task demands. These authors observed a paradoxical increase of extinction when the contralesional and the ipsilesional stimulus were identical on the
feature that was pertinent for the task. Thus, when asked to
identify the colour of two red stimuli, the patients reported
only the ipsilesional stimulus. However, when the stimuli
were of different colours, patients reported both stimuli. Attending to a particular stimulus feature thus affected contralesional extinction, suggesting a top–down modulation of
representations competing for attention. In the present study,
we compared perceptual effects of grouping by similarity in
a detection task (Experiment 1) with the effect of a criterion
change in an identification task (Experiment 2). We found
that a criterion change and a change of task demands facilitated processing of the contralesional stimulus. In Experiment 3, we used a cueing procedure to direct the patient’s
attention to a particular stimulus dimension before stimulus
presentation and observed a modulation of extinction by biased expectations of the patient.

2. Methods
2.1. Patient
AG is a 45-year-old man who suffered a small right
hemispheric cerebral haemorrhage due to a probable arterioveneous malformation 4 months before the present study,

Fig. 1. T2-weighted MRI-scan of the patient showing a right periventricular white matter lesion.

resulting in a left hemiplegia and severe sensory loss of the
left arm. A MRI scan documented the presence of subcortical damage of the right putamen, the external capsule, and
the superior part of the internal capsule (Fig. 1). At the time
of study, AG presented signs of moderate neglect in visual
exploration tasks (five left omissions in the Bells-test [13]),
and line bisection (mean 19.2% right deviation on 15–20 cm
lines). The visual fields of the patient were intact on confrontation testing. Whereas he correctly detected stationary
or moving unilateral visual stimuli, he would only report
the right stimulus on bilateral stimulation. This severe left
visual extinction was further examined in the present study.
AG gave written consent before participating in the experiments. This study was approved by the ethical committee
of the University of Geneva.
2.2. Experiment 1
2.2.1. Material and procedure
Experiment 1 was designed to measure the rate of visual
extinction in AG and its possible modulation by perceptual
grouping. Stimuli were green or red, squares or diamonds.
Squares were 3.5 cm wide; diamonds were created by rotating the squares by 45◦ . Isoluminance of the colours was
determined with a light meter. The stimuli were presented
on black background with the inner border presented at
1 cm from fixation (visual angle ∼1◦ ). Four conditions
were created for the bilateral trials: stimulus pairs with the
same colour and the same form (condition SC/SF), different
colour—same form pairs (DC/SF), same colour—different
form pairs (SC/DF) and different colour—different form
pairs (DC/DF). Bottom-up grouping by similarity would
predict the least contralesional extinction in the SC/SF condition and most extinction in the DC/DF condition. Based
on the findings of Baylis et al. [1] we could also expect the

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205

reverse, namely, increased extinction with similar stimuli,
though only if increased extinction with similar stimuli
does not depend on the task.
There were 36 unilateral left and right trials, 44 bilateral
trials in either of the SC/SF or DC/SF conditions, 24 bilateral trials in either of the SC/DF or DC/DF conditions as
well as 26 catch trials with no stimulus present. Stimuli were
presented for 150 ms in pseudo-random order (no more than
three unilateral or bilateral presentations in succession). The
patient sat at a distance of ∼60 cm from the screen in a dim
room. Every trial was initiated by the appearance of a fixation cross in the middle of the screen. One experimenter
controlled fixation and announced every trial verbally. After
having checked that the patient maintained stable fixation,
the experimenter released a stimulus presentation by pressing the mouse button. The patient indicated whether he saw
two, one left, one right or no stimulus; colour and form were
therefore irrelevant. Before starting the experiment, AG was
given 50 practice trials in which one experimenter controlled
the fixation and gave verbal feedback whenever the patient
made an eye movement. After a few practice trials gaze fixation was reliable and eye movements were no longer observed. In this way it was assured that a stimulus would only
be presented when AG fixated the central cross.
2.2.2. Results
The patient neither misidentified stimuli nor produced
false positive responses on catch trials. All errors were
misses. AG detected 28/36 unilateral left stimuli and all 36
unilateral right stimuli. The important aspect of his performance concerns his poor detection of left items in bilateral
displays (13/136 trials). The critical comparison between
the detection of left stimuli in unilateral left (22.2% missed)
and in bilateral displays (89.7% missed) was highly significant (χ 2 (1) = 73.0, P < 0.0001), indicating severe
left extinction. Contralesional detection performance did
not differ between the four conditions (SC/SF = 4/44;
DC/SF = 3/44; SC/DF = 3/24; DC/DF = 3/24; Fisher’s
exact test, all P = n.s.), suggesting that similarity by colour
or form did not afford significant grouping when these
features were irrelevant to the task.

