eps mid-career award 2014brainb.psyc.bbk.ac.uk/wp-content/uploads/2019/05/eimer...eps mid-career...

EPS Mid-Career Award 2014The control of attention in visual search: Cognitive and

neural mechanisms

Martin Eimer

Department of Psychological Sciences, Birkbeck College, University of London, London, UK

(Received 24 March 2015; accepted 30 May 2015; first published online 28 July 2015)

In visual search, observers try to find known target objects among distractors in visual scenes where thelocation of the targets is uncertain. This review article discusses the attentional processes that are activeduring search and their neural basis. Four successive phases of visual search are described. During theinitial preparatory phase, a representation of the current search goal is activated. Once visual input hasarrived, information about the presence of target-matching features is accumulated in parallel acrossthe visual field (guidance). This information is then used to allocate spatial attention to particularobjects (selection), before representations of selected objects are activated in visual workingmemory (rec-ognition). These four phases of attentional control in visual search are characterized both at the cognitivelevel and at the neural implementation level. It will become clear that search is a continuous process thatunfolds in real time. Selective attention in visual search is described as the gradual emergence of spatiallyspecific and temporally sustained biases for representations of task-relevant visual objects in corticalmaps.

Keywords: Attention; Visual system; Cognitive control; Electrophysiology; Neuroimaging.

The visual world is rich and complex. Many visualobjects and events are simultaneously received bythe visual system, but only some of these are

linked to current intentions and action goals. Tofacilitate adaptive behaviour, these task-relevantobjects need to be preferentially processed so

Correspondence should be addressed to Martin Eimer, Department of Psychological Sciences, Birkbeck College, University of

London, Malet Street, London WC1E 7HX, UK. E-mail: [email protected]

I would like to thank the members of my research group, and in particular Anna Grubert, for many helpful comments on previous

versions of this article.

This work was supported by the Economic and Social Research Council (ESRC), UK [grant number ES/L016400/1].

© 2015 The Experimental Psychology Society 2437

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015

Vol. 68, No. 12, 2437–2463, http://dx.doi.org/10.1080/17470218.2015.1065283

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

mailto:[email protected]

that they can be rapidly detected and recognized.Such goal-dependent modulations of visual pro-cesses and their effects on perception and actionare described as “selective attention”. Attentionalmechanisms affect the perception and recognitionof visual objects in different task contexts and havebeen investigated with various experimental pro-cedures. Many visual attention experiments haveemployed spatial cues that inform observersabout the likely locations of upcoming targetobjects (e.g., Posner, Snyder, & Davidson,1980). Such advance spatial information can beused to move focal spatial attention to particularvisual field locations in anticipation of task-rel-evant objects at these locations, which facilitatesthe detection and recognition of these objects.However, there are many other situations wherethe selective processing of visual input is required,but no precise advance spatial information aboutthe location of task-relevant objects is available.Trying to find the missing car keys in a clutteredoffice room is a challenge when we cannot remem-ber where we left them a few minutes before. Ineveryday life, there are many such instanceswhere visual search is required to find a knowntarget object at an unknown location. Althoughsearch in the real world will sometimes benefitfrom contextual information about the likelylocation of particular objects (e.g., alarm clocksare frequently found on bedside tables; seeHenderson, 2003), many lab-based visual searchexperiments require participants to find specifictarget objects or features at random and thusentirely unpredictable locations (e.g., Treisman& Gelade, 1980). Under certain conditions, thisis very easy: When a target object has a uniquevisual feature (e.g., when the target is the onlyred object among green distractors), it will “popout” from its surroundings and can be detectedrapidly (e.g., Müller, Heller, & Ziegler, 1995;Treisman, 1988). In many other situations, asearch target cannot be found on the basis of itsperceptual salience alone, and search becomesharder. It is generally believed that the successfuldetection and recognition of target objects insuch situations depends on selective attention.But what role does attention play during visual

search, and how do such attentional processesoperate? This article discusses the cognitive andneural mechanisms that are responsible for ourability to find known target objects at uncertainlocations. Some of the ideas developed here havebeen previously summarized in a brief reviewarticle (Eimer, 2014).

Several models of attentional processing such asFeature Integration Theory (e.g., Treisman, 1988)and guided search (Wolfe, 1994, 2007) have beendeveloped specifically to explain behavioural per-formance in various visual search tasks. Moregeneral theories of selective attention such as thebiased competition account (e.g., Desimone &Duncan, 1995) and the neural theory of visualattention (Bundesen, Habekost, & Kyllingsbaek,2005) also have direct implications for visualsearch. Some of the ideas put forward by thesemodels are considered below. To understand thecognitive and neural mechanisms involved invisual search, it is important to keep in mind thatsearch is a process that unfolds in real time. Evenan informal description of this process can readilydistinguish different stages that operate sequen-tially when an observer searches for a knowntarget object at an unknown location. First, theobserver has to form an intention to find a specificobject and activate some form of mental represen-tation of this object. Once visual input hasarrived, possible target objects have to be localizedamong other irrelevant objects in the visual field.Next, attention can be selectively focused on oneor more of these objects. Finally, particularobjects are recognized as targets or nontargets. Inthis review article, this informal description offour component processes involved in visualsearch is used to characterize these search processesat the cognitive level and to discuss how these cog-nitive operations are implemented neurally. Thefour-stage structure of visual search proposed hereis illustrated in Figure 1. Preparation, guidance,selection, and recognition are distinguished as sep-arate phases of visual search that jointly contributeto the detection and recognition of particularsearch targets. Each of these four stages is describedat the cognitive level in terms of their functionalroles and at the neural level with respect to the

2438 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015, 68 (12)

EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

brain processes that implement these particularfunctions.

In the subsequent sections of this article, each ofthese four phases of visual search are considered inturn. During the preparation phase, specific searchgoals are activated in visual working memory(“Preparatory attentional templates and visualworking memory”). Guidance refers to theaccumulation of information about goal-relevantfeatures during the initial parallel processing ofvisual input (“Attentional guidance and feature-based attention”). Selection operates through theallocation of focal attention to possible targetobjects at particular locations (“Object selectionand focal spatial attention”). Finally, the recog-nition of selected objects takes place once theseobjects are encoded into visual working memory(“Object recognition and working memory”).This four-phase processing model is illustrated inFigure 1 with a box-and-arrow diagram. Such dia-grams are frequently used in cognitive psychologyto define temporally and functionally discretestages of information processing. In visual search,such a stage-based model can be heuristicallyuseful to distinguish different aspects of thesearch process. However, this does not imply thatthe underlying cognitive and neural mechanismsoperate in a strictly modular and discrete fashion.When considering the four phases of visualsearch and their interactions, it will become clearthat search is a continuous process where specificaspects of attentional selectivity emerge graduallyin real time.

PREPARATORY ATTENTIONALTEMPLATES AND VISUALWORKING MEMORY

Before searching for a particular target objectamong distractors, observers first have to decidewhich object or object feature to look for. Next,they have to form a mental representation of thesearch target, which can be activated in a prepara-tory fashion before a search display is presented,and visual input is processed. The central role ofsuch mental representations for the goal-directedallocation of attention was highlighted by James(1890/1981), who referred to “the anticipatorypreparation from within of the ideational centresconcerned with the objects to which attention ispaid” (p. 411). According to James, “… theimage in the mind is the attention; the prepercep-tion… is half of the perception of the looked-forthing” (p. 419). In current models of visual atten-tion and visual search, James’s “images in themind” are described as “attentional templates”(Duncan & Humphreys, 1989; Olivers, Peters,Houtkamp, & Roelfsema, 2011). Such search tem-plates are representations of task-relevant objects orfeatures in visual working memory that are acti-vated while observers prepare for a search task.Attentional templates are set up prior to the presen-tation of search displays and then help to directfocal attention towards the location of candidatetargets in these displays.

How could such preparatory attentional tem-plates be implemented at the neural level?

Figure 1. The four phases of attentional control during visual search discussed in this article. Preparation, guidance, selection, and recognition

refer to different functions of selective attention that emerge at specific points in time during a search process. These functions are described at the

cognitive level (white boxes) and in terms of their implementation at the neural level (dark grey boxes).

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015, 68 (12) 2439

ATTENTIONAL CONTROL IN VISUAL SEARCH

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

Because search templates are assumed to be rep-resentations in visual working memory, an answerto this question needs to take into accountcurrent views about the neural basis of workingmemory. It is often assumed that lateral prefrontalcortex plays a critical role in the storage and main-tenance of visual information. In line with thishypothesis, neurons in monkey prefrontal cortexshow sustained delay activity during the retentionperiod of working memory tasks (e.g., Fuster &Alexander, 1971). However, more recent findingshave cast doubt on the hypothesis that prefrontalareas are the primary locus for working memorystorage. Human neuroimaging studies have foundmemory-related delay activity in brain regionsoutside prefrontal cortex, and in particular inhigher level visual areas in inferior temporalcortex (e.g., Ranganath, Cohen, Dam, &D’Esposito, 2004), suggesting that visual–percep-tual cortical regions are also involved in the activeshort-term maintenance of visual information.This emerging “sensory recruitment” model ofvisual working memory (Awh & Jonides, 2001;D’Esposito, 2007; Harrison & Tong, 2009;Postle, 2006; Sreenivasan, Curtis, & D’Esposito,2014) proposes that posterior visual brain areasthat are activated during the perception of visualstimuli are also the primary locus for the temporarymaintenance of these stimuli in working memory.According to this model, prefrontal areas havemore generic top-down control functions, such asregulating access to working memory and main-taining memory representations in an active statethrough the allocation of focal attention.

If preparatory attentional templates for particu-lar search targets are representations in visualworking memory, the sensory recruitment modelof working memory predicts that these templatesshould be implemented by sustained target-specificactivation patterns in visual cortex that are similarto the patterns observed when the same target isperceptually processed. This prediction has beeninvestigated in single-unit recording experimentswith monkeys and in human functional neuroima-ging studies. In these experiments, neural activitywas recorded prior to the presentation of search dis-plays while observers prepared for a particular visual

search task after being instructed to find a specifictarget stimulus. The first evidence for a neural cor-relate of preparatory attentional templates wasfound in a study by Chelazzi, Duncan, Miller,and Desimone (1998). Monkeys were shown apicture of a specific target object, which was thenfollowed after a delay period by a search displaythat could contain the target object and an irrele-vant distractor object or two task-irrelevantobjects. The target object had to be retained inworking memory during the delay period becausethe monkeys had to make an eye movementtowards this object when it appeared in the sub-sequent search display. Neurons in inferotemporalcortex that were selectively activated by the targetobject during its initial presentation maintainedthis activation in a sustained fashion prior to thepresentation of the search display, suggesting thata representation of the search target was keptactive during the delay period. Further evidencefor such preparatory “baseline shifts” of neuralactivity in visual cortex was obtained in human neu-roimaging studies. In these studies, where observersprepared to find search targets that were defined bya particular colour or motion, activity in colour- ormotion-selective visual brain areas increased duringthe preparation period prior to the presentation ofvisual input (e.g., Chawla, Rees, & Friston, 1999;Giesbrecht, Weissman, Woldorff, & Mangun,2006). More recent functional magnetic resonanceimaging (fMRI) studies analysed spatially distribu-ted brain activation patterns (multivoxel patternanalysis, MVPA) to identify preparatory atten-tional templates in visual cortex. Preparation for aspecific target shape elicited a shape-selectivepattern of neural activation in lateral occipitalcortex during the interval before the target was pre-sented (Stokes, Thompson, Nobre, & Duncan,2009). Even search for category-defined targetobjects in natural visual scenes (e.g., people orcars) was found to be preceded by preparatory cat-egory-selective activation patterns in visual cortex(Peelen & Kastner, 2011).

