Abstract

Using event-related functional magnetic resonance imaging (fMRI), the neural correlates of memory encoding can be studied by contrasting item-related activity elicited in a study task according to whether the items are remembered or forgotten in a subsequent memory test. Previous studies using this approach have implicated the left prefrontal cortex in the successful encoding of verbal material into episodic memory when the study task is semantic in nature. In the current study, we asked whether the neural correlates of episodic encoding differ depending on type of study task. Seventeen volunteers participated in an event-related fMRI experiment in which at study, volunteers were cued to make either animacy or syllable judgements about words. A recognition memory test followed after a delay of ~15 min. For the animacy task, words that were subsequently remembered showed greater activation in left and medial prefrontal regions. For the syllable task, by contrast, successful memory for words was associated with activations in bilateral intraparietal sulcus, bilateral fusiform gyrus, right prefrontal cortex and left superior occipital gyrus. These findings suggest that the brain networks supporting episodic encoding differ according to study task.

Introduction

With the advent of event-related functional magnetic resonance imaging (fMRI) (Dale and Buckner, 1997; Josephs et al., 1997; Zarahn et al., 1997), it has become possible to study the neural correlates of the encoding into memory of individual items. The typical approach, also used in event-related brain potential studies of memory encoding (Wagner et al., 1999), has been to record neural activity while volunteers study a sequence of items, after which memory for the items is tested. The neural responses (as indexed by blood oxygenation level dependent signal change) elicited by the items at study are then contrasted according to whether these items were remembered or forgotten in the subsequent memory test. Differences between the responses associated with subsequently remembered and forgotten items (subsequent memory effects) are interpreted as putative neural correlates of memory encoding.

Several previous event-related fMRI studies have used subsequent memory effects to study the episodic encoding of words (Wagner et al., 1998b; Henson et al., 1999; Kirchhoff et al., 2000;Baker et al., 2001;Buckner et al., 2001;Davachi et al., 2001;Otten et al., 2001). These studies have implicated a role for the left prefrontal cortex in successful episodic encoding. Relative to words that were subsequently forgotten, words that were subsequently remembered showed greater fMRI signals in ventral (Wagner et al., 1998b; Kirchhoff et al., 2000;Baker et al., 2001;Buckner et al., 2001;Davachi et al., 2001;Otten et al., 2001) and/or dorsal (Wagner et al., 1998b; Henson et al., 1999;Baker et al., 2001;Buckner et al., 2001;Davachi et al., 2001;Otten et al., 2001) regions of the left inferior frontal gyrus. It has been suggested that the cognitive operations supported by the left prefrontal cortex might be related to ‘semantic working memory’ (Gabrieli et al., 1998) — the temporary storage, manipulation and selection of an item's semantic attributes. According to this hypothesis (Buckner and Koutstaal, 1998; Wagner et al., 1998b, 1999), the more a study item engages semantic working memory, the more likely it is that its semantic features will be incorporated into a representation of the study episode and the more likely it is that the episode will be accessible in a subsequent memory test.

In addition to the left prefrontal cortex, previous studies have implicated a role for the medial temporal lobe in word encoding. Subsequent memory effects have been found both in parahippocampal regions (Wagner et al., 1998b; Kirchhoff et al., 2000) and in the hippocampus proper (Kirchhoff et al., 2000;Davachi et al., 2001;Otten et al., 2001). One possibility is that the products of the semantic working memory processes supported by the left prefrontal cortex are relayed to the medial temporal lobe (Buckner and Koutstaal, 1998; Wagner et al., 1998b). Further processing in this region may include an integration of different elements of the encoding episode, such as item and contextual information, into an episodic representation.

Most of the above studies of subsequent memory effects used an encoding task that required volunteers to engage in relatively ‘deep’, semantic analysis of study items, though there are exceptions (Baker et al., 2001;Davachi et al., 2001; Otten et al., 2001). A question of interest concerns the degree to which subsequent memory effects differ according to type of study task. The question is important because it bears on whether episodic encoding for a given class of items relies on a single neural network irrespective of task, or whether, instead, encoding is supported by multiple, task-specific, networks. In addition, knowing whether subsequent memory effects vary according to the nature of the encoding task will shed light on the functional role of the neural networks associated with successful episodic encoding. For example, the functional interpretation described above for the role of the prefrontal cortex in episodic encoding presumes a level of analysis that includes semantic aspects of the stimuli. The primary aim of the current experiment was to examine whether subsequent memory effects for words vary according to the nature (semantic versus phonological) of the encoding task.

We have already addressed the possible influence of type of encoding task on fMRI subsequent memory effects in a previous experiment (Otten et al., 2001). In that experiment, volunteers were cued to make either animacy (does the word refer to a living thing?) or alphabetical (are the first and last letters of the word in alphabetical order?) decisions about study words. A recognition memory test followed after a delay of 15 min. For words studied in the animacy task, subsequent memory effects were found in bilateral (greater on the left) inferior frontal gyrus and left anterior and posterior hippocampus. For alphabetically studied words, subsequent memory effects were found in part of the same left prefrontal region that manifested animacy subsequent memory effects, as well as in the left anterior hippocampus. Thus, the areas associated with episodic encoding in the alphabetical task were a subset of those associated with encoding in the animacy task; there was no evidence that successful episodic encoding during the alphabetical study task depended upon regions different from those that supported the encoding of semantically studied words. Baker et al .(Baker et al.,2001) arrived at the same conclusion when comparing fMRI subsequent memory effects across semantic (abstract/concrete judgement) and nonsemantic (upper/lower case judgement) encoding tasks.

There are at least two reasons why the failure in the Otten et al .(Otten et al., 2001) study to find evidence for qualitatively different neural correlates of episodic encoding for animacy and alphabetical tasks should be treated with caution. First, the power to detect subsequent memory effects in the alphabetical task was lower than that in the animacy task because many fewer words from the alphabetical task were correctly recognized. Second, the type of processing needed for the alphabetical task may not have been conducive to subsequent recognition memory judgements. It is possible that, in the absence of additional processing, words encoded in an alphabetical judgement task are not accessible in a subsequent recognition memory test. According to this account, the alphabetically encoded words given a correct recognition judgement in the Otten et al .(Otten et al., 2001) study were those that received incidental semantic processing at study. As a consequence, the neural correlates of episodic encoding in the alphabetical task were a weak reflection of those seen for the animacy task.

In the current study, we again investigated fMRI subsequent memory effects across different encoding tasks. We used the same animacy task as before, but contrasted this with a task that, while ostensibly nonsemantic, would yield a level of recognition memory intermediate between the alphabetical and animacy tasks employed in our previous study. As shown by previous studies of semantic and phonological processing [see e.g. Challis et al .(Challis et al., 1996)], a phonological processing task satisfies these criteria. We therefore chose a phonological task to contrast with the animacy task in the expectation that when items are encoded in the phonological task, the resulting memory is a byproduct of phonological processing rather than, or in addition to, semantic processing incidental to the task.

