www.elsevier.com/locate/ynimg

NeuroImage 32 (2006) 1432 – 1440

The relationship between fMRI adaptation and repetition priming

Tzvi Ganel,a,b,* Claudia L.R. Gonzalez,a Kenneth F. Valyear,a Jody C. Culham,a

Melvyn A. Goodale,a and Stefan Kohlera

aDepartment of Psychology and CIHR Group on Action and Perception, University of Western Ontario, London, ON, Canada N6A 5C2bDepartment of Behavioral Sciences, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel

Received 14 October 2005; revised 19 May 2006; accepted 21 May 2006

Available online 18 July 2006

Neuroimaging investigations of the cortically defined fMRI adaptation

effect and of the behaviorally defined repetition priming effect have

provided useful insights into how visual information is perceived and

stored in the brain. Yet, although both phenomena are typically

associated with reduced activation in visually responsive brain regions

as a result of stimulus repetition, it is presently unknown whether they

rely on common or dissociable neural mechanisms. In an event-related

fMRI experiment, we manipulated fMRI adaptation and repetition

priming orthogonally. Subjects made comparative size judgments for

pairs of stimuli that depicted either the same or different objects; some

of the pairs presented during scanning had been shown previously and

others were new. This design allowed us to examine whether object-

selective regions in occipital and temporal cortex were sensitive to

adaptation, priming, or both. Critically, it also allowed us to test

whether any region showing sensitivity to both manipulations dis-

played interactive or additive effects. Only a partial overlap was found

between areas that were sensitive to fMRI adaptation and those

sensitive to repetition priming. Moreover, in most of the object-selective

regions that showed both effects, the reduced activation associated with

the two phenomena were additive rather than interactive. Together,

these findings suggest that fMRI adaptation and repetition priming can

be dissociated from one another in terms of their neural mechanisms.

D 2006 Elsevier Inc. All rights reserved.

Keywords: Visual processing; Implicit memory; Object recognition; Lateral

occipital area; Fusiform gyrus; Visual systems

Introduction

A previous encounter with a stimulus usually leads to improved

performance when the same stimulus is encountered again. Such

improvement, referred to as the repetition priming effect, is often

observed in perceptual and conceptual tasks that require repeated

1053-8119/$ - see front matter D 2006 Elsevier Inc. All rights reserved.

doi:10.1016/j.neuroimage.2006.05.039

* Corresponding author. Department of Behavioral Sciences, Ben-Gurion

University of the Negev, Beer-Sheva 84105, Israel. Fax: +972 8 6472072.

E-mail address: tganel@bgu.ac.il (T. Ganel).

Available online on ScienceDirect (www.sciencedirect.com).

processing of a stimulus without explicit reference to the first

encounter (Schacter, 1987).

In a typical repetition priming paradigm, subjects are initially

exposed to a list of items in a study phase. Priming is assessed in a

subsequent test phase, in which these old items are intermixed with

new items. Functional neuroimaging studies using this paradigm

with perceptual tasks have typically found decreases in activation

in occipital and inferotemporal regions for old as compared to new

visual stimuli. This reduction in activation, although not observed

under all experimental conditions (e.g., Henson et al., 2000), is

widely regarded as the neural correlate of repetition priming

(Schacter et al., 2004; Wiggs and Martin, 1998).

Effects of stimulus repetition have also been the focus of

neuroimaging studies probing the nature of visual representations

at various levels of the cortical visual system. In these studies, pairs

of same or different stimuli are presented sequentially with a very

short delay and no intervening stimuli between the first and second

presentation. An increase in activation for stimuli that were

changed as compared to unchanged on a given dimension is

referred to as fMRI adaptation (Grill-Spector et al., 1999); it is

typically interpreted as evidence that the region showing this

increase is involved in processing the relevant dimension. Put

another way, fMRI adaptation can also be seen as a phenomenon in

which repeated (unchanged) as compared to new (changed) stimuli

show a decrease in activation. This decrease in activation is

‘frequently, although not always and not by definition (see Sayres

and Grill-Spector, 2006), associated with improved performance

on adaptation trials. Thus, the cortically defined fMRI adaptation

effect and the behaviorally defined priming effect have both been

associated with improved performance and a decrease in activation

in similar visual areas as a result of stimulus repetition in past

functional neuroimaging research. Given this commonality, the

question arises as to whether or not the two phenomena reflect the

same underlying neural mechanism (for a discussion, see Henson,

2003).

There is some evidence from behavioral (Bentin and Mosco-

vitch, 1988) and electrophysiological priming studies (Nagy and

Rugg, 1989) that immediate and long-term repetition effects can be

dissociated from one another. However, it is yet unknown whether

T. Ganel et al. / NeuroImage 32 (2006) 1432–1440 1433

the immediate repetition effects reflected in fMRI adaptation rely

on the same neural mechanisms that mediate repetition priming

(Henson, 2003).