2201

criterion than in Experiment 1. Fixation was controlled in
the same way as in the previous experiment. There were 96
presentations with stimuli grouped by colour (48 in conditions SC/SF and SC/DF) and 96 presentations with stimuli
grouped by form (48 in conditions SC/SF and DC/SF). Note
that in the SC/SF condition stimuli grouped by colour and
by form. In Experiment 2a, immediately after each stimulus
presentation a word indicating the side of the display (left
or right) was presented at fixation, and AG was prompted to
report both features (colour and form) on the relevant side
(e.g. “LEFT?”). Since the patient always indicated both features, he identified 48 times the colour and 48 times the form
on either display side. AG did not know in advance whether
he had to identify the left or right stimulus. The stimulus
to be identified was determined pseudo-randomly with no
more than three times the same side in succession. AG was
allowed to guess when feeling uncertain. In all other aspects
the procedure was the same as in Experiment 1.
Experiment 2b followed the same procedure as Experiment 2a except that immediately after stimulus presentation
a cue indicating the side (left or right) and the feature (colour
or form) of the display was presented at fixation, and the
patient was prompted to report the relevant feature on the
relevant side (e.g. “LEFT COLOUR?”). Since the patient
always indicated only one feature, he identified 24 times the
colour and 24 times the form on either display side.
2.3.2. Results
Table 1 shows the patient’s performance for the grouped
and ungrouped presentations and Fig. 2 summarises the results of Experiments 2a and 2b.
Experiment 2a: AG identified in 68/96 trials the colour of
the left stimulus correctly, which is significantly better than
chance (χ 2 (1) = 8.7, P < 0.01). In contrast, form identification of left-sided stimuli was correct in only 54/96 trials,
which is not significantly different from chance (χ 2 (1) =
0.8). As Fig. 2 indicates, there was an advantage in performance of displays grouped by colour or form over ungrouped
displays. In fact, only the grouped displays yielded a performance significantly better than chance (grouped: χ 2 (1) =
8.7, P < 0.01; ungrouped: χ 2 (1) = 0.8). Importantly, the
features of grouped stimuli were significantly better recog-

2.3. Experiment 2
2.3.1. Material and procedure
Experiment 2 addressed the question of whether a grouping effect would emerge when AG was forced to characterise the perceptual features of the contralesional stimulus
rather than indicating only its presence. A second aim was
to test whether a perceptual grouping effect would depend
on the number of features reported by the patient (one or
two features). The patient was explicitly told that this time
always two stimuli would be presented, and that he would
be asked to identify features of the left or right stimulus even
if he doubted whether he had seen anything. In this way
we hoped that AG would adopt a less conservative decision

Table 1
Results of Experiments 2a and 2ba
Identified
feature

Experiment 2a

Experiment 2b

LVF

RVF

LVF

RVF

Colour
Grouped
Ungrouped

38/48 (79.2)
30/48 (62.5)

48/48 (100)
48/48 (100)

19/24 (79.2)
12/24 (50)

24/24 (100)
24/24 (100)

Form
Grouped
Ungrouped

30/48 (62.5)
24/48 (50)

33/48 (68.8)
33/48 (68.8)

15/24 (62.5)
12/24 (50)

18/24 (75)
19/24 (79.2)

LVF/RVF: left/right visual field.
a Percent correct is given in parenthesis.