An additional aim of the current study was to assess whether the brain regions associated with episodic encoding in a given task are the same as the regions selectively engaged by the task overall. As mentioned earlier, an ubiquitous finding in previous fMRI studies of episodic encoding has been activation of the left prefrontal cortex. Many previous studies have implicated left prefrontal regions in semantic processing (Poldrack et al., 1999). Two studies (Wagner et al., 1998b; Otten et al., 2001) directly contrasted the left prefrontal regions associated with episodic encoding and semantic processing and found that the regions overlapped. Thus, at least for semantic tasks, the same brain regions associated with engagement in the task also appear to be associated with episodic encoding within the task.

The question of interest here is whether, similar to semantic study tasks, phonological tasks will show a correspondence between the brain regions associated with task engagement and episodic encoding within the task. Several studies (Poldrack etal., 1999) have suggested that phonological and semantic processing recruit different regions of left prefrontal cortex. Specifically, while both semantic and phonological processing engage the dorsal region of left prefrontal cortex, only semantic processing engages the ventral region. If episodic encoding in a task benefits from ‘amplification’ of one or more components of the processes engaged by the task, we might expect that episodic encoding in a phonological study task engages the same left prefrontal regions that are activated by phonological processing per se.

We addressed the foregoing issues in a study in which volunteers performed two interleaved incidental encoding tasks: the animacy task employed in our previous experiment (Otten et al., 2001) and a phonological discrimination task (to decide whether the number of syllables in a word is odd or even). A recognition memory test followed after a delay of ~15 min. To address the question of whether the neural correlates of episodic encoding differ depending on the nature of the encoding task, we compared the regions that showed fMRI subsequent memory effects across the animacy and syllable tasks. We also determined which of the regions associated with episodic encoding in a particular study task were a subset of those associated with task engagement. To do this, we compared the regions that showed fMRI subsequent memory effects with the regions that showed greater fMRI signals in that task relative to both the other study task and a ‘low-level’ baseline condition.

Materials and Methods

Subjects

The experimental procedures were approved by the National Hospital for Neurology and Neurosurgery and Institute of Neurology Medical Ethics Committee. Eighteen volunteers were recruited via local advertisements and paid to participate in the experiment. One volunteer was excluded from the analyses because of a history of clinical depression. Mean age of the 17 remaining volunteers (six men) was 25 (range 19–33) years. All volunteers were native speakers of English and gave their informed consent prior to the experiment. All volunteers were right handed according to self-report. All volunteers claimed to be in good health and to be free from neurological and psychiatric problems.

Stimulus Materials

Three sets of 140 words each were selected from a pool of words of four to nine letters, ranging in frequency between 1 and 30 per million (Kucera and Francis, 1967). Words were selected at random, with the restriction that (i) across sets, the distribution of word lengths would be identical and (ii) within each set, 35 words would be animate with an odd number of syllables, 35 animate with an even number of syllables, 35 inanimate with an odd number of syllables and 35 inanimate with an even number of syllables. Across volunteers, these sets were rotated across the animacy, syllable and new conditions.

Each study list consisted of a random sequence of the 140 words used for animacy decisions and the 140 words used for syllable decisions. These 280 critical events were interspersed with 140 ‘low-level’ events (see Tasks description below). The study list was divided into two blocks of 210 events each and one filler word was added to the beginning and middle of each block. Each test list consisted of a random sequence of the 280 studied words and 140 new words. An additional 36 words were selected from the word pool to create practice lists for the study and test tasks (the words from the study practice were re-presented along with new words in the test practice).

Tasks

The experiment consisted of an incidental study task followed by a recognition memory test after a delay of ~15 min. At study, volunteers viewed 280 critical words, presented one at a time. Each word was preceded by a prestimulus cue, which consisted of the presentation of the letter O or the letter X. When an O preceded a word, volunteers had to decide whether or not the word was animate (in the sense that it referred to the property of a living entity). When an X preceded a word, they had to count the number of syllables in the word and decide whether this number was odd or even. Animacy and syllable decisions were equiprobable and randomly intermixed. Responses were given with the left and right thumbs and the hand used to signal each of the different response options in the two tasks was counterbalanced across volunteers. Both speed and accuracy were stressed.

The 280 critical words in the study task were randomly interspersed with 140 low-level events, which served as the baseline against which to characterize the fMRI responses elicited in each experimental condition. These events consisted of the presentation of a neutral cue (the symbol +), with no subsequent word presentation. Volunteers were told that when they saw these neutral cues, they should simply wait until the next cue arrived.

Study words were presented visually for 300 ms in a white uppercase Helvetica 48 point font on a black background. The time between successive cue onsets was ~4.1 s. At a viewing distance of ~30 cm, words subtended an approximate vertical visual angle of 1.4° and a horizontal visual angle ranging between 4 and 10°. The prestimulus cue was presented for 1.5 s and measured 1.4 × 1.2° of visual angle. There was a 100 ms blank period between the offset of the prestimulus cue and the onset of the word.

The recognition memory test consisted of the re-presentation of the 140 semantically judged words and 140 syllable-counted words from the study task, along with 140 words not seen before during the experiment. A plus sign presented before each word served as a fixation point and warning stimulus. For each word, volunteers had to decide whether they had seen the word before during the experiment (old/new judgement), indicating whether or not they were confident about their decision. One of four keys had to be depressed according to whether the word was confidently judged to be old, nonconfidently judged to be old, confidently judged to be new, or nonconfidently judged to be new. Responses were given with the middle and index fingers of the left and right hands. The assignment of old responses to the left or right hand was counter-balanced across subjects, but confident responses were always given with the middle fingers. Subjects were instructed to respond as fast as possible without sacrificing accuracy.

Test words were presented visually for 300 ms in a white uppercase Helvetica 24 point font on a black background. The time between successive word onsets was 4.8 s. At a viewing distance of ~50 cm, words subtended an approximate vertical visual angle of 0.5° and a horizontal visual angle ranging between 1.5 and 3.5°. The warning stimulus was presented for 2 s and measured 0.5 × 0.4° of visual angle. There was a 100 ms blank period between the offset of the warning stimulus and the onset of the word.

Procedure

Scanning took place during the study task only. Before entering the scanner, volunteers were given an explanation of the study task and practised with a short list to familiarize themselves with the task. Scanning began with a 15 min structural scan. Volunteers then performed two blocks, each of 212 trials (~17 min), of the study task, during which the functional scans were acquired. Volunteers were not informed about the subsequent memory test. A short rest was given in the middle of each block and between the blocks. The words were projected onto a mirror in direct view of the reclining volunteer and responses were given with a hand-held response box.