The goal of the current study was to directly test the relationship

between fMRI adaptation and repetition priming. We examined the

effects of priming for familiar objects in the context of a typical fMRI

adaptation paradigm, focusing primarily on occipito-temporal regions

involved in object perception. To determine whether the two

phenomena are independent or share the same underlying mechanism,

we applied a logic widely used in behavioral studies (Sternberg, 1969,

2001) and recently in neuroimaging (Norris et al., 2004; Pinel et al.,

2001). According to this logic, fMRI adaptation and repetition priming

would be considered independent only if their combined effects were

additive and not interactive. Here, we built on this logic and used a

factorial design, in which we manipulated adaptation and priming

orthogonally (Fig. 1). This allowed us to examine whether any given

object-selective region was sensitive to adaptation, priming, or both.

Critically, it also allowed us to test whether any region showing

sensitivity to both manipulations showed interactive or additive effects.

Methods

Participants

Sixteen right-handedparticipants aged23–39 (8men and8women)

participated in the experiment. All participants signed a consent form

Fig. 1. Illustration of the experimental design. During the study (pre-scanning) se

same object was presented twice (same pairs), and, in other cases, two different obj

the second object in each pair was smaller, the same, or bigger in real life than the f

in the four subsequent runs, subjects made the same judgments for pairs of objec

approved by the ethics committee at the University ofWestern Ontario.

Data from one participant were excluded from the analysis due to

excessive magnet noise.

Stimuli and experimental design

The stimuli were 256 color photos of common objects, taken from

the Hemera Photo Object collection (http://www.hemera.com). The

stimuli were cropped to equal dimensions of 800 � 600 pixels on the

screen (comprising a visual angle of approximately 15- for subjects

lying in the scanner) and were presented against a white background

(see Fig. 1). The objects were grouped into 128 pairs that were used in

the experimental runs in the real-size judgment task. To ensure that the

objects in each pair could be readily compared for their real size, we ran

a pilot behavioral study in which 12 subjects made real-life size

comparisons of the same pairs of objects that were used in the actual

experiment. Accuracy was above 90% in the pilot experiment.

The 128 object pairs were divided into two equal sets of 64. Each

participant was presented with one of the sets in the study block before

the scanning session. In the study block and in the scanning session,

participants were told that theywould be presented with pairs of objects

that would be presented rapidly one after the other. They were asked to

judge for each pair whether the second object was bigger, smaller, or the

same in real-life size as the first object presented (Fig. 1). Note that,

following our goal to determine whether fMRI adaptation reflects

priming, our design was based on a typical adaptation paradigm.

Accordingly, we used double presentations of identical objects for

ssion, subjects made size judgments for pairs of objects. In some cases, the

ects were presented (different pairs). Subjects were asked to indicate whether

irst object. During the test session in the fMRI scanner, which was presented

t that had been shown during study (old pairs), intermixed with new pairs.

T. Ganel et al. / NeuroImage 32 (2006) 1432–14401434

‘‘same’’ pairs, and single presentations of two different objects for the

‘‘different’’ pairs. Importantly, in this arrangement, the total number of

repetitions of each object was matched for the priming and adaptation

conditions.

Different response modalities were used in the study (pre-

scanning) and test (experimental) runs to rule out the possibility

that any decrease in fMRI activation for previously encountered

stimulus pairs could be the result of specific associations formed

between these stimuli and the required motor response (Dobbins et

al., 2004). During the pre-scanning session, participants made

vocal responses, which were recorded by the experimenter in the

control room. During the experimental runs, participants provided

their responses with button presses on a keypad.

Four different experimental runs were used. Each run began

and ended with 16 s fixation periods and contained 32 object

pairs (16 new; 16 old, studied pairs) depicting either the same

objects or different objects (see Fig 1). In the experimental

sessions, each stimulus was presented for 400 ms; pairs of

stimuli were separated by an inter-stimulus interval of 200 ms.

Inter-trial intervals of 11 s were used between the presentations

of pairs. In the pre-scanning session, pair combinations and

presentation parameters were identical to those used in the

scanning runs with the exception that intervals of 4 s were used

between pair presentations. Pair presentation order was pseudo-

randomized for each run, and order of runs was counterbalanced

between subjects. Two different pair combinations were used

and were counterbalanced across subjects.

To identify objects-selective brain areas, we used two ‘‘localizer’’

runs inwhich subjects passively viewed photos of faces, objects, scenes,

and scrambled versions of those same stimuli. The photos were

monochromatic and were presented in blocks, each containing 30

different photos of each stimulus category. Each stimuluswas presented

for 400 ms followed by a 100 ms inter-stimulus interval. Each run

included 4 blocks of faces, 4 blocks of objects, and 4 blocks of scenes,

all separated by blocks of scrambled objects. Different orders of blocks

were used for the two different localizer runs, which always followed

the experimental runs.