These observations suggest that attentional tem-plates are implemented by sustained preparatorychanges in the activation pattern of visual areasthat are selective to the target feature or object in


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

an upcoming visual search task and are structurallysimilar to the activation that is elicited when thesame feature or object is visually perceived. Thisscenario is not only consistent with the sensoryrecruitment model of working memory, but mayalso be seen as a twenty-first century neuroscientificconfirmation of William James’s nineteenth-century suggestion that attentional preparation isbased on “images in the mind”. Unfortunately,there is a problem with this parsimonious andintuitively appealing identification of search tem-plates in visual working memory with preparatorybaseline shifts of neural activation patterns invisual cortex. To demonstrate that target-selectivemodulations of neural processing during the prep-aration for an attentional selection task are thephysiological counterpart of search templates, it isnecessary to show that such preparatory baselineshifts result in larger sensory responses to targetobjects once a search display has been presented,and ultimately in the successful detection andidentification of these objects (e.g., Driver &Frith, 2000). However, there is so far very limitedevidence for direct links between baseline shiftsand the subsequent selective attentional processingof visual input. Some experiments have found posi-tive correlations between the strength of prepara-tory target-selective activations in visual cortexand target detection performance (Giesbrechtet al., 2006; Peelen & Kastner, 2011; Stokeset al., 2009). However, other studies have failedto observe any systematic relationships betweenanticipatory baseline shifts and subsequent target-selective modulations of visual processing or behav-ioural target selection efficiency (e.g., Fannon,Saron, & Mangun, 2007; McMains, Fehd,Emmanouil, & Kastner, 2007). The difficulty infinding strong causal links between preparatoryactivity in visual–perceptual brain areas and theselective attentional processing of targets versus dis-tractors raises doubts about the idea that these base-line shifts are the direct neural counterpart of searchtemplates. Even if such baseline shifts are linked tothe maintenance of target-related information inworking memory, they may reflect a type ofworking memory that is not suited to the functionsof attentional templates.

To understand why this might be the case, it isimportant to consider how information about visualobjects is represented in working memory. Suchrepresentations can be either position dependentor position invariant (spatially global). In pos-ition-dependent representations, the spatial layoutof visual information that is encountered duringencoding is retained. In contrast, position-invariantrepresentations contain no explicit informationabout particular object locations in the visual field.This distinction is important when consideringthe role of working memory representations asattentional templates in visual search tasks wherethe position of a target object among distractorsin the visual field is uncertain. If the function ofattentional templates is to affect the subsequentallocation of attention in a goal-selective fashion,these templates need to operate in a spatiallyglobal fashion across all possible target locationsin the visual field. In other words, attentional tem-plates should be position invariant. If workingmemory representations that are reflected by pre-paratory baseline shifts in posterior visual areaswere strongly position dependent, this would beinconsistent with their role as attentional templatesduring search for known targets at unknownlocations.

There are two reasons why working memoryrepresentations in visual cortex should be positiondependent. On the one hand, this predictionfollows directly from the sensory recruitmentaccount of working memory, which assumes thatmemorized objects are maintained in posteriorvisual regions that are also responsible for the per-ceptual analysis of incoming visual signals. Invisual cortex, information is represented in a pos-ition-dependent fashion in two-dimensional corti-cal maps (e.g., Franconeri, Alvarez, & Cavanagh,2013). In these maps, the spatial coordinates ofvisual features and objects are defined relative totheir position on the retina (retinotopic represen-tation) or in the external world (spatiotopic rep-resentation), and even higher level visual areasretain strong retinotopic biases (e.g., Desimone &Gross, 1979; Kravitz, Kriegeskorte, & Baker,2010; Op De Beeck & Vogels, 2000; see Kravitz,Saleem, Baker, Ungerleider, & Mishkin, 2013,



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

for a review). A second reason for the positiondependence of working memory representations,which is discussed in more detail in the sectionon “Object recognition and working memory” isthat memory maintenance is mediated by focalspatial attention, which necessarily operates onspace-based representations of visual objects.

Perhaps the most direct evidence for position-dependent memory representations in visualcortex comes from event-related potential (ERP)studies that showed that neural activity during thedelay period of working memory tasks is elicitedat posterior electrodes contralateral to the sidewhere the to-be-remembered objects appearedduring encoding (contralateral delay activity,CDA; see Vogel & Machizawa, 2004). In fact,there is a direct somatosensory analogue of thevisual CDA (tactile CDA component; see Katus,Grubert, & Eimer, 2014) that shows a distinctmodality-specific topography over lateral somato-sensory cortex. Such ERP results demonstratethat the spatial layout of sensory information isretained when this information is stored and main-tained in working memory (see also Gratton,Corballis, & Jain, 1997; Hornak, Duncan, &Gaffan, 2002, for additional behavioural evidencefor the position dependence of visual memory).At the neural level, such position-dependentvisual working memory representations should beimplemented by object-selective sustained activitymodulations at particular locations within visualcortical maps. If target-selective baseline shifts ofvisual activity that have been observed during thepreparation for visual search show this kind oflocation specificity, they would not be able tomodulate subsequent visual processing in a spatiallyglobal fashion. This may be the primary reason whyit has proved difficult to demonstrate causal linksbetween preparatory activity modulations in visualcortex and subsequent stages of attentionalprocessing.

Instead of being position dependent, attentionaltemplates in visual search need to represent searchtargets in a spatially global position-independentfashion. It is possible that such spatially globalsearch templates may not be found at all invisual–perceptual areas, but only in higher level

attentional control regions such as prefrontalcortex, where visual information is represented ina largely position-independent fashion. In fact, pat-terns of neural activity in prefrontal cortex that aresensitive to current search targets have indeed beenobserved during the preparation for attentionaltasks (Peelen & Kastner, 2011; Stokes et al.,2013; Warden & Miller, 2010). However, itmight be premature to completely rule out visualcortex as a possible additional locus for position-independent preparatory attentional templates. Ina study by Ester, Serences, and Awh (2009), par-ticipants memorized the orientation of a gratingin the left or right visual field during a delayperiod before matching it to a test grating.Pattern analyses of fMRI data obtained duringthe delay period found that activity in contralateralprimary visual cortex at locations that matched thememorized grating was sensitive to its orientation,as would be expected if working memory was pos-ition dependent. Critically, Ester et al. (2009)found that corresponding areas of ipsilateralprimary visual cortex were equally sensitive to thememorized orientation. This indicates that orien-tation information was maintained in a spatiallyglobal fashion and suggests that position-indepen-dent working memory representations may alsoexist in visual cortex. If this is the case, such rep-resentations could act as spatially global attentionaltemplates in visual search.

The hypothesis that visual working memoryrepresentations can be either position dependentor position invariant, and that only position-invar-iant representations can act as attentional templatesduring visual search, might also explain anotherapparent dissociation between search templatesand other types of visual working memory rep-resentations. Visual working memory has a capacityof approximately 3–4 objects (Cowan, 2001; Luck& Vogel, 1997). If search templates are stored inworking memory, it should in principle be possibleto simultaneously activate multiple attentional tem-plates for different possible target features orobjects, up to the point where memory capacity isexceeded. In fact, search for multiple targets ismuch less efficient than search for one particularobject or feature. Houtkamp and Roelfsema


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

(2009) demonstrated that the detection of targets ina rapid serial visual presentation (RSVP) stream isimpaired when observers search for one of twopossible target objects relative to a task wherethey searched for a single object. Modelling ofthese behavioural results suggested that exactlyone attentional template can be active at anygiven time (see also Menneer, Cave, & Donnelly,2009; Stroud, Menneer, Cave, Donnelly, &Rayner, 2011, for additional behavioural evidence,and Grubert & Eimer, 2013, for ERP evidencethat attentional target selectivity is less effectiveduring multiple-colour than during single-colourvisual search). If attentional templates are workingmemory representations, and if working memorycan hold several objects simultaneously, whyshould only a single search template be active at atime? One possibility is that attentional templatesand other working memory representations arestructurally equivalent, except that the currentsearch template is more strongly activated and istherefore able to bias focal attentional processingtowards template-matching objects (Olivers et al.,2011). Other working memory representationscan be simultaneously maintained, but are tempor-arily inhibited and therefore unable to affect theallocation of attention (see also Olivers & Eimer,2011). Alternatively, the difference between anattentional template and other visual workingmemory items might reflect a more fundamentalqualitative difference between position-invariantand position-dependent representations. In thiscontext, the impaired efficiency of multitarget ascompared to singletarget search would suggestthat it is difficult to maintain more than a singlespatially global attentional template at any giventime.

In summary, the research discussed in thissection has focused on the nature of preparatorysearch templates and their neural basis. It is gener-ally agreed that attentional templates are represen-tations of target objects or features that aremaintained in working memory during the prep-aration for visual search. However, identifying theneural correlates of such templates and demonstrat-ing causal links between preparatory neural activityand subsequent attentional effects have proved to

be difficult. According to the sensory recruitmentmodel of visual working memory, search templatesshould reside in visual–perceptual cortical areas, butthis is complicated by the fact that information inthese visual areas is represented in position-depen-dent cortical maps. Because search templates haveto operate in a spatially global fashion, theyshould represent search targets irrespective oftheir particular location in visual space. Such pos-ition-independent representations of targetobjects exist in prefrontal cortex, and possibly alsoin visual areas, and these representations might bethe neural counterpart of preparatory attentionaltemplates in visual search.

ATTENTIONAL GUIDANCE ANDFEATURE-BASED ATTENTION

While attentional templates are set up in prep-aration for an upcoming search task, the searchprocess itself starts once a visual search displayhas been presented. When the location of searchtargets is unknown, focal attention cannot be allo-cated to a particular region of visual space duringthe preceding preparatory phase. The selection ofpossible target objects therefore needs to be basedon the visual information that is available in thesearch display itself. According to models ofvisual search (e.g., Treisman & Sato, 1990;Wolfe, 1994, 2007), this information is accumu-lated at early stages of visual processing. Forexample, the Guided Search model (Wolfe, 1994,2007) assumes that representations of currentlytask-relevant object features are positively weightedduring the initial parallel processing of visual input,thereby increasing the probability that these fea-tures will attract focal attention. Similar ideaswere proposed in the biased competition accountof selective attention (Desimone & Duncan,1995; Duncan, 2006). According to this account,multiple simultaneously active object represen-tations compete for neural processing resourcesand the control of behaviour, and this competitionis biased in favour of currently task-relevant objectsthat are specified by attentional templates (i.e., pre-paratory baseline shifts of neural activity). These



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

attentional biases already operate at early stages ofvisual processing, although it may take considerablylonger before the competition for attentional selec-tion is resolved.

If the function of such goal-selective biases is toguide the allocation of attention during search forknown targets at unknown locations during theearly parallel processing of visual input, thesebiases need to operate in a spatially global fashionacross the visual field. There is indeed considerableevidence for the existence of such spatially globalattentional biases from research on feature-basedattention. Results from single-unit studies inmonkeys have shown that when a specific visualfeature is currently task relevant, the neural proces-sing of this feature is enhanced at the expense of theprocessing of other features in the same dimension.Crucially, these feature-based attentional modu-lations of neural activity appear to be elicited in aspatially global fashion across the entire visualfield. In a study by Martinez-Trujillo and Treue(2004), two sets of dots that both moved in thesame direction were presented in the left andright visual field, and monkeys were trained todetect small changes in the speed and direction ofone of these sets of dots. The activity of move-ment-selective neurons in the middle temporal cor-tical area with receptive fields that covered thestimuli in the currently unattended visual fieldwas strongly modulated by the direction of move-ment that the monkey was attending on the otherside. Neurons that preferred the currently task-rel-evant movement direction showed enhanced acti-vation, while the activity of neurons that preferredthe opposite direction of movement was sup-pressed. In other words, these attention-dependentactivity modulations of movement-sensitiveneurons were triggered in response to stimuli inthe task-irrelevant unattended visual field. Furtherevidence that feature-based attention operates in aspatially global fashion across the visual field wasprovided by Bichot, Rossi, and Desimone (2005)in a study where monkeys searched for colour-defined or shape-defined target objects. Neuronsin visual area V4 that were selective for the currentlytask-relevant feature increased their activity when atarget object was present in their receptive field

even when the monkey fixated a different objectand then shifted eye gaze to another location.This demonstrates that goal-dependent attentionalmodulations are elicited outside the current focus ofattention, thus providing further evidence that thistype of feature-based visual processing bias operatesin a spatially global fashion.

