Skip to main content
NCBI home page
Cerebral Cortex (New York, NY) logoLink to Cerebral Cortex (New York, NY)
. 2010 May 21;21(2):272–280. doi: 10.1093/cercor/bhq092

Medial Temporal Lobe Contributions to Episodic Sequence Encoding

Shannon Tubridy 1,✉, Lila Davachi 1,2,✉
PMCID: PMC3020579  PMID: 20494967

Abstract

Memory for the order of events within an episode requires mechanisms capable of bridging gaps between those events and laying down memory traces that will support the subsequent retrieval of episodic sequence information. It has been proposed that the hippocampus and surrounding medial temporal lobe (MTL) play a critical role in these processes. Here, using functional magnetic resonance imaging, we examined MTL activation during the encoding of sequentially presented noun triplets. Using performance on a retrieval test, encoding activations predictive of subsequent triplet recognition (“hits” vs. “misses”) as well as activations differentiating triplets that were subsequently recognized with (ordered) and without (misordered) memory for the original sequence of presentation were identified. Within the MTL, activations in bilateral hippocampal and parahippocampal cortical regions predicted subsequent order memory, with greater activations observed during encoding of triplets subsequently correctly ordered than those subsequently misordered. Interestingly, activation in these regions did not correlate with old/new triplet identification, suggesting that these MTL regions contributed specifically to encoding of episodic details supporting subsequent recovery of sequence information.

Keywords: fMRI, hippocampus, medial temporal lobe, memory encoding, sequence

Introduction

The medial temporal lobe (MTL) has been implicated in the encoding of events into memory (Cohen and Eichenbaum 1993; Squire et al. 2004). Damage to the hippocampus and surrounding MTL cortex is known to impair episodic memory (Scoville and Milner 1957), and much work has focused on understanding how MTL subregions differentially contribute to memory formation. Functional magnetic resonance imaging (fMRI) in humans has demonstrated that activation in hippocampus and adjacent parahippocampal cortex during encoding correlates with later associative memory as measured by recovery of source details or item associations as well as in the remember/know paradigm, which is considered to roughly index recollection, or memory for specific details associated with the learning episode (Davachi et al. 2003; Ranganath et al. 2003; Kirwan and Stark 2004; Uncapher et al. 2006; Diana et al. 2007; Johnson and Rugg 2007).

The bulk of previous work examining episodic encoding through objective measures of associative encoding has assessed memory based on whether participants remember an associated detail of a previous encounter with a study item, such as the context within which an item was encountered or whether 2 items were initially presented together. However, in addition to what and where things happened, episodic memories contain information about when or in what order they occurred (Clayton and Dickinson 1998; Tulving 2002; Ergorul and Eichenbaum 2004), and the ability to discriminate between past episodes composed of overlapping elements will often depend on memory for the particular temporal arrangement of elements within those episodes. However, in contrast to examination of the what and where of episodic memory, little work has focused on the encoding processes supporting memory for the temporal relations among a set of items in a discrete episode.

The neural mechanisms proposed to underlie the establishment of long-term associative memory (i.e., long-term potentiation [LTP]) depend on the coincident activation of pre- and postsynaptic neurons within a 100-ms window (Levy and Steward 1983; Wallenstein et al. 1998). Consequently, encoding of temporal relations among events separated by longer delays requires bridging temporal gaps while preserving information about relative temporal positions. Although it may be that the temporal relationships between events separated widely in time can sometimes be reconstructed as needed based on gross contextual or semantic information, it is unlikely that such processes can support order memory for sequences of events unfolding in close temporal proximity (St Jacques et al. 2008). Recent computational models have described mechanisms by which the MTL could support the encoding of such sequences of closely spaced, nonoverlapping events (Wallenstein et al. 1998; Jensen and Lisman 2005). These models propose that oscillatory dynamics support the compression and maintenance of sequence information in cortical buffers such that autoassociative mechanisms in the hippocampus are able to encode temporally discontiguous information in a manner enabling subsequent recall of temporal relationships.

