ELSEVIERAUTH

ORPROOF

3.12a0005 Visual PrimingK. Grill-Spector, Stanford University, Stanford, CA, USA

ª 2008 Elsevier Ltd. All rights reserved.

3.12.1 Introduction 13.12.2 Behavioral Aspects of Visual Priming 23.12.3 Neural Correlates of Priming 43.12.3.1 Repetition Suppression 43.12.3.1.1 Characterizing neural representations using repetition suppression 63.12.3.2 Repetition Enhancement 83.12.3.3 Repetition Effects Measured with EEG and MEG 93.12.3.4 Investigations of the Relation between RS and Priming 93.12.3.4.1 Evidence for a correlation between priming and repetition suppression 93.12.3.4.2 Evidence for dissociable effects of performance and repetition on the level of repetition

suppression 103.12.4 Neural Models of Repetition Suppression and Priming 123.12.4.1 Fatigue Model 123.12.4.2 Sharpening Model 133.12.4.3 Facilitation Model 143.12.4.4 Distinguishing the Neural Models 143.12.4.4.1 Examining the relationship between RS and stimulus selectivity 143.12.4.4.2 Examining the effect of repetition on neural tuning 153.12.4.4.3 Examining the temporal window of processing for new and repeated stimuli 153.12.5 Conclusions and Directions for Future Research 15References 16

s0005 3.12.1 Introduction

p0005 A fundamental property of the brain thatdistinguishes it from artificially constructed compu-tational devices is its ability to continuously updateits functional properties based on prior experience.This property, plasticity, is apparent in many formsof learning and memory in humans. One importantmanifestation of plasticity in the brain is priming: thebehavioral phenomenon of improved processing of astimulus following prior experience. Priming typi-cally manifests as increased accuracy and/or fasterspeed in making judgments on a stimulus that hasbeen previously encountered (Figure 1). It is thoughtto reflect an implicit form of memory and learning, asit does not involve explicit memory of the priorexperience.

p0010 This chapter is concerned with visual priming(priming related to presentation of visual stimuli)and the neural correlates underlying this phenom-ena. Visual priming is one of the most ubiquitous

manifestations of priming and has been extensivelystudied in many levels from the behavioral level tothe neural level in both humans and animals. Thus,visual priming is an excellent model to study plas-ticity in the visual system and its relation to objectperception. Studying the neural mechanisms ofvisual priming is important because it enables under-standing the neural bases of cortical representationsas well as mechanisms involved in rapid implicitlearning. In particular, recent interest for understand-ing priming and its neural correlates has beenheightened as an increasing number of scientists usepriming methods to characterize representations inthe human brain.

p0015This chapter is organized into three main sections:it begins with a review of the behavioral aspects ofvisual priming, then examines neuroimaging experi-ments of the neural correlates of priming, andconcludes with a theoretical overview of three mod-els that have been recently suggested for explainingthe neural bases of priming.

LEME 00130

1

ELSEVIERAUTH

ORPROOF

s0010 3.12.2 Behavioral Aspects of VisualPriming

p0020 In a typical priming experiment subjects are shownan initial stimulus (prime) and are required to make adecision (e.g., categorize the stimulus, Figure 1) orproduce a response (generate a word) on a subse-quent stimulus (test) which is identical or related tothe to the initial stimulus (e.g., the same object indifferent views, or a new object that is related per-ceptually, conceptually, or semantically to theprime). The priming effect (i.e., improvement in per-formance) is largest when the repeated stimulus isidentical to the initial stimulus (prime). In manybehavioral paradigms of priming, many interveningstimuli occur between the test and the prime.However, in other paradigms the test immediatelyfollows the prime. One particular striking aspect ofpriming is that it can be manifested after a singleexposure to an object and is preserved in time scalesranging from seconds to even an year (

b0085

Cave, 1997).p0025 The level of priming is modulated by several

factors such as the number of stimulus repetitions,the number of intervening stimuli and the timebetween repeats (Figure 2(a)). The magnitude ofresponse time (RT) priming increases with the num-ber of stimulus repetitions both in short time scales

(seconds/minutes) and in longer time scales (daysand weeks), and this advantage remains over week-long delays compared to single exposures of stimuli(b0050

Brown et al., 1996). Similarly, the priming effect islargest when there are no intervening stimulibetween the prime and the test stimulus and whenthe temporal interval between them is short(Figure 2(a)). Thus, immediate repetitions producea larger priming effect compared to when repetitionsoccur after several minutes or days (

b0525

van Turennoutet al., 2000;

b0445

Sayres and Grill-Spector, 2006).Interestingly, the graded nature of priming is reliableeven in patients who are unable to remember thestimuli or judge the frequency of these stimuli inexplicit tests (

b0550

Wiggs et al., 1997).p0030Another important aspect of visual priming is its

specificity, because visual priming has been used asan experimental tool to infer the characteristics ofobject representations. Visual priming is preservedeven when the appearance of objects changes acrossrepetitions. Visual priming is invariant to changes inobject size (

b0100

Cooper et al., 1992;b0165

Fiser and Biederman,1995), position (

b0100

Cooper et al., 1992), color (b0090

Caveet al., 1996), symmetry (

b0170

Fiser and Biederman, 2001),and to some degree to the viewing angle of the object(b0040

Biederman and Bar, 1999;b0035

Biederman, 2000).However, a recent experiment suggests that there

Priming Repetition suppression

1st Presentation

BrainTaskCategorize:cat or dog?

Performance

2nd Presentation

% B

OLD

% B

OLD

Figure 1f0005AU2 Schematic illustration of priming and repetition suppression/adaptation. Left: subjects are asked to perform atask, e.g., classify an object.Middle: performance is measured on the first presentation and subsequent presentations of thesame stimulus. During repeated presentations (e.g., the second presentation) performance improves (i.e., priming) asindicated by the arrow: typically, accuracy increases and response time decreases. Right: Brain activity in object-selectivecortex measured during the same experimental conditions shows lower BOLD responses for repeated presentations (blue) ofthe object compared to the initial presentation (red).

LEME 00130

2 Visual Priming

ELSEVIERAUTH

ORPROOF

may be an interaction between the effects of shapeand location on visual priming (

b0415

Newell et al., 2005).Further, visual priming is diminished, but still pre-served for new exemplars from the same category(e.g., upright piano vs. grand piano). Experiments inwhich subjects viewed stimuli which were presentedeither in the right or left visual field suggest differ-ential priming effects across the left and righthemispheres: priming effects show higher specificitywhen stimuli are presented to the left visual field, asthey do not generalize across object rotation andexemplars of a category, whereas priming effectsgeneralize across exemplars and rotation (

b0375

Marsolek,1995;

b0070

Burgund and Marsolek, 2000;b0350

Koutstaal et al.,2001;

b0530

Vuilleumier et al., 2002;b0480

Simons et al., 2003)when stimuli are presented to the right visual field.These experiments have led to suggestions that

object representations in the right hemisphere aremore specific than left-hemisphere representations,which may be more abstract in nature (

b0375

Marsolek,1995;

b0070

Burgund and Marsolek, 2000;b0350

Koutstaal et al.,2001;

b0530

Vuilleumier et al., 2002;b0480

Simons et al., 2003).p0035Several lines of research suggest that visual prim-

ing is an implicit form of learning and memory. Aparticularly important finding is that priming occursin amnesic patients even though they are unaware ofprior exposure to the primed stimulus (

b0510

Tulving et al.,1991;

b0245

Hamann and Squire, 1997) and they are sig-nificantly impaired on explicit tests (such asrecognition memory, recall, and recollection of con-textual information) on the same stimuli (

b0245

Hamannand Squire, 1997).

p0040Another striking aspect of priming is subliminalpriming (

b0110

Dehaene et al., 1998;b0020

Bar and Biederman,

750

1.4

1.2

1

0.8% S

igna

l 1.4

1.2

1

0.8% S

igna

l700

650

RT,

ms

RT,

ms

600

550

800

700

600

500

RT,

ms

800

700

600

500

1 2Presentation number

Stimuli between repeats

Time between repeats, seconds Time between repeats, seconds

Stimuli between repeats

Presentation number

Fusiform(a) (b)Behavior fMRI

LO

3 4 5 6 7 81 2 3 4 5 6 7 8

1.4

1.2

1

1

0.8

nrep 0 1–3 4–7 !8nrep 0 1–3 4–7 !8

nrep 2 4–8

10–16

!18 nrep 2 4 –

810–16

! 18 nrep 2 4 –8

10–16

! 18

% S

igna

l

1.4

1.2

0.8% S

igna

l

1

1.4

1.2

0.8%

Sig

nal

1.4

1.2

1

0.8nrep 0 1–3 4–7 !8

% S

igna

l

1 2 3 4 5 6 7 8

Figure 2f0010AU3 Effect of repetition parameters on priming and repetition suppression in object-selective cortex as measured withfMRI. (a) Mean response time. (b) BOLD response amplitude for object-selective regions in the fusiform gyrus (middle) and LO(right). Error bars indicate SEM across eight subjects. BOLD responses are averaged across hemispheres. Asterisks indicatesignificantly lower than first presentation (p< .05). Dashed line: response to the first presentation. Top: sorting by presentationnumber. Middle: sorting by intervening stimuli between repeats. Bottom: sorting by time between repeats. Adapted fromSayres R and Grill-Spector K (2006) Object-selective cortex exhibits performance-independent repetition suppression. J.Neurophysiol. 95: 995–1007.