Additional evidence for the existence of spatiallyglobal feature-based attentional modulations wasprovided by fMRI and ERP studies in humans(Saenz, Buracas, & Boynton, 2002; Serences &Boynton, 2007; Zhang & Luck, 2009). In thesestudies, observers attended to a specific task-relevantfeature in one visual field, and objects in the otherunattended hemifield triggered enhanced visualresponses when they matched the feature that wascurrently attended on the opposite side. Althoughsuch spatially global effects of feature-based atten-tion havemostly been observed for simple target fea-tures such as colour, shape, or movement direction,they may also be present during search for category-defined targets. This has been demonstrated byPeelen, Fei-Fei, andKastner (2009), who asked par-ticipants to report the presence of people or cars inbriefly presented images of real-world visual scenesat a particular location and to ignore images thatwere simultaneously presented at other irrelevantlocations. MVPA-based analyses of fMRI datarevealed neural response patterns in object-selectivevisual cortex that were sensitive to the presence ofthe currently task-relevant stimulus category in aparticular image, even when this image appeared ata to-be-ignored location. This suggests that spatiallyglobal modulations of visual processing in favour ofpossible target objects are triggered not only insimple feature-based attentional selection tasks,but also during category-based search.

If feature-based attention operates in a spatiallyglobal fashion, its utility for the control of atten-tional selectivity in visual search is obvious.Feature-based attention can bias perceptual proces-sing in favour of candidate target objects, irrespec-tive of the location that these objects occupy in thevisual field, and can thus direct spatial attention toobjects that match one or more currently task-rel-evant features. In fact, this type of spatially globalfeature-based attentional control may represent the


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

direct neural equivalent of the guidance componentpostulated in the Guided Search model (Wolfe,1994, 2007). In this model, the parallel processingof visual information across the entire visual field isselectively weighted in favour of features thatmatch current search goals, resulting in the allo-cation of focal attention towards likely targetobjects (see also Bundesen et al., 2005, for similarideas). It is interesting to note that some target-defining visual attributes are much more effectivethan others in facilitating attentional object selectionduring visual search (see Wolfe & Horowitz, 2004,for a review). If the selection of search targets isguided by spatially global feature-based attentionalmodulations, such differences could reflect differ-ences in the availability of goal-selective biasesduring the parallel analysis of visual information.Due to the modular architecture of the visualsystem, spatially global feature-specific attentionalbiases might be relatively easy to implement forsimple target features such as colour, motion, ororientation, but not for more complex target-defin-ing attributes such as line intersections or three-dimensional volume. However, as shown by thepresence of spatially global processing modulationsduring category-based search (Peelen et al., 2009),this does not necessarily imply that feature-basedattentional control is restricted to elementary visualfeatures. The possibility that feature-based atten-tional mechanisms are more readily available forsome target-defining attributes than others, andthat this determines the effectiveness of top-downattentional control in visual search, obviouslyrequires further empirical support.

What are the links between the spatially globalfeature-based attention effects that are observedduring early stages of visual processing and thetarget-selective baseline shifts that are elicitedduring the preparation for a visual search task? Inother words, how do the preparatory attentional tem-plates that were discussed in the previous sectionaffect attentional guidance processes? If attentionaltemplates represent search targets in a spatiallyglobal fashion, it is plausible to assume that thesetemplates are directly responsible for the emergenceof spatially global feature-based attentional modu-lations. Such causal links between preparation and

guidance may be either direct or more indirect. Ifattentional templates are position-independent rep-resentations of search targets in visual cortex (assuggested by recent evidence for spatially globalworking memory representations in visual cortex;Ester et al., 2009), the activation of such represen-tations during the preparation phase of visual searchmay be simply maintained once search displays arepresented. In this case, attentional templates invisual working memory that are set up during thepreparation for search and feature-based attentionalmodulations that are observed once visual input hasbeen received would be essentially two sides of thesame coin (see also Desimone & Duncan, 1995, forsimilar suggestions). The observation that feature-based attention effects can spread to currentlyempty regions of visual space (Serences & Boynton,2007) suggests that preparatory feature-selectivebiases that are already active prior to the arrival ofvisual input can persist after a search display isencountered. The sustained presence of preparatorybaseline shifts can modulate the rapid feedforwardprocessing of visual input. This will result in anenhancement of visual activation at all locations oftemplate-matching objects in the visual field, asreflected by spatially global feature-based attentioneffects described earlier. For example, the spatiallyglobal modulations of category-selective responsesobserved by Peelen et al. (2009) in response tosearch displays containing multiple images maydirectly reflect the persistence of category-specificsearch templates that were set up during the preced-ing preparation phase (e.g., Peelen&Kastner, 2011).Alternatively, if spatially global attentional templatesreside not in visual cortex, but instead in moreanterior areas such as prefrontal cortex where searchtargets are represented in a position-independentfashion, the links between preparation and guidancemay be more indirect. In this case, feature-basedattention effects may be initiated and controlledby top-down signals from prefrontal to visual areas(e.g.,Maunsell&Treue, 2006).These two alternativescenarios are not necessarily mutually exclusive, asgoal-directed visual selection processes will generallyinvolve bidirectional recurrent interactions betweenvisual–perceptual brain regions and more anteriorattentional control areas (e.g., Bundesen et al., 2005).



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

Overall, the research discussed in this sectionstrongly suggests that when observers search for aparticular target at an unknown location, evenearly parallel stages of visual information processingare alreadymodulated by search goals that have beenactivated during the preparation for search. Becausesuch feature-based attention effects operate in aspatially global fashion, they can highlight the pres-ence of potential target objects anywhere in thevisual field and provide guidance signals for the sub-sequent allocation of focal attention to particularobjects. More generally, the presence of goal-sensi-tive attentional modulations during the rapid paral-lel analysis of visual input is interesting because itquestions the assumptions of traditional two-stagemodels of visual perception and attention. Suchmodels distinguish between an initial preattentiveprocessing stage that is entirely driven in abottom-up fashion by the physical properties ofthe visual input and a second attentive stage wherevisual processing is affected by current selectionintentions. This two-stage scenario was first pro-posed by Broadbent (1958) in his filter theory ofselective attention and has been a key feature of vir-tually all theoretical accounts of visual perceptionand attention ever since. For example, Theeuwes(2010) proposed a model where the initial stage ofattentional selection is determined exclusively bybottom-up salience signals generated during preat-tentive vision, and top-down influences onlyemerge at later stages of attentional processing.The possibility that feature-based attention alreadyaffects the early parallel feedforward analysis ofvisual input in a goal-dependent fashion castsserious doubts on the existence of a distinct preat-tentive stage of visual processing that operates in agenuinely stimulus-driven nonselective fashionand thus on the validity of the fundamental distinc-tion between preattentive and attentive vision.

OBJECT SELECTION AND FOCALSPATIAL ATTENTION

Feature-based attention operates in a spatiallyglobal fashion during the initial parallel processingof visual input and highlights the presence of

potentially task-relevant objects across the visualfield. This shows that even early stages of visualinformation processing are already selective in thesense of reflecting specific search goals. However,most models of visual search (e.g., Treisman &Gelade, 1980; Wolfe, 1994, 2007) assume theexistence of a separate and distinct attentionalobject selection process. Although the concept ofselection plays a central role in theories of attention,its precise meaning is rarely made explicit. In tra-ditional two-stage models of visual processing,“selection” marks the transition from preattentiveto attentive vision. Such models explain the needfor object selection by reference to the limitedcapacity of attentional processing and characterizethe selection process as a gatekeeping mechanismthat regulates access to this limited-capacitysystem (e.g., Broadbent, 1958). This account ofselection and its link to generic cognitive capacitylimitations has been criticized by other theorists(e.g., Allport, 1993). More recently, the limitedcapacity of visual processing and the resultingneed for attentional selectivity have been describedmore specifically as a direct result of the space-based topographical organization of the visualsystem. Because visual objects are represented intwo-dimensional cortical maps, multiple objectscan share the same neuronal receptive fields.These objects will compete for representationalspace (“cortical real estate”; Franconeri et al.,2013)—that is, for the control of neural responsesat particular locations within the visual maps. Inthis scenario, capacity is limited in the sense thatonly a small number of objects can be neurally rep-resented at any given time. “Selection” refers to theoutcome of a competitive process, where a particu-lar object has succeeded in driving neural activity ata particular location of visual space (Desimone &Duncan, 1995; Duncan, 2006). In this context,attentional object selection is space based and isdefined as the emergence of spatially focal proces-sing biases in favour of particular objects, at theexpense of other simultaneously present competingobjects.

An electrophysiological signature of this type ofobject selection was described by Chelazzi et al.(1998). In this study, monkeys had to make a


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

saccade to a specific target object that was pre-viously specified by a picture cue and wasaccompanied by a nontarget distractor object onthe same side. As described in the section on“Preparatory attentional templates and visualworking memory”, the picture cues elicited sus-tained baseline shifts of visual activity during thecue–target interval that were interpreted as neuralcorrelates of preparatory attentional templates.Neural responses to the subsequent target/nontar-get displays were recorded from inferotemporalcortex, for neurons that preferred one of thesetwo objects in these displays. Response rate wasinitially high, regardless of whether this preferredobject was the saccade target or the distractor ona given trial, due to the presence of the preferredobject in the receptive field. However, from about180 ms after search display onset, neural responseswere determined by search goals. When the pre-ferred object was the saccade target, response rateremained high. In contrast, neural activitydecreased strongly on trials where the same objectserved as distractor. In line with previous obser-vations that spatial attention determines theresponse rate of visual neurons when task-relevantand irrelevant stimuli are simultaneously presentin their receptive fields (e.g., Moran &Desimone, 1985), the results of Chelazzi et al.(1998) suggest that a spatially selective bias infavour of the target object emerged within lessthan 200 ms after stimulus onset. This spatial biascan be interpreted as the neural correlate of atten-tional object selection. Very similar electrophysio-logical effects have been found in many ERPstudies of visual search in humans. When a candi-date target object in the left or right visual field ispresented together with distractors, this object trig-gers an enhanced negativity at contralateral occipi-totemporal electrodes (N2pc component; Eimer,1996; Girelli & Luck, 1997; Luck & Hillyard,1994). Similar to the target-selective modulationsof neural responses described by Chelazzi et al.(1998), the N2pc component typically emergesaround 180 ms after stimulus onset and is primarilygenerated in extrastriate ventral visual cortex (e.g.,Hopf et al., 2000). It is therefore likely that bothmeasures are linked to the same underlying neural

process. Both reflect the spatially selective enhance-ment of responses to potential target objects versusdistractors in visual areas, and both are neuralmarkers of attentional object selection processesin visual search.

