eyeing the eyes in social scenes: evidence for top … et...eyeing the eyes in social scenes:...

Eyeing the eyes in social scenes: Evidence for top-downcontrol of stimulus selection in simultanagnosia

Kirsten A. Dalrymple1, Alexander K. Gray1, Brielle L. Perler1, Elina Birmingham2,Walter F. Bischof3, Jason J. S. Barton1,4,5, and Alan Kingstone1

1Department of Psychology, University of British Columbia, Vancouver, Canada2Faculty of Education, Simon Fraser University, Burnaby, Canada3Department of Computing Science, University of Alberta, Edmonton, Canada4Department of Medicine (Neurology), University of British Columbia, Vancouver, Canada5Department of Ophthalmology & Visual Sciences, University of British Columbia, Vancouver, Canada

Simultanagnosia is a disorder of visual attention resulting from bilateral parieto-occipital lesions.Healthy individuals look at eyes to infer people’s attentional states, but simultanagnosics allocateabnormally few fixations to eyes in scenes. It is unclear why simultanagnosics fail to fixate eyes, butit might reflect that they are (a) unable to locate and fixate them, or (b) do not prioritize attentionalstates. We compared eye movements of simultanagnosic G.B. to those of healthy subjects viewingscenes normally or through a restricted window of vision. They described scenes and explicitly inferredattentional states of people in scenes. G.B. and subjects viewing scenes through a restricted windowmade few fixations on eyes when describing scenes, yet increased fixations on eyes when inferringattention. Thus G.B. understands that eyes are important for inferring attentional states and canexert top-down control to seek out and process the gaze of others when attentional states are ofinterest.

Keywords: Attention; Eye gaze; Neuropsychology; Simultanagnosia; Social scenes perception.

Dorsal simultanagnosia is a disorder of visualattention that results in an inability to see morethan one object at a time (Holmes & Horrax,1919; Rafal, 1997). It is typically the consequenceof bilateral lesions to the parieto-occipital junction(Balint, 1909; Holmes & Horrax, 1919; Riddochet al., 2010; Rizzo & Vecera, 2002) and may

occur in the context of Balint syndrome, thoughit can also occur independently (Damasio, 1985).Sometimes these patients can only see pieces ofthe objects in their visual world, unaware thatthey are seeing just one component of a largerform. Thus, when viewing single (global) formsmade up of smaller (local) elements in hierarchical

Correspondence should be addressed to Kirsten A. Dalrymple, Department of Psychological and Brain Sciences, Dartmouth

College, 6207 Moore Hall, Hanover, NH 03755, USA. (E-mail: [email protected]).

K.A.D. and E.B. were supported by the Natural Sciences and Engineering Research Council (NSERC), and the Michael Smith

Foundation for Health Research (MSFHR). A.K. was supported by NSERC, a MSFHR Senior Scholar award, the Human Early

Learning Partnership, and the Hampton Foundation. J.J.S.B. was supported by a Canada Research Chair and MSFHR Senior

Scholarship. W.F.B. was supported by NSERC. Thank you to G.B. for his time and dedication to this project.

# 2013 Taylor & Francis 25

Cognitive Neuropsychology, 2013

Vol. 30, No. 1, 25– 40, http://dx.doi.org/10.1080/02643294.2013.778234

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

stimuli (Navon, 1977), patients report the localelements but not the global aspect, a phenomenonknown as “local capture” (Clavagnier, FruhmannBerger, Klockgether, Moskau, & Karnath, 2006;Dalrymple, Bischof, Cameron, Barton, &Kingstone, 2009; Dalrymple, Kingstone,& Barton, 2007; Karnath, Ferber, Rorden, &Driver, 2000). Notably, their global perceptionimproves when local elements become smaller ormore densely packed, emphasizing the modulatingimpact of stimulus properties on attentional defi-cits (Clavagnier et al., 2006; Dalrymple et al.,2007).

Dorsal simultanagnosia has been popularlyunderstood as a restriction of object-based visualattention: Patients can only see one object at atime, at the expense of all other objects (Baylis,Driver, Baylis, & Rafal, 1994; Moreaud, 2003;Rafal, 2003). However, others have proposedthat simultanagnosia may involve a space-basedlimitation of attention. For example, someauthors have described simultanagnosia as areduction of the useful visual field (Bay, 1953;Thaiss & de Bleser, 1992; Tyler, 1968). Morerecently, Shalev, Humphreys, and Mevorach(2004) reported a restricted spatial area of atten-tion in a simultanagnosic patient, which could beenlarged by a large preceding visual prime.Similarly, Michel and Henaff (2004) used crowd-ing, counting, attentional tracking, and other teststo determine that a simultanagnosic patient had a“shrinkage” of the attentional visual field despitenormal visual fields on perimetry.

To test the idea that simultanagnosia may beexplained by a restriction of spatial attention, wepreviously restricted the visual window of healthyparticipants to only a small region of space at anyone time (Dalrymple, Bischof, Cameron, Barton,& Kingstone, 2010). Thus, we explored theability of a literal spatial constriction of vision tofunction as an analogue of the spatial constrictionof a spotlight of attention in simultanagnosia. Thiswas accomplished using a computer-generatedgaze-contingent technique that allowed healthyindividuals to see only within a small window sur-rounding their current fixation. Gaze-contingentdisplays have been used in the past with a variety

of tasks, such as reading (McConkie & Rayner,1975), visual search (Pomplun, Reingold, &Shen, 2001), face perception (Caldara, Zhou, &Miellet, 2010), and scene exploration (Loschky,McConkie, Yang, & Miller, 2005). It is importantto note that we do not suggest that a literal shrink-age of the visual field in healthy individuals isequivalent to a shrinkage of the attentionalwindow in simultanagnosia. Rather, by restrictingthe visual window in healthy individuals, we arereducing the spatial area within which they canattend to a stimulus. Thus, those viewing scenesthrough a restricted window of vision and individ-uals with simultanagnosia are similarly restrictedin terms of the spatial area from which they cangather visual information for the experimentaltask. Using this computer-generated gaze-contingent technique, we first found that healthyindividuals viewing hierarchical letters through arestricted window of vision showed a similar diffi-culty in reporting global letters as simultanagnosicpatients, and, as in simultanagnosia, their per-formance improved as the local elements becamesmaller or more densely packed (Dalrympleet al., 2010). This was the first indication thatrestricting the visual window of healthy individualscould be a useful tool for exploring at least someaspects of the simultanagnosic experience.