LEME 00130

Visual Priming 3

ELSEVIERAUTH

ORPROOF

1999;b0410

Naccache and Dehaene, 2001), i.e., primingwithout awareness of the content of the primingstimulus. For example, Bar and colleagues (

b0020

Bar andBiederman, 1999) showed subjects briefly presentedstimuli (average 47ms) which were masked. Subjects’naming performance on these stimuli was low(!14%). However, when the same stimuli were pre-sented for the second time, naming performance onprimed stimuli significantly increased (to about35%). Subliminal priming may show higher specifi-city than suprathreshold priming, as it generalizesonly to objects presented in the same hemifield.Therefore, Bar and colleagues have suggested thatsubliminal priming may be mediated by neuralmechanisms distinct from suprathreshold priming.Other priming experiments of briefly presentedmasked stimuli show that the magnitude of primingis larger for the specific items that were primedcompared to other exemplars of the category(b0185

Furmanski and Engel, 2000;b0225

Grill-Spector et al.,2000), and that priming effects increase across daysand repeated exposures (

b0225

Grill-Spector et al., 2000)and generalize across object size (

b0185

Furmanski andEngel, 2000). Further, experiments of subliminalpriming of words show generalization of primingeffects across fonts and letter size (

b0410

Naccache andDehaene, 2001).

p0045 Overall, evidence from amnesic patients and sub-liminal priming experiments suggests that awarenessmay not be necessary for priming. These experimentslead to the prevailing theory that posits that primingreflects an implicit form of memory which is distinctfrom explicit memory and relies on distinct neuraland cognitive mechanisms.

s0015 3.12.3 Neural Correlates of Priming

p0050 Many studies have investigated the neural correlatesof priming in humans using functional magneticresonance imaging (fMRI), electroencephalography(EEG), and magnetoencephalography (MEG). Underexperimental situations similar to behavioral para-digms of visual priming, the most robust andconsistent finding with fMRI is reduced brain activa-tions to repeated presentations of a stimulus relativeto the initial presentation of that stimulus (Figure 3).This reduction has been referred to as repetitionsuppression, fMRI-adaptation (

b0485

Sobotka and Ringo,1994;

b0435

Ringo, 1996;b0230

Grill-Spector and Malach, 2001),mnemonic filtering (

b0385

Miller et al., 1993), repetitionsuppression (

b0120

Desimone, 1996), decremental

responses (b0055

Brown and Xiang, 1998), and neural prim-ing (

b0370

Maccotta and Buckner, 2004). We will userepetition suppression (RS) to refer to decreasedneural responses following stimulus repetition.However, it remains mysterious how reduced corti-cal responses provide for improved performance.While the most ubiquitous cortical phenomenonrelated to visual priming is reduced responses, thereis evidence also that some aspects of visual primingare related to increased responses, or repetitionenhancement (RE) (

b0130

Dolan et al., 1997;b0195

George et al.,1999;

b0225

Grill-Spector et al., 2000;b0250

Henson et al., 2000;b0300

James et al., 2000;b0295

James and Gauthier, 2006;b0325

Kourtziet al., 2005;

b0515

Turk-Browne et al., 2006). We will con-sider both phenomena and their relation to primingin turn.

s00203.12.3.1 Repetition Suppression

p0055When stimuli are repeated, as in typical primingparadigms, neural activity is usually reduced. Thisneural repetition effect has been reported at multiplespatial scales, from the level of individual corticalneurons in monkeys (

b0355

Li et al., 1993;b0380

Miller andDesimone, 1994;

b0490

Sobotka and Ringo, 1996) to thelevel of hemodynamic changes (measuring thepooled activation of millions of neurons) in humansusing functional imaging such as fMRI (

b0065

Buckneret al., 1995;

b0115

Demb et al., 1995;b0495

Stern et al., 1996;b0220

Grill-Spector et al., 1999;b0250

Henson et al., 2000;b0315

Jianget al., 2000;

b0410

Naccache and Dehaene, 2001).Repetition-related reductions also occur at multipletemporal scales, both in their longevity – from milli-seconds (

b0490

Sobotka and Ringo, 1996) to minutes(b0250

Henson et al., 2000) and days (b0525

van Turennoutet al., 2000) – and in the latency of their expression(b0105

Dale et al., 2000;b0275

Henson et al., 2004). Therefore RSis a robust phenomenon that occurs across manytimes scales, in multiple brain regions, and across animpressively large number of experimentalconditions.

p0060In experiments when subjects view repeated pre-sentations of objects and scenes, there is robust andreproducible RS as measured with fMRI (for reviewssee

b0255

Henson, 2003;b0330

Kourtzi and Grill-Spector, 2005;b0205

Grill-Spector et al., 2006a). RS/fMRI-adaptationtypically occurs in object-selective cortex(Figure 3) including the lateral occipital complex(LOC – consisting of regions overlapping the lateraloccipital sulcus, inferior occipital gyrus, and occi-pito-temporal sulcus) as well as more ventralregions including the fusiform gyrus (Fusiform) and

LEME 00130

4 Visual Priming

ELSEVIERAUTH

ORPROOF

the parahippocampal gyrus (PHG). RS also occurs indorsal regions (Figure 3), including regions lateral toand partially overlapping V3a and regions in theposterior bank of the intraparietal sulcus (IPS).Other regions that show RS to repeated presentationof object and scene images include medial temporalcortex (

b0495

Stern et al., 1996) and frontal cortex (b0535

Wagneret al., 1997;

b0060

Buckner and Koutstaal, 1998;b0520

vanTurennout et al., 2003).

p0065 RS is not an all-or-nothing phenomenon: Themagnitude of RS in object-selective cortex increaseswith repetition number and with fewer interveningstimuli between repetitions (Figure 2(b)). Therefore,the magnitude RS in block-design fMRI experimentsis typically larger than during event-related fMRIexperiments in which many intervening stimulioccur between repetitions of the same image(Figure 4). A recent study (

b0190

Ganel et al., 2006) sug-gests that RS to immediate repetitions of identicalstimuli is more prominent in object-selective regions

of the lateral occipital cortex and fusiform cortex,whereas RS effects for stimuli which had been pre-sented several minutes previously and occur aftermany intervening stimuli, are more prominent inmore anterior and medial regions of the temporallobe. Further, they suggest that effects of immediaterepetition and long-lagged repetition with interven-ing stimuli are largely additive (except for the leftfusiform gyrus).

p0070Repetition suppression in high-level visual areashas been associated with visual priming (

b0450

Schacterand Buckner, 1998;

b0545

Wiggs and Martin, 1998) becauseboth phenomenon occur under the same experimentalconditions (Figure 1). However, it is mysterious whyreduced cortical responses provide for improved beha-vioral performance. Notably, RS measured by fMRImay be related to other factors (unrelated to priming),such as repetition effects independent of behavioralimprovements (

b0445

Sayres and Grill-Spector, 2006), atten-tional differences between conditions (

b0565

Yi and Chun,

Right hemisphere

IPS IPS

5

Non repeated > repeated

6–log10(P )

74 8

Lateral

Ventral

FusiformFusiform

Figure 3f0015AU4 Occipito-temporal regions that show reduced responses to repeated versus non repeated stimuli (p<10"4, voxellevel uncorrected). Color bar indicates statistical significance. Data are shown for a representative subject on her partiallyinflated brain. Dark regions gray regions indicate sulci, and lighter gray regions indicate gyrii. Retinotopic visual areas(delineated by black lines) were defined from independent retinotopic scans of polar angle and eccentricity and are shown forsimplicity only on the right hemisphere. PHG: parahippocampal gyrus; IPS: intraparietal sulcus.

LEME 00130

Visual Priming 5

ELSEVIERAUTH

ORPROOF

2005;b0570

Yi et al., 2006), and/or learning of a responsemapping between a stimulus and a cognitive decision(b0125

Dobbins et al., 2004;b0455

Schacter et al., 2004). Conversely,behavioral effects, such as visual priming, may be aconsequence of activity in multiple cortical regions.Thus, RS in specific cortical regions may not relatedirectly to the behavioral changes associated withpriming. While keeping these caveats in mind, thesection titled ‘Characterizing neural representationsusing repetition suppression’ describes several experi-ments which investigated the relation betweenpriming and RS, and the section titled ‘Neural modelsof repetition suppression and priming’ lays out threemodels of the neural bases of priming, providinghypotheses for the relation between reduced corticalresponses to repeated stimuli and improved perfor-mance – namely, visual priming.

s00253.12.3.1.1 Characterizing neuralrepresentations using repetitionsuppression

p0075In addition to examining the relation between RSand priming, many neuroimaging experiments useRS to probe the functional properties of neural popu-lations. This tool has been termed ‘fMRI-adaptation’(b0230

Grill-Spector and Malach, 2001) and also ‘the prim-ing method’ (

b0410

Naccache and Dehaene, 2001;b0530

Vuilleumier et al., 2002). In the basic paradigm usedin fMRI experiments, one first measures the basic RS(or adaptation) effect induced by repetitions of iden-tical stimuli. This is done by measuring the level ofRS or adaptation to repeated presentations of iden-tical stimuli relative to the response of nonrepeatedstimuli (Figure 5 – identical). Subjects are also pre-sented with repeated stimuli that vary along some

Block Rapid event-related

Non repeatedRepeated

Rapid event-related

LO

Fusiform

Block

1.5

1

0.5

0

2

0.50 5 10 15–5 20

Time (s)

1.5

1

0.5

0

2

0.50 5 10 15–5 20

Time (s)

1.5

1

0.5

0

2

–0.50 5 10 15–5 20

Time (s)

1.5

1

0.5

0

2

–0.50 5 10 15–5 20

Time (s)

% S

igna

l fro

m s

cram

bled

% S

igna

l fro

m s

cram

bled

% S

igna

l fro

m fi

xatio

n%

Sig

nal f

rom

fixa

tion

Figure 4f0020AU5 Repetition suppression in object-selective cortex: time course data. Data are shown for one representativesubject. Object-selective cortex regions were defined as regions that showed higher activation for animals than scrambledanimals with p<10"3 at the voxel level. Black: first presentation of the stimulus; Red: Repeated presentations of the samestimulus. In the block design fMRI experiment stimuli were repeated up to 12 times within a block. In the rapid event-relatedfMRI experiment stimuli were repeated up to eight times across a 4-min and 38-s experiment. Horizontal gray bar: duration inwhich stimulus was presented. Error bars indicate SEM across trials for this subject. Adapted fromGrill-Spector K, Henson R,and Martin A (2006) Repetition and the brain: neural models of stimulus-specific effects. Trends Cogn. Sci. 10: 14–23.