The memory test was administered ~15 min after the completion of the study task. Volunteers were taken to another room where they rested and conversed with the experimenter during the delay period. They were then informed about the old/new recognition memory task and performed a short practice list. Four test blocks of 106 trials, each ~8 min in duration, were undertaken with short rests between the blocks. Responses were given with a keyboard placed on a table in front of the volunteer. At the completion of the test blocks, volunteers were debriefed about the nature of the experiment and paid for their time.

MRI Scanning Methods

A 2T Siemens VISION system (Siemens, Erlangen, Germany) was used to acquire both T1-weighted anatomical volume images (1 × 1 × 1.5 mm voxels, MPRAGE sequence) and T2*-weighted echoplanar (EPI) images (64 × 64, 3 × 3 mm pixels, TE = 40 ms) with blood oxygenation level dependent (BOLD) contrast. Each EPI volume comprised 31 axial slices, 2 mm thick, separated by 1.5 mm, positioned to cover all but the most superior region of the brain and the cerebellum. Data were acquired during two sessions, each comprising 325 volumes, corresponding to the two study blocks. Volumes were acquired continuously with an effective repetition time (TR) of 2.88 s/volume. The first five volumes were discarded to allow for T1 equilibration effects.

Preprocessing

For each volunteer, all volumes in a session were realigned to the first volume and resliced using a sinc interpolation in space. To correct for their different acquisition times, the signal measured in each slice was then shifted relative to the acquisition of the middle slice using a sinc interpolation in time. Each volume was normalized to a standard EPI template volume, based on the MNI reference brain (Cocosco et al., 1997) of 3 × 3 × 3 mm voxels in the space of Talairach and Tournoux (Talairach and Tournoux, 1988) using nonlinear basis functions. Finally, the EPI volumes were smoothed with an 8 mm full-width, half-maximum isotropic Gaussian kernel to accommodate residual anatomical differences across volunteers and proportionally scaled to a global mean of 100.

Data Analysis

The data were analysed using Statistical Parametric Mapping (Friston et al., 1995), version SPM99 (Wellcome Department of Cognitive Neurology, London, UK). The volumes acquired during each session were treated as two time-series. The haemodynamic response to the onset of each event type of interest was modelled with two basis functions: a canonical haemodynamic response function (HRF) (Friston et al., 1998) and a delayed HRF (Henson et al., 2000;Otten et al., 2001), shifted to onset 2.88 s (i.e. one TR) later than the canonical HRF. The use of both an ‘early’ and a ‘late’ response function was based on suggestions from published reports (Wilding and Rugg, 1996; Schacter et al., 1997;Otten et al., 2001) that the time of maximal activation is later for some brain regions than the sensory regions on which the canonical HRF is based. The early and late response functions, when convolved with a sequence of 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 the late HRF were orthogonalized with respect to those for the early HRF so as to give priority to the early covariate (Andrade et al., 1999). Thus, loadings on the orthogonalized late covariate account for residual variance 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 a 4 s full-width, half-maximum Gaussian kernel. Parameter estimates for each covariate were calculated from the least mean squares fit of the model to the data.

Planned contrasts (specified in the Results section) were employed to test parameter estimates for both early and late covariates. The linear combination of parameter estimates for each contrast were stored as separate images for each volunteer. These contrast images were entered into one-sample t-tests to permit inferences to be drawn about condition effects across volunteers (i.e. a ‘random effects’ analysis). The images were subsequently transformed into statistical parametric maps (SPMs) of the Z statistic. Unless mentioned otherwise, contrasts were thresholded at P < 0.001, uncorrected for multiple comparisons. When reporting masked contrasts, the Z values refer to the outcome of the masked contrast only. Only activations involving contiguous clusters of at least five voxels were interpreted. The maxima of suprathreshold regions were localized by rendering them onto both the volunteers' normalized structural images and the MNI reference brain (Cocosco et al., 1997). They were labelled using the stereotactic system and nomenclature of Talairach and Tournoux (Talairach and Tournoux, 1988).

Results

Behavioral Performance

Study Task

Animacy decisions were made with an accuracy of 94% (SD = 3) and a mean reaction time (RT) of 1127 ms (SD = 169). Syllable decisions were made with an accuracy of 91% (SD = 7) and a mean RT of 1463 ms (SD = 314). Reaction times were significantly longer for syllable than animacy decisions [F(1,16) = 33.98, P < 0.001], but the accuracy with which each type of decision was made did not differ reliably. For each task, memory for a word was not reliably affected by the type of event that preceded the word. That is, the probability of remembering a word was the same regardless of whether volunteers had to switch tasks between the previous and current trial and regardless of whether the previous trial was a low-level baseline event.

To assess whether study RTs varied according to subsequent memory performance, RTs were calculated for study items subsequently recognized with high confidence, as opposed to items that were missed (duplicating the manner in which study items were classified for the purpose of subsequent memory effects in the fMRI data, see below). Study RTs were 1110 versus 1197 ms, respectively, for subsequently remembered and subsequently forgotten items in the animacy task and 1464 versus 1490 ms, respectively, in the syllable task. RT did not differ reliably according to subsequent memory performance in the syllable task [F(1,16) < 1], but was longer for subsequently forgotten than remembered items in the animacy task [F(1,16) = 7.58, P < 0.05].

Recognition Memory

Recognition memory performance is shown in Table 1. Accuracy of confident and nonconfident recognition was indexed by the discrimination measure Pr (probability hit – probability false alarm) (Snodgrass and Corwin, 1988). This measure showed a significant interaction between confidence and type of task [F(1,16) = 99.23, P < 0.001]. For confident hits, discrimination was significantly greater than zero for words from both the animacy and syllable tasks — 0.51 and 0.29, respectively [F(1,16) = 235.16 and 97.89, both P < 0.001]. Words from the animacy task, however, gave rise to better recognition than words from the syllable task [F(1,16) = 145.66, P < 0.001]. Performance for nonconfidently recognized words was not reliably >0 for items from the animacy task (Pr = 0.03), but was greater than chance for items from the syllable task [Pr = 0.09, F(1,16) = 24.59, P < 0.001].