2202

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205

only: 56.3% versus 50%, χ 2 (1) = 0.5), suggesting that AG
had a comparable performance when reporting both or when
reporting only one feature of the contralesional stimulus.
2.4. Experiment 3

Fig. 2. Percent of trials in Experiment 2 on which stimulus features in
the grouped and ungrouped displays were correctly identified in the left
(LVF) and right (RVF) visual field. In Experiment 2a AG identified on
each trial both features of one stimulus (colour and form), whereas on
each trial in Experiment 2b he identified a single feature (colour or form).
The dashed horizontal line represents chance performance.

nised than the features of ungrouped stimuli (70.8% versus
56.3%, χ 2 (1) = 4.4, P < 0.05; see Fig. 2). As Table 1
shows, this advantage of grouped over ungrouped displays
was present for colour and form. The grouping effect was
comparable between displays with identical relevant features
only (SC/DF and DC/SF) and displays with identical relevant and irrelevant features (SC/SF; 68.8% versus 72.9%;
χ 2 (1) = 0.2), indicating that similarity of the irrelevant
stimulus feature did not contribute to the observed grouping
effect.
Experiment 2b: AG identified the left colour in 31/48 trials and the left form in 27/48 trials, which is not different
from chance (colour: χ 2 (1) = 2.1; form: χ 2 (1) = 0.4).
However, as Fig. 2 shows, identification of left-sided stimuli
was better in grouped than in ungrouped displays. Only the
grouped displays yielded a performance significantly better
than chance (grouped: χ 2 (1) = 4.4, P < 0.05; ungrouped:
χ 2 (1) = 0). In addition, performance with grouped displays
was significantly better than in ungrouped displays (70.8%
versus 50%, χ 2 (1) = 4.4, P < 0.05; see Fig. 2). As in Experiment 2a, the advantage of grouped over ungrouped displays was present for both features (Table 1), and the grouping effect was comparable between displays with identical
relevant features only (SC/DF and DC/SF) and displays with
identical relevant and irrelevant features (SC/SF; 75% versus 66.7%; χ 2 (1) = 0.4).
Comparison of Experiments 2a and 2b: In order to test
whether reporting two features (as in Experiment 2a) and
reporting only one feature (Experiment 2b) yielded different
results, we performed a comparison of overall performance
and performance in the grouped and ungrouped displays between both experiments. None of these comparisons reached
significance (overall: 63.5% versus 60.4%, χ 2 (1) = 0.3;
grouped only: 70.8% versus 70.8%, χ 2 (1) = 0; ungrouped

2.4.1. Material and procedure
While the first two experiments examined the modulation
of extinction by perceptual characteristics of the display (Experiment 1) or by task demands (Experiment 2), Experiment
3 assessed whether AG’s performance would be modulated
by his attentional set, i.e. expectations induced before presenting the stimuli. The critical question was whether AG’s
identification of the contralesional stimulus would be enhanced when his attention was oriented to a single feature
before stimulus presentation.
As in Experiment 2, there were 48 bilateral presentations
of each of the four conditions SC/SF, DC/SF, SC/DF and
DC/DF, yielding 96 presentations with stimuli grouped by
colour and 96 presentations with stimuli grouped by form.
A trial was initiated by a fixation cross appearing on the
screen. The cross was followed by a cue (e.g. the word
“COLOUR?”), presented for 1000 ms at fixation. The stimulus display was flashed for 150 ms 1 s after disappearance
of the cue, yielding a stimulus-onset asynchrony (SOA) of
2000 ms. Immediately after stimulus presentation, a question appeared on the screen indicating the side and the
feature of the stimulus that the patient had to report. The
question was either coherent (e.g. “LEFT COLOUR?”) or
incoherent (e.g. “LEFT FORM?”) with the cue. On either
side, there were 32 coherent (valid) questions and 16 incoherent (invalid) questions with grouped and ungrouped
displays. AG was told that most of the cues were coherent
with the question and that he could enhance his performance
when focusing on the feature indicated by the cue.
This procedure allowed investigating two questions. The
first was whether AG would show a “validity effect”, that
is whether his extinction on valid trials would decrease
in comparison to invalid trials. The second question concerns the comparison between the influence of bottom-up
and top–down processes on contralesional extinction. If
bottom-up processes modulate extinction despite the biased expectations of the patient, grouped displays should
produce better results than ungrouped displays. However,
if the top–down validity effect is more important than the
perceptual grouping effect, there should be no difference
between grouped and ungrouped displays.
2.4.2. Results
The results are presented in Table 2 and Fig. 3. AG correctly identified the left colour in 36/48 and the left form
in 25/48 trials. Only colour identification was better than
chance (colour: χ 2 (1) = 6.4, P < 0.05; form: χ 2 (1) =
0.1). However, as Fig. 3 shows, identification of both features of the contralesional stimulus was significantly better
in valid than in invalid trials (χ 2 (1) = 5.8, P < 0.05),