LEME 00130

6 Visual Priming

ELSEVIERAUTH

ORPROOF

dimension (e.g., the same object, but different sizes).The hypothesis tested is whether the underlyingneural representation is sensitive or not to changealong this dimension. If the underlying neural repre-sentation is insensitive to the change in the stimulus,then neurons will show a reduced response torepeated transformed versions of the object, and thefMRI signal will be reduced (i.e., RS/adaptation willbe observed) similar to the reduction produced byrepetitions of identical stimuli. Alternatively, if theneurons are sensitive to the change, the level of thefMRI signal will be similar to the initial level, and noRS/adaptation will be measured.

p0080 An example of using RS/fMRI-adaptation tocharacterize neural representations is shown inFigure 5. In these experiments subjects were shownrepeated presentations of the same image of a face(identical), or images of the same individual thatvaried in size (up to threefold changes in size), posi-tion (!6# around fovea), illumination (five differentilluminations), viewpoint (rotation around the verti-cal, "90# to 90#). Activations to repeated versions ofthe same face were compared to nonrepeated pre-sentations of faces of different individuals that weretaken under the same viewing conditions (e.g., samesize, position, illumination, and view). We founddifferential effects of RS/fMRI-adaptation acrossobject-selective cortex: LO regions show RS/fMRI-adaptation for repetitions of identical images of

objects, but no RS/fMRI-adaptation when the objectvaried in position, size, illumination, or viewingangle. In contrast, more ventral regions along thefusiform and OTS showed RS/fMRI-adaptation forchanges in object position and size, but no RS/adap-tation for different illuminations or rotations of thesame object. These experiments provide evidence fordifferential sensitivity to object transformationsacross the human ventral stream.

p0085RS/fMRI-adaptation has been widely used byresearchers to examine selectivity in object-selectivecortex to object size and position (

b0220

Grill-Spector et al.,1999;

b0530

Vuilleumier et al., 2002), viewpoint (b0220

Grill-Spector et al., 1999;

b0305

James et al., 2002;b0530

Vuilleumieret al., 2002;

b0160

Epstein et al., 2003), object format(b0335

Kourtzi and Kanwisher, 2000), perceived shape(b0340

Kourtzi and Kanwisher, 2001), contrast (b0010

Avidanet al., 2002;

b0400

Murray and He, 2006), contour comple-tion (

b0345

Kourtzi et al., 2003), and face representation(b0220

Grill-Spector et al., 1999;b0005

Andrews and Ewbank,2004;

b0555

Winston et al., 2004;b0150

Eger et al., 2005;b0360

Loffleret al., 2005;

b0440

Rotshtein et al., 2005;b0310

Jiang et al., 2006).RS has also been used to probe higher-level concep-tual representations using object pictures (

b0350

Koutstaalet al., 2001) and words (

b0540

Wheatley et al., 2005).Overall, these studies have documented that RS inoccipito-temporal cortex is not limited to the iden-tical image, but also occurs, albeit to a lesser extent,to transformed versions of the same object, to

0.40

0.60

0.80

1.00

Sen

sitiv

ity to

tran

sfor

mat

ion

LO 0.64 0.92 0.88 0.85 1.01

Fusiform

Identical Position Size Illumin Rotation

*

** *

Ada

ptat

ion

ratio

0.880.870.680.690.48

Figure 5f0025AU6 Using RS to measure sensitivity to face transformations. To examine the level of adaptation to facetransformations, we measured the adaptation ratio: repeated/nonrepeated. Repeated reflects blocks in which the same facewas repeated, and nonrepeated reflects blocks that included nonrepeated presentations of different male faces in the sameview, illumination, size, and position. An adaptation ratio of 1 indicates no adaptation (solid line); Ratios that are significantlyless than 1 are indicated by asterisks. Identical: repetitions of identical images of the same face; Position: repetitions of thesame face in different retinal positions (!6# around fixation); Size: repetitions of the same face in different size (!threefold sizechange); Illumin: repetition of the same face illuminated from different directions; Rotation: Same face at different rotationsaround the vertical. Gray bars: LO; Black bars: Fusiform; Error bars indicate SEM across 14 subjects. Numbers indicateadaptation ratios for each of the conditions. Data from Grill-Spector K and Malach R (2001)AU7 fMR-adaptation: a tool forstudying the functional properties of human cortical neurons. Acta Psychol. (Amst.) 107: 293–321.

LEME 00130

Visual Priming 7

ELSEVIERAUTH

ORPROOF

different exemplars that share the same name (e.g.,two different umbrellas), and even to different wordsthat are conceptually-related (

b0540

Wheatley et al., 2005).

s0030 3.12.3.2 Repetition Enhancement

p0090 While the most ubiquitous cortical phenomenon isrepetition suppression, there is evidence also thatsome aspects of visual priming are related to repeti-tion enhancement (RE) (

b0130

Dolan et al., 1997;b0195

Georgeet al., 1999;

b0225

Grill-Spector et al., 2000;b0250

Henson et al.,2000;

b0300

James et al., 2000;b0295

James and Gauthier, 2006;b0325

Kourtzi et al., 2005;b0515

Turk-Browne et al., 2006). REeffects have been reported for improved recognitionof repeated degraded stimuli (compared to perfor-mance on their initial presentation). Repetitionenhancement was observed when repeated exposureto subthreshold, briefly presented objects led to bet-ter recognition (

b0225

Grill-Spector et al., 2000) (Figure 6),when repeated exposure to unfamiliar shapes madethem familiar (

b0250

Henson et al., 2000), when invertedcontrast faces (that were initially unrecognizable)were primed with positive contrast faces and becamerecognizable (

b0195

George et al., 1999), and when obser-vers learned to detect low-salience shapes in noisy

backgrounds (b0325

Kourtzi et al., 2005;b0515

Turk-Browneet al., 2006).

p0095One possibility is that repetition enhancement andrepetition suppression reflect dissociable forms ofvisual priming (

b0250

Henson et al., 2000;b0240

Gruber andMuller, 2005); thus, repetition suppression mayreflect suprathreshold priming, and repetitionenhancement may reflect subliminal priming (seealso

b0325

Kourtzi et al., 2005). These findings suggestthat learning of unfamiliar or degraded stimuli ismediated by increased neural activity across high-level visual areas as new representations are formedfor these previously unseen or unfamiliar stimuli. Incontrast, learning of prominent suprathreshold (and/or familiar) stimuli is mediated by the sharpening ofexisting representations, leading to sparser coding ofobjects. An alternative account suggests that there isonly one underlying mechanism, but it producesdifferential signals below and above recognitionthreshold (

b0300

James et al., 2000;b0295

James and Gauthier,2006). James and colleagues (

b0295

James and Gauthier,2006) proposed an accumulation model for recogni-tion, in which recognition occurs when sufficientevidence for identifying an object has accumulated(See 00178). Accumulation predicts a faster rise ofactivity for primed compared to unprimed stimuli

100

(b)(a)

n = 11 n = 11

60

20

100

Nor

mal

ized

act

ivat

ion

Nor

mal

ized

act

ivat

ion

60

20CoS Fus LO V1 CoS Fus LO V1

o"

40 ms pre-train40 ms post-train

40 ms novel40 ms post-train

Figure 6f0030AU8 Repetition enhancement in object-selective cortex. Subjects were first scanned in the pretrained session whenthey viewed briefly presented images for 40ms followed by a 460-ms mask. Subjects then viewed these masked imagesduring five training sessions, and then they were scanned again in a second post-training scan in which they viewed the samebriefly presented stimuli (posttrain) and another set of novel masked images shown for the same duration (40ms – novel).Initial recognition performance (40ms pretrain) was about 25% and posttraining about 60%. Activations are plotted aspercentage of activation to object stimuli shown for 120ms followed by a mask for 380ms. Recognition of 120ms stimuli wasclose to ceiling. (a) Normalized BOLD signal elicited by identical images viewed for 40ms during pretraining and posttrainingscans. X-axis, brain area; Y-axis, normalized signal compared to activation elicited by stimuli presented for 120ms. Darkbars: pretraining; Gray bars: identical stimuli posttraining. Error bars: SEM. Asterisks (p < .001) and circles (p < .03):significantly larger signal posttraining versus pretraining. (b) Average signal in posttraining scan for two sets of images (trainedand novel) shown for 40ms. White bar: novel images; Gray bars: trained images; Error bars: SEM. Square (p < .01) and circle(p < .05): significantly stronger signal for trained images compared to novel images. Abbreviations: CoS: collateral sulcus; Fus:Fusiform gyrus; LO: lateral occipital. Adapted from Grill-Spector K, Kushnir T, Hendler T, and Malach R (2000) The dynamicsof object-selective activation correlate with recognition performance in humans. Nat. Neurosci. 3: 837–843.