On the basis of these findings, words were categorized as ‘remembered’ when they attracted a confident hit during the subsequent recognition test. Only confident hits showed both accurate discrimination between old and new words and a benefit for deep processing (i.e. better memory for semantically studied versus phonologically studied items). Furthermore, previous studies have shown that subsequent memory effects are predominantly found for confident hits (Brewer et al., 1998; Wagner et al., 1998b;Otten et al., 2001). Words were categorized as ‘forgotten’ when they attracted a miss. This classification differs from that adopted in our previous study (Otten et al., 2001), in which words were categorized as forgotten when they attracted either a miss or a nonconfident hit. In contrast to that study, all volunteers in the current study made at least 12 misses in either task, allowing a more direct assessment of subsequent memory effects by including misses only while maintaining an acceptable signal-to-noise ratio.

fMRI Findings

All fMRI analyses described below were confined to study trials associated with correct animacy or syllable decisions. Because the results from the late covariate did not provide information relevant to the pre-experimental questions, for the sake of clarity and brevity we do not report those results here (results are available from the corresponding author on request).

Type of Study Task

Compared to the low-level baseline condition, words from the study tasks were associated with greater fMRI signals in bilateral visual cortex, extending into fusiform cortex, in bilateral posterior parietal regions and in left and medial prefrontal regions (see Figs 1A,B). The regions selectively engaged by each study task were identified by direct contrasts between the two tasks. fMRI signals were greater for words from the animacy than the syllable task in, among other regions, the medial frontal gyrus, the ventral extent of the left inferior frontal gyrus and left parahippocampal cortex (see Fig. 1C). Greater activations for words from the syllable than animacy tasks were found in the dorsal extent of the left inferior frontal gyrus and bilateral posterior parietal and occipital regions (see Fig. 1D).

Subsequent Memory Effects

Regions demonstrating a subsequent memory effect were identified by contrasts between remembered and forgotten items, performed separately for the animacy and syllable tasks. Table 2 lists the regions that showed greater fMRI signals for subsequently remembered than subsequently forgotten words for the two study tasks. For the animacy task, subsequent memory effects were found in the dorsal extent of the left inferior frontal gyrus and in the medial frontal gyrus. For the syllable task, subsequent memory effects were found in bilateral fusiform cortex, bilateral intraparietal sulcus, the right inferior frontal gyrus, the left cuneus and the left superior occipital gyrus. The regions showing subsequent memory effects in the animacy and syllable tasks are illustrated in Figures 2A and B, respectively.

Compared with our previous study (Otten et al., 2001), fewer regions showed reliable subsequent memory effects in the animacy task at the P < 0.001 threshold. However, when the threshold was lowered to P < 0.01, a pattern more similar to that found in our previous experiment was revealed. Most notably, lowering the threshold brought out a subsequent memory effect in a more extensive region of left inferior frontal gyrus, including a ventral region (BA 47, 24 voxels, x = –54, y = 30, z = 0, Z = 2.75). Even at this liberal threshold, however, no sign of a hippocampal effect was found.

Based on the subsequent memory contrasts performed in each study task, it appears that, in contrast to the frontal focus of the regions showing a subsequent memory effect in the animacy task, the areas showing a subsequent memory effect in the syllable task were localized primarily to posterior regions. The question of a dissociation between animacy and syllable subsequent memory effects was addressed further by comparing the parameter estimates for the subsequent memory contrasts across study tasks (see Fig. 3). These parameter estimates were obtained from the voxel of maximum activation in the regions showing reliable subsequent memory effects in each task.

As can be seen in Figure 3, the parameter estimates for subsequent memory effects in the medial frontal gyrus, right inferior frontal gyrus, bilateral intraparietal sulcus, bilateral fusiform gyrus and left superior occipital gyrus showed a reliable difference according to type of study task [F(1,16) = 14.84, 5.47, 21.94, 5.07, 19.40, respectively, all P < 0.05]. For the medial frontal gyrus, this difference took the form of a reliable subsequent memory effect in the animacy, but not the syllable, task. For the remaining regions, this pattern was reversed. The only region not to show a reliable subsequent memory difference across study tasks was the left inferior frontal gyrus [F(1,16) = 1.06, P > 0.10]. This region none the less only showed a reliable subsequent memory effect for the animacy task, the subsequent memory effect in the syllable task exhibiting substantial across-subject variability.

A potential problem with the foregoing analyses is that the interactions supporting the task dissociations between subsequent memory effects may have been positively biased. The bias originates from the fact that the voxels selected for these analyses were those demonstrating the most significant subsequent memory effect in only one of the two tasks. Thus, we supplemented these analyses by using SPM to perform two further contrasts, which identified all voxels in which subsequent memory effects were greater in one of the tasks than the other. These contrasts corroborated the conclusions presented above, in that they identified voxels which either overlapped, or were nearby, the loci of the voxels illustrated in Figure 3. Thus, regions in which subsequent memory effects were greater (albeit in two cases at the lower threshold of P < 0.005) in the syllable task included the left (x = –42, y = –39, z = 42, Z = 2.91) and right (x = 39, y = –42, z = 45, Z = 3.78) intraparietal sulcus, the right fusiform gyrus (x = 45, y = –42, z = –24, Z = 3.50), the left superior occipital gyrus (x = –21, y = –72, z = 27, Z = 3.26) and the right inferior frontal gyrus (x = 33, y = 30, z = 3, Z = 3.16). Larger effects for the animacy task were found in the medial frontal gyrus (x = –6, y = 45, z = 33, Z = 2.77). The time courses of the signal change for subsequently remembered and subsequently forgotten words (relative to the null events) are shown in Figure 4 for the left inferior and medial frontal gyrus and the left superior occipital gyrus.

Overlap between Subsequent Memory Effects and Type of Study Task

To identify overlap between the regions demonstrating a subsequent memory effect in a study task and the regions selectively involved during engagement in the task, two types of masking analysis were performed. In the first, the subsequent memory contrast for each study task was masked by the contrast comparing that study task with the low-level baseline condition. In the second, the subsequent memory contrasts were masked by the appropriate contrast between the two study tasks. All contrasts were thresholded at P < 0.001.

For the animacy task, the left inferior frontal gyrus (BA 44/45, 5 voxels, x = –48, y = 18, z = 15, Z = 3.67) was found to overlap between the subsequent memory contrast and the task versus baseline contrast. When the animacy subsequent memory contrast was masked by the animacy versus syllable contrast, overlap was found in the medial frontal gyrus (BA 9, 8 voxels, x = –3, y = 48, z = 33, Z = 3.74). For the syllable task, the masking of the subsequent memory contrast by the task versus baseline contrast resulted in overlap in left (BA 40, 14 voxels, x = –39, y = –45, z = 51, Z = 3.86) and right (BA 40, 27 voxels, x = 39, y = –45, z = 45, Z = 4.02) intraparietal sulcus, left fusiform cortex (7 voxels, x = –42, y = –48, z = –15, Z = 3.92) and left superior occipital gyrus (BA 19, 26 voxels, x = –27, y = –66, z = 33, Z = 3.94). The masking of the subsequent memory contrast with the syllable versus animacy contrast resulted in an overlap in both left (BA 40, 13 voxels, x = –39, y = –45, z = 51, Z = 3.86) and right (BA 40, 13 voxels, x = 39, y = –45, z = 45, Z = 4.02) intraparietal sulcus.

