brain depth of processing effects on neural correlates of memory...

Brain (2001), 124, 399–412

Depth of processing effects on neural correlates ofmemory encodingRelationship between findings from across- and within-taskcomparisons

Leun J. Otten,1 Richard N. A. Henson1,2 and Michael D. Rugg1

1Institute of Cognitive Neuroscience and Department of Correspondence to: Leun J. Otten, Institute of CognitivePsychology, University College London and 2Wellcome Neuroscience, 17 Queen Square, London WC1N 3AR, UKDepartment of Cognitive Neurology, Institute of Neurology, E-mail: [email protected], UK

SummaryNeuroimaging studies have implicated the prefrontalcortex and medial temporal areas in the successfulencoding of verbal material into episodic memory. Thepresent study used event-related functional MRI toinvestigate whether the brain areas associated withsuccessful episodic encoding of words in a semanticstudy task are a subset of those demonstrating depth ofprocessing effects. In addition, we tested whether thebrain areas associated with successful episodic encodingdiffer depending on the nature of the study task. At study,15 volunteers were cued to make either animacy oralphabetical decisions about words. A recognition memorytest including confidence judgements followed after adelay of 15 min. Prefrontal and medial temporal regionsshowed greater functional MRI activations forsemantically encoded words relative to alphabeticallyencoded words. Two of these regions (left anteriorhippocampus and left ventral inferior frontal gyrus)

Keywords: event-related fMRI; depth of processing; episodic encoding; hippocampus; prefrontal cortex

Abbreviations: BA � Brodmann area; BOLD � blood oxygenation level-dependent; EPI � echoplanar imaging;fMRI � functional MRI; HRF � haemodynamic response function; MNI � Montreal Neurological Institute; RT � reaction time

IntroductionNeuropsychological evidence indicates that episodic memory(explicit memory for recently experienced events) dependscritically upon a number of circumscribed, interconnectedbrain regions, including the medial temporal lobe and theprefrontal cortex (Squire and Knowlton, 2000). Whereasinvestigation of the effects of brain lesions can identifyregions whose integrity is necessary for episodic memory,the specification of the functional role of these regions isless easy. In particular, it is difficult on the basis of behaviouralevidence alone to distinguish between brain regions that play

© Oxford University Press 2001

showed greater activation for semantically encoded wordsthat were subsequently recognized confidently. However,other regions (left posterior hippocampus and rightinferior frontal cortex) demonstrated subsequent memoryeffects, but not effects of depth of processing. Successfulmemory for alphabetically encoded words was alsoassociated with greater activation in the left anteriorhippocampus and left ventral inferior frontal gyrus. Thefindings suggest that episodic encoding for words in asemantic study task involves a subset of the regionsactivated by deep relative to shallow processing. The dataprovide little evidence that successful episodic encodingduring a shallow study task depends upon regionsdifferent from those that support the encoding of deeplystudied words. Instead, the findings suggest that successfulepisodic encoding during a shallow study task relies on asubset of the regions engaged during successful encodingin a deep task.

a role in the encoding as opposed to the retrieval of memories(Fletcher et al., 1997).

The question of which neural structures are associatedwith successful memory encoding can be addressed moredirectly with functional neuroimaging methods. While thesemethods do not permit conclusions to be drawn about whethera given brain region is necessary for normal memory function(Rugg, 1999), they provide a way to determine whether theregion is active during encoding, retrieval, or both. In thestudy described in the present paper, event-related functional

400 L. J. Otten et al.

MRI (fMRI) was used to identify regions associated withsuccessful memory encoding during two tasks that differedin their semantic processing demands.

Most previous neuroimaging studies of memory encodinghave employed blocked experimental designs and searchedfor putative neural correlates of encoding by contrasting theactivities associated with different study tasks. It has beenreported consistently that tasks promoting relatively goodmemory performance, e.g. intentional learning versus reading(Kapur et al., 1996; Kelley et al., 1998), semantic versusnon-semantic classification (Kapur et al., 1994; Demb et al.,1995; Wagner et al., 1998b, experiment 1), and full versusdivided attention (Shallice et al., 1994) are associated withgreater activity in several regions of left prefrontal cortex(for review, see Buckner et al., 1999). Findings from morerecent studies, in which non-verbal as well as verbal itemswere employed, add further weight to the notion that theprefrontal cortex plays a role in memory encoding, andindicate that the lateralization of encoding-related prefrontalactivity is material-specific (e.g. Kelley et al., 1998; Wagneret al., 1998a; McDermott et al., 1999).

In contrast to findings for the prefrontal cortex,neuroimaging studies have been less consistent indemonstrating encoding-related activation in the medialtemporal lobe. Few of the studies cited above (for exceptions,see Kelley et al., 1998; Wagner et al., 1998b, experiment 1)reported that enhanced prefrontal activity was accompaniedby enhanced medial temporal activity, and in one study(Dolan and Fletcher, 1997) activity in hippocampal andprefrontal regions was found to dissociate. One procedurethat has yielded robust effects in the medial temporal lobeinvolves a contrast between blocks of trials wherein the sameitem is repeated and blocks in which items are trial-unique(Stern et al., 1996; Gabrieli et al., 1997). Blocks containingunique items elicited enhanced activity in the para-hippocampal cortex (but not the hippocampus). This findingwas interpreted as evidence for the role of theparahippocampal region in the encoding of contextually novelstimuli (for a similar argument in respect of the hippocampus,see Dolan and Fletcher, 1997; Saykin et al., 1999).

A problem in the interpretation of studies such as thosecited above arises from the use of across-task or across-blockcontrasts to reveal activity that may be related to memoryencoding. Whereas some of the effects revealed by suchcontrasts may indeed reflect differences in the effectivenesswith which information is encoded into memory, other effectsmay arise because of differences between conditions thathave nothing to do with memory encoding operations. Thusthe argument linking, say, the left prefrontal cortex withepisodic encoding is an indirect one.

An experimental approach that avoids the foregoingproblem involves contrasts not between different encodingtasks, but between different classes of item presented duringa single task. This approach was adopted in several studies(Cahill et al., 1996; Alkire et al., 1998; Fernandez et al.,1998, 1999a) in which it was demonstrated that the magnitude

of activity in one or more medial temporal structures at thetime of encoding predicted memory performance either across(Cahill et al., 1996; Alkire et al., 1998) or within (Fernandezet al., 1998, 1999a) subjects. Because data were obtainedover blocks of items, however, the findings offer little insightinto encoding processes supported by the medial temporallobe that operate at the level of individual items.

With the advent of event-related fMRI (Dale and Buckner,1997; Josephs et al., 1997; Zarahn et al., 1997) it isnow possible to obtain functional neuroimaging data at theindividual item level. To study memory encoding, item-related activity at study is contrasted according to whetheritems are remembered or forgotten in a subsequent memorytest. Differences between the activities associated withremembered and forgotten items (subsequent memory effects)are then taken as candidate neural correlates of encoding.This approach has been used in numerous event-related brainpotential studies of memory encoding over the past 20 yearsor so (for reviews, see Rugg, 1995; Wagner et al., 1999).