LEME 00130

8 Visual Priming

ELSEVIERAUTH

ORPROOF

as the evidence accumulates faster for primed stimuli.Another consideration for interpreting RE effects isthat, at close to threshold, recognition performanceon trials in which subjects recognize objects (hits) iscorrelated with a higher signal than trials in whichsubjects fail to recognize objects (misses) (

b0025

Bar et al.,2001;

b0200

Grill-Spector, 2003;b0215

Grill-Spector et al., 2004).Thus, RE associated with improved recognition ofpreviously unrecognized stimuli could reflect ahigher hit rate for repeated stimuli that is not relatedto priming per se.

p0100 While RE has been suggested to reflect a distinctform of priming than RS, to date there is no para-metric study that systematically varied factors thataffect RE and visual priming to test whether RE andvisual priming are quantitatively linked.

s0035 3.12.3.3 Repetition Effects Measured withEEG and MEG

p0105 Repetition effects have also been studied by measur-ing changes in the electrical (EEG) or magnetic(MEG) field, usually recorded above the scalp.These effects reflect changes in the amplitude and/or synchrony of local field potentials (LFPs) causedby transmembrane currents in large numbers ofneurons.

p0110 Most EEG studies examine event-related poten-tials (ERPs), which reflect changes in electricalpotential during the few hundred milliseconds fol-lowing stimulus onset, averaged across trials. Theearliest object repetition effects are typicallyobserved approximately 200ms after stimulus onset(b0155

Eimer, 2000;b0135

Doniger et al., 2001;b0460

Schendan andKutas, 2003;

b0275

Henson et al., 2004;b0470

Schweinbergeret al., 2004;

b0240

Gruber and Muller, 2005), but someexperiments show earlier repetition effects 150–170ms after stimulus onset (

b0075

Campanella et al.,2000).

b0275

Henson et al. (2004) found effects of repeatingthe same viewAU9 of an object as early as 160–190mswhen there were no intervening objects; with one ormore intervening objects, repetition effects onlyemerged from approximately 200ms onward.

p0115 Other EEG studies concentrate on changes in thepower of electrical or magnetic oscillations that areinduced by stimulus repetition (high-frequencyoscillations are not observed in ERPs if they are notphase-locked across trials). Some studies reportdecreased high frequency (>40Hz) power around220–350ms for repetition of familiar objects acrosslags of one or two intervening objects (

b0240

Gruber andMuller, 2005). Such changes in power in certain

frequency bands have been shown to correlate withthe blood-oxygen-level dependent (BOLD) changesmeasured by fMRI (

b0045

Brookes et al., 2005;b0445

Sayres andGrill-Spector, 2006).

s00403.12.3.4 Investigations of the Relationbetween RS and Priming

p0120One useful approach to examine the relation betweenpriming and RS is to parametrically manipulate fac-tors that influence the level of priming and/or RSand to examine whether the modulation of primingand RS effects covary or follow distinct profiles.Researchers have shown that many factors modulatethe level of priming and also the level of RS. Theseinclude the number of stimulus repetitions (

b0260

Hensonet al., 2003;

b0445

Sayres and Grill-Spector, 2006), fre-quency of repetition (

b0445

Sayres and Grill-Spector,2006), duration of stimulus presentation (

b0575

Zago et al.,2005;

b0445

Sayres and Grill-Spector, 2006), stimulus con-trast (

b0010

Avidan et al., 2002), amount of noise added tothe stimulus (

b0515

Turk-Browne et al., 2006), as well ashigh-level cognitive factors such as attention (

b0145

Egeret al., 2004;

b0405

Murray and Wojciulik, 2004;b0565

Yi andChun, 2005;

b0570

Yi et al., 2006), relevance to the task(b0280

Henson et al., 2002b; but seeb0315

Jiang et al., 2000),familiarity (

b0250

Henson et al., 2000) and emotion (b0290

Ishaiet al., 2004,

b0285

2006).p0125Recently, researchers have used this approach to

examine the correlation between RS and priming(b0125

Dobbins et al., 2004;b0370

Maccotta and Buckner, 2004;b0575

Zago et al., 2005;b0445

Sayres and Grill-Spector, 2006).Results of these experiments are mixed. Someexperiments suggest that some factors modulate thelevel of priming and RS in a similar way, suggesting aquantitative relation between priming and RS(b0370

Maccotta and Buckner, 2004;b0575

Zago et al., 2005).However, other factors may affect priming, but notRS (or RS but not priming), suggesting that RS inspecific occipito-temporal regions may not be adirect neural correlate of visual priming (

b0260

Hensonet al., 2003;

b0125

Dobbins et al., 2004;b0445

Sayres and Grill-Spector, 2006).

s00453.12.3.4.1 Evidence for a correlationbetween priming and repetitionsuppression

p0130Several recent studies have reported a correlationbetween the level of response time priming andRS in prefrontal cortex (PFC) (

b0125

Dobbins et al.,2004;

b0365

Lustig and Buckner, 2004;b0370

Maccotta andBuckner, 2004).

b0365

Lusting and Buckner (2004) report

LEME 00130

Visual Priming 9

ELSEVIERAUTH

ORPROOF

across-subject correlation between the level of prim-ing during a meaning-based word classification taskand PFC activity across young adults, older adults,and adults with initial signs of Alzheimer’s disease(b0365

Lustig and Buckner, 2004). Similarly,b0370

Maccota andBuckner (2004) showed that the level of priming as afunction of number repetitions of a word and thelevel of RS in PFC was correlated across subjects.Dobbins and colleagues (

b0125

Dobbins et al., 2004) alsoreported a positive within-subject correlationbetween response time priming induced by repeatedpresentations of visually presented objects and RS inPFC. In the same study they reported a negativewithin-subject correlation between priming and RSin the left fusiform cortex.

p0135 Zago and colleagues (b0575

Zago et al., 2005) showedthat both visual priming and RS vary as a function ofthe initial exposure of the stimulus which variedbetween 40 and 1900ms. The largest priming andRS effects were observed for 250-ms exposure dura-tions. Further, the average level of priming acrosssubjects is correlated with the average level of RSin occipito-temporal object-selective regions.However, Zago and colleagues do not reportwithin-subject correlations between visual primingand RS.

p0140 Together, these studies suggest that under somecircumstances there is a correlation between primingand RS. The most consistently reported corticalregion in which activation is associated with primingis PFC. Further research is necessary to examine thegenerality of these results to additional stimulusmanipulations to understand whether these effectsare modality-specific and investigate more compre-hensively whether the correlation between primingand RS can be found within individual subjects, asresponses and brain activations are likely to varyacross individuals.

s0050 3.12.3.4.2 Evidence for dissociableeffects of performance and repetition onthe level of repetition suppression

p0145 Priming effects can be reduced when the responses toa stimulus are changed across repetitions. A recentstudy examined whether changes in priming effectsand RS effects were dependent on the particularresponse/judgment made about the stimulus.Dobbins and colleagues (

b0125

Dobbins et al., 2004) mea-sured priming and RS effects as a function ofrepetition number under two experimental condi-tions – when the same judgment was made on thestimulus (indicate whether an object is larger than a

shoebox, for both initial and repeated presentations)and when different judgments were made (initiallysubjects judged whether the item was larger than ashoebox, and when the item was repeated they wereasked if it was smaller than a shoebox). Primingeffects were larger when the question was identicalin the initial and repeated conditions. However,priming was observed even when the judgment dif-fered. RS in PFC was correlated with priming effectsin both experimental conditions. In contrast, RS infusiform cortex was observed when items wererepeated and the judgment was identical, but wasabolished when the judgment changed. These datasuggest that RS in fusiform cortex was related to theability of the subject to use prior responses duringrepeated trials, rather than reflecting a priming effect.

p0150Another consideration when relating BOLDresponses to performance is that several factors maycontribute to BOLD responses measured in a specificbrain region, and performance is likely to be an out-come of activation across several brain regions.Therefore, while RS and priming occur in the sameexperimental situations, the two phenomenon maynot coincide under all experimental conditions. Forexample, RS may be driven by shorter RT (e.g.,shorter RT may produce lower BOLD responses)but not repetition. This alternative predicts lowerBOLD responses for trials with shorter RT regardlessof whether these trials contain stimuli shown for thefirst time, or contain stimuli which have been seenpreviously. Alternatively, repetition may producelower responses independent of performancechanges. This alternative suggests that repeated trialswill correspond to trials with lower BOLD response,even if there is no change in performance – i.e., amemory component to RS that is independent ofperformance.

p0155Another question is whether RS and visual prim-ing reflect changes in neural activity during therecognition process, or whether they reflect changesafter recognition has occurred. Under many experi-mental conditions the stimulus is presented forlonger periods than the minimal time necessary forrecognition (

b0210

Grill-Spector and Kanwisher, 2005)(!67–120ms), yet both performance and BOLD sig-nals are measured after recognition has occurred.

p0160In a recent study (b0445

Sayres and Grill-Spector, 2006),we examined both factors: Is RS in object-selectivecortex correlated with response time priming orrepetition independent of performance changes?Second, do RS and priming occur during or afterrecognition? To quantitatively examine the relation

LEME 00130

10 Visual Priming

ELSEVIERAUTH

ORPROOF

between priming and RS, we first measured the rela-tion between priming and RS in object-selectivecortex as we parametrically varied repetition para-meters (number of repetitions, number of interveningstimuli, time between repeats). We then measured ifthere are independent contributions of RT and repe-tition to RS effects. Second, to assess whether RS andpriming occurred during or after recognition, weexamined RS and priming effects when objects werepresented briefly (and masked for the remainingduration of a 2-s trial). Exposure durations were setfor each subject to be the minimal duration for!85%accuracy performance on a classification task (67–101ms for our subjects). We then compared RS andpriming effects for short exposures (67–101ms) andlong exposure durations (1750ms).

p0165 We found that both RT priming and RS occurredfor both long and short exposures, and these effectswere modulated by repetition parameters (Figure 2).The level of RS also varied with stimulus duration, asthe magnitude of RS was lower for short compared tolong exposure durations. In contrast, the magnitudeof RT priming was not significantly different acrossthese exposure durations. We did find an improve-ment in accuracy for short exposures, but not longexposures (perhaps due to a ceiling effect), but thisdid not depend on repetition parameters.

p0170 Importantly, we found that when exposure dura-tions were long (1750ms) there was significantcorrelation between RS in object-selective cortexand RT priming for some repetition parameters

(stimuli between repeats, time between repeats) butnot others (number of repeats). When exposures wereshort (67–101ms) we observed significant primingand significant RS, but they were not correlated.Thus, both priming and RS can occur under thesame experimental conditions, but they do notalways covary. These data suggest that RS inobject-selective cortex may not reflect improvedRT performance observed during priming.

p0175Finally, we examined whether there are separablecontributions of response time and repetition to RSby sorting our data into repeated and nonrepeatedconditions. For each condition, we ranked each sub-ject’s trials according to response time and groupedthe trials into four equally sized bins according toRT. Responses to repeated trials were lower thannonrepeated trials even when response times wereequated between conditions (Figure 7). Importantly,for both long and short image durations and allobject-selective regions of interest (ROIs), we founda significant effect of repetition independent ofresponse time. In contrast, we did not find significanteffects of response time independent of repetition.There was a weak, statistically significant effect pre-sent in LO and only for stimuli that were presentedfor 1750ms. Finally, we found no significant interac-tion between repetition and response times. Takentogether, these data reveal that RS in object-selectivecortex reflects stimulus-specific repetition, evenwhen performance is matched between repeatedand nonrepeated objects and when stimuli are

1.5

1.4

1.2

1

0.8

% S

igna

l

1

0.5

2

0800

Fusiform LO

400 1000600 800400 1000600

Nonrepeated Repeated

% S

igna

l

RT, ms RT, ms

Figure 7f0035AU10 Separable contributions of repetition and response-time. Data were sorted first into repeated (solid gray) andnonrepeated (dashed black) correct trials and then grouped into four response time bins for each subject. The first responsetime bin represents fastest quartile of correct trials for each subject. Error bars indicate SEM across seven subjects. Adaptedfrom Sayres R and Grill-Spector K (2006) Object-selective cortex exhibits performance-independent repetition suppression.J. Neurophysiol. 95: 995–1007.