Most ERP studies that have used the N2pccomponent as an electrophysiological marker ofattentional object selection in visual search haveinvestigated situations where search targets weredefined by simple visual features such as particularcolours or shapes. In such tasks, the N2pcemerges within less than 200 ms after searchdisplay onset, demonstrating that spatially specificattentional object selection processes are elicitedduring relatively early stages of visual processing.Even though one might assume that attentionalselection operates much more slowly duringsearch for category-defined targets, a series ofrecent N2pc experiments from our lab suggestthat this is not the case. In one study, participantssearched for targets that were defined with respectto their alphanumerical category (e.g., any letteramong digit distractor objects; Nako, Wu, &Eimer, 2014). N2pc components to target itemswere elicited at around 180 ms post stimulus,which is very similar to the N2pc onset latenciestypically observed in search tasks where targets aredefined by simple visual features. In anotherstudy, targets were line drawings of real-worldobjects that were defined in terms of their categorymembership (e.g., kitchen objects among items ofclothing; Nako, Wu, Smith, & Eimer, 2014).Here, the N2pc to target objects emerged slightlylater, at around 240 ms post stimulus. The obser-vation that category-based attentional selectionprocesses are triggered within less than 250 msafter a search display has been presented showsthat category search mechanisms can operateremarkably rapidly. This is consistent with theresults from fMRI studies that investigated prep-aration and guidance processes during visualsearch for category-defined real-world targetobjects. Category-selective modulations of activityin visual cortex were already observed when partici-pants prepared for a particular search episode(Peelen & Kastner, 2011), and similar effectswere elicited in a spatially global fashion during



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

the subsequent processing of visual input (Peelenet al., 2009). These findings suggest that atten-tional preparation and subsequent attentional gui-dance processes can be selectively set for specifictarget categories, which may explain the rapidemergence of category-based attentional selectionprocesses that was observed in our recent N2pcstudies.

What is the relationship between the spatiallyglobal attentional guidance mechanisms discussedin the previous section and the attentional selectionprocesses discussed here? The attentional selectionof a particular object during visual search (i.e., theemergence of a spatially selective bias in favour ofthis object) is assumed to be based on informationabout the locations of candidate target objects thatis accumulated during the preceding guidancephase (Wolfe, 1994, 2007). Different architectureshave been suggested to describe this interplaybetween guidance and selection. In hierarchicalmodels, potential target locations are representedvia priority or salience maps (e.g., Fecteau &Munoz, 2006; Itti & Koch, 2001). Informationabout the presence of potential targets is generatedin parallel for different feature dimensions and con-verges on a shared priority map where it is inte-grated (e.g., Wolfe, 2007). The priority map islocated in dedicated attentional control regionsthat are anatomically and functionally distinctfrom the visual areas where spatially selectivebiases towards target objects emerge. The frontaleye fields (FEFs), posterior parietal cortex, or thethalamus have all been considered as the potentialneural locus of such maps (e.g., Bundesen et al.,2005; Gottlieb, Kusunoki, & Goldberg, 1998;Schall, 2004). Specific locations in a priority mapare linked to spatially corresponding locations invisual cortex, so that information about likelytarget locations within this map can trigger spatiallyselective modulations of visual processing. Thishypothesis is supported by the observation thatelectrical stimulation of FEF (one of the potentialattentional control areas) modulates the activity ofspatially corresponding regions in visual area V4(Moore & Armstrong, 2003). In terms of the dis-tinction between guidance and selection, the cre-ation of a specific priority map can be described

as the result of a spatially global guidance mechan-ism, and the subsequent selective modulation ofvisual activity at particular locations as the resultingattentional selection process.

Not all accounts of attentional control duringvisual search postulate the existence of dedicatedpriority maps. According to nonhierarchical dis-tributed models, goal-sensitive biases in favour ofparticular objects or features can be generated atdifferent levels of the visual processing hierarchy.These biases are then propagated both to higherand lower levels where spatially selective competi-tive advantages for specific objects emerge (“inte-grated competition”; Duncan, Humphreys, &Ward, 1997). In these models, the transitionfrom guidance to selection during visual search isa continuous process where the competitionbetween multiple objects is gradually resolved infavour of those objects with currently task-relevantproperties.

A controversial issue in past and present debatesabout the mechanisms of attentional object selec-tion in visual search concerns the serial versus par-allel nature of such selection processes. Doesattentional selection operate sequentially for oneobject at a time, or in parallel, so that severalobjects can be selected simultaneously and inde-pendently? If selection is implemented at theneural level as the emergence of spatially specificactivity modulations of object representations atparticular locations in visual maps, the issue of par-allel versus serial selection refers to the question ofwhether such modulations will eventually berestricted to one specific location or can be main-tained simultaneously at multiple locations in thevisual field. In traditional two-stage models ofvisual perception and attention, the transitionfrom preattentive to attentive vision that is con-trolled by selection mechanisms coincides withthe transition from parallel to serial processing.This assumption is retained in current models ofvisual search that describe attentional object selec-tion as a serial process. According to FeatureIntegration Theory, focal attention is directedsequentially to individual objects, which impliesthat the attentional selection of a new object is pre-ceded by a de-allocation of attention from its


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

previous location (e.g., Treisman & Gelade, 1980).Along similar lines, the Guided Search model(Wolfe, 1994, 2007) describes the selection of indi-vidual objects during visual search as a serial atten-tional bottleneck. Perhaps the most compellingreason for assuming that object selection operatesin a serial fashion is that the allocation of spatialattention is closely linked to eye movementcontrol (e.g., Moore, Armstrong, & Fallah, 2003;Thompson & Bichot, 2005). Saccadic eye move-ments are executed sequentially and are precededby the allocation of spatial attention to the locationof the next saccade target (e.g., Deubel &Schneider, 1996). Such attention shifts thatprecede eye movements are necessarily serial.However, spatial attention can be allocated in theabsence of overt gaze shifts, and visual searchdoes not depend on eye movements (e.g.,Zelinsky & Sheinberg, 1997). For these reasons,the serial nature of oculomotor control processesis not sufficient to conclude that attentional objectselection always operates in a strictly serialfashion. In fact, several theories of attention (e.g.,Bundesen et al., 2005; Desimone & Duncan,1995) postulate that object selection processes canoperate in parallel at multiple locations in thevisual field. Along similar lines, the ability to sim-ultaneously track multiple moving objects in thevisual field has been explained by assuming thatfocal spatial attention is allocated independentlyand in parallel to these objects (Cavanagh &Alvarez, 2005).

As an electrophysiological marker of attentionalobject selection, the N2pc component can track thetime course of attentional selection processes invisual search on a millisecond-by-millisecondbasis and can therefore provide insights into theparallel versus serial nature of these processes.Indirect N2pc evidence for parallel selectioncomes from the observation that N2pc amplitudesare sensitive to the number of task-relevantobjects in a display. When observers have toreport how many colour-defined target objects arepresent among distractors on one side of a searchdisplay, N2pc amplitudes increase with thenumber of targets (Mazza & Caramazza, 2011;see also Drew & Vogel, 2008, for similar

observations). This N2pc amplitude increase hasbeen interpreted in terms of object individuationprocesses, which operate in parallel when multipleobjects have to be simultaneously distinguishedfrom distractor objects. If “object individuation” isthe same process as “object selection”, this wouldimply that the increase of N2pc amplitudes withthe number of targets in a search display reflectsattentional selection processes that operate simul-taneously and in parallel for each target object.

More direct N2pc evidence for the existence ofsuch parallel object selection processes was obtainedin a recent study from our lab (Eimer & Grubert,2014a) where the selection of one target wasmeasured independently of the selection ofanother target. In this study, two displays were pre-sented in rapid succession, with stimulus onsetasynchronies (SOAs) of either 100 ms or 10 ms(as shown in Figure 2, top panel). Both displayscontained one target object that was defined byone specific colour and was accompanied by a dis-tractor object in a different colour on the oppositeside. Participants had to identify the two targetobjects that were successively presented in display1 and display 2 and to report whether thesetargets belonged to the same alphanumerical cat-egory (two letters, two digits) or not (one letterand one digit). The target/nontarget pair in onedisplay always appeared on the horizontal meridian(to the left and right of fixation), while the stimuluspair in the other display was presented on the ver-tical meridian. This procedure was used to track theattentional selection of one of the two target objectsindependently of the selection of the other target.Because the N2pc is always elicited contralateralto the side of a target object in the left or rightvisual field, no N2pc is triggered by objects thatappear on the vertical meridian. In our study,N2pc components therefore always reflected theattentional selection of the horizontal target.When the two displays were separated by a 100-ms SOA, the N2pc to horizontal targets indisplay 1 preceded the N2pc to horizontaltargets in display 2 by almost exactly 100 ms. TheN2pc to targets in the first display was followedby a second negative peak that overlappedwith the N2pc to the targets in display 2



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

Figure 2. Top panel: Stimulus set-up employed in the study by Eimer and Grubert (2014a). On each trial, two search displays were shown that

contained a colour-defined target and a distractor on opposite sides. In the examples shown here, the red items are the targets. The two displays

were presented in rapid succession, with a stimulus onset asynchrony (SOA) of 100 ms or 10 ms. One display contained two horizontal items

and the other two vertical items. The horizontal target was equally likely to appear in display 1 or in display 2. Bottom panel: N2pc results in

blocks where the SOA between the two displays was either 100 ms or 10 ms. Event-related potentials (ERPs) at lateral posterior electrodes

contralateral and ipsilateral to the horizontal target are shown together with N2pc difference waveforms obtained by subtracting ipsilateral

from contralateral ERPs. All ERPs are plotted relative to the onset of display 1. With an SOA of 100 ms, the N2pc to a horizontal target

in display 2 (H2) emerged 100 ms after the N2pc to a horizontal target in display 1 (H1). When the SOA was 10 ms, N2pc components

to H1 and H2 targets were triggered within 10 ms of each other and overlapped in time. This shows that the two targets were selected in

parallel, with each selection process following its own independent time course. Data from “Spatial Attention can be Allocated Rapidly and

in Parallel to New Visual Objects”, by M. Eimer and A. Grubert (2014), Current Biology, 24, p. 196. Copyright 2014 by the authors.

Reproduced in a different format with permission.


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

(Figure 2, bottom panel, SOA 100). This secondpeak reflects the initial phase of the sustained con-tralateral negativity that is associated with theencoding of task-relevant stimuli into workingmemory (see “Object recognition and workingmemory”). Critically, when the two displaysappeared within 10 ms of each other, the N2pc tohorizontal targets in display 2 emerged 10 mslater than the N2pc to targets in display 1 (Figure2, SOA 10). These two N2pc components wereequal in size and overlapped in time, demonstratingthat spatial attention was allocated rapidly and inparallel to both target objects, with each selectionprocess following its own independent timecourse (see also Khayat, Spekreijse, & Roelfsema,2006, for corresponding evidence for temporallyoverlapping attentional selection processes frommonkey neurophysiology). Similar results wereobtained in another set of studies that employedthe same procedures, except that the two targetobjects were now defined by two different colours(Grubert & Eimer, 2015), so that their selectioncould no longer be controlled by attentional prep-aration and guidance processes that are set for asingle target colour. In spite of this fact, the tem-poral pattern of N2pc components was similar tothe pattern observed in our initial experiment(Eimer & Grubert, 2014a). These N2pc resultsare difficult to reconcile with a strictly serial selec-tion account, which implies that the selection of anew object can only commence once attention iswithdrawn from its previous locus. They suggestinstead that multiple attentional selection processescan operate independently and in parallel.