Few studies have employed the event-related fMRItechnique to investigate subsequent memory effects forindividual words (Wagner et al., 1998b; Henson et al., 1999;Kirchhoff et al., 2000; Buckner et al., 2001; for a similarstudy involving pictures, see Brewer et al., 1998). Hensonand colleagues reported that study words subsequently judgedas ‘remembered’ elicited greater left prefrontal activity[Brodmann area (BA) 9/44] than did words judged as ‘known’(Henson et al., 1999), a finding consistent with the view thatthe left prefrontal cortex supports processes important forepisodic encoding. Similar findings were reported by Bucknerand colleagues (Buckner et al., 2001) and Kirchhoff andcolleagues (Kirchhoff et al., 2000).

Wagner and colleagues employed a concrete/abstractdiscrimination at encoding and tested recognition memorysome 20 min later (Wagner et al., 1998b). Relative to wordsthat were subsequently unrecognized, words recognized withhigh confidence were associated at encoding with activationof the left prefrontal cortex (BA 44, 45 and 47) and the leftparahippocampal/fusiform cortex (BA 35/36 and 37). Theseregions were a subset of the regions activated in a separate,blocked, experiment in which the same study task wascontrasted with a non-semantic task (case judgement). In thelight of the overlap between the regions activated in thesemantic versus non-semantic contrast, and the regionsdemonstrating a within-task subsequent memory effect,Wagner and colleagues argued that many of the cognitiveoperations mediating the ‘depth of processing’ effect onmemory performance were the same as those supportingeffective encoding during the performance of ‘deep’ encodingtasks (Wagner et al., 1998b; see also Tulving et al., 1994).

The present study takes as its starting point the findingsof Wagner and colleagues (Wagner et al., 1998b). Twoprincipal questions are at issue. First, if subsequent memoryand depth of processing effects are compared within subjectsduring the same experimental procedure, do the former effectsstill represent a subset of the latter? Secondly, to what degree

Neural correlates of episodic encoding 401

do subsequent memory effects differ according to the natureof the encoding task used to reveal them? Specifically, do theeffects found previously for semantic study tasks generalize tonon-semantic tasks such as those typically used to engage‘shallow’ encoding operations, or is it the case that the neuralactivity supporting effective encoding in the two classes oftask differs qualitatively? This question, which has not beenaddressed systematically in previous studies, is importantbecause it bears on whether episodic encoding for a givenclass of items relies on a single neural system irrespectiveof study task, or whether encoding is supported by multiple,task-specific, systems.

We addressed these questions with an experimental designin which event-related fMRI images were obtained whilesubjects performed two interleaved incidental encoding tasks,one of which required a semantic discrimination and theother a non-semantic discrimination. To identify overlapbetween depth of processing and subsequent memory effects,we compared the outcome of the contrast between the twotasks (the depth of processing effect) with the contrastbetween semantically encoded words that were subsequentlyremembered versus forgotten (the deep subsequent memoryeffect). To investigate the dependence of subsequent memoryeffects on the nature of the encoding task, we contrastedsemantic (deep) with non-semantic (shallow) subsequentmemory effects.

MethodsParticipantsThe experimental procedures were approved by the NationalHospital for Neurology and Neurosurgery and Institute ofNeurology Medical Ethics Committee. Fifteen volunteers(nine women) were recruited via local advertisements. Meanage was 26 (range 20–31) years. All volunteers were nativespeakers of English and gave their informed consent prior tothe experiment. All but one volunteer was right-handedaccording to self-report. All subjects claimed to be in goodhealth and to be free from neurological and psychiatricproblems.

Stimulus materialsThe stimulus lists were constructed from a pool of 560 wordsranging in frequency between 1 and 30 per million (Kuceraand Francis, 1967) and in length between four and nineletters. Three sets of 140 words each were selected at randomfrom this pool, with the restriction that (i) the distribution ofword lengths was identical across the three sets, and (ii)within each set, 35 words were animate and ‘alphabetical’(i.e. their first and last letters were in alphabetical order), 35animate and non-alphabetical, 35 inanimate and alphabetical,and 35 inanimate and non-alphabetical. These sets were usedto form three different study–test blocks by rotating the sets

across the animacy, alphabetical and new conditions. Avolunteer saw one of the three possible blocks.

For each study list, a random sequence was generatedfrom the 140 words used for animacy decisions and 140words used for alphabetical decisions. The study sequencewas divided into two blocks of 140 trials, and two fillerwords were added to the beginning of each block. For eachtest list, a random sequence was generated from the 140words used for animacy decisions, 140 words used foralphabetical decisions, and 140 new words. The test sequencewas divided into four blocks of 105 words each, and twofiller words were added at the beginning of each block. Anadditional 24 words were selected from the word pool tocreate a practice list for the study task.

TaskThe experiment was composed of an incidental study taskfollowed by a recognition memory test after a delay of15 min. During the study task, volunteers saw a series of280 critical words, presented one at a time. Each word waspreceded by a prestimulus cue, which consisted of thepresentation of either the letter ‘O’ or the letter ‘X’. Thetype of cue indicated whether the decision about the upcomingword should be based on the semantic (animacy decision) ornon-semantic (alphabetical decision) properties of the word;this decision was signalled by a button press. Animacy andalphabetical decisions were equiprobable and were intermixedrandomly. The words were presented visually for 300 ms ina white upper-case Helvetica 48-point font on a blackbackground. The time between successive word onsets was4.8 s. At a viewing distance of ~30 cm, words subtended avertical visual angle of ~1.4°, and a horizontal visual angleof 4–10°. The prestimulus cue was presented for 2 s andmeasured 1.4 � 1.2° of visual angle. There was a 100-msblank period between the offset of the cue and the onset ofthe word.

The recognition memory test consisted of the re-presentation of the 140 semantically judged words and 140alphabetically judged words from the study task, along with140 words not seen before during the experiment. A plussign presented before each word served as a fixation pointand warning stimulus. For each word, volunteers had todecide whether they had seen the word before during theexperiment (old/new judgement), indicating whether theywere confident or non-confident about their decision. Thewords were presented visually for 300 ms in a white upper-case Helvetica 24-point font on a black background. Thetime between successive word onsets was 4.8 s. At a viewingdistance of ~50 cm, words subtended a vertical visual angleof ~0.5° and a horizontal visual angle of 1.5–3.5°. Thewarning stimulus was presented for 2 s and measured0.5 � 0.4° of visual angle. There was a 100-ms blank periodbetween the offset of the warning stimulus and the onset ofthe word.