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205
Table 2
Results of Experiment 3a
Identified feature

LVF

RVF

Grouped
Valid
Invalid

25/32 (78.1)
8/16 (50)

27/32 (84.4)
15/16 (93.8)

Ungrouped
Valid
Invalid

21/32 (65.6)
7/16 (43.8)

27/32 (84.4)
14/16 (87.5)

LVF/RVF: left/right visual field.
a Percent correct is given in parenthesis.

Fig. 3. Percent correct identifications of features of the left (LVF) or right
(RVF) stimulus after valid or invalid cueing (Experiment 3). The dashed
horizontal line represents chance performance.

consistent with a validity effect. Only performance in valid
trials was significantly different from chance (valid: χ 2 (1) =
6.4, P < 0.05; invalid: χ 2 (1) = 0.1). A validity effect was
found for both features (size of the validity effect: 28.1%
for colour and 21.9% for form) and for grouped as well as
ungrouped displays (size of the validity effect: 28.2% for
grouped and 21.9% for ungrouped displays).
We also examined the influence of bottom-up grouping
in comparing the grouped with ungrouped displays. The
performance in both of these was not significantly different from chance, although it approached significance in the
grouped condition (grouped: χ 2 (1) = 3.5, P < 0.07; ungrouped: χ 2 (1) = 0.7). Importantly, there was no significant
difference between performance in grouped and ungrouped
displays (χ 2 (1) = 1.1). In sum, these results suggest that
bottom-up grouping by perceptual factors was not of the
same importance as in Experiment 2 when processing of
relevant features was enhanced by top–down cues.

3. Discussion
Previous studies have demonstrated that visual extinction
could be reduced by bottom-up grouping factors such as
similarity of shape [31] or polarity [14]. Our patient’s left

2203

visual extinction was not affected by similarity of colour and
shape in a task requiring detection and localisation of the
contralesional stimulus (Experiment 1). In contrast, AG had
reduced contralesional extinction in the identification task of
Experiment 2 when both stimuli grouped by colour or form.
It is unlikely that this effect reflects a bias to systematically
report the features of the ipsilesional stimulus since such a
bias would lead to a performance worse than chance in the
condition with different ipsilesional and contralesional features (i.e. the ungrouped condition of Experiment 2). Since
AG’s performance in the ungrouped condition was at chance
or slightly better (61.4% in Experiment 2a and 50% in Experiment 2b), the results rather reflect a grouping effect than
the tendency to report ipsilesional stimulus features. Grouping by similarity was observed for stimuli grouped by colour
and stimuli grouped by form (see Table 1), although form
identification was much more difficult for AG (even for stimuli presented to his ipsilesional visual field), reflecting the
severe visual-spatial deficits of the patient.
Previous studies reporting evidence for perceptual grouping by factors like collinearity [14,20,25] or polarity [14]
used detection by localisation or counting as performance
measure. A failure to report a contralesional stimulus in
these paradigms is likely to reflect complete extinction of
that stimulus from awareness. In contrast, a failure to report
a contralesional feature in an identification task (Experiment
2) does not necessarily reflect complete absence of knowledge. Degraded information may still be represented, and
expectations of the patient may affect further processing of
this degraded knowledge. Except for the absence of unilateral displays, Experiment 2 used exactly the same (bilateral)
displays as Experiment 1. Therefore, the observed grouping effect cannot be explained by perceptual factors only.
An explanation in terms of a subjective criterion shift may
better account for these findings. Our results suggest that a
weak representation of the contralesional stimulus was not
sufficient to generate a response in Experiment 1. In contrast, in Experiment 2 the patient was explicitly told that always two stimuli would be presented, and that he would be
asked to identify features of the left or right stimulus even
if he doubted whether he had seen anything. This advance
knowledge and the fact that he gave forced-choice answers
may have facilitated AG’s attempts to identify weakly represented contralesional stimuli. The contrast between the results of Experiment 1 and 2, using the same displays, but
differing with respect to the instructions and task demands,
suggests that the observed grouping effect was due to a shift
of AG’s subjective response criterion. Comparable adjustments of response criteria have been reported in healthy subjects expecting a pre-specified number of targets [15]. The
present experiments specifically tested grouping by similarity, but it would be interesting to examine whether a comparable modulation of extinction could be observed with other
grouping factors.
The present results contrast with the observations of
Baylis et al. [1] and Vuilleumier and Rafal [30], who