This possibility has led to a new set of questions:How do simultanagnosics view more complex andnatural stimuli like social scenes, and how well doesa restricted window of vision replicate theirbehaviour in these settings? We asked a simultanag-nosic patient (S.L.) to describe social scenes while wemonitored her eye movements (Dalrymple,Birmingham, Bischof, Barton, & Kingstone,2011a). Unlike healthy subjects, who allocate alarge proportion of fixations to the eyes of people inthese scenes, S.L. distributed her gaze more evenlyacross the scenes, on objects, and on the heads andbodies of people. This contrasts with patients withparieto-occipital brain damage but no simultanagno-sia, who show the normal tendency to fixate the eyesof people in scenes (Birmingham, Bischof, &Kingstone, 2007, 2008a, 2008b; Dalrymple et al.,2011a; Smilek, Birmingham, Cameron, Bischof, &Kingstone, 2006). Critically, reduced fixations on

26 Cognitive Neuropsychology, 2013, 30 (1)

DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

the eyes that was apparent in S.L.’s behaviour wasalso found in our gaze-contingent simulation ofsimultanagnosia in healthy subjects: Individualsviewing social scenes through a restricted windowof vision made abnormally few fixations on the eyesof people in the scenes.

While a restricted window of vision againappeared to capture the behavioural pattern ofthe restricted window of attention in simultanag-nosia, it is unclear why these constraints shouldlead to reduced fixations on the eyes. One straight-forward explanation is that it reflects a problemwith the allocation of spatial attention in therestricted window condition and in simultanagno-sia—a limitation that restricts the ability to locateand fixate the eyes of the people in the scenes. Analternative explanation is more in line with a sug-gestion by Birmingham et al. (2008b) who pro-posed that observers look at eyes of people inorder to understand their attentional states. Thisaccount was tested by Birmingham et al. (2008b)by asking observers to either describe scenes thatcontained people or to infer the attention ofpeople in scenes. The results indicated that obser-vers made a significantly greater number of fix-ations on the eyes in the infer attention conditionthan when they were asked simply to describe thescenes. Thus an alternative explanation for whysimultanagnosic patients and healthy individualsin the gaze-contingent paradigm reduce their fix-ations on the eyes is that the attentional states ofthe people in the scene are not prioritized whenthey are asked to describe a scene when much ofthe scene is not perceived. In other words, patientsand healthy gaze-contingent participants canlocate and fixate the eyes, but when scene percep-tion is restricted, the eyes of people in the scenesare not prioritized because they provide little infor-mation about the content of the scene.

To distinguish between these two alternatives,in this study we monitored the eye movements ofsimultanagnosic patient G.B., healthy individualsviewing scenes through a gaze-contingent

window, and healthy individuals viewing scenesin an unrestricted viewing condition, while theyperformed two separate tasks (as in Birminghamet al., 2008b). In the first task they describedsocial scenes, in part to replicate our previous find-ings that patients with simultanagnosia andhealthy subjects with a restricted viewing windowtend not to look at the eyes of people in socialscenes. Replication here is important in and ofitself because it confirms the unusual looking be-haviour in a second patient with simultanagnosiaand demonstrates the reliability of our findingswith healthy subjects using the gaze-contingentdisplay.1 In the second task, which is novel tothis study and included specifically to determinewhy simultanagnosics exhibit this unusual behav-iour, the same individuals were asked to infer theattentional states of people in the scenes.Birmingham, Bischof, and Kingstone (2009)have demonstrated that the tendency to fixateeyes of people in scenes is not due to bottom-upsaliency cues, but rather due to a goal-directedsearch for social attention information (see alsoBirmingham & Kingstone, 2009, for a fullreview on social attention and the importance ofgaze information). Thus, this task was chosenbecause explicitly asking individuals to infer atten-tional states of people in social scenes increases theimportance of the eyes to the normal healthyobserver. If patients with simultanagnosia cannotfind the eyes to use them to infer attention ofpeople in scenes, they should make equally lowproportions of fixations on the eyes of people inscenes in either task. However, if simultanagnosicpatients are capable of locating and using the eyesto infer attention, but do not usually prioritizeattentional states in scene viewing, then weshould see an increase in fixations when thepatients are asked explicitly to infer attentionalstates compared to when asked to describe thescenes. Finally, if the simultanagnosic behaviouralresults are related to an inability to process morethan a small portion of the scenes at one time,

1 The importance of replication has never been more important in the field of cognitive and social neuroscience, a point that has

recently been driven home in the Special Issue of the Perspectives on Psychological Science (2012) Volume 7, Number 6.

Cognitive Neuropsychology, 2013, 30 (1) 27

TOP-DOWN CONTROL IN SIMULTANAGNOSIA

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

there should be strong parallels with the behaviourof healthy subjects when they view scenes througha restricted viewing window.

The results will have important implications forour understanding of the nature of the attentionaldeficit in simultanagnosia. First, they will providean index of the top-down control of selection ofvisual information in simultanagnosia.Traditionally, simultanagnosics have beendescribed as having visual elements enter and exitawareness with little or no control over theirvisual experience (Holmes & Horrax, 1919; Rizzo& Hurtig, 1987). This claim is consistent withthe possibility that simultanagnosics allocate fewfixations on the eyes of people in scenes whenasked to describe scenes because they are unableto locate the eyes. However, increased fixationson the eyes of people in scenes when explicitlyinferring attentional states would indicate thatsimultanagnosics do in fact have top-downcontrol over their selection of visual information,a finding that would contradict the view that theirvisual experience is mostly controlled by stimulus-driven information. It would also indicate that,unlike healthy individuals, social attention is notprioritized in simultanagnosia when merelydescribing a scene. Second, testing the validityof a spatially restricted window of vision as ananalogue for simultanagnosia across differenttasks will speak to the debate between object-based versus space-based theories of simultanag-nosia. Similarities between simultanagnosic scan-ning behaviour and that of healthy individualsviewing scenes through a spatially restrictedwindow of vision would support a space-basedtheory of simultanagnosia. Critically, it wouldalso suggest that a restricted visual field for atten-tional selection may contribute to a decline inprioritization of social attention in simultanagno-sia when other task demands compete for limitedattentional resources.

Method

ParticipantsPatient G.B.. G.B. is a 32-year-old man with pos-terior reversible leukoencephalopathy (PRES).