ProcedureScanning took place during the study task only. Beforeentering the scanner, volunteers were given an explanationabout the study task. They were told that they would seewords, presented one at a time, and that each word wouldbe preceded either by the letter ‘O’ or by the letter ‘X’.When an ‘O’ preceded a word, they had to decide whetheror not the word was animate or referred to the property of aliving entity. When an ‘X’ preceded a word, they had todecide whether or not the first and last letters of the wordwere in alphabetical order. Half of the volunteers were askedto make a left thumb response if a word was animate oralphabetical, and a right thumb response if a word wasinanimate or non-alphabetical. For the other half, this responseassignment was reversed. Both speed and accuracy werestressed. Volunteers undertook a short practice session tofamiliarize themselves with the study task. They were notinformed about the nature of the task that would be performedsubsequently outside the scanner.

Scanning began with a 15-min structural scan. Volunteersthen performed two blocks of 142 trials each of the studytask (~12 min each), during which the functional scans wereacquired. A short rest was given between the blocks. Thewords were projected on to a mirror in direct view of thereclining volunteer, and responses were given with a hand-held response box.

The memory test was administered ~15 min after thecompletion of the study task. Volunteers were taken to anotherroom, where they rested and conversed with the experimenterduring the delay period. They were then informed about theold/new recognition memory task. It was explained that theywould again see a sequence of words, presented one at atime, and that some of the words had been presented duringthe task they had performed in the scanner. For each word,volunteers had to decide whether or not the word had beenpresented during the study task. In addition, they wererequired to indicate whether they were confident or non-confident about their decision. One of four keys had to bedepressed according to whether the word was confidentlyjudged to be old, non-confidently judged to be old, confidentlyjudged to be new, or non-confidently judged to be new. Nospecific instruction was given about how confident someoneshould be before pressing the ‘confident’ key. Responseswere given with the middle and index fingers of the left andright hands, which rested on the ‘r’, ‘g’, ‘k’, and ‘o’ keys ofa keyboard placed on a table in front of the volunteer.Confident responses were always given with the middlefingers (‘r’ and ‘o’ keys). The assignment of old responsesto the left or right hand was counterbalanced across subjects.Volunteers were instructed to respond as fast as possiblewithout sacrificing accuracy.

Four test blocks of 107 trials, each ~8 min in duration,were undertaken with short rests between the blocks. At thecompletion of the test blocks, volunteers were debriefedabout the nature of the experiment and paid for their time.

MRI scanning methodsA 2 T Siemens Vision system (Siemens, Erlangen, Germany)was used to acquire both T1-weighted anatomical volumeimages [1 � 1 � 1.5 mm voxels, MPRAGE (magnetization-prepared, rapid acquisition gradient echo) sequence] and T2*-weighted echoplanar (EPI) images (64 � 64, 3 � 3 mmpixels, echo time � 40 ms) with blood oxygenation level-dependent (BOLD) contrast. Each EPI volume comprised 31axial slices, 2 mm thick and separated by 1.5 mm, positionedto cover all but the most superior region of the brain and thecerebellum. Data were acquired during two sessions, eachcomprising 250 volumes, corresponding to the two studyblocks. Volumes were acquired continuously with an effectiverepetition time of 2.85 s/volume. The first five volumes werediscarded to allow for T1 equilibration effects. The constantinterstimulus interval of 4.8 s allowed an effective samplingrate of the haemodynamic response of 6.7 Hz.

PreprocessingFor each volunteer, all volumes in a session were realignedto the first volume and resliced using a sinc interpolation inspace. To correct for their different acquisition times, thesignal measured in each slice was then shifted relative to theacquisition of the middle slice using a sinc interpolation intime. Each volume was normalized to a standard EPI templatevolume [based on the Montreal Neurological Institute (MNI)reference brain (Cocosco et al., 1997)] of 3 � 3 � 3 mmvoxels in Talairach and Tournoux space (Talairach andTournoux, 1988), using non-linear basis functions. Finally,the EPI volumes were smoothed with an 8 mm full-widthhalf-maximum isotropic Gaussian kernel to accommodateresidual anatomical differences across volunteers, andproportionally scaled to a global mean of 100.

Data analysisThe data were analysed by statistical parametric mapping(Friston et al., 1995), using the SPM99b program (WellcomeDepartment of Cognitive Neurology, London, UK). Thevolumes acquired during each session were treated as twotime series. The haemodynamic response to the onset of eachevent type of interest was modelled with two basis functions:a canonical haemodynamic response function (HRF) (Fristonet al., 1998) and a delayed HRF (Henson et al., 2001),shifted to onset 2.85 s (i.e. one repetition time) later than thecanonical HRF. The use of both an early and a late responsefunction was based on suggestions from published reportsthat the time of maximal activation is later for some brainregions (e.g. the prefrontal cortex) than the sensory regionson which the canonical HRF is based (Wilding and Rugg,1996; Schacter et al., 1997; Buckner et al., 1998). The earlyand late response functions, when convolved with a sequenceof delta functions representing the onset of each event,comprised the covariates in a general linear model, together


with a constant term for each session. The covariates for thelate HRF were orthogonalized with respect to those for theearly HRF using a Gram–Schmidt procedure so as to givepriority to the early covariate (Andrade et al., 1999). Thisorthogonalization attributes variance common to the earlyand late covariates to the early covariate; loadings on theorthogonalized late covariate account for part of the residualvariance in the data not explained by the early covariate.The data were high-pass filtered to a maximum of 1/120 Hz,and both model and data were smoothed temporally with a4 s full-width half-maximum Gaussian kernel. Parameterestimates for each covariate were calculated from the leastmean squares fit of the model to the data.

Planned contrasts (specified in the Results section) wereemployed to test parameter estimates for both early and latecovariates. The linear combinations of parameter estimatesfor each contrast were stored as separate images for eachparticipant. These contrast images were entered into one-sample t-tests to permit inferences about condition effectsacross subjects (i.e. a random effects analysis). The imageswere subsequently transformed into statistical parametricmaps of the Z statistic. Unless mentioned otherwise, contrastswere thresholded at P � 0.001, uncorrected for multiplecomparisons. When reporting masked contrasts, the Z valuesrefer to the outcome of the masked contrast only. Onlyactivations involving contiguous clusters of at least fivevoxels were interpreted. We report the results from the latecovariate only when they add meaningfully to the findingsfrom the early covariate. The maxima of suprathresholdregions were localized by rendering them onto both thevolunteers’ normalized structural images and the MNIreference brain (Cocosco et al., 1997). They were labelledusing the stereotactic system and nomenclature of Talairachand Tournoux (Talairach and Tournoux, 1988).

ResultsBehavioural performanceStudy taskAnimacy decisions were made with an accuracy of 94%(SD � 3) and a mean reaction time (RT) of 955 ms(SD � 159). Alphabetical decisions were made with anaccuracy of 92% (SD � 4) and a mean RT of 1380 ms(SD � 288). RTs were significantly longer for alphabeticalthan animacy decisions [F(1,14) � 79.78, P � 0.001], butthe accuracy with which each type of decision was made didnot differ reliably.