Additional evidence for this conclusion comesfrom a visual search study from our lab (Eimer &Grubert 2014b) where participants searched for atarget that was defined by a specific conjunctionof colour and shape (e.g., a blue circle). Thistarget was presented together with two task-irrele-vant distractors and with an additional nontargetobject that matched one of the two target-definingfeatures (e.g., a blue square). On different trials, thetarget object was presented on the horizontal mer-idian and the partially matching nontarget objecton the vertical meridian, or vice versa, so thatN2pc components could be measured

independently to both types of objects. Accordingthe Guided Search model (Wolfe, 2007), attentionshould always be allocated to the target object,because this object has both task-relevant featuresand will therefore trigger the strongest activationon the priority map. If this is correct, N2pc com-ponents should be elicited only by targets, but notby a partially matching nontarget object in thesame display. In fact, reliable N2pc componentswere observed not only for targets, but also for par-tially matching nontargets, even though the targetobject was simultaneously present. This showsthat attention was allocated in parallel and indepen-dently to all features in the display that matched thecurrent target attributes (see also Andersen,Hillyard, & Müller, 2008, for corresponding evi-dence for the parallel selection of target featuresfrom steady-state visual evoked potentials).

Overall, these N2pc results show that atten-tional object selection processes can be elicited inparallel for multiple objects with target-matchingfeatures. Such observations are difficult to reconcilewith the widely held view that selection operatesserially in visual search (e.g., Treisman & Gelade,1980;Wolfe, 2007). But does the N2pc componentexclusively reflect processes that operate during theattentional selection phase? It is possible that theN2pc might also be sensitive to processes thattake place during the earlier spatially global atten-tional guidance stage. In fact, the pattern of N2pcresults observed in our recent studies where differ-ent objects with target-matching features were pre-sented simultaneously or in rapid succession (Eimer& Grubert, 2014a, 2014b; Grubert & Eimer,2015) appears to be similar to the spatially globalmodulations of visual processing produced byfeature-based attention that were described in theprevious section as the neural correlate of atten-tional guidance. Although such feature-basedattention effects typically emerge earlier than theN2pc component (e.g., Hopf, Boelmans,Schoenfeld, Luck, & Heinze, 2004; Zhang &Luck, 2009), the question remains whether theN2pc results discussed earlier reflect spatiallyglobal feature-based attentional guidance or parallelspace-based attentional selection processes. Thisquestion assumes that there is a strict separation



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

between spatially global guidance processes(feature-based attention) and spatially specificselection mechanisms (spatial attention), and thatthis dichotomy describes functionally and tem-porally discrete stages of attentional processingduring visual search. In fact, the distinctionbetween guidance and selection is a heuristicallyuseful way of conceptualizing different aspects ofattentional selectivity, but does not reflect theessentially continuous nature of attentional mech-anisms at the neural level. Because attentionalselectivity develops gradually during the processingof visual input, the transition from spatially globalattentional guidance to spatially focal attentionalobject selection is a continuous process, whereearly feature-based spatially global biases of visualprocessing gradually develop into object-basedspatially selective processing modulations.

In summary, attentional object selection wasdescribed in this section as the emergence ofspatially specific visual processing biases in favourof particular objects that result from the compe-tition for representational space in visual corticalmaps. These biases are typically elicited withinless than 200 ms after stimulus onset when searchtargets are defined by simple visual features andcan also emerge rapidly during category-basedvisual search. They are the result of informationabout the presence of possible target objects thatis accumulated by attentional guidance processes.Spatially selective modulations of visual processingcan be triggered simultaneously and independentlyat different locations in the visual field, whichimplies that attentional selection processes canoperate in parallel for different objects.

OBJECT RECOGNITION ANDWORKING MEMORY

The neural basis of attentional object selection wasdescribed as the emergence of a spatially specificbias of visual processing within visual corticalmaps in favour of a particular task-relevant object.However, the presence of such object-selectiveattentional modulations does not imply thatselected objects are instantly recognized. During

the attentional tracking of multiple visual objects,access to the features and identity of these objectsis remarkably poor (Horowitz et al., 2007), demon-strating that the allocation of focal attention tospecific objects is not sufficient for their recog-nition. In visual search, the selection and identifi-cation of target objects are separable mechanisms(e.g., Ghorashi, Enns, Klein, & Di Lollo, 2010).Many models of visual attention and visual searchmake an explicit distinction between object selec-tion and object recognition. Object selection isdescribed as a stage where particular objects areindividuated via the allocation of focal attention.Object recognition is assumed to take place at asubsequent stage where the features of theseobjects are integrated, and their identity becomesaccessible (e.g., Huang & Pashler, 2007; Wolfe,2007; Xu & Chun, 2009). In line with these sug-gestions, recent studies from our lab have alsofound ERP evidence for the transition betweenan early phase of attentional selectivity where atten-tion is rapidly allocated to candidate target objectsand a later phase where information about the attri-butes of selected objects is integrated across featuredimensions (Eimer & Grubert, 2014b; Kiss,Grubert, & Eimer, 2013).

Selection and recognition are sensitive to differ-ent experimental factors: The efficiency of targetselection is determined by the number of compet-ing nontarget objects in a search display and theirsimilarity to the target (e.g., Duncan &Humphreys, 1989), whereas recognition processesare primarily affected by target complexity (e.g.,Franconeri et al., 2013; Xu & Chun, 2009).When two target objects are presented successivelyand without competing distractors in the samedisplay, so that the demands on spatial selectivityare minimal, identification of the second target isoften strongly impaired (“attentional blink”; e.g.,Duncan, 1980; Duncan, Ward, & Shapiro, 1994;Raymond, Shapiro, & Arnell, 1992; see Wyble,Bowman, & Nieuwenstein, 2009, for an accountof the attentional blink in terms of competitivemechanisms in working memory). This suggeststhat attentional capacity limitations can arisespecifically at object recognition stages that followthe spatial selection of target objects.


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

How can the transition from object selection toobject recognition be described at the cognitive andneural levels? The recognition of a particular objectis based on the activation of a working memory rep-resentation of this object (e.g., Bundesen et al.,2005; Chun & Johnson, 2011). Such workingmemory representations are actively maintainedby a sustained focus of spatial attention (e.g.,Awh, Vogel, & Oh, 2006; Oberauer, 2002).Object selection was previously defined as the allo-cation of spatial attention to particular objects, asreflected by a spatially selective modulation ofneural responses to these objects in visual corticalmaps. The encoding of a selected object intoworking memory during the recognition phasecan thus be described as the active maintenanceof an attentional focus that was established duringobject selection, or, in other words, as spatial atten-tion that is sustained internally over time (e.g.,Chun, 2011).

This critical role of spatial attention for visualworking memory was shown in behavioural andERP studies that found spatially selective visualprocessing enhancements for locations that werecurrently maintained in memory (Awh, Anllo-Vento, & Hillyard, 2000; Awh, Jonides, &Reuter-Lorenz, 1998). According to the sensoryrecruitment model of working memory (e.g.,Postle, 2006), memorized visual objects are storedin visual areas that are also activated during theperceptual processing of visual input. The atten-tion-based maintenance of visual objects shouldtherefore take place within cortical maps in visualareas and should operate via spatially selectiveactivity enhancements for visual object represen-tations at particular locations within these maps,because spatial attention necessarily operates in aspace-based fashion. This essential involvement ofspatial attention in the maintenance of workingmemory representations is one of the reasons whyvisual working memory representations are stronglyposition dependent (see section on “Preparatoryattentional templates and visual working memory”).

If object selection and working memory are bothmediated by spatial attention, this should bereflected by functional links between the N2pccomponent (which marks attentional object

selection; see section on “Object selection andfocal spatial attention”) and the subsequent CDAcomponent (which is elicited during workingmemory maintenance). As mentioned in thesection on “Preparatory attentional templates andvisual working memory”, CDA components are eli-cited during the delay period of working memorytasks at posterior electrodes contralateral to theside where memorized objects were presentedduring encoding (Vogel & Machizawa, 2004). Ina study where observers had to memorize betweenone and six display objects for subsequent recall(Anderson, Vogel, & Awh, 2011), CDA ampli-tudes increased with memory set size, up to thepoint where individual working memory capacitywas exceeded. Importantly, N2pc componentsthat were triggered during the initial attentionalselection of the to-be-remembered objects showedexactly the same sensitivity to memory set size.Such parallel effects of memory set size on N2pcand CDA amplitudes should indeed be observedif separate independent foci of spatial attentionare established during the attentional selection ofmemorized objects (N2pc component), and arethen sustained over time during the maintenanceof these objects in working memory (CDAcomponent).

The hypothesis that attentional object selectionand working memory maintenance are both basedon spatially selective modulations of visual proces-sing at one or several locations in the visual fieldyields another interesting prediction: Individualdifferences in the ability to maintain multipleobjects in visual working memory should belinked to individual differences in the ability to allo-cate spatial attention simultaneously to multipleobjects during visual search. The existence of sucha link between memory capacity and search per-formance was demonstrated by Anderson, Vogel,and Awh (2013). Individuals with high workingmemory capacity performed more efficiently thanlow-capacity participants in a difficult search taskwhere targets and distractors were very similar, sothat each item had to be focally attended in orderto be recognized as target or nontarget. Theseobservations suggest that object selection duringvisual search tasks and working memory capacity



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

are determined by a common underlying factor—the ability to select and maintain multiple represen-tations of individuated objects in a spatiallyselective fashion.

The critical role of focal spatial attention duringthe selection and subsequent active maintenance ofobjects in visual working memory is illustrated by arecent ERP study from our lab (Towler, Kelly, &Eimer, 2015) that investigated visual facememory. Memory displays that contained twodifferent faces in the left and right visual fieldwere followed after a brief delay period by test dis-plays with a single face at fixation (see Figure 3, leftpanel). Participants’ task was to encode and main-tain both faces in the memory display and toreport whether the test face matched one of thesetwo faces or was a different face. Performance wassurprisingly poor in this task and suggested thatonly one of the two faces in the memory displayswas encoded into working memory on most trials.Which of these two faces was maintained wasdetermined by spatial attention, and this was

revealed by the ERPs recorded in response to thememory displays, prior to the arrival of the sub-sequent test displays. Figure 3 (right panel) showsthese ERPs for trials where one of the twomemory display faces was later repeated in thetest display. On trials where participants detectedthis face repetition correctly and rapidly, N2pcand CDA components were found contralateralto the face in the memory display that was thenrepeated. On these trials, focal attention was allo-cated to the “correct” face (i.e., the face thatwould reappear as the test face), and this spatialfocus was then maintained during the delayperiod, resulting in the rapid detection of an iden-tity match between the memorized face and the testface. On trials where participants failed to detect aface repetition, N2pc and CDA components wereinstead elicited contralateral to the face in thememory display that was not repeated. On thesetrials, focal attention was evidently allocated tothe “wrong” (i.e., nonrepeated) face, and this facewas then retained in working memory, at the

Figure 3. Left panel: Stimulus set-up employed by Towler et al. (2015). Participants had to encode the two faces in a memory display, compare

them to a centrally presented face in a test display, and decide whether one of the two memorized faces was repeated. The delay period between

memory display offset and test display onset was very brief (200 ms). Right panel: Event-related potentials (ERPs) measured at lateral

posterior electrodes in response to memory displays in the 500-ms interval after display onset, for trials where one of the two faces in the

memory display was later repeated in the test display. On trials where this identity repetition was detected correctly and rapidly, N2pc and

contralateral delay activity (CDA) components were elicited contralateral to the face that would reappear as the test face. On trials where

participants failed to report the face repetition, N2pc and CDA components were elicited contralateral to the face in the memory display

that was not repeated. These results show that attention was selectively allocated to one of the two faces in the memory displays. This focus

of spatial attention determined which face was retained in working memory and predicted the success or failure of face identity matching on

individual trials. Data from “The Focus of Spatial Attention Determines the Number and Precision of Face Representations in Working

Memory” by Towler et al. (2015), Cerebral Cortex. Copyright 2015 by Oxford University Press. Reproduced in a different format with

permission.


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

expense of the face that would reappear as the testface. These ERP results show that the success orfailure of maintaining a visual representation of anindividual face in working memory is determinedby the allocation of spatial attention, and that atten-tion can only be allocated to one face at a time.