Magnetic resonance imaging (MRI) revealed loca-lized abnormal hyperintense fluid-attenuatedinversion recovery (FLAIR) signals in medialoccipital lobes and parietal lobes, mainly involvingsubcortical U fibres (see Figure 1). Goldmannperimetry initially revealed paracentral rightinferior homonymous relative scotoma involvingthe central 5 degrees. He had a dressing apraxiaand some problems with visual search tasks.Reading was normal in fluency. He did well on asubset of items from the famous faces test rulingout the possibility of prosopagnosia. His perform-ance on a selection of items from the HooperVisual Organization Test revealed a few errorssuggesting inattention to detail. He could copygeometric designs but showed poor planning inclock drawing, with numbers placed inside andoutside the ring. Previous neuropsychologicaltesting on the Wechsler Adult IntelligenceScale–Third Edition (WAIS–III) suggestedaverage to superior premorbid intelligence, basedon verbal comprehension, vocabulary, andgeneral knowledge. G.B. did well on numericalcalculations and repetition, but had problemswith left–right orientation and visual memory.Scores from his neuropsychological evaluationcan be found in the supplementary onlinematerials.

Standard diagnostic tools for simultanagnosiaare the Boston Cookie Theft picture (Goodglass& Kaplan, 1983), which patients tend to describein a piecemeal manner (Rizzo & Vecera, 2002),and the overlapping figures test, which consistsof overlapping line drawings of familiar objects.Simultanagnosics tend to report seeing only oneof the overlapping figures at one time (Rafal,2003). Consistent with previous reports of simul-tanagnosia, G.B. described the Boston CookieTheft picture in a piecemeal manner, slowly, andlaboriously, despite having seen the picturebefore. For the overlapping figures test, G.B. haddifficulty identifying more than one element at atime. For example, he said he saw a basket andthat he thought that there were items in thebasket, but that he could not identify the items.Eventually he identified a clock (which was actu-ally a wrist watch) and much later a key, but


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

much of this was done through inferences aboutshapes that he could discriminate. He also ident-ified a cup, but eventually decided that it may infact be something else (it was actually a pipe).He never identified the light bulb.

Recent empirical work has revealed consistenciesacross simultanagnosics on other tasks. For example,when viewing hierarchical letters, which are globalletters made up of several repetitions of a localletter, patients with simultanagnosia tend to report

seeing the local letters, but not the global letters(Clavagnier et al., 2006; Karnath et al., 2000;Shalev et al., 2004). Performance on global letterstends to improve when interelement spacing isreduced (Dalrymple et al., 2009; Dalrymple et al.,2007). We tested G.B. with hierarchical letters ofdifferent sizes and densities. He performed poorlyat identifying the global form when the letterswere constructed of sparse local elements, but wasnear perfect when they were made up of local

Figure 1. Magnetic resonance imaging (MRI) scans of patient G.B.



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

elements that were densely packed. He performedperfectly at identifying the local letters for all size/density configurations (Table 1).

G.B. also performed like a previously reportedsimultanagnosic (S.L.) on a hierarchical facestest. Hierarchical faces are faces made up of aseries of objects (e.g., a face made up of fruitwith a pear for a nose and berries for eyes).When viewing these faces, S.L. reported seeingjust the face, or the objects that make up theface, but not both (Dalrymple et al., 2007).Similarly, G.B. identified either the elementsthat made up the face (e.g., berries), or the face,but never identified both in the same image.

Despite his simultanagnosia, G.B. performednormally when asked to name real objects, bodyparts, and line drawings of inanimate andanimate objects. He had average abilities for dis-crimination of visual detail, pattern recognition,and picture arrangement. He did well on degradedstimuli tasks and on object vision generally. At thetime of testing, his visual field defects were muchreduced (Figure 2). He still had visuospatial orien-tation difficulties and dressing apraxia.

Comparison groups. Our primary research questionsare specific and concern (a) the change, if any, inG.B.’s fixations on the eyes during the describetask versus the infer attention task, and (b)whether G.B.’s eye movement behaviour duringthese tasks can be mimicked by restricting thespatial extent of the visual window of healthy indi-viduals performing the same tasks. To anticipateour results, the answer is that healthy individualswill appear “simultanagnosic” on the present taskwhen the spatial extent of the visual window isconstrained. Thus, we included two comparisongroups in this study. The “full-view group” con-sisted of healthy individuals who viewed thescenes naturally and allowed for a measure ofnormal fixation allocation in each task. The“gaze-contingent group” consisted of healthy

Table 1. G.B.’s accuracy for naming global and local levels of

hierarchical letters

Density

Name Size Sparse Medium Dense

Global Small 45 64 91

Medium 64 91 82

Large 82 91 82

Local Small 100 100 100

Medium 100 100 100

Large 100 100 100

Note: Accuracy in %.

Figure 2. Goldmann perimetry for G.B., showing small homonymous paracentral scotoma in the right lower quadrant.


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

individuals who viewed the scenes through a com-puter-generated aperture and allowed for theevaluation of this spatial restriction of vision asan analogue for simultanagnosia across tasks. It isworth noting that in a previous investigation ofeye movement behaviour in simultanagnosia(Dalrymple et al., 2011a), we tested brain-damaged control participants and found that it issimultanagnosia, and not parietal damage per se,that leads to reduced fixations on the eyes whendescribing scenes.

Full-view group. Participants (n ¼ 12, 5 male)were undergraduate students at the University ofBritish Columbia who ranged in age from 19 to28 years (mean ¼ 23 years). All participants inthis and the subsequent group reported normalor corrected-to-normal vision and gave informedconsent prior to participation in the experiments,which were performed in accordance with theethical guidelines of the University of BritishColumbia.

Gaze-contingent group. Participants (n ¼ 14, 6male) were undergraduate students at theUniversity of British Columbia who ranged inage from 19 to 56 years (mean ¼ 25.6 years).

Stimuli and apparatusFull colour images were taken with a digitalcamera in different rooms in the Psychology build-ing at the University of British Columbia. Imagesize was 36.5 × 27.5 (cm) corresponding to40.18 × 30.88 at the viewing distance of 50 cm,and image resolution was 800 × 600 pixels.Sixteen scenes were used in the present exper-iment. Each scene contained three persons. Allscenes were comparable in terms of their basiclayout: each room had a table, chairs, objects,and background items (e.g., see Figure 3a).