Anticipating the manner in which study items would becategorized for the purpose of analysing subsequent memoryeffects in the fMRI data (see below), RTs were calculatedfor study items that were subsequently recognized withhigh confidence as opposed to items recognized with lowconfidence or missed. These RTs were 954 and 951 ms,respectively, in the animacy task, and 1441 and 1362 ms,respectively, in the alphabetical task. RT did not differ reliably

according to subsequent memory performance in the animacytask [F(1,14) � 1], but was slightly longer for subsequentlyremembered than forgotten items in the alphabetical task[F(1,14) � 10.94, P � 0.01].

Recognition memoryRecognition memory performance is shown in Table 1.Accuracy of confident and non-confident recognition wasindexed by the discrimination measure Pr [probability of ahit (Phit) minus probability of a false alarm (Pfalse alarm)](Snodgrass and Corwin, 1988). This measure showed asignificant interaction between confidence and type ofencoding [F(1,14) � 49.89, P � 0.001]. For confident hits,discrimination was significantly greater than zero for wordsfrom both the animacy and the alphabetical task [0.48 and0.19, respectively; F(1,14) � 146.43 and 50.99, both P �0.001]. Words from the animacy task, however, gave riseto better recognition than did alphabetically judged words[F(1,14) � 63.79, P � 0.001]. Performance for non-confidently recognized words was not reliably greater thanzero for items from the animacy task [Phit – Pfalse alarm �0.02; F(1,14) � 1], but was slightly greater than chance foralphabetically judged items [Phit – Pfalse alarm � 0.07;F(1,14) � 13.32, P � 0.01].

On the basis of these findings, words were categorizedas ‘remembered’ when they attracted a confident hit duringthe subsequent recognition test. Words were categorized as‘forgotten’ when they attracted a miss or a non-confident hit.Only confident hits were classified as remembered itemsbecause accurate discrimination between old and new wordswas carried primarily and the depth of processing effect wascarried exclusively by confident judgements. Furthermore,previous studies (Brewer et al., 1998; Wagner et al., 1998b)have shown that subsequent memory effects are foundpredominantly for confidently recognized items. Accordingly,we maximized the signal-to-noise ratio for contrasts involvingsubsequent memory effects by contrasting confident hits withall other responses to old words. Because discrimination wasabove chance for alphabetically encoded words accorded anon-confident decision, we also computed a subsequentmemory comparison for these items by contrasting confidenthits with misses. The outcome of this comparison did notdiffer from that found when non-confident hits were includedas forgotten items along with misses.

fMRI findingsAll analyses of fMRI depth of processing and subsequentmemory effects described below were confined to study trialsassociated with correct animacy or alphabetical decisions.

Depth of processingCollapsed over subsequent memory performance, fMRIsignals were greater for words from the animacy than for


Table 1 Recognition memory performance for new words and old words requiring an animacyor alphabetical decision during study

Word type Recognition judgement

Sure old Unsure old Sure new Unsure new

Proportion of responsesOld

Animacy 0.58 (0.19) 0.19 (0.11) 0.07 (0.06) 0.15 (0.09)Alphabetical 0.29 (0.15) 0.24 (0.10) 0.20 (0.19) 0.27 (0.15)

New 0.10 (0.07) 0.17 (0.07) 0.36 (0.24) 0.35 (0.19)

Mean reaction time (ms)Old

Animacy 1094 (163) 1536 (317) 1275* (181) 1589 (407)Alphabetical 1221 (258) 1573 (371) 1342 (263) 1552 (369)

New 1255 (347) 1579 (322) 1297 (221) 1544 (360)

Values are across-subject means (standard deviation). *Mean reaction time for sure new judgements toold words requiring animacy decisions during study is based on 14 volunteers. One volunteer did notmake any such judgement.

Fig. 1 Maximum-intensity projections detailing regions that showed significant fMRI signal increases loading on the early covariate forwords studied in the animacy task versus the alphabetical task (A) and for the reverse subtraction (B). Increases that exceeded athreshold of P � 0.001 are shown.

words from the alphabetical task in widespread regions ofthe brain, including the frontal cortex, hippocampus and theadjacent medial temporal cortex (Fig. 1A). The fMRI signalswere greater for words from the alphabetical than the animacytask in, among others, the bilateral parietal and posteriorprefrontal cortex (Fig. 1B). All activations loaded on theearly covariate.

Subsequent memory effectsRegions demonstrating a subsequent memory effect wereidentified by contrasts between remembered and forgottenitems, performed separately for each study task. Data fromone volunteer were excluded from the subsequent memorycontrast for alphabetically judged words because too few(�12) words were confidently recognized.


Table 2 Regions showing significant (P � 0.001) signal increases on the early covariate foranimacy-judged words that were subsequently remembered versus forgotten

Region BA Location Peak Z score(x, y, z) (no. voxels)

Left inferior frontal gyrus 47 –36, 36, –9 4.91 (140)Left inferior frontal gyrus 9/44 –51, 12, 21 3.81 (25)Left inferior frontal gyrus 45 –51, 27, 18 3.70 (52)Left medial superior frontal gyrus 8 –3, 36, 45 3.43 (9)Right inferior frontal gyrus 47 33, 36, –9 4.45 (74)Right inferior frontal gyrus 9/44 57, 12, 21 3.67 (10)Right inferior frontal gyrus 45 39, 27, 9 3.60 (20)Right superior frontal gyrus 6 27, –3, 57 3.83 (10)Left anterior hippocampus –27, –15, –12 3.99 (38)Left posterior hippocampus –30, –42, 0 3.88 (7)Left inferior temporal/fusiform gyrus 37 –48, –54, –18 4.45 (28)Left lateral parietal cortex 40 –45, –39, 33 4.47 (9)Left occipital cortex 17/18 –18, –93, –6 3.54 (22)Right anterior inferior temporal gyrus 20 33, 3, –45 3.39 (8)Right subcentral gyrus 4 48, –6, 18 3.63 (7)Right lateral occipital cortex 19 12, –87, 24 3.42 (5)Right calcarine cortex 17 18, –81, 12 3.62 (10)Brainstem 9, –24, –15 3.62 (6)Left cerebellum –30, –42, –27 3.56 (16)

Location is with respect to the system of Talairach and Tournoux (Talairach and Tournoux, 1988).Z scores refer to the peak of the activated cluster, the size of which is indicated in parentheses.

Regions showing a subsequent memory effect for itemsencoded in the animacy task are given in Table 2. All ofthese regions were identified from contrasts performed onparameter estimates derived from the early covariate. Amongthese regions were the left inferior frontal gyrus (BA 47, 45,9/44), along with a less extensive homotopic region on theright, and the anterior and posterior left hippocampus. Theseregions are illustrated in Figs 2A and 3A.