Imaging parameters

Imaging data were collected at the Robarts Research Institute

(London, ON, Canada) using a 4 T, whole-body MRI system

(Varian/Siemens) with a full-volume head coil. Each imaging

session consisted of 6 functional runs (4 experimental and 2

localizer runs) and a single high-resolution anatomical acquisi-

tion. Functional images were collected using a T2*-weighted,

segmented (navigator corrected), interleaved SPIRAL acquisition

(TE = 150 ms, TR = 2000 ms, flip angle = 45-, two segments/

plane) for blood-oxygen-level-dependent (BOLD)-based imaging

(Ogawa et al., 1992). The field of view was 22 by 22 cm with

an in-plane matrix size of 64 by 64 pixels. Each volume

comprised of 17 6-mm-thick pseudo-axial slices, resulting in a

voxel size of 3.4 � 3.4 � 6.0 mm. Volume acquisition time

was 2000 ms. High-resolution T1-weighted anatomical volumes

were acquired using 3D magnetization-prepared FLASH acqui-

sition (TI = 1300 ms, TE = 30 ms, TR = 50 ms, FA = 20).

Imaging analysis

The imaging data were preprocessed and analyzed using Brain

Voyager QX software (Brain Innovation, Maastricht, The Nether-

lands). The anatomical data from each subject were transformed

into Talairach space (Talairach and Tournoux, 1998). Functional

volumes underwent linear trend removal and were then aligned to

the transformed anatomical volumes, thereby transforming the

functional data into a common brain space across subjects.

The imaging data were analyzed using a GLM (general linear

model) procedure. This procedure allows the correlation of

predictor functions with the recorded activation data. Predictor

functions were c functions (D = 2.5, s = 1.25), designed to estimate

hemodynamic response properties (Boynton et al., 1996), spaced in

time to coincide with each stimulus event.

For each subject, the averaged functional volumes from the

localizer scans were used to identify object selective ROIs within

occipotemporal regions. We used the balanced contrast objects vs.

(faces + houses) to identify three regions of interest within each

hemisphere that have previously been shown to be involved in the

processing of objects and that have shown repetition-based

reduction in activation effects (e.g., Henson et al., 2004); these

regions included the lateral occipital cortex (area LO), the mid-

fusiform gyrus, and the parahippocampal cortex. The ROIs were

selected as the most significantly activated 27 voxels (3 � 3 � 3

voxels), the Fhotspot_ of activity, within each region. Fig. 3 shows

an example of the distribution of objects selective areas in one

representative subject. The timepoints that were used for the

statistical analysis were the peak BOLD activations in each of the

experimental condition, which occurred approximately 5 s

following stimulus presentation.

Results

Behavioral results

Reaction times (RTs) and accuracy were recorded for each

participant during the experimental runs. RTs were analyzed only

for correct responses using a two-way ANOVA with Fpair status_(same object, different objects) and Fstudy status_ (old objects, new

objects) as within-subject variables. All p levels were computed

based on two-tailed tests. As can be seen in Fig. 2, RTs for old

pairs were significantly faster than RTs for new pairs, F(1,14) =

36.3, P < .001. Planned comparisons revealed that the effect was

significant both for same, t(14) = 3.6, P < .01, and for different

pairs, t(14) = 4.6, P < .01. Thus, a significant repetition priming

effect was found in the experiment at the behavioral level. In

addition, RTs for same pairs were faster than for different pairs,

F(1,14) = 119.7, P < .001. The interaction between pair and study

status was not significant, F(1,14) = 2.3, P > .1.

Accuracy levels in the size-comparison task were high for all

conditions. For same trials, it was 99.4% for old items and 99.3%

for new items; for different trials, it was 84.1% for old items and

80.2% for new items. In agreement with the reaction time data, an

ANOVA revealed a repetition priming effect, with higher accuracy

for old as compared to new items, F(1,14) = 6.4, P < .05. The main

effect of pair status was also significant, F(1,14) = 48.4, P < .001,

as was the interaction between pair and study status, F(1,14) = 5.5,

P < .05.

fMRI results

Our main analysis focused on the pattern of activation within

objects-selective areas in occipital and temporal cortex. These

Fig. 2. Effects of stimulus repetition on task performance. Reaction times in

the size judgment task during the fMRI test session were significantly faster

for old as compared to new items. This priming effect was significant for

both same and different pairs. In addition, performance for Fsame_ pairs, in

which the same object was immediately repeated, was significantly faster as

compared to Fdifferent_ pairs.

T. Ganel et al. / NeuroImage 32 (2006) 1432–1440 1435

ROIs were identified, separately for each subject, based on the

pattern of activation in the localizer scans. The object-selective

areas identified with these scans included the left and right lateral

occipital area (LO), bilateral fusiform gyrus (FG), and bilateral

parahippocampal gyrus (PG). The pattern of activation in the

experimental runs within each of these regions, averaged across

subjects, is presented in Fig. 3. We performed a two-way repeated

measures ANOVA on the activation data in each of the ROIs,

considering only trials with correct responses. The two indepen-

dent variables were Fpair status_ (same vs. different objects), which

was used as a measure of immediate adaptation, and Fstudy status_(old vs. new objects), which was used as a measure of repetition

priming.