Eye movements were monitored using thedesk-mounted EyeLink 1000 eye trackingsystem (SR Research Ltd., www.eyelinkinfo.com). The Eyelink 1000 has a temporal resol-ution of 1 ms and a spatial resolution of 0.58. Itrecords data indicating the location of gaze inpixel coordinates. Before any analysis was

carried out, these data were parsed into fixationand saccade events (and blinks) using theEyeLink software. The event parser identifiesepochs in the data file where a saccade is occur-ring by calculating the distance between gazeposition in different samples and implementingmotion, velocity, and acceleration thresholds.The online saccade detector of the eye trackerwas set to detect saccades with an amplitude ofat least 0.158, using an acceleration threshold of8,0008 s2 and a velocity threshold of 308 s. A fix-ation is defined as any event that was not a

Figure 3. Example of scene stimuli (a) and regions of interest (b).



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

http://www.eyelinkinfo.com

http://www.eyelinkinfo.com

saccade or a blink. Two computers were used inthe experimental set-up and were connected toeach other via Ethernet, allowing for real-timetransfer of saccade and gaze position data. Theexperimenter’s computer collected the data fromthe eye tracker and displayed an image of the par-ticipant’s eye and calibration information. Thedisplay computer presented the stimuli andrecorded keypresses.

Subjects in the gaze-contingent group viewedthe scenes through a 28 × 28 square aperturecentred on their current fixation point. This sizewas chosen based on previous studies modellingsimultanagnosic behaviour with a gaze-contingentaperture (Dalrymple et al., 2011a; Dalrymple,Birmingham, Bischof, Barton, & Kingstone,2011b; Dalrymple et al., 2010). Outside of theportion of the scene visible through the aperture,the screen was white. As the subject’s fixationmoved, the aperture moved with the fixationpoint. Subjects in this group underwent a shorttraining session to familiarize them with thegaze-contingent window: They were instructedto start from a circle at the centre of the screenlabelled “Start” and to follow a line with thegaze-contingent window from that circle untilthey reached a second circle labelled “End”. Theywere then instructed to freely search the screenfor a hidden object on the screen. This task wasirrelevant to the experimental task and wasdesigned to teach them how to control the gaze-contingent aperture. Once they located thehidden object and felt comfortable with the appar-atus, the experiment began. Eye-monitoringstudies can be demanding for participants, in par-ticular those in the gaze-contingent condition. Sixparticipants were excluded because they failed tocomplete the study (four from the gaze-contingentcondition), and two because their fixation pro-portions fell more than 2 standard deviationsfrom the mean.

ProcedureSubjects were seated 50 cm from the screen of thedisplay computer with their chin supported by arest. Eye movements were recorded from the lefteye. The eye monitor was calibrated and validated

using a 9-dot array. Subjects were then asked tofixate a dot at centre-screen while the exper-imenter corrected for drift in gaze position. Oncethe dot was fixated, the experimenter initiatedthe onset of the scene image by a keypress.

There were two tasks (describe and infer) andtwo sets of eight scenes (A and B). Scenes were ran-domly assigned to Set A or B. In the describe task,subjects were asked to verbally describe each scene(i.e., “Please describe the scenes out loud”). In theinfer task, subjects were asked to verbally describewhere people in the picture were directing theirattention (i.e., “Please describe out loud where thepeople in the scenes are directing their attention”).Trials were blocked by task, and each task was per-formed with both scene sets, for a total of fourblocks (and a total of 32 trials). Subjects alternatedbetween tasks (e.g., describe Set A, infer Set B,describe Set B, infer Set A), with the assignmentof the first task counterbalanced across subjects.G.B. first performed the describe condition.Although the order of scene sets was fixed (i.e.,ABBA or BAAB), scenes were presented inrandom order within a block. All verbal responseswere given concurrently with scene viewing andwere recorded using a digital voice recorder.Subjects terminated their trials by keypress, initiat-ing the next trial. They had a maximum of 3 minutesto view each scene, but rarely reached this time limit.

AnalysisFor each image, an outline was drawn around eachregion of interest (e.g., “eyes”), and each region’spixel coordinates and area were recorded. Wedefined the following regions: eyes, head (excludingeyes), body (including arms, torso, and legs), fore-ground objects (e.g., tables, chairs, objects on thetable) and background (e.g., walls, shelves, itemson the walls). Figure 3b illustrates these regionsfor one scene. To compensate for the differentsizes of these regions, we computed area-normal-ized fixation proportions (Birmingham et al.,2008b; Smilek et al., 2006), by first dividing thenumber of fixations in each region by the area ofthe region, separately for each image and each par-ticipant, and then computing proportions based onthese normalized data.


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

Given our specific research question and a priorihypotheses, we focused our analysis on the eyesregion only. We used two-tailed paired t tests tocompare proportions of fixations on the eyes inthe describe task to those in the infer task separ-ately for each group (full-view, gaze-contingent,and G.B.). We then used Crawford, Garthwaite,and Howell (2009) modified t tests usingSINGLIMS software (Crawford et al., 2009;Sokal & Rohlf, 1995) to compare G.B. to thefull-view and gaze-contingent groups, respect-ively, in each task (describe and infer). All pvalues were compared to ≤.05.

Results

Full-view, gaze-contingent, and G.B.Figure 4 shows the mean proportions of fixationson the eyes region in the describe and infer con-ditions for each group. Two-tailed paired t testsrevealed that all groups made a significantlylarger proportion of fixations on the eyes in theinfer condition than in the describe condition:full-view, t(7) ¼ –4.37, p ¼ .003; gaze-contingent, t(8) ¼ –3.57, p ¼ .007; G.B., t(15)¼ –5.44, p , .001. Figure 5 shows representativescan patterns for each group and task.

Full-view versus gaze-contingentWe used a two-way analysis of variance (ANOVA)with factors of group (full-view vs. gaze-contin-gent) and task (infer vs. describe) to compare thefull-view group to the gaze-contingent group interms of their fixations on the eyes. There was amain effect of group, F(1, 35) ¼ 11.83, MSE ¼.173, p ¼ .003, and a main effect of task, F(1, 35)¼ 24.23, MSE ¼ .042, p , .001, but no Group ×Task interaction, F(1, 35) ¼ 1.97, MSE ¼ .003, p¼ .180. The effect of group was due to a larger pro-portion of fixations on the eyes in the full-view thanin the gaze-contingent group. The effect of task wasdue to both groups looking at the eyes more wheninferring the attention of people in the scenesthan when asked to describe the scenes. The lackof interaction indicates that fixations on the eyesfor both groups varied by task in the same way.