To identify those regions that were not only sensitive todepth of processing but also manifested a subsequent memoryeffect in the animacy task, the subsequent memory contrastfor that task was masked by the depth of processing effect(animacy task minus alphabetical task). Both the mask andsubsequent memory contrasts were thresholded at P � 0.001,and were computed for the early covariate only. As illustratedin Fig. 2B, two regions survived the masking procedure: theleft ventral inferior frontal gyrus (BA 47, 57 voxels, x �–36, y � 36, z � –9; Z � 4.91) and the left anteriorhippocampus (10 voxels, x � –27, y � –12, z � –12;Z � 3.96).

For alphabetically judged words, the only area to show asubsequent memory effect was the left anterior hippocampus(10 voxels, x � –27, y � –15, z � –12; Z � 4.45; Fig. 3B).In contrast to the activations described in the foregoingsections, this effect loaded on the late rather than the earlycovariate. When masked in the same manner as describedabove for the animacy task, this left anterior hippocampalarea was found not to overlap with the areas that showed agreater fMRI signal for words from the alphabetical than theanimacy task.

In a subsequent set of analyses, the question of the overlapbetween animacy and alphabetical subsequent memory effectswas addressed further. Subsequent memory effects for thealphabetical task were recomputed with a statistical thresholdof P � 0.01, and were masked by the animacy subsequentmemory effect (thresholded as before at P � 0.001). Theseanalyses permit the identification, at a high level of sensitivity,of regions in which alphabetical subsequent memory effectsoverlap with animacy subsequent memory effects, whilemaintaining an acceptable type I error rate (as the twocontrasts are orthogonal and independent, the probability oftype I error is given by the multiple of their respectivethresholds, i.e. 0.01 � 0.001). For the early covariate, themasked contrast revealed an alphabetical subsequent memoryeffect in left inferior frontal gyrus (BA 47, 17 voxels,x � –36, y � 36, z � –9; Z � 2.88) (Fig. 3C).

Figures 4 and 5 summarize quantitatively the mostimportant of the effects described above. In the case ofFig. 4, it can be seen that subsequent memory effects areessentially absent for the alphabetical task in the rightprefrontal cortex, and in the more dorsal of the two leftfrontal regions manifesting an effect in the animacy task. Bycontrast, subsequent memory effects for the two tasks are ofsimilar magnitude [F(1,13) � 1.50, P � 0.20] in the ventralleft frontal cortex. Figure 5 shows the relative magnitudesand variabilities of the parameter estimates obtained in theleft anterior and left posterior hippocampus. The differentialloadings of the animacy and alphabetical subsequent memoryeffects in the anterior hippocampus on the early and latecovariates are clearly evident, as is the exclusivity to the


Fig. 2 (A) Significant clusters of signal increases for remembered versus forgotten words from theanimacy task in the inferior frontal and medial temporal regions (P � 0.001). The activations arerendered onto the MNI reference brain. (B) The same contrast as for A, but masked by the regions thatshowed significant signal increases for the animacy versus alphabetical contrast (contrast and mask boththresholded at P � 0.001).


Fig. 3 Regions showing significant subsequent memory effects for words from the animacy and alphabetical tasks (threshold P � 0.001).The activations are rendered onto the MNI reference brain. (A) Effects from the animacy task (see legend to Fig. 2). (B) Effects fromthe alphabetical task in the left anterior hippocampus, loading on the late covariate. (C) Additional left frontal subsequent memory effect(threshold P � 0.01) in the alphabetical task, revealed by a mask (threshold P � 0.001) derived from the animacy subsequent memoryeffect.

animacy task of the subsequent memory effects in theposterior hippocampus.

DiscussionThis study addressed two principal issues: (i) the extent towhich there is overlap between regions that are more active

during a deep (animacy decision) than a shallow (alphabeticaldecision) encoding task, and regions where activity predictssubsequent memory for the deeply encoded items; (ii) whethersubsequent memory effects dissociate according to the studytask (animacy versus alphabetical ) in which they are obtained.With respect to the first issue, the findings indicated that asubset of the regions revealed by the depth of processing


Fig. 4 Parameter estimates for the early covariate for subsequentmemory effects (i.e. differences between remembered andforgotten words) in the animacy (filled columns) and alphabetical(open columns) tasks for the left and right ventral (BA 47) andleft and right dorsal (9/44) inferior frontal gyrus (for coordinates,see Table 2). Error bars show the standard error of the mean.

Fig. 5 Parameter estimates for early and late covariates forsubsequent memory effects in the animacy (filled columns) andalphabetical (open columns) tasks for the left anterior and leftposterior hippocampus (for coordinates, see Table 2). Error barsshow the standard error of the mean. Note that the scale for theparameter estimates is specific to the covariate employed to obtainthe estimates.

contrast did indeed manifest subsequent memory effects inthe animacy task; however, effects were also found in regionsthat were insensitive to the depth of processing manipulation(Fig. 2). With regard to the second issue, we found littleevidence that alphabetical and animacy subsequent memoryeffects were anatomically dissociable. Rather, the regionswhere effects were found for the alphabetical task were asubset of those showing effects for the animacy task. Below,after commenting on the behavioural data, we discuss therelevance of these fMRI findings to current ideas about thefunctional and neural bases of memory encoding.

Behavioural performanceAs expected, recognition memory was substantially moreaccurate for items that were encoded in the animacy task

rather than the alphabetical task. The advantage for deeplyencoded items was observed despite the fact that RTs were�400 ms shorter in the animacy task, a clear demonstrationthat depth of processing effects in memory cannot beattributed to differences in factors such as time on task ortask difficulty. The same argument applies with respect tothose brain regions in which activity was greater duringperformance of the animacy task (Demb et al., 1995).

Comparison of hit and false alarm rates showed that non-confident recognition of items subjected to animacy decisionswas at chance. By contrast, non-confident hits foralphabetically studied words marginally exceeded thecorresponding false alarm rate, indicating the possibility ofabove-chance recognition of these items. However, the sizeof this effect was small (Phit � 0.24, Pfalse alarm � 0.17),meaning that the majority (~70%) of non-confident hits toalphabetically studied words were, like their counterpartsfrom the animacy task, based on guesses. Together withprevious findings (Brewer et al., 1998; Wagner et al., 1998b),these results emphasize the importance of employingconfidence judgements when recognition memory is used toidentify subsequent memory effects. Without the data fromsuch judgements, the subsequent memory effects would havebeen diluted (to a greater degree in the alphabetical than inthe animacy task) by the inclusion of trials in which therewas no veridical memory of the study episode, and in whichsuccessful recognition was the result of guessing.