If object selection is implemented through thespatially selective modulation of neural responsesto particular objects in cortical maps, and objectrecognition is based on the active maintenance ofthis spatial bias over time, the question ariseswhether these two successive stages can be disso-ciated. The maintenance of a particular object inworking memory during its recognition shouldalways depend on the prior attentional selectionof this object. But are there situationswhere objects are selected without subsequentlybeing encoded in working memory?Electrophysiological evidence for selectionwithout memory-based recognition was found ina visual search study from our lab (Mazza,Turatto, Umiltà, & Eimer, 2007) where observershad to select a colour singleton target amonguniform distractor objects in two different task con-ditions. In a localization task, they simply had toreport whether the target appeared in the left orright visual field. In a discrimination task, thespecific shape of the colour-defined target objecthad to be identified. Identical N2pc componentswere triggered in both tasks, demonstrating thatthe initial attentional selection of targets was unaf-fected by the difference in task demands. In con-trast, the subsequent sustained contralateralnegativity that marks the activation of a target rep-resentation in working memory mediated by focalspatial attention was only elicited when participantshad to discriminate the target shape, but not in thelocalization task. These observations demonstratethat object selection and recognition are indeedseparable stages in visual search, and that the acti-vation of a sustained working memory represen-tation of an object is not an automatic andinevitable consequence of its previous attentionalselection. Even though they are dissociable, selec-tive attention and working memory are usuallyclosely linked. For example, the current contentof working memory can affect the allocation of

attention during visual search even when thiscontent is irrelevant for the search task. Whenobservers are asked to memorize a particularcolour for subsequent recall before performingan independent visual search task, the presence ofa distractor that matches the memorized colourimpairs search performance (e.g., Downing,2000; Olivers & Eimer, 2011; Olivers, Meijer, &Theeuwes, 2006; Soto, Heinke, Humphreys, &Blanco, 2005). This suggests that spatial attentioncan be biased towards memory-matching but cur-rently task-irrelevant objects (see Olivers et al.,2011, for further discussion).

Sustaining a spatial focus of attention over timeduring the maintenance of visual object represen-tations in working memory requires recurrent feed-back from higher order attentional control areas(e.g., Bundesen et al., 2005; Luck & Vogel, 2013;Xu & Chun, 2009; see also Bar, 2003; Hochstein& Ahissar, 2002, for the importance of recurrentfeedback signals during object identification). Inthe absence of such recurrent feedback loops, anyspatially specific enhancement of object processingthat is triggered in visual cortex during the initialobject selection stage is assumed to remain transi-ent and fade rapidly, as was observed in the localiz-ation task of our ERP study (Mazza et al., 2007).Regions in the intraparietal sulcus that are sensitiveto working memory load and individual capacitylimits (e.g., Todd & Marois, 2004; Xu & Chun,2006) may play a central role in sustaining visualworking memory representations of target objectsin visual cortex when these objects have to be recog-nized. Why would object recognition require aspatially selective focus of attention that remainsactive for an extended period of time? Sustainedfocal attention may be needed to facilitate thebinding of individual features within object rep-resentations in visual working memory (e.g.,Wheeler & Treisman, 2002), in particular whenthese objects are no longer perceptually present. Itmay also be needed because object recognition pro-cesses involve comparisons between workingmemory representations of currently selectedobjects and stored object representations in long-term memory. Such comparison processes operatein real time and may therefore require a sustained



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

focus of spatial attention on particular workingmemory representations. More generally, recog-nition and identification processes are unlikely tobe based exclusively on object representations invisual cortical maps, but will usually also involveinteractions between these representations andother areas where semantic or episodic informationabout particular objects is stored (e.g., Sreenivasanet al., 2014). According to global neuronal work-space models of cognitive processing (e.g.,Dehaene & Naccache, 2001), such long-distanceinteractions between different cortical regions arelikely to be based on neural activation patternsthat are maintained in a stable fashion over anextended period of time.

The research discussed in this section againhighlights the continuous nature of the neuralmechanisms that are active during visual search.Although cognitive models of visual attentionemphasize the distinction between discrete selec-tion and recognition stages, current ideas aboutthe neural basis of attentional object selection andworking memory maintenance suggest that thesetwo stages are both based on spatially selectivemodulations of visual processing that differ primar-ily with respect to their temporal duration. Objectselection is implemented by fast and transientspatially specific processing enhancements withincortical maps. Memory-based object recognitioncan take place when these spatial biases are activelysustained over time. Because time is a continuousvariable, the transition from object selection toobject recognition—that is, the transition from per-ceptual attention to working memory—is also likelyto be a continuous process.

SUMMARY AND CONCLUSIONS

This article has reviewed the attentional processesthat are responsible for our ability to find knowntarget objects at unknown locations in visualsearch. These processes can be studied at the cogni-tive level and at the neural implementation level,and one aim of this review was to provide linksbetween empirical findings and theoretical ideasat these two levels. The picture that has emerged

from this discussion is quite different from tra-ditional conceptualizations of selective attentionthat are based on the fundamental distinctionbetween preattentive and attentive stages of percep-tual processing. In such two-stage models, atten-tional selection mechanisms are located at theintersection between these two stages and regulatethe access of visual information to a centrallimited-capacity system. This two-stage architec-ture was proposed by Donald Broadbent in 1958and has remained highly influential ever since.Indeed, as Jon Driver remarked in his review ofattention research in the twentieth century,Broadbent’s ideas may “have been almost too influ-ential; once exposed to them, it becomes hard tothink about attentional issues in any other way”(Driver, 2001, p. 56). The description of the atten-tional mechanisms involved in visual search thatwas developed in this review goes beyond such tra-ditional two-stage models. It stresses the functionaland temporal continuity of attentional processesand their neural basis and questions the existenceof a genuinely preattentive (i.e., goal-unselective)processing stage in visual search. The idea thatsearch goals are represented by attentional tem-plates that are activated during the preparation forsearch (“Preparatory attentional templates andvisual working memory”) and the hypothesis thatthese preparatory processes produce spatiallyglobal goal-selective modulations during the sub-sequent attentional guidance phase (“Attentionalguidance and feature-based attention”) do not sitcomfortably with a fundamental separationbetween preattentive and attentive processing anda distinct locus of attentional selection at the inter-face between these two stages.

Because attentional mechanisms are implementedby neural processes that unfold gradually in real time,they do not lend themselves easily to traditional dis-crete stage models of information processing.Nevertheless, it is still conceptually and heuristicallyuseful to distinguish successive phases of attentionalselectivity during visual search, such as the fourphases described here (see Figure 1). Preparatoryattentional templates can be set up prior to thearrival of visual input and are reflected by goal-selec-tive sustained baseline shifts of neural activity during


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

the preparation for an upcoming attentional selectiontask (“Preparatory attentional templates and visualworking memory”). Attentional guidance processesstart once visual input has been presented and arebased on goal-sensitive modulations of visual proces-sing (feature-based attention) that operate in aspatially global fashion across the entire visual field(“Attentional guidance and feature-based attention”).If preparatory attentional templates in visual areasrepresent search goals in a position-independentfashion, the transition from preparation to guidanceis likely to be continuous rather than discrete.“Selection” was defined as the emergence of spatiallyspecific modulations of neural activity in visual corti-cal maps that facilitate the processing of task-relevantfeatures or objects at particular locations (“Objectselection and focal spatial attention”). Again, thetransition from spatially global feature-based atten-tion to attentional object selection is best understoodas a continuous process where spatially selectivefeature- and object-specific biases gradually developacross time. Finally, object recognition has beenlinked to the encoding of particular selected objectsin visual working memory (“Object recognition andworking memory”). Because working memorydepends on spatially specific processing biases thatwere initially established during object selection andare then sustained over time, the transition betweenselection and maintenance is also continuous ratherthan discrete.

At the most general level, the attentional mech-anisms described here can be characterized as theemergence and subsequent maintenance of spatiallyselective biases of visual processing in favour ofpotentially task-relevant objects. Such spatiallyfocal biases are set up and sustained via recurrentinteractions between visual cortex and higherorder control areas, develop within the first 200ms after visual input has been presented, andremain active until target objects have been success-fully identified. The gradual transition fromspatially global to spatially focused goal-selectiveneural activation patterns in cortical visual mapsdescribed here does not imply that this processwill always eventually result in a single unitaryfocus of spatial attention. If working memorymaintenance and the ability to track moving

objects depend on spatial attention, the fact thatmultiple objects can be simultaneously tracked(e.g., Cavanagh & Alvarez, 2005) and held invisual working memory (e.g., Cowan, 2001; Luck& Vogel, 1997) suggests that multiple independentattentional foci can be maintained in parallel whenrequired by current task demands. In other taskcontexts, a single focus of spatial attention may berequired. This may be the case when complexvisual objects such as individual faces have to beselected and maintained in working memory (e.g.,Towler et al., 2015). Another example is theactive exploration of visual scenes where the eyesmove rapidly between different objects. In suchsituations, the selection of the next saccade targetis based on allocating focal attention to one particu-lar object in the visual field.

This review has described the attentional pro-cesses that are activated in visual search taskswhere known targets have to be found at uncertainlocations. In the absence of precise advance spatialinformation, the allocation of spatial attention topossible target objects will depend on spatiallyglobal attentional guidance processes. In taskswhere the location of an upcoming task-relevantobject is specified in advance (e.g., Posner et al.,1980), this type of guidance is not required,because observers can prepare for a particulartarget location before the arrival of visual input,and attention can then be allocated rapidly to thislocation. Preparatory baseline shifts of neuralactivity at prespecified target locations haveindeed been observed in spatial cueing experimentsprior to the presentation of visual input (e.g., Luck,Chelazzi, Hillyard, & Desimone, 1997). ERPstudies of cued visual–spatial attention (e.g.,Eimer, 1994; Mangun & Hillyard, 1991) havefound enhancements of visual P1 components forstimuli at attended versus unattended locationsthat start around 90 ms after stimulus onset. Thisshows that when target locations are known inadvance, spatially specific attentional selection pro-cesses are triggered very rapidly. In contrast, theN2pc component that marks the selection of atarget object at a previously unknown location invisual search tasks emerges nearly 100 ms later.This onset difference between spatially selective



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

attentional effects in spatial cueing and visualsearch experiments fits perfectly with the sugges-tion that target selection has to be preceded by aspatially global attentional guidance process invisual search, whereas no such guidance is requiredin spatial cueing tasks where target locations areknown in advance.

The account of the attentional processesinvolved in visual search outlined in this reviewcan provide a general framework that helps to inte-grate the results of behavioural and neurosciencestudies of selective attention with diverse exper-imental procedures and measurement techniques.However, it does not provide a complete descrip-tion of the complexities of visual search. In difficultvisual search tasks, the target object is unlikely to befound on the basis of a single attentional episodethat involves guidance, selection, and object recog-nition. When a target is hard to find, several iter-ations of these processes will be required that aretriggered by a mismatch between visual object rep-resentations currently activated in working memoryand an attentional template of the search target.The efficiency of visual search varies greatly acrosstasks (Wolfe, 1998), and such differences mayreflect the number of iterations of a search cyclethat are required before the target has been success-fully recognized.