Discussion

The present data demonstrate a task-based dissociation in the loci of subsequent memory effects. Effects in the animacy task were localized to left and medial prefrontal cortex, whereas in the syllable task, with the exception of the right prefrontal cortex, effects were confined predominantly to posterior regions. The findings suggest that episodic encoding of visually presented words is supported by different cortical regions depending on the nature of the processing engaged by the study task. Thus, it appears that episodic encoding for words is not supported by a single neural network that operates regardless of the circumstances in which encoding occurs. The present findings therefore complement those from previous studies in which it was demonstrated that the loci of encoding-related activations vary according to the nature of the to-be-remembered material, for example words versus faces (Kelley et al., 1998; Wagner et al., 1998a; McDermott et al., 1999).

The present findings differ from those of our previous fMRI study (Otten et al., 2001). In that study, the regions showing subsequent memory effects in an alphabetical study task were merely a subset of the regions showing effects in the same animacy task as that employed here, but see Otten and Rugg (Otten and Rugg, 2001) for electrophysiological evidence of a qualitative difference between the subsequent memory effects associated with animacy versus alphabetical tasks. As noted in the Introduction, there are at least two possible reasons why the present findings differed from those found previously. First, more words were remembered from the syllable task than the alphabetical task employed previously, yielding greater power for the detection of subsequent memory effects. Second, and more likely, memory for words presented in the syllable task was supported by products of the processing engaged by the task itself, rather than semantic processing incidental to it. Below, we discuss the pattern of findings for each study task in more detail.

Collapsed over memory performance, the regions activated during engagement in the animacy task, as revealed by contrasts with either the baseline condition or the syllable task, resembled those reported in previous positron emission tomography (PET) and fMRI studies employing similar contrasts (Poldrack et al., 1999; Price, 2000). Engagement in semantic discrimination tasks has consistently been associated with activation in a distributed neural network that includes the left inferior frontal gyrus and multiple regions of the temporal lobe. A notable addition to this pattern seen in the present experiment was activation of a medial frontal region (BA 8/9), revealed by the semantic versus syllable task contrast. Activation of the same region was reported by Scott and colleagues (Scott et al., 1999) for a contrast very similar to that employed here (can a word apply to a human being? versus does the word have three syllables?). Together, these findings seemingly suggest a role for this region in cognitive operations related to semantic discrimination.

Two further aspects of the findings regarding the medial frontal region are noteworthy. First, task-sensitivity of this region interacted with subsequent memory. As illustrated in Figure 4, only those words in the animacy task that were subsequently remembered elicited responses distinct from those elicited by phonologically judged items. Given that study performance on the animacy task differed little between words that were subsequently remembered as opposed to forgotten, it is unlikely that the greater medial prefrontal activity observed in the semantic relative to the syllable task reflects processing necessary for successful semantic discrimination. That said, whatever the nature of the processing reflected by this difference in medial prefrontal activity, it appears to have conferred an advantage on semantically judged items in respect of their subsequent memorability.

Second, Figure 4 shows that the medial frontal subsequent memory effect (and, to a lesser extent, the effects in left inferior frontal and superior occipital regions also) developed prior to stimulus onset. A plausible explanation for this observation, which is consistent with findings from our previous ERP study (Otten and Rugg, 2001) is that the prestimulus cues that were employed to signal the task requirement on each trial permitted the initiation of processes important for subsequent memory performance. These processes may be related to preparation for the task signalled by the prestimulus cue, reflecting the fact that some preparatory states are more conducive to effective encoding than others. Using a design which permits separate estimation of cue- and item-evoked responses (Ollinger et al., 2001), it will be of interest in future research to determine how such prestimulus effects interact with encoding processes initiated by the study items themselves.

In addition to the medial prefrontal region just discussed, subsequent memory effects in the animacy task involved a left prefrontal region, which corresponded fairly closely to that identified in our previous study (Otten et al., 2001). Unlike in that study, however, there was no sign in the present case of a subsequent memory effect in the left hippocampus. Our original finding of subsequent memory effects in left hippocampus was replicated recently in another laboratory in a study employing pleasantness judgements as an encoding task (P. Fletcher, personal communication). See Kirchhoff et al .and Davachi et al . (Kirchhoff et al., 2000; Davachi et al., 2001) for similar findings. It therefore seems unlikely that our previous finding represents a type I error, but this does nothing to explain why no hippocampal subsequent memory effect was observed in the animacy task of the present study. In this regard it may be of relevance to note that the mean RT for animacy decisions in the current experiment was 1127 ms, compared to 955 ms in Otten et al .(Otten et al., 2001) [F(1,30) = 29.85, P < 0.01]. The difference in RTs suggests that performance in the animacy task is affected by its ‘context’, i.e. the nature of the task with which animacy judgements are intermixed. How or why such context effects would modulate encoding-related hippocampal activity is, however, unclear.

The robustness of subsequent memory effects in the left inferior (Wagner et al., 1998b; Henson et al., 1999; Kirchhoff et al., 2000; Baker et al., 2001;Buckner et al., 2001;Davachi et al., 2001;Otten et al., 2001) and medial (Otten et al., 2001) (present findings) frontal cortex suggests that these prefrontal regions show subsequent memory effects in semantic study tasks independently of the context in which the task is performed. We have commented already on our findings for the medial frontal cortex. In the case of the left inferior frontal regions, the present findings are consistent with previous suggestions (Buckner and Koutstaal, 1998; Gabrieli et al., 1998; Wagner et al., 1998b, 1999) that the left prefrontal subsequent memory effect reflects the extent to which an item engages semantic working memory operations. In turn, this influences the likelihood that the item will be effectively encoded.

Collapsed across memory performance, engagement in the syllable task was associated primarily with activation in the dorsal extent of the left inferior frontal gyrus and in a variety of posterior cortical regions. These findings closely resemble those reported previously for similar phonological processing tasks (Price et al., 1997; Mummery et al., 1998; Poldrack et al, 1999.). As was the case for the animacy task, the regions showing subsequent memory effects in the syllable task were, in the main, a subset of those activated by engagement in the task overall.