It is important to note that, in the animacy task, RT didnot predict whether items would be recognized subsequently.This argues against the possibility that the subsequent memoryeffects obtained in the fMRI data for that task are correlatesof trial-by-trial fluctuations in such factors as arousal,attention and time devoted to processing each item. Thispossibility is further diminished by the exclusion of fMRIdata associated with study trials on which incorrect responseswere made from the analyses of subsequent memory effects.A small difference in RT between subsequently rememberedand forgotten items was found in the alphabetical task. Itseems unlikely, however, that this difference was responsiblefor the subsequent memory effects observed in the fMRIdata from this task; the regions displaying such effects werea subset of those identified in the animacy task, in which noRT differences were found.

fMRI findingsThe depth of processing contrast revealed signal increasesfor deeply studied words in a variety of areas, including theprefrontal cortex and regions of the medial temporal lobe.With respect to the prefrontal cortex particularly, thesefindings broadly replicate earlier findings (e.g. Kapur et al.,1994; Demb et al., 1995; Fletcher et al., 1995; Wagner et al.,1998b, experiment 1; for review, see Poldrack et al., 1999).The subsequent memory contrast for deeply studied wordsrevealed subsequent memory effects in two regions thatoverlapped with those activated by the depth of processing


contrast, namely the left ventral inferior prefrontal cortexand the left anterior hippocampus. These findings supportthe claim by Wagner and colleagues, made on the basis of acomparison between data obtained with different groups ofsubjects and experimental procedures, that effective memoryencoding in a deep study task involves activation of someof the same regions that are revealed by contrasts betweentasks requiring deep versus shallow processing (Wagneret al., 1998b).

The overlap between regions activated by the depth ofprocessing and deep subsequent memory effects implies theexistence of cognitive operations that are engageddifferentially both by semantic versus non-semanticprocessing and by effective versus less effective episodicencoding in a semantic task. It has been suggested that theoperations supported by the left inferior prefrontal cortexmight contribute to ‘semantic working memory’ (Gabrieliet al., 1998)—the temporary storage, manipulation andselection of an item’s semantic attributes. According to thishypothesis (see also Buckner and Koutstaal, 1998; Wagneret al., 1998b, 1999), the more a study item engages semanticworking memory, the more likely it is that its semanticfeatures will be incorporated in a representation of the studyepisode, and thus the more likely it is that the episode willbe accessible in a subsequent memory test. It is easy to seehow this hypothesis can encompass the findings frombetween-task comparisons, as in the case of the depth ofprocessing manipulation in the present study. Additionalassumptions are needed, however, to allow the hypothesis toaccount for left prefrontal subsequent memory effects when,as in the present study, the behavioural data provide noevidence that subsequently remembered and forgotten itemswere processed differently. One might imagine, for example,that the words placing the heaviest load on semantic workingmemory would be those that were most difficult to classifysemantically, leading to a positive relationship between RTand the probability of subsequent recognition. There is,however, no hint of such a relationship. To reconcile theworking memory hypothesis with these findings, it isnecessary to assume that the engagement of semantic workingmemory goes beyond what is required to perform the studytask, and that it is the extent of this additional processingthat is particularly important for memory encoding.

Subsequent memory effects were observed during theanimacy study task in the right as well as the left prefrontalcortex. These right-sided effects were localized to regionsbroadly homotopic with those found on the left but, unliketheir left-sided counterparts, they were not localized toregions sensitive to the main effect of depth of processing.Thus, right prefrontal subsequent memory effects may reflectprocesses that benefit episodic encoding in ways other thanthrough cognitive operations associated with semanticprocessing. In this regard, it is noteworthy that studiesemploying non-verbal material, such as textures, pictures andfaces, have reported encoding-related activity in the vicinityof the right prefrontal regions identified in the present study

(Brewer et al., 1998; Kelley et al., 1998; Wagner et al.,1998a; McDermott et al., 1999). The present findings suggestthat the (presumably non-linguistic) processes reflected bysuch right frontal activity can, at least under somecircumstances, operate on information derived fromostensibly verbal input. The nature of these processes isunclear. Nonetheless, together with the previous findings justmentioned, the present results argue against the view (Tulvinget al., 1994) that prefrontally mediated episodic encodingoperations are invariably left-lateralized.

In addition to the left inferior prefrontal cortex, the leftanterior hippocampus was sensitive to both the depth ofprocessing and animacy subsequent memory contrasts.Greater left anterior hippocampal activation during deeprather than shallow processing was also reported by Wagnerand colleagues (Wagner et al., 1998b), but, for reasons thatare unclear, these authors failed to observe subsequentmemory effects in this region (but see Kirchhoff et al., 2000).The present findings are consistent with the large body ofevidence from animal and human studies implicating thehippocampus in episodic memory, and provide direct evidencethat the hippocampus plays a role in memory encoding (seealso Fernandez et al., 1999b). While the exact contributionto memory of the hippocampus is still debated, it seemslikely that it includes some kind of integrative function,whereby disparate elements of an encoding episode, e.g. itemand contextual information, are bound together to form anepisodic representation (see Amaral, 1999, and followingpapers). One possibility is that hippocampal subsequentmemory effects reflect differences in the processing associatedwith such binding operations, which in turn reflect the amountand nature of the item information (in the present case,largely semantic in origin) available to the hippocampus fromworking memory systems supported by prefrontal cortex (forsimilar views, see Buckner and Koutstaal, 1998; Wagneret al., 1998b). According to this hypothesis, encoding-relatedactivity in the prefrontal cortex and hippocampus should beorganized serially, the former activity onsetting before thelatter. The poor temporal resolution of the BOLD signal,however, together with difficulties in interpreting inter-regional differences in the time course of such signals(Rajapakse et al., 1998), make the question of the relativetiming of prefrontal and hippocampal encoding-relatedactivity difficult to address with fMRI. The question mayprove more amenable to investigation with methods basedon electroencephalography or magnetoencephalography.

Whereas the left anterior hippocampus demonstratedeffects of both depth of processing and deep subsequentmemory, effects in a more posterior left hippocampal regionwere restricted to the subsequent memory contrast. Thesediffering patterns of activity in the anterior and posteriorhippocampus add to the evidence that this structure isfunctionally heterogeneous (Fernandez et al., 1998; Lepageet al., 1998; Strange et al., 1999). The region identified in thepresent study as the posterior hippocampus (peak activation atx � –30, y � –42, z � 0) is quite close to the left


medial temporal region reported by Wagner and colleaguesto demonstrate a subsequent memory effect (peak activationat x � –31, y � –46, z � –12) (Wagner et al., 1998b).Although the region was identified by these authors as theparahippocampal cortex, it is possible, given the proximityof the respective peaks, that the two areas overlap tosome extent.

We turn now to the second issue that motivated the presentstudy, namely whether subsequent memory effects differaccording to the nature of the encoding task used to revealthem. As already discussed, for words studied in the animacytask, the subsequent memory contrast revealed activations inthe bilateral inferior prefrontal cortex and the left anteriorand posterior hippocampus. For alphabetically studied words,the subsequent memory contrast revealed activations in partof the same region of the left inferior prefrontal cortex thatmanifested deep subsequent memory effects, as well as inthe left anterior hippocampus. No regions were found thatdemonstrated shallow, but not deep, effects. Thus, the areasassociated with episodic encoding in the alphabetical taskwere a subset of those associated with encoding in theanimacy task.