LEME 00130

Visual Priming 11

ELSEVIERAUTH

ORPROOF

presented close to the minimum time required forrecognition. This suggests that there is a perfor-mance-independent component to RS in object-selective cortex which may be an implicit form ofmemory.

p0180 Overall, our experiments show that both primingand RS effects depend on repetition parameters.However, different factors have dissociable effectson priming and RS. Further, these experimentsunderscore the importance of conducting futureexperiments using parametric designs, systematicallyvarying factors that modulate priming and RS toquantitatively measure the relation between thesephenomena.

s0055 3.12.4 Neural Models of RepetitionSuppression and Priming

p0185 Three models have been suggested for the neuralmechanisms underlying repetition suppression thatmay explain priming effects (

b0205

Grill-Spector et al.,

2006a) (Figure 8): (1) the Fatigue model, wherebythe amplitude of firing of stimulus-responsive neu-rons decreases (

b0380

Miller and Desimone, 1994;b0230

Grill-Spector and Malach, 2001), (2) the Sharpeningmodel, whereby fewer neurons respond (

b0355

Li et al.,1993;

b0120

Desimone, 1996;b0545

Wiggs and Martin, 1998),and (3) the Facilitation model, whereby the latency(

b0295

James and Gauthier, 2006) and/or duration ofneural activity is shortened (

b0490

Sobotka and Ringo,1996;

b0270

Henson and Rugg, 2003). An important con-sideration to keep in mind is how each of thesemodels increases processing efficiency.

s00603.12.4.1 Fatigue Model

p0190According to this model, all neurons initially respon-sive to a stimulus show a proportionally equivalentreduction in their response to repeated presentationsof the same stimulus. As a consequence the meanpopulation firing rate declines, but there are nochanges in the pattern of relative responses acrossneurons, or in the temporal window in which neurons

Firing rate

Spi

kes

Spi

kes

Time

Time Time Time

max

BOLD

0

maxBOLD

0

Fatigue modelLess overall activation

Sharpening modelLess neurons

Repeated presentation

First presentation

Facilitation modelLess processing time

Figure 8f0040AU11 Models for repetition suppression. Left: Neural responses. The top panel indicates neural responses in a putativebrain region to the initial presentation of a stimulus. The bottom panels indicate responses in this region to repeated stimuli asposited by each of the three models. Circles indicate the peak spiking rate of neurons in this region (darker circles indicatehigher spiking rates). Blue time courses illustrate the spiking rate as a function of time for the most responsive neurons(indicated by black circles). Right: BOLD response in this region of cortex. Since the BOLD signal integrates neuronal activityover time, all of these models predict reduced BOLD responses for repeated stimuli, but due to different reasons: Fatigue:lower firing rates across the entire time window; Sharpening: fewer neurons respond, but those which remain active showsimilar spiking as for the first presentation; Facilitation: shorter duration of neural processing, but the peak firing is similar tothe initial presentation. Adapted from Grill-Spector K, Henson R, and Martin A (2006) Repetition and the brain: neural modelsof stimulus-specific effects. Trends Cogn. Sci. 10: 14–23.

LEME 00130

12 Visual Priming

ELSEVIERAUTH

ORPROOF

are responding. One mechanism for fatigue may befiring-rate adaptation, in which the reduction in aneuron’s firing rate is proportional to its initialresponse (

b0355

Li et al., 1993;b0010

Avidan et al., 2002) (similarto a gain mechanism (

b0080

Carandini and Ferster, 1997)).However, this mechanism does not explain the spe-cificity of RS, i.e., why the neuron’s response isreduced to some stimuli, yet resumes high firingrates to other stimuli. An alternative mechanism isreduced synaptic efficacy of specific synapses fromconnected neurons (synaptic depression). In thismanner, only specific patterns of presynaptic inputto the neuron (which are stimulus dependent) reduceits firing rate. This type of mechanism has beenimplicated in early visual cortex and usually occurswith prolonged repetitive stimulation.

p0195 One prediction from this model is that theamount of RS will be greater in neurons that respondoptimally to a stimulus than in other neurons. Asa result, the sensitivity of the system to stimulithat are different from the repeating stimulus isincreased, thereby providing a mechanism for‘novelty detection.’ Reducing the firing rate mayalso help prevent saturation of the neural responsefunction by increasing its dynamic range. Anotheradvantage hypothesized for such a mechanism isthat it reduces redundancies in the neural code,which increases the efficiency of information encod-ing (

b0395

Muller et al., 1999).p0200 However, it is not immediately clear how reduced

firing rates can account for increased speed andaccuracy of processing repeated stimuli (in additionto increased sensitivity to novel stimuli), as arises inpriming. One explanation is provided in a computa-tional model by Gotts (2002)AU12 , in which a reduction inthe mean and variance of firing rates allows forgreater synchrony of neural responses. Since greatersynchrony of presynaptic input is believed to bemore effective in triggering a postsynaptic response(b0175

Fries et al., 2001), this would imply more rapidtransmission of information throughout the network,resulting in faster responses (priming). A key predic-tion of this model is that synchrony can increasewhile stimulus-specific firing rates decrease. Whilethere is evidence in support of this possibility(b0500

Stopfer and Laurent, 1999), others have argued forthe opposite effect (

b0095

Chawla et al., 1999). An increasein synchrony may also be difficult to reconcile withobservations of reduced oscillatory power followingstimulus repetition (

b0240

Gruber and Muller, 2005),though it is possible that decreased amplitude of

local field potentials outweighs the increased syn-chrony of those potentials.

s00653.12.4.2 Sharpening Model

p0205b0120

Desimone (1996) andb0545

Wiggs and Martin (1998) havesuggested that repetition results in a sparser repre-sentation of stimuli. According to this model, some,but not all neurons that initially responded to astimulus will show RS to subsequent presentation ofthat stimulus. Importantly, it is the neurons that codefeatures irrelevant to identification of a stimulus thatexhibit RS (Figure 8). Thus, repetition-relatedchanges are viewed primarily as a learning process,in which representations (e.g., tuning curves) are‘sharpened,’ and as a consequence, the distributedrepresentation becomes sparser, resulting in fewerresponsive neurons in total. An important differencebetween the Sharpening and Fatigue models is that,for Sharpening, many of the neurons that are opti-mally tuned to the repeating stimulus should showlittle or no response reduction, rather than exhibit thegreatest response reduction, as in the Fatigue model.

p0210Sparser representations clearly have adaptivevalue in terms of a reduced metabolic cost. Also,because the representation becomes sharper (tuningcurves become narrower), the neurons become moresensitive to change. Sparser representations may alsoallow for more efficient or faster processing, thoughthis depends on the manner in which their informa-tion is read out by downstream neurons. Because theSharpening model suggests a changed and improvedrepresentation for repeated stimuli, this model hasbeen widely used to explain priming (

b0545

Wiggs andMartin, 1998;

b0270

Henson and Rugg, 2003;b0575

Zago et al.,2005). However, a recent study suggests that RS inobject-selective cortex may reflect response-learn-ing, and implies that object representations do notnecessarily reorganize as a consequence of repetition(b0125

Dobbins et al., 2004).p0215The mechanism underlying the formation of spar-

ser representations is unknown, but could reflectinhibition from lateral connections between neuronswithin a population. For example,

b0420

Norman andO’Reilly (2003) used a competitive Hebbian learningrule to simulate the sharpening of representationswith repetition (within medial temporal cortex).Initially, many neurons respond weakly to a distrib-uted input pattern representing the stimulus.Through competitive interactions, the neurons withthe strongest initial response get ‘stronger’ and inhibitthe ‘weaker’ neurons. Thus some neurons show

LEME 00130

Visual Priming 13

ELSEVIERAUTH

ORPROOF

increased firing rates following repetition, whereasothers show decreased firing rates. By assuming thatthe number of ‘strong’ units is less than the number of‘weak’ units, the population response decreases withrepetition because there are more neurons showingreduced activity than showing increased activity. Ifinformation only from those neurons with high firingrates is ‘read out’ by downstream neurons, theirincreased firing rate following repetition (despitethe global decrease) could explain the faster proces-sing of repeated stimuli.

s0070 3.12.4.3 Facilitation Model

p0220 At its simplest, this model predicts that repetitioncauses faster processing of stimuli, i.e., shorter laten-cies or shorter durations of neural firing, and therebymay explain faster response times observed duringpriming. One example is the ‘accumulation’ model ofb0295

James and Gauthier (2006), in which stimulus infor-mation is accrued faster following repetition. Giventhat the hemodynamic signal measured by fMRIintegrates over several seconds of neural activity, ashorter duration of activity results in a decreasedamplitude of the fMRI signal. A shorter duration ofneural activity is also consistent with earlier peakingof the fMRI response (

b0265

Henson et al., 2002a) andmight explain why decreases in firing rate can appearto arise after the initial visual response (

b0435

Ringo, 1996):the neurons initially fire robustly to both first andrepeated presentations, but this firing stops sooner forrepeated presentations.