2204

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205

observed a significant increase of extinction by stimulus
similarity when patients identified features of contralesional
and ipsilesional stimuli. The difference in results reflects
how task demands may affect the processing of stimuli
competing for attention. Baylis et al. [1] presented red or
green letters and asked their patients to report the colour or
the form of both stimuli. When both stimuli were identical (e.g. both red) with respect to the feature that patients
attempted to identify (e.g. colour) the patients showed increased contralesional extinction in comparison to when
the stimuli were different (e.g. red and green). Vuilleumier
and Rafal [30] instructed their patients to count the number
of stars present among one, two, or four stars or triangles.
Similarly, these authors observed increased extinction when
a contralesional star was paired with an ipsilesional star in
comparison to when the ipsilesional stimulus was a triangle.
In contrast to these two studies our patient identified only
one stimulus on each trial. Thus, both studies observing
increased extinction with identical items required the patient to identify the ipsilesional and contralesional stimulus
on each trial, whereas in our study decreased extinction
was associated with processing limited to the contralesional
stimulus. The identity of an ipsilesional stimulus which the
patient attempts to identify thus seems to interfere with
the identification of the contralesional stimulus, whereas it
facilitates its identification when the patient is not required
to identify the ipsilesional stimulus. These results suggest
that preattentive parsing by stimulus similarity may affect
performance positively only when patients do not explicitly
process the ipsilesional attributes involved in the parsing
process. This conclusion is in accord with findings of decreased extinction when patients performed detection or
localisation tasks [14,19–30,31] and with the observation
of decreased extinction when patients were instructed to
ignore the ipsilesional stimulus [21].
A further finding of this study is that AG’s identification
of a contralesional stimulus feature was better when he expected to identify this feature than when he expected to identify the other feature (Experiment 3). The cueing paradigm
used in Experiment 3 resembles the paradigm originally applied by Posner et al. [26] to study spatial orienting of attention. These authors used spatial cues in a simple reaction
task to provoke covert ipsilesional or contralesional shifts of
attention before the appearance of the target and observed
an increased cue-validity effect (i.e. an advantage of ipsilateral cueing over contralateral cueing) in patients with parietal damage when targets appeared on the contralesional
side after ipsilesional cueing. In contrast to the paradigm
of Posner et al., the cues in the present Experiment 3 did
not provoke a spatial shift of attention, but rather facilitated
processing of relevant object attributes in comparison to attributes that had not been cued (feature-validity effect). A
possible mechanism of this effect is that focusing of attention on a stimulus feature primed the perceptual trace of this
feature compared to a condition where the patient’s attention was divided between two features, again suggesting a