The failure to find evidence for subsequent memory effectsassociated uniquely with the alphabetical encoding task must,like all null results, be treated with caution. This is especiallyso given that the power to detect subsequent memory effectswas lower for the alphabetical than the animacy task, aconsequence of the fewer confident hits made toalphabetically studied words, and the need to reject onevolunteer’s data from the relevant analyses. Thus, the issueof whether there are subsequent memory effects specific tonon-semantic study tasks remains open.

As just noted, comparison of the outcome of the subsequentmemory contrasts for the two tasks revealed no regions whereeffects were unique to the alphabetical task. There wereregions—notably the right prefrontal cortex and the leftposterior hippocampus—where subsequent memory effectswere found only for the animacy task. This finding mightindicate a dissociation between the neural substrates ofencoding in the two tasks. The difference in the power ofthe two subsequent memory contrasts (see above) means,however, that it is not possible to reject the possibility thatthe finding is a consequence of regional differences insensitivity to subsequent memory contrasts rather thandifferences related more specifically to the two encodingtasks.

The anatomical overlap between the subsequent memoryeffects in the animacy task and the two alphabeticalsubsequent memory effects that did prove reliable suggeststhat a common set of cognitive processes acted to facilitateencoding in the two tasks. As already mentioned, oneplausible account is that subsequent memory effects in theleft prefrontal cortex and the left anterior hippocampus reflectthe benefit of incorporating relatively elaborate semanticinformation about an item in a representation of its studyepisode. By this account, even when the task is ostensibly

non-semantic, it is the items that receive the greater semantic-level processing that are encoded the most effectively. Onepossibility is that semantic processing, and effective episodicencoding, occurred in alphabetical trials in which volunteersconfused the two encoding tasks. This is unlikely, however,as study trials on which errors were committed were notincluded in the analyses of the fMRI data. It seems morelikely that any semantic processing accorded these itemswas incidental, and perhaps subsequent, to the processingnecessary for alphabetical classification. For example, forany given volunteer, words may have differed in their capacityto capture or hold attention beyond what was needed for thealphabetical decision. A delay in the engagement of semanticprocessing in the alphabetical task relative to the animacytask may account for the finding that the left anteriorhippocampal subsequent memory effect loaded on the laterof the two covariates employed to detect event-related fMRIsignal change (Fig. 5).

In summary, the present findings confirm previoussuggestions (Wagner et al., 1998b) that some of the brainregions sensitive to depth of processing manipulations playa role in verbal memory encoding when the study task isheld constant. Chief among these regions are the left inferiorprefrontal cortex and left anterior hippocampus. The findingsalso demonstrate a role in the episodic encoding of deeplyprocessed words for regions, such as the right prefrontalcortex and left posterior hippocampus, that are insensitive todepth of processing. There was no evidence that successfulepisodic encoding during a non-semantic study task dependedupon regions different from those that supported the encodingof semantically studied words. Instead, the findings suggestthat successful non-semantic encoding relies on a subset ofthe regions engaged during successful semantic encoding.

AcknowledgementsWe wish to thank the radiography staff of the WellcomeDepartment of Cognitive Neurology and Jane Herron fortheir help with data acquisition. The authors and their researchare supported by the Wellcome Trust.

ReferencesAlkire MT, Haier RJ, Fallon JH, Cahill L. Hippocampal, but notamygdala, activity at encoding correlates with long-term, free recallof nonemotional information. Proc Natl Acad Sci USA 1998; 95:14506–10.

Amaral DG. Introduction: what is where in the medial temporallobe? Hippocampus 1999; 9: 1–6.

Andrade A, Paradis AL, Rouquette S, Poline JB. Ambiguous resultsin functional neuroimaging data analysis due to covariate correlation.Neuroimage 1999; 10: 483–6.

Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD. Makingmemories: brain activity that predicts how well visual experiencewill be remembered. Science 1998; 281: 1185–7.


Buckner RL, Koutstaal W. Functional neuroimaging studies ofencoding, priming, and explicit memory retrieval. [Review]. ProcNatl Acad Sci USA 1998; 95: 891–8.

Buckner RL, Koutstaal W, Schacter DL, Dale AM, Rotte M, RosenBR. Functional–anatomic study of episodic retrieval. II. Selectiveaveraging of event-related fMRI trials to test the retrieval successhypothesis. Neuroimage 1998; 7: 163–75.

Buckner RL, Kelley WM, Petersen SE. Frontal cortex contributesto human memory formation. [Review]. Nat Neurosci 1999; 2:311–4.

Buckner RL, Wheeler ME, Sheridan M. Encoding processes duringretrieval tasks. J Cogn Neurosci. In press 2001.

Cahill L, Haier RJ, Fallon J, Alkire MT, Tang C, Keator D, et al.Amygdala activity at encoding correlated with long-term, free recallof emotional information. Proc Natl Acad Sci USA 1996; 93:8016–21.

Cocosco CA, Kollokian V, Kwan RKS, Evans AC. BrainWeb:online interface to a 3D MRI simulated brain database. Neuroimage1997; 5 (4 Pt 2): S425.

Dale AM, Buckner RL. Selective averaging of rapidly presentedindividual trials using fMRI. Hum Brain Mapp 1997; 5: 329–40.

Demb JB, Desmond JE, Wagner AD, Vaidya CJ, Glover GH,Gabrieli JD. Semantic encoding and retrieval in the left inferiorprefrontal cortex: a functional MRI study of task difficulty andprocess specificity. J Neurosci 1995; 15: 5870–8.

Dolan RJ, Fletcher PC. Dissociating prefrontal and hippocampalfunction in episodic memory encoding. Nature 1997; 388: 582–5.

Fernandez G, Weyerts H, Schrader-Bolsche M, Tendolkar I, SmidHG, Tempelmann C, et al. Successful verbal encoding into episodicmemory engages the posterior hippocampus: a parametricallyanalyzed functional magnetic resonance imaging study. J Neurosci1998; 18: 1841–7.

Fernandez G, Brewer JB, Zhao Z, Glover GH, Gabrieli JD. Levelof sustained entorhinal activity at study correlates with subsequentcued-recall performance: a functional magnetic resonance imagingstudy with high acquisition rate. Hippocampus 1999a; 9: 35–44.

Fernandez G, Effern A, Grunwald T, Pezer N, Lehnertz K,Dumpelmann M, et al. Real-time tracking of memory formation inthe human rhinal cortex and hippocampus. Science 1999b; 285:1582–5.

Fletcher PC, Frith CD, Grasby PM, Shallice T, Frackowiak RS,Dolan RJ. Brain systems for encoding and retrieval of auditory–verbal memory: an in vivo study in humans. Brain 1995; 118:401–16.