Attentional guidance in visual search may alsobe more complex than described in this article.While spatially global guidance processes areimportant when target locations are unknown,search does not always operate without any priorinformation about the likely position of task-rel-evant objects. During the processing of real-worldvisual scenes, context-sensitive spatial expectationscan play an important role for the allocation ofspatial attention (e.g., Hollingworth, 2009). Suchcontextual spatial information can also affect thedetection of targets in simple search displays (“con-textual cueing”; see Chun, 2000). In such situ-ations, attentional guidance mechanisms may notoperate in a spatially global fashion across theentire visual field, but could be restricted to specificareas that are most likely to contain the targetobjects. This suggests that context-dependentspatial expectations can be explicit or implicit

parts of preparatory attentional templates. Inaddition, the allocation of attention in visualsearch tasks is not always exclusively guided in agoal-directed fashion by representations of target-defining features, but can also be affected by theperceptual salience of visual objects (e.g., Itti &Koch, 2001). This is acknowledged in theconcept of priority-driven attentional guidance(e.g., Horowitz, Wolfe, Alvarez, Cohen, &Kuzmova, 2009; Wolfe, 2007), which is deter-mined jointly by top-down information abouttarget-defining features, context-dependent spatialexpectations, and bottom-up salience signals.There may also be situations where search operateswithout goal-selective guidance. In these cases, theallocation of spatial attention to particular objectswould be determined either randomly, or exclu-sively by bottom-up salience signals. Workingmemory representations of target objects wouldstill be required, because the recognition of selectedobjects as targets or nontargets depends on theircomparison with stored representations of searchgoals. The original version of Feature IntegrationTheory (Treisman & Gelade, 1980) describes sucha scenario where search operates without guidance.

The localization and recognition of targetobjects among distractors in visual search arebased on processes that unfold in real time. Theattentional mechanisms that are involved in visualsearch and their neural basis have been studied fordecades, and this review has described the outlinesof a general processing model. In this model,“attention” is not characterized as a distinct stageof visual processing or as a dedicated top-downcontrol system. Selective attention in visual searchrefers to the gradual and goal-dependent emer-gence of spatially selective processing biases for par-ticular object representations in cortical visualmaps.

REFERENCES

Allport, A. (1993). Attention and control: Have we beenasking the wrong questions? A critical review oftwenty-five years. In Attention and performance XIV:

Synergies in experimental psychology, artificial


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

intelligence, and cognitive neuroscience (pp. 183–218).Cambridge, Mass: MIT Press.

Andersen, S. K., Hillyard, S. A., & Müller, M. M.(2008). Attention facilitates multiple stimulus fea-tures in parallel in human visual cortex. Current

Biology, 18, 1006–1009.Anderson, D. E., Vogel, E. K., & Awh, E. (2011).

Precision in visual working memory reaches a stableplateau when individual item limits are exceeded.Journal of Neuroscience, 31, 1128–1138.

Anderson, D. E., Vogel, E. K., & Awh, E. (2013). Acommon discrete resource for visual workingmemory and visual search. Psychological Science, 24,929–938.

Awh, E., Anllo-Vento, L., & Hillyard, S. A. (2000).The role of spatial selective attention in workingmemory for locations: Evidence from event-relatedpotentials. Journal of Cognitive Neuroscience, 12,840–847.

Awh, E., & Jonides, J. (2001). Overlapping mechanismsof attention and spatial working memory. Trends inCognitive Sciences, 5, 119–126.

Awh, E., Jonides, J., & Reuter-Lorenz, P. A. (1998).Rehearsal in spatial working memory. Journal of

Experimental Psychology: Human Perception and

Performance, 24, 780–790.Awh, E., Vogel, E. K., & Oh, S. H. (2006). Interactions

between attention and working memory. Neuroscience,139, 201–208.

Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visual object recognition. Journalof Cognitive Neuroscience, 15, 600–609.

Bichot, N. P., Rossi, A. F., & Desimone, R. (2005).Parallel and serial neural mechanisms for visualsearch in macaque area V4. Science, 308, 529–534.

Broadbent, D. E. (1958). Perception and communication.London: Pergamon Press.

Bundesen, C., Habekost, T., & Kyllingsbæk, S. (2005).A neural theory of visual attention: Bridging cogni-tion and neurophysiology. Psychological Review, 112,291–328.

Cavanagh, P., & Alvarez, G. A. (2005). Tracking mul-tiple targets with multifocal attention. Trends in

Cognitive Sciences, 9, 349–354.Chawla, D., Rees, G., & Friston, K. J. (1999). The phys-

iological basis of attentional modulation in extrastriatevisual areas. Nature Neuroscience, 2, 671–676.

Chelazzi, L., Duncan, J., Miller, E. K., & Desimone, R.(1998). Responses of neurons in inferior temporalcortex during memory-guided visual search. Journalof Neurophysiology, 80, 2918–2940.

Chun, M. M. (2000). Contextual cueing of visual atten-tion. Trends in Cognitive Sciences, 4, 170–178.

Chun, M. M. (2011). Visual working memory as visualattention sustained internally over time.Neuropsychologia, 49, 1407–1409.

Chun, M. M., & Johnson, M. K. (2011). Memory:Enduring traces of perceptual and reflective attention.Neuron, 72, 520–535.

Cowan, N. (2001). The magical number 4 in short-termmemory: A reconsideration of mental storagecapacity. Behavioral and Brain Sciences, 24, 87–114.

Dehaene, S., & Naccache, L. (2001). Towards acognitive neuroscience of consciousness: Basicevidence and a workspace framework. Cognition,79, 1–37.

Desimone, R., & Duncan, J. (1995). Neural mechanismsof selective visual attention. Annual Review of

Neuroscience, 18, 193–222.Desimone, R., & Gross, C. G. (1979). Visual areas in the

temporal cortex of the macaque. Brain Research, 178,363–380.

D’Esposito, M. (2007). From cognitive to neural modelsof working memory. Philosophical Transactions of theRoyal Society of London Series B: Biological Sciences,362, 761–772.

Deubel, H., & Schneider, W. X. (1996). Saccade targetselection and object recognition: Evidence for acommon attentional mechanism. Vision Research, 36,1827–1837.

Downing, P. E. (2000). Interactions between visualworking memory and selective attention.Psychological Science, 11, 467–473.

Drew, T., & Vogel, E. K. (2008). Neural measures ofindividual differences in selecting and tracking mul-tiple moving objects. Journal of Neuroscience, 28,4183–4191.

Driver, J. (2001). A selective review of selective attentionresearch from the past century. British Journal of

Psychology, 92, 53–78.Driver, J., & Frith, C. (2000). Shifting baselines in atten-

tion research. Nature Reviews Neuroscience, 1,147–148.

Duncan, J. (1980). The locus of interference in the per-ception of simultaneous stimuli. Psychological

Review, 87, 272–300.Duncan, J. (2006). EPS Mid-Career Award 2004: Brain

mechanisms of attention. The Quarterly Journal of

Experimental Psychology, 59, 2–27.Duncan, J., Humphreys, G., & Ward, R. (1997).

Competitive brain activity in visual attention.Current Opinion in Neurobiology, 7, 255–261.



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

Duncan, J., & Humphreys, G. W. (1989). Visual searchand stimulus similarity. Psychological Review, 96,433–458.

Duncan, J., Ward, R., & Shapiro, K. (1994). Directmeasurement of attentional dwell time in humanvision. Nature, 369, 313–315.

Eimer, M. (1994). “Sensory gating” as a mechanism forvisuospatial orienting: Electrophysiological evidencefrom trial-by-trial cuing experiments. Perception &

Psychophysics, 55, 667–675.Eimer, M. (1996). The N2pc component as an indicator

of attentional selectivity. Electroencephalography and

Clinical Neurophysiology, 99, 225–234.Eimer, M. (2014). The neural basis of attentional control

in visual search. Trends in Cognitive Sciences, 99,225–234.

Eimer, M., & Grubert, A. (2014a). Spatial attention canbe allocated rapidly and in parallel to new visualobjects. Current Biology, 24, 193–198.

Eimer, M., & Grubert, A. (2014b). The gradual emer-gence of spatially selective target processing in visualsearch: From feature-specific to object-based atten-tional control. Journal of Experimental Psychology:

Human Perception and Performance, 40, 1819–1831.Ester, E. F., Serences, J. T., & Awh, E. (2009). Spatially

global representations in human primary visual cortexduring working memory maintenance. Journal of

Neuroscience, 29, 15258–15265.Fannon, S. P., Saron, C. D., & Mangun, G. R. (2007).

Baseline shifts do not predict attentional modulationof target processing during feature-based visual atten-tion. Frontiers in Human Neuroscience, 1, 7–18.

Fecteau, J. H., & Munoz, D. P. (2006). Salience, rel-evance, and firing: A priority map for target selection.Trends in Cognitive Sciences, 10, 382–390.

Franconeri, S. L., Alvarez, G. A., & Cavanagh, P.(2013). Flexible cognitive resources: Competitivecontent maps for attention and memory. Trends inCognitive Sciences, 17, 134–141.

Fuster, J. M., & Alexander, G. E. (1971). Neuronactivity related to short-term memory. Science, 173,652–654.

Ghorashi, S., Enns, J. T., Klein, R. M., & Di Lollo, V.(2010). Spatial selection and target identification areseparable processes in visual search. Journal of Vision,10, 1–12.

Giesbrecht, B., Weissman, D. H., Woldorff, M. G., &Mangun, G. R. (2006). Pre-target activity in visualcortex predicts behavioral performance on spatialand feature attention tasks. Brain Research, 1080,63–72.

Girelli, M., & Luck, S. J. (1997). Are the same atten-tional mechanisms used to detect visual searchtargets defined by color, orientation, and motion?Journal of Cognitive Neuroscience, 9, 238–253.

Gottlieb, J. P., Kusunoki, M., & Goldberg, M. E.(1998). The representation of visual salience inmonkey parietal cortex. Nature, 391, 481–484.

Gratton, G., Corballis, P. M., & Jain, S. (1997).Hemispheric organization of visual memories.Journal of Cognitive Neuroscience, 9, 92–104.

Grubert, A., & Eimer, M. (2013). Qualitative differencesin the guidance of attention during single-color andmultiple-color visual search: Behavioral and electro-physiological evidence. Journal of Experimental

Psychology: Human Perception and Performance, 39,1433–1442.

Grubert, A., & Eimer, M. (2015). Rapid parallel atten-tional target selection in single-colour and multiple-colour visual search. Journal of Experimental


Harrison, S. A., & Tong, F. (2009). Decoding reveals thecontents of visual working memory in early visualareas. Nature, 458, 632–635.

Henderson, J. M. (2003). Human gaze control duringreal-world scene perception. Trends in Cognitive

Sciences, 7, 498–504.Hochstein, S., & Ahissar, M. (2002). View from the top:

Hierarchies and reverse hierarchies in the visualsystem. Neuron, 36, 791–804.

Hollingworth, A. (2009). Two forms of scene memoryguide visual search: Memory for scene context andmemory for the binding of target object to scenelocation. Visual Cognition, 17, 273–291.

Hopf, J. M., Boelmans, K., Schoenfeld, M. A., Luck, S.J., & Heinze, H. J. (2004). Attention to features pre-cedes attention to locations in visual search: Evidencefrom electromagnetic brain responses in humans. TheJournal of Neuroscience, 24, 1822–1832.

Hopf, J. M., Luck, S. J., Girelli, M., Hagner, T.,Mangun, G. R., Scheich, H., et al. (2000). Neuralsources of focused attention in visual search.Cerebral Cortex, 10, 1233–1241.

Hornak, J., Duncan, J., & Gaffan, D. (2002). The role ofthe vertical meridian in visual memory for objects.Neuropsychologia, 40, 1873–1880.

Horowitz, T.S.,Klieger, S.B., Fencsik,D.E.,Yang,K.K.,Alvarez,G.A.,&Wolfe, J.M. (2007).Tracking uniqueobjects. Perception & Psychophysics, 69, 172–184.

Horowitz, T. S., Wolfe, J. M., Alvarez, G. A., Cohen,M. A., & Kuzmova, Y. I. (2009). The speed of free


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

will. The Quarterly Journal of Experimental Psychology,62, 2262–2288.