p0225 An extension of the Facilitation model assumesthat the cause of this faster processing is synapticpotentiation between neurons following an initialstimulus presentation, and that this potentiation canoccur at many levels in the processing stream. As aconsequence, information flows through the streammore rapidly, and hence identification of a repeatedstimulus occurs faster. In terms of attractor neuralnetwork models, synaptic potentiation might beviewed as deepening the basin of attraction, resultingin a shorter time for the network to settle on a stablepattern corresponding to identification of the stimu-lus. An example of such a dynamical, network modelis sketched by

b0180

Friston (2005). The key idea behindthis model is that the firing rate of the long-rangeexcitatory (output) neurons in a population codes‘prediction error’ (

b0430

Rao and Ballard, 1999), which isthe difference between bottom-up input (‘evidence’)and top-down input (‘prediction’). The dynamics ofthe network are such that prediction error decreases

over time after stimulus onset, and synaptic changesserve to accelerate this decrease when the stimulus isrepeated (i.e., repetition improves prediction).

p0230This emphasis on recurrent activity betweenmany levels of the processing stream is consistentwith the spatiotemporal pattern of repetition effectsemerging from MEG/EEG data. Moreover, if inter-regional interactions require an initial volley ofactivity through the network (

b0505

Sugase et al., 1999),this model could further explain the relatively lateonset of long-lag repetition effects recorded withEEG/MEG. However, such a model would notexplain why much earlier repetition effects havebeen observed in some neurons (e.g., 75–100ms,which is thought to be too early for feedback (

b0560

Xiangand Brown, 1998)), and further, this model does notnecessarily predict decreases in the peak firing rate ofindividual neurons.

p0235Each of the above models would clearly benefitfrom further elaboration, including instantiation asdetailed computational models. It is possible thatdifferent models may apply in different brain regionsand under different experimental conditions (e.g.,different paradigms/tasks). Nevertheless, specificneural mechanisms matter, because interpretationand design of experiments depend on the nature ofthe underlying neural model. For example, modelsdiffer as to whether RS reflects quantitative or qua-litative changes in representations. One importantpossibility is that there are multiple models thatvary in their relevance across space, time, and task,which may parallel the multiplicity of potentialneural/synaptic mechanisms. Finally, it is yetunknown whether the same or different mechanismsoperate in different brain regions.

s00753.12.4.4 Distinguishing the Neural Models

p0240There are three main directions in which these mod-els can be distinguished: (1) examining therelationship between RS and stimulus selectivity,(2) examining the effect of repetition on the tuningof cortical responses along a stimulus dimension, and(3) examining the temporal window in which proces-sing occurs for new and repeated stimuli.

s00803.12.4.4.1 Examining the relationshipbetween RS and stimulus selectivity

p0245The models differ in their predictions on whether RSis strongest for the preferred stimulus, or for nonpre-ferred stimuli. The Sharpening model predicts thatneurons showing little or no RS to a repeated

LEME 00130

14 Visual Priming

ELSEVIERAUTH

ORPROOF

stimulus are highly selective for that stimulus. Incontrast, both the Fatigue and Facilitation modelspredict that RS is proportional to the initial response.Thus, neurons that respond optimally for a stimulusshould show the largest suppression. These hypoth-eses can be tested with single-cell recording.

s0085 3.12.4.4.2 Examining the effect ofrepetition on neural tuning

p0250 Another way to distinguish the models would be tofind a single dimension (e.g., motion, orientation),along which stimuli differ, and examine the effect ofrepetition on the tuning curves of different neuronsalong that dimension. The models differ in theirprediction of how repetition will change the shapeof neuronal tuning. According to the Fatigue model,repetition reduces the response in proportion to theinitial response, but the tuning width does notchange. Most likely, the reduction will be maximalfor tuning curves centered on the location of therepeating stimulus along the stimulus dimension,and lesser for tuning curves centered further away.This is consistent with physiological data in V1 fororientation (

b0140

Dragoi et al., 2002), spatial frequency(b0390

Movshon and Lennie, 1979), and motion direction(b0320

Kohn and Movshon, 2004). In contrast, according tothe Sharpening model, repetition sharpens tuningcurves. This is consistent with studies of learning-related changes in IT cortex and V4 (

b0015

Baker et al.,2002;

b0475

Sigala and Logothetis, 2002;b0425

Rainer et al., 2004).Finally, the Facilitation model does not suggest anyparticular effect on tuning curves. Indeed, even awidening of the curves might be possible if repetitionenlarged the attractor basin in an attractor networkmodel.

s0090 3.12.4.4.3 Examining the temporalwindow of processing for new andrepeated stimuli

p0255 The models may also be distinguished in thetemporal domain. In particular, the Facilitationmodel suggests that the latency and/or durationof the response to repeated items will beshorter than to first presentations. The Fatigue andSharpening models do not suggest a difference in thetemporal processing window for repeated stimuli.The latency and duration of processing might beexamined via single-cell recordings and/or EEG/MEG techniques.

s00953.12.5 Conclusions and Directionsfor Future Research

p0260Visual priming is one of the most-studied cognitiveprocesses, as it is a window to understanding theunderlying representations and mechanisms of rapidimplicit learning and memory in the human brain.Progress has been made using priming to infer thenature of representations in different cortical regions,or as a marker for increased processing efficiency,without a complete understanding of its neural basis.Nevertheless, specific neural mechanisms matter,because interpretation of experimental data dependson the nature of the underlying neural mechanisms.

p0265Clearly, many questions remain regarding theneural basis of visual priming. For example, how dothe fatigue and sharpening models account forimproved performance during priming? Do differentmechanisms occur in different time scales (e.g.,immediate priming vs. priming with many interven-ing stimuli)? Do fundamentally different mechanismsunderlie RE and RS? How does the specificity ofpriming correlate with particular representations indifferent visual regions? Are there different mechan-isms underlying subliminal and suprathreshold visualpriming?

p0270Notably, any empirical data relevant to the mod-els presented here is likely to depend on otherfactors, such as the lag between repetitions. One ofthe central outstanding questions is whether differentmodels apply at different time scales. One possibilityis that the mechanisms related to the Fatigue modeloperate during immediate repetitions of a stimuluswithin a few hundred milliseconds and reflect tran-sient stimulus-specific effects that onset rapidly,whereas the effects of repetition across many inter-vening stimuli may be more consistent with theSharpening or Facilitation models and reflect long-term learning that leads to changes in the spatialpattern of stimulus-selective responses and/ordynamics of those responses. Also models differ asto whether priming is associated with quantitative orqualitative changes in cortical representations. Onepossibility is that there are multiple mechanisms thatvary across space, time, and task. Finally, it is yetunknown whether the same or different mechanismsoperate in different brain regions.

p0275Progress will be aided by integrating data using acombination of behavioral and neuroimaging meth-ods such as fMRI and EEG/MEG (providedimportant differences between these measurements

LEME 00130

Visual Priming 15

ELSEVIERAUTH

ORPROOF

are kept in mind), linking between electrophysiologydata in animals and neuroimaging data in humans,and improvements in the spatial resolution of fMRI(b0030

Beauchamp et al., 2004;b0465

Schwarzlose et al., 2005;b0235

Grill-Spector et al., 2006b). Future experimentswill yield important empirical data that willvalidate (or refute) current theoretical predictions.Understanding the neural bases of priming will becritical for understanding whether priming reflectsquantitative or qualitative changes in the brain andwill allow a fundamental understanding of implicitlearning and memory in the adult brain.

Acknowledgments

This research was supported by National ScienceFoundation Grant 0345920 to Kalanit Grill-Spector.

References

b0005 Andrews TJ and Ewbank MP (2004) Distinct representations forfacial identity and changeable aspects of faces in the humantemporal lobe. Neuroimage 23: 905–913.

b0010 Avidan G, Hasson U, Hendler T, Zohary E, and Malach R (2002)Analysis of the neuronal selectivity underlying low fMRIsignals. Curr. Biol. 12: 964–972.

b0015 Baker CI, Behrmann M, and Olson CR (2002) Impact of learningon representation of parts and wholes in monkeyinferotemporal cortex. Nat. Neurosci. 5: 1210–1216.

b0020 Bar M and Biederman I (1999) Localizing the cortical regionmediating visual awareness of object identity. Proc. Natl.Acad. Sci. USA 96: 1790–1793.

b0025 Bar M, Tootell RB, Schacter DL, et al. (2001) Corticalmechanisms specific to explicit visual object recognition.Neuron 29: 529–535.

b0030 Beauchamp MS, Argall BD, Bodurka J, Duyn JH, and Martin A(2004) Unraveling multisensory integration: Patchyorganization within human STS multisensory cortex. Nat.Neurosci. 7: 1190–1192.

b0035 Biederman I (2000) Recognizing depth-rotated objects: Areview of recent research and theory. Spat. Vis. 13: 241–253.

b0040 Biederman I and Bar M (1999) One-shot viewpoint invariance inmatching novel objects. Vision Res. 39: 2885–2899.

b0045 Brookes MJ, Gibson AM, Hall SD, et al. (2005) GLM-beamformer method demonstrates stationary field, alphaERD and gamma. Neuroimage 26: 302–308.

b0050 Brown AS, Jones TC, andMitchell DB (1996) Single andmultipletest repetition priming in implicit memory. Memory 4:159–173.

b0055 Brown MW and Xiang JZ (1998) Recognition memory: Neuronalsubstrates of the judgement of prior occurrence. Prog.Neurobiol. 55: 149–189.

b0060 Buckner RL and Koutstaal W (1998) Functional neuroimagingstudies of encoding, priming, and explicit memory retrieval.Proc. Natl. Acad. Sci. USA 95: 891–898.

b0065 Buckner RL, Petersen SE, Ojemann JG, Miezin FM, Squire LR,and Raichle ME (1995) Functional anatomical studies ofexplicit and implicit memory retrieval tasks. J. Neurosci. 15:12–29.