Overall, the majority of object-selective regions showed a

significant decrease in activation following both adaptation and

priming (see Fig. 3). The right LO (Talairach x y z coordinates

(SE), 39 T 2 �72 T 2 �6 T 1), the right FG (31 T 1 �54 T 2 �15 T1), the right PG (24 T 1 �31 T 2 �17 T 1), and the left PG (�35 T 1�25 T 2 �17 T 1) showed a main effect of adaptation (for rLO,

F(1,14) = 14.5, P < .01; for rFG, F(1,14) = 24.3, P < .001; for

rPG, F(1,14) = 6.7, P < .05, for lPG, F(1, 13) = 10.9, P < .01).

Furthermore, all four regions also showed a main effect of priming

(for rLO, F(1,14) = 6.8, P < .05; for rFG, F(1,14) = 8.9, P < .01;

for rPG, F(1,14) = 8.1, P < .05, for lPG, F(1, 13) = 8.5, P < .05).

Most importantly, there was no significant difference in the size of

the priming effect and the adaptation effect (P > .05), and there

was no interaction between the effects of adaptation and priming

(P > .05) in any of these four regions. In other words, the effects of

priming and adaptation were additive (rather than interactive) and

comparable in size in rLO, rFG, rPA, and lPA, arguing in favor of

the idea that these two phenomena are mediated by independent

neural mechanisms.

In the left LO (�46 T 1 �69 T 2 �5 T 2), a decrease in

activation was found only for adaptation, F(1,14) = 8.9, P < .01,

but not for priming, P > .05, again with no significant interaction,

P > .05. Furthermore, when compared directly, the main effect of

adaptation in the left LO was significantly larger than that of

priming, t(14) = 3.1, P < .01. Thus, this pattern of activation in the

left LO cannot be interpreted as a Fthreshold effect_. Instead, itreflects a selective sensitivity of this region to adaptation,

providing additional support for the idea that adaptation and

priming rely on different mechanisms.

The only region that showed an interaction between adaptation

and priming was the left FG (�36 T 2 �49 T 2 �15 T 1), F(1,14) =4.6, P < .05. Furthermore, a decrease of activation in this region

was found both for adaptation, F(1,14) = 69.3, P < .001, and for

priming, F(1,14) = 7.3, P < .05. As in the left LO, the main effect

of adaptation in the left FG was significantly larger than that of

priming, t(14) = 4.1, P < .01. The significant interaction between

adaptation and priming in left FG indicates that adaptation and

priming are not independent in all object-selective brain regions.

In a secondary analysis, we expanded the focus of our fMRI

analysis beyond those regions that respond selectively to objects

and that were included in our ROI analyses. For this purpose, we

applied a voxelwise random-effects GLM to the entire occipital

and temporal cortex to determine which regions, if any, showed

any evidence of an interaction between adaptation and priming.

This analysis revealed that the only region displaying a significant

interaction between adaptation and priming was the portion of left

fusiform gyrus that we had already identified to show an

interaction in our ROI analyses (see Fig. 4).

For exploratory purposes, we also used a voxelwise GLM

analysis to determine the extent of overlap in occipital and

temporal cortex between regions that are sensitive to adaptation

and those sensitive to priming (as reflected in the corresponding

main effect). In agreement with the results from our ROI analyses,

we found partial overlap between adaptation- and priming-

sensitive regions in fusiform gyrus, parahippocampal gyrus, and

lateral occipital cortex. Outside of these regions, adaptation-related

reductions in activation were more pronounced in posterior regions

in occipital cortex whereas priming-related reductions were more

pronounced in anterior regions in the parahippocampal gyrus and

the hippocampus (see Fig. 5).

In two final sets of analyses, we returned to our ROIs for

object-selective regions in order to address potential confounds in

our design. Notably, it could be argued that any differences in

activation that we observed across adaptation and priming

conditions might simply reflect the differing numbers of stimulus

repetitions across cells in our 2 � 2 factorial design. Such

differences are inevitable given that we crossed the two factors of

adaptation and priming independently, aiming to determine

whether adaptation and priming show additive or interactive

effects. To discount the possibility that stimulus repetition can

account for any of the differences in brain activity we observed

across adaptation and priming conditions in our ROIs, we first

counted the number of stimulus presentations for each stimulus in

each trial type. According to such a counting schema, Fnew/same_stimuli were presented for the first and second time (2 presenta-

tions per stimulus) while Fnew/different_ stimuli were presented

once each (1 presentation per stimulus). Fold/same_ stimuli were

presented for the third and fourth time (4 presentations per

stimulus) and Fold/different_ stimuli were presented for the second

time each (2 presentations per stimulus).

Assuming that the number of presentations is negatively

correlated with BOLD activation (i.e., that an increase in

repetitions is associated with a decrease in activation), it may

seem, at first glance, that the general trend across cells in some of

Fig. 3. Effects of fMRI adaptation and repetition priming in object-selective areas. Object-selective areas were identified separately for each subject using the

localizer runs. The middle panel displays cortical regions showing object-selective activation in one representative subject. These areas included regions within

bilateral LO (A, B), fusiform gyrus (C, D), and parahippocampal gyrus (E, F). All regions, with the exception of the right LO (A), showed a significant

reduction in activation following both immediate repetition (same vs. different items) and priming (old vs. new items). In the majority of regions, the combined

effects of fMRI adaptation and priming were additive (B, D, E, F). The only region showing a significant interaction between both effects was the left fusiform

gyrus (C). Error bars represent standard errors in each condition for peak activation data.