performance cost due to any additional processing with limited attentional resources.
Similar top–down effects on processing of neglected or
extinguished information have been observed in other studies. For example, Smania et al. [27] observed that predictability of the location of visual stimuli speeded up detection times of stimuli both in the unaffected as well as the
affected visual field, indicating spared voluntary shifting of
attention. Duncan et al. [10] found that neglect patients were
able to use colour as top–down cue to identify contralesional
targets in a partial report task despite being unable to identify the colour of contralesional stimuli explicitly. Concerning extinction studies, consider again Baylis et al. [1] observation of a paradoxical increase of extinction when the ipsilesional and contralesional stimulus were coherent on the
feature that the patient was attempting to identify. In this
study, whether the colour of the two stimuli was identical
or not, did not affect identification of form (and vice versa),
suggesting that once attention was engaged with processing of a particular feature (form), the other feature (colour)
did not interfere with processing of the display. In another
study, Danckert et al. [4] found that flankers in the neglected
visual field of a patient with parietal damage affected his
identification of a central target only when they were identical with the target on the attended feature. In control participants, Corbetta et al. [3] found that attention focused on
one of three stimulus features enhanced sensitivity to detect
a slight change in this feature in comparison to a condition
in which attention was divided between all three features.
These results are consistent with our observation that contralesional identification of a stimulus feature was enhanced
only, when attention had been focused on this feature prior
to stimulus presentation.
Our results have two implications. First, they demonstrate
that expectations of the patient may influence his identification of the contralesional stimulus. In situations of double
simultaneous stimulation, contralesional stimuli may activate a weakened representation rather than being completely
“extinguished”. This weakened representation might be revealed when using a more sensitive task [30] or when modulating the decision criterion of the patient, suggesting that
competition between stimuli is not only influenced by perceptual characteristics of the display but also by task demands and expectations of the patient.
Second, the validity effect observed in Experiment 3 suggests that advance information may facilitate processing of
specific object attributes. This effect implies that processors of particular stimulus dimensions (e.g. form or colour)
may be primed to enhance the identification of this dimension even if the stimulus location is not a priori known. The
facilitation of contralesional recognition when attention is
focused on a single feature might reflect the liberation of
resources initially engaged in processing of the uncued feature. Comparative effects of enhancement of processing by
selective attention have been found in recent neurophysiological studies. For example, when an animal directs his at-

R. Ptak et al. / Neuropsychologia 40 (2002) 2199–2205

tention to a preferred stimulus of a cell in V4 or the inferior
temporal lobe, the activity of the cell will increase whereas
it will decrease with a shift of attention to a non-preferred
stimulus, although both stimuli are within the receptive field
of the cell [2,23,28]. Since V4 and the inferior temporal
lobe are involved in form, pattern, and colour analysis, such
attention-dependent modulations of cell activity suggest a
possible neurophysiological mechanism for attentional selection of objects that are pertinent to a task. Functional
brain imaging studies [17,18,24] found activity associated
with spatial cueing in the superior frontal cortex, the superior temporal cortex, and the posterior cingulate cortex, suggesting that a network of regions outside the parietal lobe is
involved in top–down modulation of attention. Further, recent studies suggest that non-spatial allocation of attentional
resources specifically activates regions of the parietal lobe
and thalamic pulvinar, structures that were not activated by
spatial cues [18]. It is possible that the preservation of these
regions in our patient was crucial for the occurrence of a
non-spatial cueing-effect.
In summary, our study shows that bottom-up modulation
may not always be strong enough to enhance contralesional
detection. We observed a recovery of extinction based on a
criterion shift of the patient (Experiment 2) or on the expectation of a specific stimulus feature (Experiment 3). These
results demonstrate that attention may enhance processing
of specific object features and thus diminish a severe spatial
attentional bias.

Acknowledgements
We thank C. Greber and P. Vuilleumier for their helpful
comments on an earlier version of this paper.

References
[1] Baylis GC, Driver J, Rafal RD. Visual extinction and stimulus
repetition. Journal of Cognitive Neuroscience 1993;5:453–66.
[2] Chelazzi L, Duncan J, Miller EK, Desimone R. Responses of
neurones in inferior temporal cortex during memory-guided visual
search. Journal of Neurophysiology 1998;80:2918–40.
[3] Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Petersen SE.
Selective and divided attention during visual discriminations of
shape, colour, and speed: functional anatomy by positron emission
tomography. Journal of Neuroscience 1991;11:2383–402.
[4] Danckert J, Maruff P, Kinsella G, de Graaff S, Currie J. Attentional
modulation of implicit processing of information in spatial neglect.
NeuroReport 1999;10:1077–83.
[5] Desimone R, Duncan J. Neural mechanisms of selective visual
attention. Annual Review of Neuroscience 1995;18:193–222.
[6] Driver J, Baylis GC. Attention and visual object segmentation. In:
Parasuraman R, editor. The attentive brain. Cambridge (MA): MIT
Press; 2000. p. 299–325.
[7] Driver J, Mattingley JB, Rorden C, Davis G. Extinction as a paradigm
measure of attentional bias and restricted capacity following brain
injury. In: Thier P, Karnath H-O, editors. Parietal lobe contributions
to orientation in 3D space. Heidelberg: Springer; 1997. p. 401–29.