b0070Burgund ED and Marsolek CJ (2000) Viewpoint-invariant andviewpoint-dependent object recognition in dissociableneural subsystems. Psychon. Bull. Rev. 7: 480–489.

b0075Campanella S, Hanoteau C, Depy D, et al. (2000) Right N170modulation in a face discrimination task: An account forcategorical perception of familiar faces. Psychophysiology37: 796–806.

b0080Carandini M and Ferster D (1997) A tonic hyperpolarizationunderlying contrast adaptation in cat visual cortex. Science276: 949–952.

b0085Cave CB (1997) Very long-lasting priming in picture naming.Psychol. Sci. 8: 322–325.

b0090Cave CB, Bost PR, and Cobb RE (1996) Effects of color andpattern on implicit and explicit picture memory. J. Exp.Psychol. Learn. Mem. Cogn. 22: 639–653.

b0095Chawla D, Lumer ED, and Friston KJ (1999) The relationshipbetween synchronization among neuronal populations andtheir mean activity levels. Neural Comput. 11: 1389–1411.

b0100Cooper LA, Schacter DL, Ballesteros S, and Moore C (1992)Priming and recognition of transformed three-dimensionalobjects: Effects of size and reflection. J. Exp. Psychol. Learn.Mem. Cogn. 18: 43–57.

b0105Dale AM, Liu AK, Fischl BR, et al. (2000) Dynamic statisticalparametric mapping: Combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron 26: 55–67.

b0110Dehaene S, Naccache L, Le Clec HG, et al. (1998) Imagingunconscious semantic priming. Nature 395: 597–600.

b0115Demb JB, Desmond JD, Wagner AD, Vaidya CD, Glover GH, andGabrieli JD (1995) Semantic encoding and retrieval in the leftinferior prefrontal cortex: A functional MRI study of taskdifficulty and process specificity. J. Neurosci. 15: 5780–5788.

b0120Desimone R (1996) Neural mechanisms for visual memory andtheir role in attention. Proc. Natl. Acad. Sci. USA 93:13494–13499.

b0125Dobbins IG, Schnyer DM, Verfaellie M, and Schacter DL (2004)Cortical activity reductions during repetition priming canresult from rapid response learning. Nature 428: 316–319.

b0130Dolan RJ, Fink GR, Rolls E, et al. (1997) How the brain learns tosee objects and faces in an impoverished context. Nature389: 596–599.

b0135Doniger GM, Foxe JJ, Schroeder CE, Murray MM, Higgins BA,and Javitt DC (2001) Visual perceptual learning in humanobject recognition areas: A repetition priming study usinghigh-density electrical mapping. Neuroimage 13: 305–313.

b0140Dragoi V, Sharma J, Miller EK, and Sur M (2002) Dynamics ofneuronal sensitivity in visual cortex and local featurediscrimination. Nat. Neurosci. 5: 883–891.

b0145Eger E, Henson RN, Driver J, and Dolan RJ (2004) BOLDrepetition decreases in object-responsive ventral visualareas depend on spatial attention. J. Neurophysiol. 92:1241–1247.

b0150Eger E, Schweinberger SR, Dolan R, and Henson RN (2005)Familiarity enhances invariance of face representations inhuman inferotemporal cortex: fMRI evidence. Neuroimage26: 1128–1139.

b0155Eimer M (2000) Event-related brain potentials distinguishprocessing stages involved in face perception andrecognition. Clin. Neurophysiol. 111: 694–705.

b0160Epstein R, Graham KS, and Downing PE (2003) Viewpoint-specific scene representations in human parahippocampalcortex. Neuron 37: 865–876.

b0165Fiser J and Biederman I (1995) Size invariance in visual objectpriming of gray-scale images. Perception 24: 741–748.

b0170Fiser J and Biederman I (2001) Invariance of long-term visualpriming to scale, reflection, translation, and hemisphere.Vision Res. 41: 221–234.

b0175Fries P, Reynolds JH, Rorie AE, and Desimone R (2001)Modulation of oscillatory neuronal synchronization byselective visual attention. Science 291: 1560–1563.

LEME 00130

16 Visual Priming

ELSEVIERAUTH

ORPROOF

b0180 Friston K (2005) A theory of cortical responses. Philos. Trans. R.Soc. Lond. B Biol. Sci. 360: 815–836.

b0185 Furmanski CS and Engel SA (2000) Perceptual learning in objectrecognition: Object specificity and size invariance. VisionRes. 40: 473–484.

b0190 Ganel T, Gonzalez CL, Valyear KF, Culham JC, Goodale MA,and Kohler S (2006) The relationship between fMRIadaptation and repetition priming. Neuroimage 32:1432–1440.

b0195 George N, Dolan RJ, Fink GR, Baylis GC, Russell C, and Driver J(1999) Contrast polarity and face recognition in the humanfusiform gyrus. Nat. Neurosci. 2: 574–580.

b0200 Grill-Spector K (2003) The neural basis of object perception.Curr. Opin. Neurobiol. 13: 159–166.

b0205 Grill-Spector K, Henson R, and Martin A (2006a) Repetition andthe brain: Neural models of stimulus-specific effects. TrendsCogn. Sci. 10: 14–23.

b0210 Grill-Spector K and Kanwisher N (2005) Visual recognition: Assoon as you know it is there, you know what it is. Psychol.Sci. 16: 152–160.

b0215 Grill-Spector K, Knouf N, and Kanwisher N (2004) The fusiformface area subserves face perception, not generic within-category identification. Nat. Neurosci. 7: 555–562.

b0220 Grill-Spector K, Kushnir T, Edelman S, Avidan G, Itzchak Y, andMalach R (1999) Differential processing of objects undervarious viewing conditions in the human lateral occipitalcomplex. Neuron 24: 187–203.

b0225 Grill-Spector K, Kushnir T, Hendler T, and Malach R (2000) Thedynamics of object-selective activation correlate withrecognition performance in humans. Nat. Neurosci. 3:837–843.

b0230 Grill-Spector K and Malach R (2001) fMR-adaptation: A tool forstudying the functional properties of human cortical neurons.Acta Psychol. (Amst.) 107: 293–321.

b0235 Grill-Spector K, Sayres R, and Ress D (2006b) High-resolutionimaging reveals highly selective nonface clusters in thefusiform face area. Nat. Neurosci. 1177–1185.

b0240 Gruber T and Muller MM (2005) Oscillatory brain activitydissociates between associative stimulus content in a repetitionpriming task in the human EEG. Cereb Cortex 15: 109–116.

b0245 Hamann SB and Squire LR (1997) Intact perceptual memory in theabsence of consciousmemory.Behav. Neurosci. 111: 850–854.

b0250 Henson R, Shallice T, and Dolan R (2000) Neuroimagingevidence for dissociable forms of repetition priming. Science287: 1269–1272.

b0255 Henson RN (2003) Neuroimaging studies of priming. Prog.Neurobiol. 70: 53–81.

b0260 Henson RN, Goshen-Gottstein Y, Ganel T, Otten LJ, Quayle A,and Rugg MD (2003) Electrophysiological andhaemodynamic correlates of face perception, recognitionand priming. Cereb. Cortex 13: 793–805.

b0265 Henson RN, Price CJ, Rugg MD, Turner R, and Friston KJ(2002a) Detecting latency differences in event-related BOLDresponses: Application to words versus nonwords and initialversus repeated face presentations. Neuroimage 15: 83–97.

b0270 Henson RN and Rugg MD (2003) Neural response suppression,haemodynamic repetition effects, and behavioural priming.Neuropsychologia 41: 263–270.

b0275 Henson RN, Rylands A, Ross E, Vuilleumeir P, and Rugg MD(2004) The effect of repetition lag on electrophysiological andhaemodynamic correlates of visual object priming.Neuroimage 21: 1674–1689.

b0280 Henson RN, Shallice T, Gorno-Tempini ML, and Dolan RJ(2002b) Face repetition effects in implicit and explicitmemory tests as measured by fMRI. Cereb. Cortex 12:178–186.

b0285 Ishai A, Bikle PC, and Ungerleider LG (2006) Temporaldynamics of face repetition suppression. Brain Res. Bull. 70:289–295.

b0290Ishai A, Pessoa L, Bikle PC, and Ungerleider LG (2004)Repetition suppression of faces is modulated by emotion.Proc. Natl. Acad. Sci. USA 101: 9827–9832.

b0295James TW and Gauthier I (2006) Repetition-induced changes inBOLD response reflect accumulation of neural activity. Hum.Brain Mapp. 27: 37–46.

b0300James TW, Humphrey GK, Gati JS, Menon RS, and GoodaleMA (2000) The effects of visual object priming on brainactivation before and after recognition. Curr. Biol. 10:1017–1024.

b0305James TW, Humphrey GK, Gati JS, Menon RS, and GoodaleMA (2002) Differential effects of viewpoint on object-drivenactivation in dorsal and ventral streams. Neuron 35:793–801.

b0310Jiang X, Rosen E, Zeffiro T, Vanmeter J, Blanz V, andRiesenhuber M (2006) Evaluation of a shape-based model ofhuman face discrimination using fMRI and behavioraltechniques. Neuron 50: 159–172.

b0315Jiang Y, Haxby JV, Martin A, Ungerleider LG, and ParasuramanR (2000) Complementary neural mechanisms for trackingitems in human working memory. Science 287: 643–646.

b0320Kohn A and Movshon JA (2004) Adaptation changes thedirection tuning of macaque MT neurons. Nat. Neurosci. 7:764–772.

b0325Kourtzi Z, Betts LR, Sarkheil P, and Welchman AE (2005)Distributed neural plasticity for shape learning in the humanvisual cortex. PLoS Biol. 3: e204.

b0330Kourtzi Z and Grill-Spector K (2005) fMRI adaptation: A tool forstudying visual representations in the primate brain. In:Rhodes G and Clifford CWG (eds.) Fitting the Mind to theWorld: Adaptation and After-Effects in High-Level Vision,pp. 173–188. Oxford, Oxford University Press.