An unexpected observation in the syllable task was the absence of reliable left prefrontal subsequent memory effects. This was true even for the left frontal region in the vicinity of BA 44 that demonstrated greater activity than during animacy judgement (mean parameter estimate for the syllable subsequent memory effect at the voxel displaying the peak inter-task maximum was 0.025, with a standard error of 0.035). One possibility is that the high level of engagement of this region necessitated by the demands of the syllable task ‘saturated’ the BOLD signal and obscured any encoding-related variation. As shown in Figures 3and 4, the region in which an animacy subsequent memory effect was found did show a trend toward a syllable subsequent memory effect, albeit in the face of high inter-subject variability. In light of this trend, it would be premature to conclude that left prefrontal cortex does not demonstrate subsequent memory effects in phonological study tasks such as that employed here. The inter-subject variation observed in this region for the syllable task may reflect differences across subjects in the degree to which words received incidental semantic processing.

The most novel aspect of the present data concerns the finding of subsequent memory effects specific to the syllable task in posterior cortex. These effects were found in bilateral intraparietal sulcus, the left cuneus/superior occipital gyrus and right fusiform cortex. Posterior subsequent memory effects are, however, not without precedent. Kirchhoff et al .(Kirchhoff etal., 2000) described subsequent memory effects in right intraparietal sulcus (as well as the posterior cingulate) for pictorial stimuli and Davachi et al. (Davachi et al., 2001) reported subsequent memory effects in superior parietal cortex for words subjected to rote rehearsal. Fusiform subsequent memory effects were reported by Wagneret al .(Wagner et al., 1998b) and Baker et al .(Baker et al., 2001).

The regions identified in the present experiment as left and right intraparietal sulcus (x = –39, y = –45, z = 51 and x = 39, y= –45, z = 45, respectively) are slightly superior to those identified by Price and colleagues (Price et al., 1997) (x = –40, y = –46, z = 32 and x = 34, y = –50, z = 36, respectively) for a contrast between similar syllable and semantic judgement tasks to those employed here. Similarly, the left cuneus and the region identified as the left superior occipital gyrus in the present experiment (x = –24, y = –75, z = 33) are close to the occipital region identified by Price et al .(x = –14, y = –80, z = 20) in the same contrast. Price et al .(Price et al., 1997) suggested that the parietal activations they observed may reflect short-term phonological memory processes engaged by the syllable task, perhaps related to its counting component (Paulesu et al., 1993; Démonet et al., 1994). They further suggested that the lateral occipital activation was associated with the use of a word's visual attributes (such as its length) to arrive at the correct count. The present findings therefore suggest that the greater the engagement of phonological working memory by an item, the more likely is the item to undergo effective episodic encoding. Furthermore, the existence of subsequent memory effects in different posterior areas suggests that the representational basis of the memories supporting the recognition memory of words from the syllable task may have relied on the products of multiple, distinct cognitive operations.

In each study task, the majority of the regions showing subsequent memory effects were also selectively activated by task engagement. Thus, it appears that variation in effectiveness of episodic encoding is determined by variation in the activity of regions supporting task-specific cognitive operations. This implies that, at the cortical level, episodic encoding of a stimulus event is a ‘byproduct’ of whatever processes are active in the service of online processing of the event (Lockhart, 1992). In both the present study and other experiments that investigated the question (Wagner et al., 1998b;Baker et al., 2001;Otten et al., 2001), only a subset of the regions exhibiting task specificity showed subsequent memory effects; see, however, Davachi et al .(Davachi et al., 2001) for an exception. This observation raises the possibility that the products of a relatively constrained set of cognitive operations have a privileged role in the formation of durable memory representations. Alternatively, the observation may merely reflect inhomogeneity in the sensitivity with which subsequent memory effects can be detected in different cortical regions. According to this second possibility, given a sufficiently sensitive experiment, all regions engaged by a study task would manifest such effects.

The only region to show a subsequent memory effect, but no task effect, was the dorsal extent of the right inferior frontal gyrus (BA 45) in the syllable task. A similar region of right prefrontal cortex (albeit in the animacy task) showed a subsequent memory effect in the Otten et al .(Otten et al., 2001) study, again with no overlap with task effects. This right prefrontal area may therefore reflect processes that benefit episodic encoding in ways other than through cognitive operations associated with task engagement, perhaps by supporting the encoding of nonverbal stimulus attributes (Brewer et al., 1998; Kelley et al., 1998; Wagner et al., 1998a; McDermott et al., 1999).

It is important to emphasize that the subsequent memory approach used here and in similar previous studies identifies brain regions that demonstrate an association with effective memory encoding. The approach does not, however, permit strong conclusions as to which, if any, of the regions play a causal role in the formation of durable memories. Such conclusions require evidence from studies in which the candidate loci of memory encoding have been rendered dysfunctional (for example, through the effects of a lesion or, perhaps, transcranial magnetic stimulation). The present findings identify some of these candidates and indicate that they vary according to encoding task. Whether other manipulations, most notably type of retrieval task, also influence the regions that are associated with episodic encoding, remains an interesting question for future research.

Notes

The authors and their research are supported by the Wellcome Trust. MRI scanning took place at the Wellcome Department of Cognitive Neurology, London, UK. We thank the radiography staff and Dimitris Tsivilis for their help with data acquisition, Rik Henson and Jorge Armony for their advice on data analysis and two anonymous reviewers for their comments.

Table 1

Recognition memory performance for new words, and old words requiring an animacy or syllable decision during study

Word type Recognition judgement 
 Sure old Unsure old Sure new Unsure new 
Values are across-volunteer means (SD). 
a,b,cThe mean reaction times for sure new judgements to old animacy-judged, old syllable-judged and new words are based on 14, 15 and 16 volunteers, respectively. The remaining volunteers did not make any such judgements. 
 Proportion of responses 
Old     
    Animacy 0.59 (0.15) 0.21 (0.08) 0.06 (0.08) 0.14 (0.09) 
    Syllable 0.37 (0.14) 0.27 (0.08) 0.13 (0.14) 0.24 (0.13) 
New 0.07 (0.06) 0.19 (0.11) 0.33 (0.26) 0.41 (0.21) 
 
 Mean reaction time (ms) 
Old     
    Animacy 1194 (198) 1712 (341) 1564a (396) 1778 (409) 
    Syllable 1233 (182) 1657 (328) 1554b (301) 1807 (392) 
New 1359 (323) 1720 (362) 1521c (329) 1719 (362) 
Word type Recognition judgement 
 Sure old Unsure old Sure new Unsure new 
Values are across-volunteer means (SD). 
a,b,cThe mean reaction times for sure new judgements to old animacy-judged, old syllable-judged and new words are based on 14, 15 and 16 volunteers, respectively. The remaining volunteers did not make any such judgements. 
 Proportion of responses 
Old     
    Animacy 0.59 (0.15) 0.21 (0.08) 0.06 (0.08) 0.14 (0.09) 
    Syllable 0.37 (0.14) 0.27 (0.08) 0.13 (0.14) 0.24 (0.13) 
New 0.07 (0.06) 0.19 (0.11) 0.33 (0.26) 0.41 (0.21) 
 