T. Ganel et al. / NeuroImage 32 (2006) 1432–14401436

our ROIs can be linked to the number of stimulus presentations

(see Fig. 3). However, a closer examination of the pattern of

activation in specific cells rules against this possibility. First,

according to a repetitions account, there should be no differences in

activation between the Fold/different_ trials and the Fnew/same_trials. However, the observed pattern of activation revealed a

difference in activation in all ROIs, with numerically larger

activation in the Fold/different_ than in the Fnew/same_ trials.

Indeed, this difference was significantly larger in the rLO (t(14) =

2, P < .05, one tailed), in the lLO (t(14) = 3.1, P < .01), and in the

lFG (t(14) = 4.07, P < .005). The second piece of evidence against

a simple repetitions account comes from an examination of the

difference in activation between the Fold/different_ and Fold/same_trials, as compared to the difference between Fnew/different_ andFnew/same_ trials. If number of presentations were the most critical

variable that mediated our results, then the two contrasts should be

different, leading to an interaction between adaptation and priming.

However, in five out of our six ROIs, we did not observe a

significant interaction. Moreover, even in the one ROI that did

show an interaction between adaptation and priming (i.e., the lFG,

Fig. 3C), the specific pattern of activation did not match what

would be expected based on a simple repetitions account; notably,

activation in this region was significantly larger, rather than smaller

or equal, for Fold/different_ as compared to Fnew/same_ trials.

Together, these results argue strongly against a simple account of

the presented results in terms of variations in the number of

repetitions across priming and adaptation conditions. If variations

in repetitions played any role, one would have to assume that they

interacted with the repetition lag in order to account for the full

pattern of data observed.

Given that BOLD responses in several ROIs mirrored the

pattern of RTs observed in participants_ behavior, a second

potential confound in our design might be that the reported BOLD

responses in these areas were driven by generic, non-specific

variations in RTs, rather than reflecting the two specific types of

repetition effects manipulated. To explore this possibility, we re-

analyzed the fMRI data by adding non-specific RT as an additional

independent variable. Considering RTs separately for every

subject, every run, and in every experimental condition (i.e., old/

same, old/different, new/sale, new/different), we categorized trials

Fig. 4. A voxelwise GLM analysis of interaction between adaptation and priming. To complete our regions of interest analysis, we performed a voxelwise GLM

analysis of interaction between adaptation and priming (random-effects group analysis, P < .0005, uncorrected). The only area that showed a pattern of

interaction between adaptation and priming within occipital and temporal object-selective regions was within the left fusiform gyrus (Talairach x y z coordinates

�32, �37, �19). This finding is in agreement with the regions of interest analysis, presented in Fig. 3.

T. Ganel et al. / NeuroImage 32 (2006) 1432–1440 1437

as Ffast_ versus Fslow_ based on the corresponding median RT. This

categorization of trials allowed us to enter RT as a third

independent factor into the ANOVA-based analyses for all ROIs.

We then determined whether RT was related to the BOLD effect in

any region, either by way of a statistical main effect or an

interaction. Importantly, we also examined whether the adaptation

and priming effects revealed with our previous analyses remained

significant once variations in RT unrelated to the experimental

repetition manipulations were taken into consideration.

For five of the six ROIs examined, we found neither a main

effect of RT nor an interaction between RT and the other two

factors. Importantly, the effects obtained for adaptation and

priming in these regions were virtually identical to those obtained

in the previous set of analyses, which did not take RT into account.

The only region that showed an effect of RT (higher activation for

slow as compared to fast responses) was the lFG, in which a

significant main effect of RT, F(1,14) = 5.9, P < .05, was observed.

Even in this region, however, the effects for the other factors

paralleled those reported in the previous analyses, most notably

reflected in the significant interaction between adaptation and

priming, F(1,14) = 5.7, P < .05. Thus, variations in RT that were

unrelated to the experimental repetition manipulations did not

interact with adaptation or priming in any of the ROIs examined,

suggesting that they did not confound the corresponding fMRI

results (see Sayres and Grill-Spector, 2006 for similar conclusions).

However, the significant main effect of RT observed for lFG

suggests that, in this part of the ventral visual pathway, adaptation,

priming, and general task demands cannot easily be distinguished.

We will return to the significance of this finding in the Discussion

section.

Discussion

The purpose of the present study was to test the relationship

between fMRI adaptation and repetition priming in occipital and

temporal object-selective regions, in which repetition-related

reductions in the hemodynamic response have previously been

observed (Henson, 2003). All regions that we examined showed

fMRI adaptation effects, and all regions except the left LO showed

Fig. 5. Brain areas sensitive to fMRI adaptation and to priming within occipital and temporal cortex. Brain areas sensitive to fMRI adaptation (blue) and

repetition priming (yellow) were identified using a voxelwise GLM (random-effects group analysis, for presentation reasons, the P level displayed in the figure

for both adaptation and priming was set to P < .005). Although some overlap (green) between adaptation- and priming-sensitive regions was found in fusiform

and parahippocampal gyri, fMRI adaptation effects were more pronounced in posterior regions and repetition priming effects were more pronounced in anterior

medial regions.