b0335Kourtzi Z and Kanwisher N (2000) Cortical regions involved inperceiving object shape. J. Neurosci. 20: 3310–3318.

b0340Kourtzi Z and Kanwisher N (2001) Representation of perceivedobject shape by the human lateral occipital complex.Science 293: 1506–1509.

b0345Kourtzi Z, Tolias AS, Altmann CF, Augath M, and Logothetis NK(2003) Integration of local features into global shapes:Monkey and human FMRI studies. Neuron 37: 333–346.

b0350Koutstaal W, Wagner AD, Rotte M, Maril A, Buckner RL, andSchacter DL (2001) Perceptual specificity in visual objectpriming: Functional magnetic resonance imaging evidencefor a laterality difference in fusiform cortex.Neuropsychologia 39: 184–199.

b0355Li L, Miller EK, and Desimone R (1993) The representation ofstimulus familiarity in anterior inferior temporal cortex. J.Neurophysiol. 69: 1918–1929.

b0360Loffler G, Yourganov G, Wilkinson F, and Wilson HR (2005) fMRIevidence for the neural representation of faces. Nat.Neurosci. 8: 1386–1390.

b0365Lustig C and Buckner RL (2004) Preserved neural correlates ofpriming in old age and dementia. Neuron 42: 865–875.

b0370Maccotta L and Buckner RL (2004) Evidence for neural effectsof repetition that directly correlate with behavioral priming. J.Cogn. Neurosci. 16: 1625–1632.

b0375Marsolek CJ (1995) Abstract visual-form representations in theleft cerebral hemisphere. J. Exp. Psychol. Hum. Percept.Perform. 21: 375–386.

b0380Miller EK and Desimone R (1994) Parallel neuronal mechanismsfor short-term memory. Science 263: 520–522.

b0385Miller EK, Li L, and Desimone R (1993) Activity of neurons inanterior inferior temporal cortex during a short-term memorytask. J. Neurosci. 13: 1460–1478.

b0390Movshon JA and Lennie P (1979) Pattern-selective adaptationin visual cortical neurones. Nature 278: 850–852.

b0395Muller JR, Metha AB, Krauskopf J, and Lennie P (1999) Rapidadaptation in visual cortex to the structure of images.Science 285: 1405–1408.

LEME 00130

Visual Priming 17

ELSEVIERAUTH

ORPROOF

b0400 Murray SO and He S (2006) Contrast invariance in the humanlateral occipital complex depends on attention. Curr. Biol.16: 606–611.

b0405 Murray SO and Wojciulik E (2004) Attention increases neuralselectivity in the human lateral occipital complex. Nat.Neurosci. 7: 70–74.

b0410 Naccache L and Dehaene S (2001) The priming method:Imaging unconscious repetition priming reveals an abstractrepresentation of number in the parietal lobes. Cereb. Cortex11: 966–974.

b0415 Newell FN, Sheppard DM, Edelman S, and Shapiro KL (2005)The interaction of shape and location-based priming inobject categorization: Evidence for a hybrid ‘‘what $ where’’representation stage. Vision Res. 45: 2065–2080.

b0420 Norman KA and O’Reilly RC (2003) Modeling hippocampal andneocortical contributions to recognition memory: Acomplementary-learning-systems approach. Psychol. Rev.110: 611–646.

b0425 Rainer G, Lee H, and Logothetis NK (2004) The effect of learningon the function of monkey extrastriate visual cortex. PLoSBiol. 2: E44.

b0430 Rao RP and Ballard DH (1999) Predictive coding in the visualcortex: A functional interpretation of some extra-classicalreceptive-field effects. Nat. Neurosci. 2: 79–87.

b0435 Ringo JL (1996) Stimulus specific adaptation in inferior temporaland medial temporal cortex of the monkey. Behav. BrainRes. 76: 191–197.

b0440 Rotshtein P, Henson RN, Treves A, Driver J, and Dolan RJ(2005) Morphing Marilyn into Maggie dissociates physicaland identity face representations in the brain. Nat. Neurosci.8: 107–113.

b0445 Sayres R and Grill-Spector K (2006) Object-selective cortexexhibits performance-independent repetition suppression.J. Neurophysiol. 95: 995–1007.

b0450 Schacter DL and Buckner RL (1998) Priming and the brain.Neuron 20: 185–195.

b0455 Schacter DL, Dobbins IG, and Schnyer DM (2004) Specificity ofpriming: A cognitive neuroscience perspective. Nat. Rev.Neurosci. 5: 853–862.

b0460 Schendan HE and Kutas M (2003) Time course of processesand representations supporting visual object identificationand memory. J. Cogn. Neurosci. 15: 111–135.

b0465 Schwarzlose RF, Baker CI, and Kanwisher NK (2005) SeparateFace and Body Selectivity on the Fusiform Gyrus. J.Neurosci. 25: 11055–11059.

b0470 Schweinberger SR, Huddy V, and Burton AM (2004) N250r: Aface-selective brain response to stimulus repetitions.Neuroreport 15: 1501–1505.

b0475 Sigala N and Logothetis NK (2002) Visual categorization shapesfeature selectivity in the primate temporal cortex. Nature415: 318–320.

b0480 Simons JS, Koutstaal W, Prince S, Wagner AD, and SchacterDL (2003) Neural mechanisms of visual object priming:Evidence for perceptual and semantic distinctions infusiform cortex. Neuroimage 19: 613–626.

b0485 Sobotka S and Ringo JL (1994) Stimulus specific adaptation inexcited but not in inhibited cells in inferotemporal cortex ofmacaque. Brain Res. 646: 95–99.

b0490 Sobotka S and Ringo JL (1996) Mnemonic responses of singleunits recorded from monkey inferotemporal cortex,

accessed via transcommissural versus direct pathways: Adissociation between unit activity and behavior. J. Neurosci.16: 4222–4230.

b0495Stern CE, Corkin S, Gonzalez RG, et al. (1996) The hippocampalformation participates in novel picture encoding: Evidencefrom functional magnetic resonance imaging. Proc. Natl.Acad. Sci. USA 93: 8660–8665.

b0500Stopfer M and Laurent G (1999) Short-term memory in olfactorynetwork dynamics. Nature 402: 664–668.

b0505Sugase Y, Yamane S, Ueno S, and Kawano K (1999) Global andfine information coded by single neurons in the temporalvisual cortex. Nature 400: 869–873.

b0510Tulving E, Hayman CA, and Macdonald CA (1991) Long-lastingperceptual priming and semantic learning in amnesia: A caseexperiment. J. Exp. Psychol. Learn. Mem. Cogn. 17:595–617.

b0515Turk-Browne NB, Yi DJ, Leber AB, and Chun MM (2006) Visualquality determines the direction of neural repetition effects.Cereb. Cortex 17: 425–433.

b0520van Turennout M, Bielamowicz L, and Martin A (2003)Modulation of neural activity during object naming: Effects oftime and practice. Cereb. Cortex 13: 381–391.

b0525van Turennout M, Ellmore T, and Martin A (2000) Long-lastingcortical plasticity in the object naming system.Nat. Neurosci.3: 1329–1334.

b0530Vuilleumier P, Henson RN, Driver J, and Dolan RJ (2002)Multiple levels of visual object constancy revealed byevent-related fMRI of repetition priming. Nat. Neurosci. 5:491–499.

b0535Wagner AD, Gabrieli JD, and Verfaellie M (1997) Dissociationsbetween familiarity processes in explicit recognition andimplicit perceptual memory. J. Exp. Psychol. Learn. Mem.Cogn. 23: 305–323.

b0540Wheatley T, Weisberg J, Beauchamp MS, and Martin A (2005)Automatic priming of semantically related words reducesactivity in the fusiform gyrus. J. Cogn. Neurosci. 17:1871–1885.

b0545Wiggs CL and Martin A (1998) Properties and mechanisms ofperceptual priming. Curr. Opin. Neurobiol. 8: 227–233.

b0550Wiggs CL, Martin A, and Sunderland T (1997) Monitoringfrequency of occurrence without awareness: Evidence frompatients with Alzheimer’s disease. J. Clin. Exp.Neuropsychol. 19: 235–244.

b0555Winston JS, Henson RN, Fine-Goulden MR, and Dolan RJ(2004) fMRI-adaptation reveals dissociable neuralrepresentations of identity and expression in faceperception. J. Neurophysiol. 92: 1830–1839.

b0560Xiang JZ and Brown MW (1998) Differential neuronal encodingof novelty, familiarity and recency in regions of the anteriortemporal lobe. Neuropharmacology 37: 657–676.

b0565Yi DJ and Chun MM (2005) Attentional modulation of learning-related repetition attenuation effects in humanparahippocampal cortex. J. Neurosci. 25: 3593–3600.

b0570Yi DJ, Olson IR, and Chun MM (2006) Shape-specificperceptual learning in a figure-ground segregation task.Vision Res. 46: 914–924.

b0575Zago L, Fenske MJ, Aminoff E, and Bar M (2005) Therise and fall of priming: How visual exposure shapescortical representations of objects. Cereb. Cortex 11:1655–1665.

LEME 00130

18 Visual Priming

ELSEVIER

AUTH

ORPROOF

Author’s Contact Information

Kalanit Grill-SpectorAU1

Department of Psychology and Neuroscience InstituteStanford UniversityStanford, CA 94305kalanit@psych.stanford.edu

Keywords: adaptation; BOLD; fMRI; human visual cortex; implicit memory; objectrecognition; priming; repetition enhancement; repetition suppression; ventral stream;visual priming

AbstractVisual priming is the behavioral phenomenon of improved processing of a visual stimulus followingprior experience and is thought to reflect an implicit form of memory. Interest for understandingpriming and its neural correlates has been heightened as an increasing number of scientists usepriming methods to characterize representations in the human brain. Here we review the behavioraland neural correlates of visual priming and describe three models for the underlying neural mechan-isms. Understanding the neural bases of visual priming is critical for a fundamental understanding ofthe characteristics of visual representations and for the interpretation of neuroimaging data andlinking it to behavior.

LEME 00130