Fletcher PC, Frith CD, Rugg MD. The functional neuroanatomy ofepisodic memory. [Review]. Trends Neurosci 1997; 20: 213–8.

Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD,Frackowiak RSJ. Statistical parametric maps in functional imaging:a general linear approach. Hum Brain Mapp 1995; 2: 189–210.

Friston KJ, Fletcher P, Josephs O, Holmes A, Rugg MD, Turner R.Event-related fMRI: characterizing differential responses.Neuroimage 1998; 7: 30–40.

Gabrieli JDE, Brewer JB, Desmond JE, Glover GH. Separate neural

bases of two fundamental memory processes in the human medialtemporal lobe. Science 1997; 276: 264–6.

Gabrieli JDE, Poldrack RA, Desmond JE. The role of left prefrontalcortex in language and memory. [Review]. Proc Natl Acad Sci USA1998; 95: 906–13.

Henson RN, Rugg MD, Shallice T, Josephs O, Dolan RJ.Recollection and familiarity in recognition memory: an event-relatedfunctional magnetic resonance imaging study. J Neurosci 1999; 19:3962–72.

Henson RN, Rugg MD, Shallice T, Dolan RJ. Confidence inrecognition memory for words: dissociating right prefrontal rolesin episodic retrieval. J Cogn Neurosci. In press 2001.

Josephs O, Turner R, Friston K. Event-related fMRI. Hum BrainMapp 1997; 5: 243–8.

Kapur S, Craik FI, Tulving E, Wilson AA, Houle S, Brown GM.Neuroanatomical correlates of encoding in episodic memory: levelsof processing effect. Proc Natl Acad Sci USA 1994; 91: 2008–11.

Kapur S, Tulving E, Cabeza R, McIntosh AR, Houle S, Craik FI.The neural correlates of intentional learning of verbal materials: aPET study in humans. Brain Res Cogn Brain Res 1996; 4: 243–9.

Kelley WM, Miezin FM, McDermott KB, Buckner RL, RaichleME, Cohen NJ, et al. Hemispheric specialization in human dorsalfrontal cortex and medial temporal lobe for verbal and nonverbalmemory encoding. Neuron 1998; 20: 927–36.

Kirchhoff BA, Wagner AD, Maril A, Stern CE. Prefrontal–temporalcircuitry for episodic encoding and subsequent memory. J Neurosci2000; 20: 6173–80

Kucera H, Francis WN. Computational analysis of present-dayAmerican English. Providence (RI): Brown University Press; 1967.

Lepage M, Habib R, Tulving E. Hippocampal PET activations ofmemory encoding and retrieval: the HIPER model. Hippocampus1998; 8: 313–22.

McDermott KB, Buckner RL, Petersen SE, Kelley WM, SandersAL. Set- and code-specific activation in frontal cortex: an fMRIstudy of encoding and retrieval of faces and words. J Cogn Neurosci1999; 11: 631–40.

Poldrack RA, Wagner AD, Prull MW, Desmond JE, Glover GH,Gabrieli JD. Functional specialization for semantic and phonologicalprocessing in the left inferior prefrontal cortex. Neuroimage 1999;10: 15–35.

Rajapakse JC, Kruggel F, Maisog JM, von Cramon DY. Modelinghemodynamic response for analysis of functional MRI time-series.Hum Brain Mapp 1998; 6: 283–300.

Rugg MD. ERP studies of memory. In: Rugg MD, Coles MGH,editors. Electrophysiology of mind. Oxford: Oxford UniversityPress; 1995. p. 132–70.

Rugg, MD. Functional neuroimaging in cognitive neuroscience. In:Brown CM, Hagoort P, editors. The neurocognition of language.Oxford: Oxford University Press; 1999. p. 15–36.

Saykin AJ, Johnson SC, Flashman LA, McAllister TW, SparlingM, Darcey TM, et al. Functional differentiation of medial temporaland frontal regions involved in processing novel and familiar words:an fMRI study. Brain 1999; 122: 1963–71.


Schacter DL, Buckner RL, Koutstaal W, Dale AM, Rosen BR. Lateonset of anterior prefrontal activity during true and false recognition:an event-related fMRI study. Neuroimage 1997; 6: 259–69.

Shallice T, Fletcher P, Frith CD, Grasby P, Frackowiak RS, DolanRJ. Brain regions associated with acquisition and retrieval of verbalepisodic memory. Nature 1994; 368: 633–5.

Snodgrass JG, Corwin J. Pragmatics of measuring recognitionmemory: applications to dementia and amnesia. J Exp Psychol1988; 117: 34–50.

Squire LR, Knowlton BJ. The medial temporal lobe, thehippocampus, and the memory systems of the brain. In: GazzanigaMS, editor. The new cognitive neurosciences. 2nd ed. Cambridge(MA): MIT Press; 2000. p. 765–79.

Stern CE, Corkin S, Gonzalez RG, Guimaraes AR, Baker JR,Jennings PJ, et al. The hippocampal formation participates in novelpicture encoding: evidence from functional magnetic resonanceimaging. Proc Natl Acad Sci USA 1996; 93: 8660–5.

Strange BA, Fletcher PC, Henson RN, Friston KJ, Dolan RJ.Segregating the functions of human hippocampus. Proc Natl AcadSci USA 1999; 96: 4034–9.

Talairach J, Tournoux P. Co-planar stereotaxic atlas of the humanbrain. Stuttgart: Thieme Verlag; 1988.

Tulving E, Kapur S, Craik FI, Moscovitch M, Houle S. Hemispheric

encoding/retrieval asymmetry in episodic memory: positronemission tomography findings. [Review]. Proc Natl Acad Sci USA1994; 91: 2016–20.

Wagner AD, Poldrack RA, Eldridge LL, Desmond JE, Glover GH,Gabrieli JDE. Material-specific lateralization of prefrontal activationduring episodic encoding and retrieval. Neuroreport 1998a; 9:3711–7.

Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, DaleAM, et al. Building memories: remembering and forgetting of verbalexperiences as predicted by brain activity. Science 1998b; 281:1188–91.

Wagner AD, Koutstaal W, Schacter DL. When encoding yieldsremembering: insights from event-related neuroimaging. [Review].Philos Trans R Soc Lond B Biol Sci 1999; 354: 1307–24.

Wilding EL, Rugg MD. An event-related potential study ofrecognition memory with and without retrieval of source. Brain1996; 119: 889–905.

Zarahn E, Aguirre GK, D’Esposito M. Empirical analyses of BOLDfMRI statistics: 1. Spatially unsmoothed data collected under null-hypothesis conditions. Neuroimage 1997; 5: 179–97.

Received April 7, 2000. Revised October 10, 2000.Accepted October 18, 2000

brain depth of processing effects on neural correlates of memory...

Documents