 Mean reaction time (ms) 
Old     
    Animacy 1194 (198) 1712 (341) 1564a (396) 1778 (409) 
    Syllable 1233 (182) 1657 (328) 1554b (301) 1807 (392) 
New 1359 (323) 1720 (362) 1521c (329) 1719 (362) 
Table 2

Regions showing significant (P < 0.001) signal increases for remembered versus forgotten words in the animacy and syllable study tasks

Study task Location (x,y,zPeak Z (no. of voxels) Region Brodmann area 
The location is with respect to the system of Talairach and Tournoux (Talairach and Tournoux, 1988). Z values refer to the peak of the activated cluster, the size of which is indicated in parentheses. 
Animacy –51, 18, 15 3.76 (16) left inferior frontal gyrus 44 
 –3, 48, 33 3.74 (20) medial frontal gyrus  8/9 
 
Syllable –39,–45, 51 3.86 (14) left intraparietal sulcus 
 39,–45, 45 4.02 (45) Right intraparietal sulcus 
 –42,–48,–15 3.92 (21) left fusiform gyrus 37 
 48,–48,–21 3.34 (9) right fusiform gyrus 37 
 –24,–75, 33 3.99 (58) left superior occipital gyrus 19 
 –27,–84, 24 3.89 (6) left cuneus 18 
 36, 36, 3 3.47 (7) right inferior frontal gyrus 45 
Study task Location (x,y,zPeak Z (no. of voxels) Region Brodmann area 
The location is with respect to the system of Talairach and Tournoux (Talairach and Tournoux, 1988). Z values refer to the peak of the activated cluster, the size of which is indicated in parentheses. 
Animacy –51, 18, 15 3.76 (16) left inferior frontal gyrus 44 
 –3, 48, 33 3.74 (20) medial frontal gyrus  8/9 
 
Syllable –39,–45, 51 3.86 (14) left intraparietal sulcus 
 39,–45, 45 4.02 (45) Right intraparietal sulcus 
 –42,–48,–15 3.92 (21) left fusiform gyrus 37 
 48,–48,–21 3.34 (9) right fusiform gyrus 37 
 –24,–75, 33 3.99 (58) left superior occipital gyrus 19 
 –27,–84, 24 3.89 (6) left cuneus 18 
 36, 36, 3 3.47 (7) right inferior frontal gyrus 45 
Figure 1.

Maximum intensity projections detailing regions that showed significant (P < 0.001) signal increases loading on the early covariate for: (A) words studied in the animacy study task relative to the low-level baseline condition; (B) words studied in the syllable study task relative to the low-level baseline condition; (C) words studied in the animacy versus syllable study tasks; and (D) words studied in the syllable versus animacy study tasks. Also shown are surface renderings of the activations on the Montreal Neurological Institute (MNI) reference brain.

Figure 1.

Maximum intensity projections detailing regions that showed significant (P < 0.001) signal increases loading on the early covariate for: (A) words studied in the animacy study task relative to the low-level baseline condition; (B) words studied in the syllable study task relative to the low-level baseline condition; (C) words studied in the animacy versus syllable study tasks; and (D) words studied in the syllable versus animacy study tasks. Also shown are surface renderings of the activations on the Montreal Neurological Institute (MNI) reference brain.

Figure 2.

Signal increases for subsequently remembered versus subsequently forgotten words loading on the early covariate (P < 0.001).The activations are rendered onto the Montreal Neurological Institute (MNI) reference brain. (A) Effects in the animacy task in left inferior frontal and medial frontal gyrus. (B) Effects in the syllable task in bilateral intraparietal sulcus, bilateral fusiform cortex, right inferior frontal cortex and left superior occipital gyrus.

Figure 2.

Signal increases for subsequently remembered versus subsequently forgotten words loading on the early covariate (P < 0.001).The activations are rendered onto the Montreal Neurological Institute (MNI) reference brain. (A) Effects in the animacy task in left inferior frontal and medial frontal gyrus. (B) Effects in the syllable task in bilateral intraparietal sulcus, bilateral fusiform cortex, right inferior frontal cortex and left superior occipital gyrus.

Figure 3.

Parameter estimates on the early covariate for subsequent memory effects (i.e. differences between remembered and forgotten words) in the animacy and syllable tasks for the left inferior frontal gyrus, medial frontal gyrus, right inferior frontal gyrus, intraparietal sulcus (averaged across the left and right hemispheres), fusiform gyrus (averaged across the left and right hemispheres) and left superior occipital gyrus (see Table 2 for coordinates). Error bars show the standard error of the mean.

Parameter estimates on the early covariate for subsequent memory effects (i.e. differences between remembered and forgotten words) in the animacy and syllable tasks for the left inferior frontal gyrus, medial frontal gyrus, right inferior frontal gyrus, intraparietal sulcus (averaged across the left and right hemispheres), fusiform gyrus (averaged across the left and right hemispheres) and left superior occipital gyrus (see Table 2 for coordinates). Error bars show the standard error of the mean.

Figure 4.

Event-related responses (relative to null events) time-locked to onset of subsequently remembered and forgotten words in the two tasks, averaged across all subjects. Responses were obtained for each subject with a finite impulse response model (Ollinger et al., 2001). Responses are shown for the voxel of maximum activation in the subsequent memory cluster in: (A) left inferior frontal gyrus (x = –51, y = 18, z = 15); (B) medial frontal gyrus (x = –3, y = 48, z = 33); and (C) left superior occipital gyrus (x = –24, y = –75, z = 33).

Figure 4.

Event-related responses (relative to null events) time-locked to onset of subsequently remembered and forgotten words in the two tasks, averaged across all subjects. Responses were obtained for each subject with a finite impulse response model (Ollinger et al., 2001). Responses are shown for the voxel of maximum activation in the subsequent memory cluster in: (A) left inferior frontal gyrus (x = –51, y = 18, z = 15); (B) medial frontal gyrus (x = –3, y = 48, z = 33); and (C) left superior occipital gyrus (x = –24, y = –75, z = 33).