T. Ganel et al. / NeuroImage 32 (2006) 1432–14401438

reduced activation following repetition priming. The fact that fMRI

adaptation but not repetition priming was observed in the left LO is

the first piece of evidence suggesting that these two phenomena

can be dissociated from one another.

A second piece of evidence supporting the idea that the two

phenomena are independent comes from the observation that the

reduced activation effects in all but one of the regions that showed

priming effects and fMRI adaptation were additive rather than

interactive. Together, these two findings suggest that the neural

mechanisms underlying fMRI adaptation and those underlying

repetition priming are, at least in some brain regions, different from

each another. One possible reason for why different neural

mechanisms might be invoked relates to differences in the temporal

characteristics of the two phenomena. There is evidence from

event-related potential studies showing that the ERP component

related to immediate repetition occurs earlier than the component

related to long-term repetition (Henson et al., 2004; Nagy and

Rugg, 1989). The changes in activation associated with fMRI

adaptation in our study could reflect this early component, whereas

those associated with repetition priming could reflect the later

component. Indeed, it may be the case that fMRI adaptation

reflects a rapid and short-lasting effect related to changes in

neuronal firing whereas repetition priming engages slower and

more stable changes in synaptic efficacy of the neuronal population

(see also Henson, 2003). At present, however, this must remain a

speculation given that fMRI does have neither sufficient temporal

nor spatial resolution to determine the timing and exact nature of

the neural mechanisms underlying changes in activation. Never-

theless, the additive factor logic that we applied (Sternberg, 1969)

does allow us to conclude that the neural mechanisms underlying

fMRI adaptation and repetition priming are largely independent

across occipital and temporal cortex.

It is informative to compare the present findings with those

reported by Henson et al. (2004), who found larger repetition-

related decreases in activation in a number of occipito-temporal

regions for short as compared to long lags in a priming study.

Although Henson et al.’s experimental design differed from the one

we used in several respects, it is worth noting that some aspects of

the present results could indeed be interpreted as ‘‘lag’’ effects in

line with Henson et al.’s findings. In particular, we found

significantly larger effects of adaptation, involving short lags, as

compared to priming, involving longer lags, in the left LO and the

left FG. Further evidence that can be considered to be in line with

Henson’s findings comes from our more descriptive whole-brain

analysis. Fig. 5 illustrates that regions in the occipotemporal cortex

that showed sensitivity to adaptation were larger in extent than

those that showed sensitivity to priming. Together, these similar-

ities in findings across the two studies suggest that the repetition

lag, i.e., the length of the delay between a first and a repeated

stimulus encounter, is an important factor to consider in

understanding the differences in mechanisms that underlie priming

and fMRI adaptation.

Only one object-selective region, which was located in left FG,

showed interactive rather than additive effects of fMRI adaptation

and repetition priming. It has previously been suggested that

memory representations of objects in this region (but not in right

FG nor in LO in either hemisphere) represent the semantic

attributes of visual stimuli (Koutstaal et al., 2001, Simons et al.,

2003). Because the size judgment task we used in the present

experiment required semantic processing of visual objects and

because such processing was arguably more critical for different

than for same trials (see below), it is possible to interpret the

interaction we observed in left FG as being related to a differential

involvement of semantic processes across task conditions. How-

ever, it is worth noting that the left FG was also the only region that

showed a significant main effect of RT in the current study. Thus,

the pattern of activation we observed in this region may

alternatively be linked to non-specific effects of task difficulty.

Although the present results do not allow us to rule out this latter

interpretation, recent research by Sayres and Grill-Spector (2006)

T. Ganel et al. / NeuroImage 32 (2006) 1432–1440 1439

suggests that repetition-related reductions in activation in FG can

clearly be distinguished from general task demands reflected in RT

in the context of other semantic classification tasks. Nevertheless,

these authors did not report their data separately for left and right

FG. Therefore, further research will be necessary to determine with

more certainty the relationship between the neural mechanisms of

fMRI adaptation, repetition priming, and RT variability that is

unrelated to repetition in left FG.

The results of the whole-volume analysis in which we

compared the effects of fMRI adaptation and priming across the

temporal and occipital cortex more broadly converge with the

findings from our ROI analysis, providing additional support for

the idea that the two phenomena are dissociable. Specifically, an

interaction effect between adaptation and priming was only found

in the left FG. Furthermore, despite the observation of overlap in

several regions, fMRI adaptation was more pronounced in occipital

regions, extending from occipital and medial fusiform areas to

more anterior parahippocampal areas. Repetition priming effects,

on the other hand, were more pronounced in anterior occipotem-

poral regions, extending from medial fusiform areas to more

anterior parahippocampal and hippocampal areas.