G.B. versus full-view groupWe used Crawford et al.’s (2009) modified t tests(Crawford et al., 2009; Sokal & Rohlf, 1995) tocompare G.B. to the full-view group. These testsrevealed that G.B. had a significantly smaller pro-portion of fixations on the eyes than the full-viewgroup in the describe condition, t(7) ¼ –2.62, p ¼.034, but that he did not differ from them in theinfer condition, t(7) ¼ –1.17, p ¼ .280. That is,when the task involved inferring the direction ofattention of the people in the scenes, G.B. wasjust as likely to look at the eyes as were the full-view group.

G.B. versus gaze-contingent groupCrawford et al.’s (2009) modified t test revealedthat G.B. did not differ significantly from thegaze-contingent group in either the describe, t(8)¼ –2.18, p ¼ .061, or the infer conditions, t(8) ¼0.02, p ¼ .984.

Verbal reportsVerbal reports from G.B. and participantsfrom the gaze-contingent and full-view groups(Table 2) were transcribed and inspected toensure that participants were capable of perform-ing the tasks and that they understood the

Figure 4. Fixation proportions on the eyes for each group for the

describe and infer tasks. Fixation proportions are normalized to

the size of each region. Error bars represent standard error.



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

difference between them. These reports clearlyindicate that when asked to describe a scene,participants were able to synthesize meaningfrom the elements that they perceived. They alsoindicate that participants understood the tasks:Their descriptions during the “infer attention”task show an understanding that attention can beinferred by looking at the eyes of the individualsin scene. These reports also corroborate the eyemovement measures: Participants look at anddiscuss the eyes more during the “infer attention”task than during the “describe the scenes” task.

Trial durationsOn average, the full-view group spent 25.96 s (SD¼ 11.81) describing scenes and 28.51 s (SD ¼14.01) inferring the attentional states of peoplein the scenes. The gaze-contingent group spenton average 53.37 s (SD ¼ 24.54) describing thescenes and 49.26 s (SD ¼ 10.62) inferring theattentional states of people in the scenes. G.B.spent on average 60.21 s describing the scenesand 78.68 s inferring the attentional states ofpeople in the scenes. The gaze-contingent grouphad significantly longer trial durations overall than

Figure 5. Representative scan paths for the full-view and gaze-contingent groups and for G.B. for both the describe and infer tasks. Circles

represent fixations. Lines represent movements from one fixation to the next. To view a colour version of this figure, please see the online issue of

the Journal.


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

the full-view group, F(1, 35) ¼ 13.00, p ¼ .002,but there was no significant effect of task, F(1,35) ¼ 0.04, p ¼ .836, or Task × Group inter-action, F(1, 35) ¼ 0.80, p ¼ .383, indicating thatboth groups had equal trial durations across tasks.Similar to the gaze-contingent group, G.B.’s trialdurations were significantly longer than those ofthe full-view group for both tasks: describe, t(8)¼ 2.75, p ¼ .025; infer, t(8) ¼ 3.40, p ¼ .009.However, while G.B. did not differ from thegaze-contingent group in terms of time spentdescribing scenes, t(8) ¼ 0.26, p ¼ .798, he didspend significantly longer inferring the attentionalstates of people in the scenes than this group, t(8)¼ 2.63, p ¼ .030. Verbal reports (Table 2) confirmthat all participants are capable of performing thetasks, so increased trial durations for G.B. and thegaze-contingent participants relative to the full-view group appear to reflect the increased timeneeded to gather information when restricted toseeing only a small portion of the scene at one time.

Discussion

Here we highlight three main aspects of ourresults. First, we replicate a previous findingfrom another simultanagnosic patient S.L. (e.g.,from Dalrymple et al., 2011a, 2011b): Whendescribing social scenes, patient G.B. makes fewfixations on the eyes of people in the scenes.This is important because it supports thesuggestion that this unusual behaviour is relatedto simultanagnosia, rather than being an anoma-lous finding from a single case. Second, whenengaged in a task that explicitly requires G.B. toinfer attention, a task that encourages eye fixationsin healthy subjects, he does fixate them, and with aproportion of fixations that is no different fromthat of healthy controls performing the sametask. Thus, like healthy subjects, G.B. shows an

understanding that the eyes are an importantsource of information for attentional states anduses this understanding to allocate his fixationsin a goal-directed way. G.B.’s understanding ofthe importance of the eyes for inferring attentionalstates is evident in his verbal reports (Table 2). Hisbehaviour indicates that he can locate and use theeyes to infer attention, but when simply describingscenes he does not give others’ attentional statespriority to the degree that healthy individuals do.The notion has been raised that, when asked todescribe social scenes, G.B. may be using parts ofthe scenes other than the eyes to gather infor-mation about the attentional states of people inthe scenes. However, G.B.’s tendency to increasefixations on the eye regions when explicitly askedto infer attentional states suggests that, likehealthy individuals, he does in fact gather infor-mation about attentional states by looking ateyes. Thus, his reduced fixations on the eyeswhen describing scenes suggests that he is priori-tizing information other than attentional statesduring this task. These findings also indicate thatG.B. can execute top-down control over his selec-tion of visual information.

The third main aspect of our results is thatG.B.’s behaviour is largely reproduced by healthysubjects viewing the scenes through a restrictedwindow of vision. Compared to subjects viewingscenes without any restriction, these subjects, likeG.B., showed fewer fixations on the eyes whendescribing the scenes. However, as with subjectsviewing without restriction, subjects in the gaze-contingent window condition showed an increasein fixations on eyes when inferring the directionof attention of people in scenes. This discoverylends new and converging support to a restrictedwindow of vision as a useful model for exploringsimultanagnosic behaviour with healthy individ-uals.2 Importantly, this speaks to the nature of theattentional deficit in simultanagnosia, supporting

2 The present finding is especially remarkable because the gaze-contingent group consisted of young controls (mean age 25.6

years), demonstrating that visual restriction, and not age, is the critical factor. We have previously shown that the fixation behaviour

of healthy individuals who were matched in age to the simultanagnosic patient S.L. (49 years) did not differ from the fixation

behaviour of a group of young controls (mean age 22 years), further suggesting that age is not critical to our results (Dalrymple

et al., 2011a).



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

Table 2. Verbal reports from G.B. and randomly selected participants from the full-view and gaze-contingent groups

Group Describe Infer

Full-view Three students, probably in the [psychology building], are drinking tea or

something, and they’re doing cheers. They look happy.

Ummm . . . this picture looks like the young man looking . . . looking at the

woman in the green shirt and the lady in the green shirt . . . hard to say . .