References

Andrade A, Paradis AL, Rouquette S, Poline JB (
1999
) Ambiguous results in functional neuroimaging data analysis due to covariate correlation.
NeuroImage
 
10
:
483
–486.
Baker JT, Sanders AL, Maccotta L, Buckner RL (
2001
) Neural correlates of verbal memory encoding during semantic and structural processing tasks.
NeuroReport
 
12
:
1251
–1256.
Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JDE (
1998
) Making memories: brain activity that predicts how well visual experience will be remembered.
Science
 
281
:
1185
–1187.
Buckner RL, Koutstaal, W (
1998
) Functional neuroimaging studies of encoding, priming, and explicit memory retrieval.
Proc Natl Acad Sci USA
 
95
:
891
–898.
Buckner RL, Wheeler ME, Sheridan M (
2001
) Encoding processes during retrieval tasks.
J Cogn Neurosci
 
13
:
406
–415.
Challis BH, Velichkovsky BM, Craik FIM (
1996
) Levels-of-processing effects on a variety of memory tasks: new findings and theoretical implications.
Consciousness Cogn
 
5
:
142
–164.
Cocosco CA, Kollokian V, Kwan RKS, Evans AC (
1997
) BrainWeb: online interface to a 3D MRI simulated brain database.
NeuroImage
 
5
:
S425
.
Dale A, Buckner R (
1997
) Selective averaging of rapidly presented individual trials using fMRI.
Hum Brain Mapp
 
5
:
329
–340.
Davachi L, Maril A, Wagner AD (2001) When keeping in mind supports later bringing to mind: neural markers of phonological rehearsal predict subsequent remembering. J Cogn Neurosci (in press).
Démonet JF, Price CJ, Wise RW, Frackowiak RSJ (
1994
) Differential activation of right and left posterior sylvian regions by semantic and phonological tasks: a positron emission tomography study in normal subjects.
Neurosci Lett
 
182
:
25
–28.
Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RSJ (
1995
) Statistical parametric maps in functional imaging: a general linear approach.
Hum Brain Mapp
 
2
:
189
–210.
Friston KJ, Fletcher PC, Josephs O, Holmes A, Rugg MD, Turner R (
1998
) Event-related fMRI: characterising differential responses.
NeuroImage
 
7
:
30
–40.
Gabrieli JDE, Poldrack RA, Desmond JE (
1998
) The role of left prefrontal cortex in language and memory.
Proc Natl Acad Sci USA
 
95
:
906
–913.
Henson RNA, Rugg MD, Shallice T, Josephs O, Dolan RJ (
1999
) Recollection and familiarity in recognition memory: an event-related functional magnetic resonance imaging study.
J Neurosci
 
19
:
3962
–3972.
Henson RNA, Rugg MD, Shallice T, Dolan RJ (
2000
) Confidence in recognition memory for words: dissociating right prefrontal roles in episodic retrieval.
J Cogn Neurosci
 
12
:
913
–923.
Josephs O, Turner R, Friston KJ (
1997
) Event-related fMRI.
Hum Brain Mapp
 
5
:
243
–248.
Kelley W, Miezin F, McDermott K, Buckner RL, Raichle ME, Cohen NJ et al .(
1998
) Hemispheric specialization in human dorsal frontal cortex and medial temporal lobe for verbal and nonverbal memory encoding.
Neuron
 
20
:
927
–936.
Kirchhoff BA, Wagner AD, Maril A, Stern CE (
2000
) Prefrontal-temporal circuitry for episodic encoding and subsequent memory.
J Neurosci
 
20
:
6173
–6180.
Kucera H, Francis WN (1967) Computational analysis of present-day American English. Providence, RI: Brown University Press.
Lockhart RS (1992) Levels of processing. In: Encyclopedia of learning and memory (Squire LR, ed.), pp. 106–108. New York: Macmillan.
McDermott KB, Buckner RL, Petersen SE, Kelley WM, Sanders AL (
1999
) Set- and code-specific activation in frontal cortex: an fMRI study of encoding and retrieval of faces and words.
J Cogn Neurosci
 
11
:
631
–640.
Mummery CJ, Patterson K, Hodges JR, Price CJ (
1998
) Functional neuroanatomy of the semantic system: divisible by what?
J Cogn Neurosci
 
10
:
766
–777.
Ollinger JM, Shulman, GL, Corbetta, M (
2001
) Separating processes within a trial in event-related functional MRI: I. The method.
NeuroImage
 
13
:
210
–217.
Otten LJ, Henson RNA, Rugg MD (
2001
) Depth of processing effects on neural correlates of memory encoding: relationship between findings from across- and within-task comparisons.
Brain
 
124
:
399
–412.
Otten LJ, Rugg MD (
2001
) Electrophysiological correlates of memory encoding are task-dependent.
Cogn Brain Res
 
12
:
11
–18.
Paulesu E, Frith CD, Frackowiak RSJ (
1993
) The neural correlates of the verbal component of working memory.
Nature
 
362
:
342
–344.
Poldrack RA, Wagner AD, Prull MW, Desmond JE, Glover GH, Gabrieli JDE (
1999
) Functional specialization for semantic and phonological processing in the left inferior prefrontal cortex.
NeuroImage
 
10
:
15
–35.
Price CJ (
2000
) The anatomy of language: contributions from functional neuroimaging.
J Anat
 
197
:
335
–359.
Price CJ, Moore CJ, Humphreys GW, Wise RJS (
1997
) Segregating semantic from phonological processes during reading.
J Cogn Neurosci
 
9
:
727
–733.
Schacter DL, Buckner RL, Koutstaal W, Dale AM, Rosen BR (
1997
) Late onset of anterior prefrontal activity during retrieval of veridical and illusory memories: a single trial fMRI study.
NeuroImage
 
6
:
259
–269.
Scott SK, Leff A, Wise R (
1999
) Can a human be a lemon? A neural system for thinking about meaning.
NeuroImage
 
9
:
S341
.
Snodgrass JG, Corwin J (
1988
) Pragmatics of measuring recognition memory: applications to dementia and amnesia.
J Exp Psychol
 
117
:
34
–50.
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas of the human brain. Stuttgart: Thieme.
Wagner AD, Poldrack RA, Eldridge LL, Desmond JE, Glover GH, Gabrieli JDE (
1998
) Material-specific lateralization of prefrontal activation during episodic encoding and retrieval.
NeuroReport
 
9
:
3711
–3717.
Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Dale AM et al .(
1998
) Building memories: remembering and forgetting of verbal experiences as predicted by brain activity.
Science
 
281
:
1188
–1191.
Wagner AD, Koutstaal W, Schacter DL (
1999
) When encoding yields remembering: insights from event-related neuroimaging.
Phil Trans R Soc Lond B Biol Sci
 
354
:
1307
–1324.
Wilding EL, Rugg MD (1996) An event-related potential study of recognition memory with and without retrieval of source.
Brain
 
119
:
889–
905. [Erratum: Brain 1996;119:1416.]
Zarahn E, Aguirre GK, D'Esposito M (
1997
) A trial-based experimental design for fMRI.
NeuroImage
 
5
:
179
–197.