At first glance, it may appear surprising that we found hippocampal

activation in association with repetition priming, given that priming, as

a form of implicit memory, is generally thought to be preserved in

patients suffering from amnesia followingmedial temporal damage (for

a review, see Schacter and Buckner, 1998). However, functional

neuroimaging research and studies in non-human species have

implicated the hippocampus in processes related to novelty assessment

that are thought to take place even under conditions inwhich no explicit

memory judgments are required (for a discussion, see Kohler et al.,

2005; Nyberg, 2005); the invocation of such novelty detection

processes could be reflected in the pattern of activation we observed.

Our finding that this region was not sensitive to the immediate effects

associated with fMRI adaptation suggests that the putative novelty

assessment processes in the hippocampus operate over longer time

scales. Thus, this finding provides further support for the involvement

of the hippocampus in long-term memory (Milner et al., 1998). Due to

the exploratory nature of the analysis that revealed this finding,

however, such a conclusion must remain tentative.

Alternatively, the hippocampal activation we observed could be

related to the influence of explicit contamination, i.e., the explicit

recognition of primed items in the context of our size judgment task

during scanning. Behavioral evidence from cognitive studies

suggests that such explicit contamination is frequently present and

cannot easily be prevented on purportedly implicit priming tasks

(e.g., Bowers and Schacter, 1990). Because the issue has not been

systematically explored in functional neuroimaging research as of

yet, however, little is known about the impact of conscious

contamination on the neural correlates of repetition priming. To

the extent that the hippocampal activation we observed does indeed

reflect explicit contamination, the present results highlight the

importance of considering that priming tasks, as well as the fMRI

activations associated with them, are typically not process-pure.

This notion has long been recognized in the cognitive literature but

has not been given sufficient consideration in fMRI studies (see

Henson, 2003). We think, however, that it does not invalidate the

main conclusion of the present study, namely that the neural

mechanisms involved in priming, inasmuch as they are captured

with a typical fMRI priming paradigm, can be dissociated from those

involved in fMRI adaptation. Rather, a consideration of explicit

contamination processes promises to provide a deeper understand-

ing of the significance of such differential neural mechanisms at the

cognitive level in the present and in future research.

It is important to note that a major challenge we faced in

comparing adaptation and priming was to develop a paradigm in

which the conditions were as similar as possible to the conditions

used in earlier fMRI experiments that examined these phenomena

separately. Thus, because we addressed both priming and

adaptation in the same experiment, we could not unpack issues

such as whether or not the priming effect occurs at the level of

individual objects or pairs of objects. Additional research, in which

objects are presented in different combinations of pairs, is required

to fully address this issue.

An additional methodological aspect of our paradigm to

consider is the role of task difficulty, which may have modulated

adaptation effects in the present study. One could argue that

‘‘same’’ trials are easier to perform than ‘‘different’’ trials given that

answers for the former can be based on perceptual information

present in the stimuli whereas the latter cannot. We used a variant

of the same/different task in our study because same/different tasks

are the most widely used tasks employed in the fMRI adaptation

literature (e.g., Grill-Spector and Malach, 2001; Kourtzi et al.,

2003); using it allowed for the most valid comparisons between our

results and those reported by others in this literature. There are

several arguments that can be made against an account of the

presented adaptation findings in terms of task difficulty. First, it has

been shown elsewhere that valid adaptation effects for objects can

be found even when attention is not directed to the objects

themselves (Murray and Wojciulik, 2004). Second, there are

noticeable similarities between the effects of adaptation in the

present study and those of repetition at short lags in previous

research in which other tasks were used (Henson et al., 2004).

Finally, it is important to note that even accepting task difficulty as

a factor that may have influenced our fMRI adaptation results

would not put into question our main conclusions. This is because

they rely on the examination of the combined effects of adaptation

and priming rather than on a comparison of the magnitude of each

effect in isolation. The relative size of the adaptation and priming

effects, which are not central to our conclusions, should be treated

more carefully given that they may be the result of specific task

demands including task difficulty. Regardless, the issue of task

difficulty clearly requires further systematic investigation in future

studies that aim to compare the neural correlates of repetition

priming and fMRI adaptation in more detail.

To summarize, our study provides novel evidence that fMRI

adaptation can be dissociated from the neural correlates of

repetition priming captured with typical fMRI priming paradigms.

The strongest support for this conclusion comes from our finding

that the combined effects of fMRI adaptation and repetition

priming were additive in the majority of object-selective brain

regions. Furthermore, we found that some brain regions show one

effect but not the other. We conclude, therefore, that, although

fMRI adaptation and repetition priming are both typically reflected

in reduced activation in occipital and temporal cortex regions, this

does not imply that they rely on the same neural mechanism.

Acknowledgments

This work was supported by grants from the Canadian Institutes for

Health Research (CIHR) to SK andMAG and a grant from the Natural

Sciences and Engineering Research Council of Canada to SK.

T. Ganel et al. / NeuroImage 32 (2006) 1432–14401440

References

Bentin, S., Moscovitch, M., 1988. The time course of repetition effects for

words and unfamiliar faces. J. Exp. Psychol. Gen. 117, 148–160.

Bowers, J.C., Schacter, D.L., 1990. Implicit memory and test awareness. J.