. looks like she’s looking . . . maybe at the—at the young man . . . and

then the la- the girl in the middle looks like she’s looking at the cups in

general.

Gaze-

contingent

K, there’s . . . uh (long pause) two women, one’s wearing a green shirt, it looks

like . . . um . . . smiling . . . (long pause) . . . holding a cup . . . uh . . . in a

room . . . with a green door, looks like (long pause). Um . . . (long pause).

There’s a man as well, I think they’re cheering or something, putting their

cups together. Uh . . . (long pause). There’s a lot of green. So it looks like

there’s three people in the room—that’s all I see, anyway.

So I see three cu-cups coming together so they look like they’re . . . banging

their cups together so that’s what their direction is atten- towards, all

three of them it seems. K . . . yeah . . . yeah, they’re all looking towards

the same spot.

G.B. (Long pause) There’s three people doing cheers, with um . . . coffee cups and

there’s um a coffee (pause) carafe on the table . . . and umm . . . (pause)

three people, there’s a woman in a red- a green shirt (pause) and another

woman (pause) and um . . . (pause) a man and (pause) a green door (pause)

to the room that’s propped open (pause) and there’s stuff written on a

whiteboard, behind um . . . (pause)(inaudible) . . . I can’t see what it is . . .

(long pause).

(Long pause) OK so I, I . . . I see who’s directing their attention at (pause) at

. . . what or whom and I see (inaudible) in fact (pause) um . . . (pause) it

looks to me like all three people are, are directing their attention at the . .

. at their cups, but, we- they’re doing cheers and they’re all looking at

their . . . at their cups (pause) and um . . . (pause) that’s because I’m

looking at their eyes (pause) and they don’t look like they are angled in a

way so they could be looking at making um eye contact with . . . any of

the other people . . . (pause) and I also think it’s just normal when you do

cheers you . . . watch the cups so you don’t spill anything or . . . knock it

too hard.

Note: Reports were recorded during the trials depicted in Figure 5.

36

Cogn

itive

Neu

ropsychology,

20

13

,3

0(1

)

DA

LR

YM

PL

EE

TA

L.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

space-based, as opposed to object-based, account ofthe disorder.

The data from our full-view control subjects fitnicely with previous findings by Birmingham et al.(2008a) who found that task modulated the fix-ation behaviour of healthy subjects in a similarway to what we found in the present study. Thisreplication of the Birmingham et al. results is par-ticularly interesting because it occurred despitesome key methodological differences betweenour study and theirs. Birmingham et al. allowedsubjects to view the scenes for only 15 seconds,at which point they were asked to describe whatthey saw or infer the attention of people in thescenes (i.e., when the scene was no longervisible). In contrast, subjects in our study viewedscenes for up to 3 minutes and were asked todescribe the scenes or infer attention while theyviewed the scene. This means that subjects allocatetheir fixations in a similar way for both relativelyshort and prolonged scene viewings, and whetherthey provided immediate or delayed report. Thisattests to the robustness of the tendency for sub-jects with normal vision to look at eyes in socialscenes and supports the idea that the findings inthis paradigm are capturing a behavioural profilethat generalizes across situational complexities.

Collectively, our findings suggest that althoughpatient G.B. is capable of looking at the eyes whennecessary to the task (i.e., during the infer task), hetends not to look at the eyes when it is not expli-citly required of him (i.e., during the describetask). This supports the hypothesis that the atten-tional states of others are not a high priority forsimultanagnosics when they are describing ascene. The present data, combined with previousand current findings from modelling simultanag-nosia in healthy individuals, can offer insightinto why these patients may reprioritize theimportance of the eyes. We previously suggestedthat some aspects of simultanagnosia may be mod-elled as a restriction of the spatial window of atten-tion, a type of “attentional tunnel vision”. Thisnotion dovetails with initial reports documentinga reduction of the useful visual field in simultanag-nosia (Bay, 1953; Thaiss & de Bleser, 1992; Tyler,1968), as well as more recent findings suggesting a

“shrinkage” of the attentional visual field in simul-tanagnosia despite normal visual fields (Michel &Henaff, 2004), and the enlargement of an abnor-mally small spatial area of attention in simultanag-nosia through the use of large preceding visualprimes (Shalev et al., 2004). Our prior work hasshown that healthy subjects viewing scenesthrough a restricted window of vision behavemuch like simultanagnosic patients in processinghierarchical letters (Dalrymple et al., 2010) andwhen describing social scenes (Dalrymple et al.,2011a, 2011b). Here we demonstrate that theparallels between limiting the spatial extent ofthe visual window in healthy subjects and theattentional restriction in simultanagnosia extendto the task of inferring the attention of people insocial scenes.

In line with results from a previous study(Dalrymple et al., 2011a), the describe task datafrom G.B. and from our gaze-contingent groupsuggest that when only one element of a scenecan be processed at a time, the most efficient wayto describe the scene is to scan more elements ofthe scene rather than revisiting a single regionsuch as the eyes. This contrasts with healthy sub-jects with unrestricted vision, who can allocatethe majority of their fixations to the eyes ofpeople and still describe the rest of the sceneusing gist, or vision at a glance from informationin their parafoveal vision (Hochstein & Ahissar,2002). However, when the task shifts to inferringattention, all groups realize that the primary sourceof information lies in the eyes, which can still belocated and fixated despite restrictions in thespan of either vision or attention. Critically, thisfinding indicates that, like the non-brain-damaged participants who viewed scenes througha restricted window of vision, G.B. maintainssome top-down control over his fixations.

This conclusion is consistent with previouswork that has shown top-down processing strat-egies in another individual (patient J.J.) withsimultanagnosia (Jackson, Swainson, Mort,Husain, & Jackson, 2009). For example, in apop-out search task, where a target with uniquefeatures is presented among distractors that donot contain those features, J.J. used a systematic,



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

item-by-item search strategy to find the target.Our findings, and those of Jackson et al. (2009),provide convergent evidence against previous sug-gestions that the visual experience of simultanag-nosics is largely controlled by bottom-up factorssuch as object saliency (Holmes & Horrax, 1919;Rizzo & Hurtig, 1987).