Houtkamp, R., & Roelfsema, P. R. (2009). Matching ofvisual input to only one item at any one time.Psychological Research Psychologische Forschung, 73,317–326.

Huang, L., & Pashler, H. (2007). A Boolean map theoryof visual attention. Psychological Review, 114,599–631.

Itti, L., & Koch, C. (2001). Computational modelling ofvisual attention. Nature Reviews Neuroscience, 2,194–203.

James,W. (1890/1981).The principles of psychology.Volume1. Cambridge, Mass.: Harvard University Press.

Katus, T., Grubert, A., & Eimer, M. (2014).Electrophysiological evidence for a sensory recruit-ment model of somatosensory working memory.Cerebral Cortex. Online publication; doi:10.1093/cercor/bhu153

Khayat, P. S., Spekreijse, H., & Roelfsema, P. R. (2006).Attention lights up new object representations beforethe old ones fade away. Journal of Neuroscience, 26,138–142.

Kiss, M., Grubert, A., & Eimer, M. (2013). Top-downtask sets for combined features: Behavioural and elec-trophysiological evidence for two stages in attentionalobject selection. Attention, Perception, & Psychophysics,75, 216–228.

Kravitz, D. J., Kriegeskorte, N., & Baker, C. I. (2010).High-level visual object representations are con-strained by position. Cerebral Cortex, 20, 2916–2925.

Kravitz, D. J., Saleem, K. S., Baker, C. I., Ungerleider, L.G., & Mishkin, M. (2013). The ventral visualpathway: an expanded neural framework for the pro-cessing of object quality. Trends in Cognitive Sciences,17, 26–49.

Luck, S. J., Chelazzi, L., Hillyard, S. A., & Desimone,R. (1997). Neural mechanisms of spatial selectiveattention in areas V1, V2, and V4 of macaque visualcortex. Journal of Neurophysiology, 77, 24–42.

Luck, S. J., & Hillyard, S. A. (1994). Spatial filteringduring visual search: Evidence from human electro-physiology. Journal of Experimental Psychology:

Human Perception and Performance, 20, 1000–1014.Luck, S. J., & Vogel, E. K. (1997). The capacity of visual

working memory for features and conjunctions.Nature, 390, 279–281.

Luck, S. J., & Vogel, E. K. (2013). Visual workingmemory capacity: From psychophysics and neurobiol-ogy to individual differences. Trends in Cognitive

Sciences, 17, 391–400.

Mangun, G. R., & Hillyard, S. A. (1991). Modulationsof sensory-evoked brain potentials indicate changes inperceptual processing during visual-spatial priming.Journal of Experimental Psychology: Human Perception

and Performance, 17, 1057–1074.Martinez-Trujillo, J. C., & Treue, S. (2004). Feature-

based attention increases the selectivity of populationresponses in primate visual cortex. Current Biology,14, 744–751.

Maunsell, J. H., & Treue, S. (2006). Feature-basedattention in visual cortex. Trends in Neurosciences,29, 317–322.

Mazza, V., & Caramazza, A. (2011). Temporal braindynamics of multiple object processing: the flexibilityof individuation. PLoS One, 6, e17453. doi:10.1371/journal.pone.0017453

Mazza, V., Turatto, M., Umiltà, C., & Eimer, M.(2007). Attentional selection and identification ofvisual objects are reflected by distinct electrophysio-logical responses. Experimental Brain Research, 181,531–536.

McMains, S. A., Fehd, H. M., Emmanouil, T. A., &Kastner, S. (2007). Mechanisms of feature-andspace-based attention: Response modulation andbaseline increases. Journal of Neurophysiology, 98,2110–2121.

Menneer, T., Cave, K. R., & Donnelly, N. (2009). Thecost of search for multiple targets: Effects of practiceand target similarity. Journal of Experimental

Psychology: Applied, 15, 125–139.Moore, T., & Armstrong, K. M. (2003). Selective gating

of visual signals by microstimulation of frontal cortex.Nature, 421, 370–373.

Moore, T., Armstrong, K. M., & Fallah, M. (2003).Visuomotor origins of covert spatial attention.Neuron, 40, 671–683.

Moran, J., & Desimone, R. (1985). Selective attentiongates visual processing in the extrastriate cortex.Science, 229, 782–784.

Müller, H. J., Heller, D., & Ziegler, J. (1995). Visualsearch for singleton feature targets within and acrossfeature dimensions. Perception & Psychophysics, 57,1–17.

Nako, R., Wu, R., & Eimer, M. (2014). Rapid guidanceof visual search by object categories. Journal of


Performance, 40, 50–60.Nako, R., Wu, R., Smith, T. J., & Eimer, M. (2014).

Item and category-based attentional control duringsearch for real-world objects: Can you find thepants among the pans? Journal of Experimental



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

http://dx.doi.org/10.1093/cercor/bhu153

http://dx.doi.org/10.1093/cercor/bhu153

http://dx.doi.org/10.1371/journal.pone.0017453

http://dx.doi.org/10.1371/journal.pone.0017453


Oberauer, K. (2002). Access to information in workingmemory: Exploring the focus of attention. Journal ofExperimental Psychology: Learning, Memory, and

Cognition, 28, 411–421.Olivers, C. N., Meijer, F., & Theeuwes, J. (2006).

Feature-based memory-driven attentional capture:Visual working memory content affects visual atten-tion. Journal of Experimental Psychology: Human

Perception and Performance, 32, 1243–1265.Olivers, C. N. L., & Eimer, M. (2011). On the differ-

ence between working memory and attentional set.Neuropsychologia, 49, 1553–1558.

Olivers, C. N. L., Peters, J., Houtkamp, R., &Roelfsema,P. R. (2011). Different states in visual workingmemory: When it guides attention and when it doesnot. Trends in Cognitive Sciences, 15, 327–334.

Op De Beeck, H., & Vogels, R. (2000). Spatial sensi-tivity of macaque inferior temporal neurons. The

Journal of Comparative Neurology, 426, 505–518.Peelen, M. V., Fei-Fei, L., & Kastner, S. (2009). Neural

mechanisms of rapid natural scene categorization inhuman visual cortex. Nature, 460, 94–97.

Peelen, M. V., & Kastner, S. (2011). A neural basis forreal-world visual search in human occipitotemporalcortex. Proceedings of the National Academy of

Sciences, 108, 12125–12130.Posner, M. I., Snyder, C. R., & Davidson, B. J. (1980).

Attention and the detection of signals. Journal of

Experimental Psychology: General, 109, 160–174.Postle, B. R. (2006). Working memory as an emergent

property of the mind and brain. Neuroscience, 139,23–38.

Ranganath, C., Cohen, M. X., Dam, C., & D’Esposito,M. (2004). Inferior temporal, prefrontal, and hippo-campal contributions to visual working memorymaintenance and associative memory retrieval.Journal of Neuroscience, 24, 3917–3925.

Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992).Temporary suppression of visual processing in anRSVP task: An attentional blink? Journal of


Performance, 18, 849–860.Saenz, M., Buracas, G. T., & Boynton, G. M. (2002).

Global effects of feature-based attention in humanvisual cortex. Nature Neuroscience, 5, 631–632.

Schall, J. D. (2004). On the role of frontal eye field inguiding attention and saccades. Vision Research, 44,1453–1467.

Serences, J. T., & Boynton, G. M. (2007). Feature-basedattentional modulations in the absence of direct visualstimulation. Neuron, 55, 301–312.

Soto, D., Heinke, D., Humphreys, G.W., & Blanco, M.J. (2005). Early, involuntary top-down guidance ofattention from working memory. Journal of


Performance, 31, 248–261.Sreenivasan, K. K., Curtis, C. E., & D’Esposito, M.

(2014). Revisiting the role of persistent neural activityduring working memory. Trends in Cognitive Sciences,18, 82–89.

Stokes, M., Thompson, R., Nobre, A. C., & Duncan, J.(2009). Shape-specific preparatory activity mediatesattention to targets in human visual cortex.Proceedings of the National Academy of Sciences, 106,19569–19574.

Stokes, M. G., Kusunoki, M., Sigala, N., Nili, H.,Gaffan, D., & Duncan, J. (2013). Dynamic codingfor cognitive control in prefrontal cortex. Neuron,78, 364–375.

Stroud, M. J., Menneer, T., Cave, K. R., Donnelly, N.,& Rayner, K. (2011). Search for multiple targets ofdifferent colours: Misguided eye movements reveal areduction of colour selectivity. Applied Cognitive

Psychology, 25, 971–982.Theeuwes, J. (2010). Top–down and bottom–up

control of visual selection. Acta Psychologica, 135,77–99.

Thompson, K. G., & Bichot, N. P. (2005). A visual sal-ience map in the primate frontal eye field. Progress inBrain Research, 147, 249–262.

Todd, J. J., & Marois, R. (2004). Capacity limit of visualshort-term memory in human posterior parietalcortex. Nature, 428, 751–754.

Towler, J., Kelly, M., & Eimer, M. (2015). The focus ofspatial attention determines the number and precisionof face representations in working memory. CerebralCortex. doi:10.1093/cercor/bhv083

Treisman, A. (1988). Features and objects: The four-teenth Bartlett memorial lecture. The Quarterly

Journal of Experimental Psychology, 40, 201–237.Treisman, A., & Gelade, G. (1980). A feature-inte-

gration theory of attention. Cognitive Psychology, 12,97–136.

Treisman, A., & Sato, S. (1990). Conjunction searchrevisited. Journal of Experimental Psychology: Human

Perception and Performance, 16, 459–478.Vogel, E. K., & Machizawa, M. G. (2004).

Neural activity predicts individual differences in


EIMER

Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

http://dx.doi.org/10.1093/cercor/bhv083

visual working memory capacity. Nature, 428,748–751.

Warden, M. R., & Miller, E. K. (2010). Task-dependent changes in short-term memory in theprefrontal cortex. Journal of Neuroscience. 30,15801–15810.

Wheeler, M. E., & Treisman, A. M. (2002). Binding inshort-term visual memory. Journal of Experimental

Psychology: General, 131, 48–64.Wolfe, J. M. (1994). Guided Search 2.0: A revised model

of visual search. Psychonomic Bulletin and Review, 1,202–238.

Wolfe, J. M. (1998). What can 1 million trials tell usabout visual search? Psychological Science, 9, 33–39.

Wolfe, J. M. (2007). Guided search 4.0: Current progresswith a model of visual search. In W. Gray (Ed.),Integrated models of cognitive systems (pp. 99–119).New York: Oxford.

Wolfe, J. M., & Horowitz, T. S. (2004). Opinion: Whatattributes guide the deployment of visual attention

and how do they do it? Nature Reviews Neuroscience,5, 495–501.

Wyble, B., Bowman, H., & Nieuwenstein, M. (2009).The attentional blink provides episodic distinctive-ness: Sparing at a cost. Journal of Experimental


Xu, Y., & Chun, M. M. (2006). Dissociable neuralmechanisms supporting visual short-term memoryfor objects. Nature, 440, 91–95.

Xu, Y., & Chun, M. M. (2009). Selecting and perceivingmultiple visual objects. Trends in Cognitive Sciences,13, 167–174.

Zelinsky, G. J., & Sheinberg, D. L. (1997). Eye move-ments during parallel–serial visual search. Journal ofExperimental Psychology: Human Perception and

Performance, 23, 244–262.Zhang, W., & Luck, S. J. (2009). Feature-based atten-

tion modulates feedforward visual processing.Nature Neuroscience, 12, 24–25.



Dow

nloa

ded

by [

Bir

kbec

k C

olle

ge]

at 0

1:02

13

Nov

embe

r 20

15

eps mid-career award 2014brainb.psyc.bbk.ac.uk/wp-content/uploads/2019/05/eimer...eps mid-career...

Documents