Exper. Psychol., Learn., Mem., Cogn. 16, 404–416.

Boynton, G.M., Engel, S.A., Glover, G.H., Heeger, D.J., 1996. Linear

systems analysis of functional magnetic resonance imaging in human

V1. J. Neurosci. 16, 4207–4221.

Dobbins, I.G., Schnyer, D.M., Verfaellie, M., Schacter, D.L., 2004. Cortical

activity reductions during repetition priming can result from rapid

response learning. Nature 428, 316–319.

Grill-Spector, K., Malach, R., 2001. fMR-adaptation: a tool for studying the

functional properties of human cortical neurons. Acta Psychol. (Amst.)

107, 232–293.

Grill-Spector, K., Kushnir, T., Edelman, S., Avidan, G., Itzchak, Y.,

Malach, R., 1999. Differential processing of objects under various

viewing conditions in the human lateral occipital complex. Neuron

24, 187–203.

Henson, R.N.A., 2003. Neuroimaging studies of priming. Prog. Neurobiol.

70, 53–81.

Henson, R.N.A., Shallice, T., Dolan, R., 2000. Neuroimaging evidence for

dissociable forms of repetition priming. Science 287, 1269–1272.

Henson, R.N.A., Rylands, A., Ross, E., Vuilleumeir, P., Rugg, M.D., 2004.

The effect of repetition lag on electrophysiological and haemodynamic

correlates of visual object priming. NeuroImage 21, 1674–1689.

Kohler, S., Danckert, S., Gati, J.S., Menon, R.S., 2005. Novelty responses

to relational and non-relational information in the hippocampus and the

parahippocampal region: a comparison based on event-related fMRI.

Hippocampus 15, 763–774.

Kourtzi, Z., Erb, M., Grodd, W., Bulthoff, H.H., 2003. Representation of

the perceived 3-D object shape in the human lateral occipital complex.

Cereb. Cortex 13, 911–920.

Koutstaal, W., Wagner, A.D., Rotte, M., Maril, A., Buckner, R.L., Schacter,

D.L., 2001. Perceptual specificity in visual object priming: functional

magnetic resonance imaging evidence for a laterality difference in

fusiform cortex. Neuropsychologia 39, 184–199.

Milner, B., Squire, L.R., Kandel, E.R., 1998. Cognitive neuroscience and

the study of memory. Neuron 20, 445–468.

Murray, S.O., Wojciulik, E., 2004. Attention increases neural selectivity in

the human lateral occipital complex. Nat. Neurosci. 7, 70–74.

Nagy, M.E., Rugg, M.D., 1989. Modulation of event-related potentials by

word repetition: the effects of inter-item lag. Psychophysiology 26,

431–436.

Norris, C.J., Chen, E.E., Zhu, D.C., Small, S.L., Cacioppo, J.T., 2004. The

interaction of social and emotional processes in the brain. J. Cogn.

Neurosci. 16, 1818–1829.

Nyberg, L., 2005. Any novelty in hippocampal formation and memory?

Curr. Opin. Neurol. 18, 424–428.

Ogawa, S., Tank, D.W., Menon, R., Ellermann, J.M., Kim, S.G., Merkle,

H., Ugurbil, K., 1992. Intrinsic signal changes accompanying sensory

stimulation: functional brain mapping with magnetic resonance imag-

ing. Proc. Natl. Acad. Sci. U. S. A. 89, 5951–5955.

Pinel, P., Dehaene, S., Riviere, D., LeBihan, D., 2001. Modulation of

parietal activation by semantic distance in a number comparison task.

NeuroImage 14, 1013–1026.

Sayres, R., Grill-Spector, K., 2006. Object-selective cortex exhibits

performance-independent repetition suppression. J. Neurophysiol. 95,

995–1007.

Schacter, D.L., 1987. Implicit memory: history and current status. J. Exp.

Psychol. Learn. Mem. Cogn. 13, 501–518.

Schacter, D.L., Buckner, R.L., 1998. Priming and the brain. Neuron 20,

185–195.

Schacter, D.L., Dobbins, I.G., Schnyer, D.M., 2004. Specificity of

priming: a cognitive neuroscience perspective. Nat. Rev., Neurosci. 5,

853–862.

Simons, J.S., Koutstaal, W., Prince, S., Wagner, A.D., Schacter, D.L., 2003.

Neural mechanisms of visual object priming: evidence for perceptual

and semantic distinctions in fusiform cortex. NeuroImage 19, 613–626.

Sternberg, S., 1969. The discovery of processing stages: extensions of

Donders’ method. Acta Psychol. 30, 276–315.

Sternberg, S., 2001. Separate modifiability, mental modules and the use of

pure and composite measures to reveal them. Acta Psychol. 106, 147–

246.

Talairach, J., Tournoux, P., 1998. Co-Planar Stereotaxic Atlas of the Human

Brain. 3-Dimensional Proportional System: An Approach to Imaging.

Thieme, New York.

Wiggs, C.L., Martin, A., 1998. Properties and mechanisms of perceptual

priming. Curr. Opin. Neurobiol. 8, 227–233.