Data from the gaze-contingent group showthat restricting the useful visual field of healthyindividuals leads to similar fixation patterns tothose that are recorded with simultanagnosicpatients who can only see one thing at a time,whether describing social scenes or inferring thedirection of attention of people in social scenes.However, it is important to be clear about thelimitations of the restricted window of vision as amodel of simultanagnosia. We do not claim thata restricted window of vision models all aspect ofthe simultanagnosic deficit. Rather, we are usingthe restricted window of vision to reduce thespatial area to which healthy subjects can attend,akin to the reduction of the visual windowwithin which attentional selection can operate insimultanagnosia. In fact, one strength of thismethod is that this simple manipulation can leadto complex scanning behaviours in healthy adultsthat are similar to those seen in simultanagnosiaacross a variety of tasks (e.g., global/local proces-sing, Dalrymple, et al., 2010; recovery from simul-tanagnosia, Dalrymple et al., 2011b; describingscenes, Dalrymple et al., 2011a; and inferringattention of people in scenes as reported here).

It is important to note that although G.B.and the gaze-contingent group fixate the eyesto a similar degree when describing the scenes,there was a trend for these groups to differfrom each other in this task, with the gaze-contingent group fixating the eyes slightly morethan G.B. However, we have previously shownthat the degree of fit between simultanagnosicbehaviour and the gaze-contingent model isinfluenced by the size of the window used inthe gaze-contingent paradigm (Dalrympleet al., 2011b)—a smaller window tends to leadto fewer fixations on the eyes and might havebeen a better fit for patient G.B.’s behaviour.Ultimately, it is not the exact proportion of

fixations that is of critical importance here.Rather, it is the tendency for both G.B. andthe gaze-contingent group to show abnormallylow fixation proportions on the eyes whendescribing scenes and to increase that proportionof fixations when inferring attention that makesthis a useful model of social scene viewing insimultanagnosia.

One question that may arise from this and pre-vious studies of complex scene viewing in simulta-nagnosia is to what degree simultanagnosics areable to describe and understand the scenes thatthey are viewing. Verbal descriptions from G.B.(Table 2) and from previous scene-viewing exper-iments with simultanagnosia patient S.L.(Dalrymple et al., 2011a, 2011b) indicate thatpatients with simultanagnosia are capable ofpiecing together elements of a scene to describethe scene as a whole. This process is often slowand laborious, and not always perfectly accurate,but may result in a surprising degree of coherence.Rizzo and Vecera (2002) articulate how it is poss-ible for an individual with simultanagnosia toreport multiple elements of a scene in a seeminglycoherent fashion: “ . . . object recognition canproceed economically from just a few key features,and identifying a person or object is not the sameas seeing him, her, or it all at once” (p. 164).Patients with simultanagnosia suffer from avisual disorder that may (and often does) leavetheir cognitive faculties intact (e.g., see G.B.’scase report). Thus patients are able to compensatefor their visual disorder by consciously synthesiz-ing meaning from what they see.

We would also like to acknowledge that ananonymous reviewer flagged for our attention thefact that these data give rise to a separate but inter-esting question for future investigation concerningwhether G.B.’s top-down attention to the eyeswhen inferring attentional states interferes withhis ability to derive meaning from the scene.Jackson et al. (2009) suggest that goal-directed be-haviour in simultanagnosia comes at the expense ofcompeting stimuli that fall outside of the immedi-ate task demands. Applied here, this suggests thatwhen explicitly asked to determine the attentionalstates of people in the scenes, G.B. may have


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

sacrificed some understanding of the scene as awhole in order to perform the task. However, itis also possible that a comparable loss may haveoccurred for G.B. during the “describe thescenes” task, and that this loss may not be specificto simultanagnosia: It is well known in attentionresearch that selection for some items comes at acost of reduced processing for others (Broadbent& Broadbent, 1987). Indeed, the content ofG.B.’s verbal reports from the infer attention anddescribe the scenes tasks were convergent withthose provided by the healthy participants inboth the full-view and the gaze-contingent con-ditions. Future investigations using different taskinstructions could be used to determine to whatextent goal-oriented behaviour impedes additionalinformation processing in simultanagnosia.

Importantly for our primary research question,our results indicate that patients strategize theirscene exploration based on task demands. Whenasked to describe a scene, patients use goal-directedbehaviour to distribute their fixations across thescene to gather information about the scene.When asked to infer the attentional states ofpeople in the scenes, patients, like healthy controls,increase fixations on the eyes. These strategies werealso employed by healthy individuals who viewedthe scenes though a restricted window of vision,further indicating that it is likely to be the restrictedwindow of visual information, not brain damage perse, that is driving this behaviour.

Seeing one object at a time, as in simultanagno-sia and in gaze-contingent viewing, drasticallyaffects how one visually explores the world.What our results suggest is that what could pre-viously have been interpreted as a lack of controlover exploratory eye movements (leading to anavoidance of the eyes of people in scenes whendescribing the scenes) can now be understood asa top-down effort to derive information from ascene. Ultimately we can conclude that simulta-nagnosic eye movements are task dependent.How they vary with the task is informative aboutthe overall visual experience of patients withthis profound disorder and about the directionof attention in situations with limited visualinformation.

Supplementary material

Supplementary Material is available via the“Supplementary” tab on the article’s online page(http://dx.doi.org/10.1080/02643294.2013.778234).

Manuscript received 4 September 2012

Revised manuscript received 13 February 2013

Revised manuscript accepted 14 February 2013

First published online 27 March 2013

REFERENCES

Balint, R. (1909). Seelenlahmung des “Schauens”, optischeAtaxia, raumliche Storung der Aufmerksamkeit[Psychic paralysis of gaze, optic ataxia, and spatial dis-order of attention]. Monatsschrift fur Psychiatrie und

Neurologie, 25, 51–81.Bay, E. (1953). Disturbances of visual perception and

their examination. Brain, 76, 515–550.Baylis, G., Driver, J., Baylis, L., & Rafal, R. (1994).

Reading of letters and words in a patient withBalint’s syndrome. Neuropsychologia, 32, 1273–1286.

Birmingham, E., Bischof, W., & Kingstone, A. (2007).Why do we look at people’s eyes? Journal of Eye

Movement Research, 1, 1–6.Birmingham, E., Bischof, W., & Kingstone, A. (2008a).

Gaze selection in complex social scenes. Visual

Cognition, 16, 341–355.Birmingham, E., Bischof, W., & Kingstone, A. (2008b).

Social attention and real world scenes: The rolesof action, competition and social content. Quarterly

Journal of Experimental Psychology, 61, 986–998.Birmingham, E., Bischof, W., & Kingstone, A. (2009).

Saliency does not account for fixations to eyes withinsocial scenes. Vision Research, 49, 2992–3000.