2205

[8] Driver J, Vuilleumier P. Perceptual awareness and its loss in unilateral
neglect and extinction. Cognition 2001;79:39–88.
[9] Duncan J. Selective attention and the organisation of visual
information. Journal of Experimental Psychology: General
1984;113:501–17.
[10] Duncan J, Bundesen C, Olson A, Humphreys G, Chavda S, Shibuya
H. Systematic analysis of deficits in visual attention. Journal of
Experimental Psychology: General 1999;128:450–78.
[11] Duncan J, Humphreys G, Ward R. Competitive brain activity in
visual attention. Current Opinion in Neurobiology 1997;7:255–
61.
[12] Farah MJ, Monheit MA, Wallace MA. Unconscious perception
of “extinguished” visual stimuli: reassessing the evidence.
Neuropsychologia 1991;29:949–58.
[13] Gauthier L, Dehaut F, Joanette Y. The Bells Test: a quantitative and
qualitative test for visual neglect. International Journal of Clinical
Neuropsychology 1989;11:49–54.
[14] Gilchrist ID, Humphreys GW, Riddoch MJ. Grouping and extinction:
evidence for low-level modulation of visual selection. Cognitive
Neuropsychology 1996;13:1223–49.
[15] Gorea A, Sagi D. Failure to handle more than one internal
representation in visual detection tasks. Proceedings of the National
Academy of Sciences 2000;97:12380–4.
[16] Heilman KM, Watson RT, Valenstein E. Neglect and related disorders.
In: Heilman KM, Valenstein E, editors. Clinical neuropsychology.
New York: Oxford University Press; 1993. p. 279–336.
[17] Hopfinger JB, Buonocure MH, Mangun GR. The neural mechanisms
of top–down attentional control. Nature Neuroscience 2000;3:284–
91.
[18] Hopfinger JB, Woldorff MG, Fletcher EM, Mangun GR. Dissociating
top–down attentional control from selective perception and action.
Neuropsychologia 2001;39:1277–91.
[19] Humphreys GW. Neural representation of objects in space: a dual
coding account. Philosophical Transactions of the Royal Society of
London B 1998;353:1341–51.
[20] Humphreys GW, Romani C, Olson A, Riddoch MJ, Duncan J.
Non-spatial extinction following lesions of the parietal lobe in
humans. Nature 1994;372:357–9.
[21] Karnath H-O. Deficits of attention in acute and recovered visual
hemineglect. Neuropsychologia 1988;26:27–43.
[22] Marzi CA, Girelli M, Natale E, Miniussi C. What exactly is
extinguished in unilateral visual extinction? Neurophysiological
evidence. Neuropsychologia 2001;39:1354–66.
[23] Moran J, Desimone R. Selective attention gates visual processing in
the extrastriate cortex. Science 1985;229:782–4.
[24] Nobre AC. The attentive homunculus: now you see it, now
you don’t. Neuroscience and Biobehavioral Reviews 2001;25:477–
96.
[25] Pavlovskaya M, Sagi D, Soroker N, Ring H. Visual extinction and
cortical connectivity in human vision. Cognitive Brain Research
1997;6:159–62.
[26] Posner MI, Walker JA, Friedrich FJ, Rafal RD. Effects of parietal
injury on covert orienting of attention. Journal of Neuroscience
1984;4:1863–74.
[27] Smania N, Martini MC, Gambina G, Tomelleri G, Palamara A, Natale
E, et al. The spatial distribution of visual attention in hemineglect
and extinction patients. Brain 1998;121:1759–70.
[28] Treue S. Neural correlates of attention in primate visual cortex.
Trends in Neuroscience 2001;24:295–300.
[29] Vuilleumier P, Rafal R. “Both” means more than “two”:
localizing and counting in patients with visuospatial neglect. Nature
Neuroscience 1999;2:783–4.
[30] Vuilleumier PO, Rafal RD. A systematic study of visual extinction.
Brain 2000;123:1263–79.
[31] Ward R, Goodrich S, Driver J. Grouping reduces visual extinction:
neuropsychological evidence for weight-linkage in visual selection.
Visual Cognition 1994;1:101–29.