Birmingham, E., & Kingstone, A. (2009). Human socialattention: A new look at past, present, and futureinvestigations. Annual New York Academy of

Sciences, 1156, 118–140.Broadbent, D. E., & Broadbent, M. H. P. (1987). From

detection to identification: Response to multipletargets in rapid serial visual presentation. Perception

& Psychophysics, 42(2), 105–113.Caldara, R., Zhou, X., & Miellet, S. (2010). Putting

culture under the “spotlight” reveals universal infor-mation use for face recognition. PLoS One, 5(3), e9708.

Clavagnier, S., Fruhmann Berger, M., Klockgether, T.,Moskau, S., & Karnath, H.-O. (2006). Restricted



Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

http://dx.doi.org/10.1080/02643294.2013.778234

ocular exploration does not seem to explain simulta-nagnosia. Neuropsychologia, 44, 2330–2336.

Crawford, J., Garthwaite, P., & Howell, D. (2009). Oncomparing a single case with a control sample: An alter-native perspective. Neuropsychologia, 47, 2690–2695.

Dalrymple, K. A., Birmingham, E., Bischof, W.,Barton, J., & Kingstone, A. (2011a). Experiencingsimultanagnosia through windowed viewing ofsocial scenes. Brain Research, 1367, 265–277.

Dalrymple, K. A., Birmingham, E., Bischof, W.,Barton, J., & Kingstone, A. (2011b). Opening awindow on attention: Documenting and simulatingrecovery from simultanagnosia. Cortex, 47, 787–799.

Dalrymple, K. A., Bischof, W., Cameron, D., Barton, J.,& Kingstone, A. (2009). Global perception insimultanagnosia is not as simple as a game ofconnect-the-dots. Vision Research, 49, 1901–1908.

Dalrymple, K. A., Bischof, W., Cameron, D., Barton, J.,& Kingstone, A. (2010). Simulating simultanagno-sia: Spatially constricted vision mimics local captureand the global processing deficit. Experimental

Brain Research, 202, 445–455.Dalrymple, K. A., Kingstone, A., & Barton, J. (2007).

Seeing trees OR seeing forests in simultanagnosia:Attentional capture can be local or global.Neuropsychologia, 45, 871–875.

Damasio, A. (1985). Disorders of complex visual proces-sing: Agnosias, achromatopsia, Balint’s syndrome,and related difficulties of orientation and construc-tion. In M.-M. Mesulam (Ed.), Principles of

behavioral neurology (pp. 259–288). Philadelphia,PA: F.A. Davis Company.

Goodglass, H., & Kaplan, E. (1983). The assessment of

aphasia and related disorders. Philadephia, PA: Leaand Febiger.

Hochstein, S., & Ahissar, M. (2002). View from thetop: Hierarchies and reverse hierarchies in thevisual system. Neuron, 36, 791–804.

Holmes, G., & Horrax, G. (1919). Disturbances ofspatial orientation and visual attention, with loss ofstereoscopic vision. Archives of Neurology and

Psychiatry, 1, 385–407.Jackson, G., Swainson, R., Mort, D., Husain, M., &

Jackson, S. (2009). Attention, competition, and theparietal lobes: Insights from Balint syndrome.Psychological Research, 73(2), 263–270.

Karnath, H.-O., Ferber, S., Rorden, C., & Driver, J.(2000). The fate of global information in dorsalsimultanagnosia. Neurocase, 6, 295–306.

Loschky, L., McConkie, G., Yang, J., & Miller, M.(2005). The limits of visual resolution in

natural scene viewing. Visual Cognition, 12(6),1057–1092.

McConkie, G., & Rayner, K. (1975). The span of theeffective stimulus during a fixation in reading.Perception and Psychophysics, 17, 578–586.

Michel, F., & Henaff, M.-A. (2004). Seeing withoutthe occipito-parietal cortex: Simultanagnosia as ashrinkage of the attentional visual field. Behavioural

Neurology, 15, 3–13.Moreaud, O. (2003). Balint syndrome. Archives of

Neurology, 60, 1329–1331.Navon, D. (1977). Forest before trees: The precedence

of global features in visual perception. Cognitive

Psychology, 9, 353–383.Pomplun, M., Reingold, E., & Shen, J. (2001).

Investigating the visual span in comparative search:The effects of task difficulty and divided attention.Cognition, 81, B57–B67.

Rafal, R. (1997). Balint syndrome. In T. Feinberg &M. Farah (Eds.), Behavioral neurology and neuropsy-

chology (pp. 337–356). New York, NY: McGraw-Hill.Rafal, R. (2003). Balint’s syndrome: A disorder of

visual cognition. In M. D’Esposito (Ed.), Neurological

foundations of cognitive neuroscience (pp. 27–40).Cambridge, MA: MIT Press.

Riddoch, M., Chechlacz, M., Mevorach, C., Mavritsaki,E., Allen, H., & Humphreys, G. (2010). The neuralmechanisms of visual selection: The view from neu-ropsychology. Annals of the New York Academy of

Sciences, 1191, 156–181.Rizzo, M., & Hurtig, R. (1987). Looking but not

seeing: Attention, perception, and eye movementsin simultanagnosia. Neurology, 37(10), 1642–1648.

Rizzo, M., & Vecera, S. (2002). Psychoanatomical sub-strates of Balint’s syndrome. Journal of Neurology,

Neurosurgery & Psychiatry, 72, 162–178.Shalev, L., Humphreys, G., & Mevorach, C. (2004).

Global processing of compound letters in a patientwith Balint’s syndrome. Cognitive Neuropsychology,22, 737–751.

Smilek, D., Birmingham, E., Cameron, D., Bischof,W., & Kingstone, A. (2006). Cognitive ethologyand exploring attention in real-world scenes. Brain

Research, 1080, 101–119.Sokal, R., & Rohlf, J. (1995). Biometry. San Francisco,

CA: W.H. Freeman.Thaiss, L., & de Bleser, R. (1992). Visual agnosia: A case of

reduced attentional “spotlight”? Cortex, 28, 601–621.Tyler, R. (1968). Abnormalities of perception with

defective eye movements (Balint’s syndrome).Cortex, 4, 154–171.


DALRYMPLE ET AL.

Dow

nloa

ded

by [

Uni

vers

ity C

olle

ge L

ondo

n] a

t 07:

58 0

3 Se

ptem

ber

2013

eyeing the eyes in social scenes: evidence for top … et...eyeing the eyes in social scenes:...

Documents