Lesion work in rats has shown that the hippocampus is necessary for memory for the order of a sequence of odors but not for simple odor recognition (Fortin et al. 2002; Kesner et al. 2002). In humans, it has recently been shown that during encoding, the hippocampus plays a role in the binding of episodic details separated in time (Staresina and Davachi 2009), but participants in that study were required to report only whether 2 pieces of information occurred within the same event and not the order in which they occurred. Experiments examining sequence encoding have shown hippocampal and/or MTL activation correlating with learning of sequences of visuomotor mappings (Schendan et al. 2003) and faces (Kumaran and Maguire 2006). However, these studies assessed activation across learning of repeated sequences and are consequently limited in their ability to specify MTL contributions to fast (i.e., single trial) temporal encoding within episodes. A recent fMRI study reports activation of hippocampus and parahippocampal cortex during retrieval of temporal order information (Lehn et al. 2009), but MTL contributions to the encoding of information enabling such memory have yet to be demonstrated.

In the study reported here, we used event-related fMRI to assess blood oxygenation level–dependent (BOLD) signal while participants viewed single presentations of triplets of sequentially presented words. During retrieval, participants engaged in a 2-step retrieval test in which they were first queried for their memory for whether a set of 3 words had previously been presented together and then, for recognized triplets, were asked to report the order of presentation of the 3 words during encoding. This paradigm enabled us to assess MTL activations supporting the encoding of temporal relations in discrete episodes and to distinguish such signals from activations contributing to event recognition memory without order information.

Materials and Methods

Participants

Eighteen (13 females and 5 males) healthy, right-handed, native English-speaking volunteers with normal or corrected-to-normal vision participated in the experiment for remuneration after providing informed consent obtained in a manner approved by the University Committee on Activities Involving Human Subjects (UCAIHS) of New York University. Of the 18 participants, 1 female subject was removed from analysis due to image normalization issues. Two female participants completed only 3 of 6 runs of the experiment, but because the resultant number of trials from those 3 runs was sufficient for data analysis purposes, they remained in the subject pool.

Stimuli and Behavioral Procedures

All subjects were familiarized with the tasks prior to scanning and were given an opportunity to practice the tasks until they were comfortable with the timing and nature of the required responses.

Scanned Encoding Task

Over 6 encoding runs, 180 triplets (30 in each run) of common nouns matched for length, frequency, and imageability were presented intermixed with a baseline fixation period (white plus sign). During each 7500-ms trial, each noun from a triplet was presented in the center of the display for 750 ms with a 500-ms blank screen separating the presentation of each word (Fig. 1). Following the third word in the triplet, subjects were presented with a cue (red asterisk) to “imagine” during which time they were to imagine the referents of each of the nouns interacting in the order in which they were presented. The red asterisk remained on the screen for 3500 ms to allow participants sufficient time to associate the presented elements. After 3500 ms, the red asterisk turned to a green question mark signaling that subjects were to indicate their success at having completed the task. Specifically, they were instructed to report if 1) they had successfully imagined the referents of all 3 nouns and were able to do this in the order of presentation of the nouns, 2) they had successfully imagined the referents of all 3 nouns but were not able to do this in the order of presentation of the nouns, or 3) they were unsuccessful in imagining the referents of all 3 nouns. The encoding trials were interspersed with variable duration baseline fixation displays whose length was designed to optimize the recovery of the BOLD response (OPTSEQ, http://surfer.nmr.mgh.harvard.edu/optseq; Dale 1999).

Figure 1.

Figure 1.

Open in a new tab

Schematic of behavioral tasks. During each 7500 ms, scanned encoding trial participants saw 3 nouns presented sequentially after which they had 3500 ms to create a mental image of a scene incorporating the 3 words (asterisk) and 750 ms to report their degree of success at creating the mental image (question mark). Each unscanned retrieval trial consisted of a probe for participants’ memory of whether a noun triplet was presented during encoding (“old or new?”) followed by a prompt to reconstruct the order of presentation of the words in a noun triplet during encoding (“order?”). See Stimuli and behavioral procedures for more details.

Unscanned Retrieval Test

Each of the 6 encoding runs was followed immediately by an unscanned retrieval test block. During each test block, subjects were presented with the 30 noun triplets from the immediately preceding study block along with an additional 15 previously unseen distractor noun triplets. Test blocks were self-paced. On each test trial, all 3 words of a noun triplet (target or distractor) were presented simultaneously in a row across the middle of the screen (Fig. 1). The spatial position of each word on the screen was counterbalanced across trials and as such gave no information regarding the ordinal position of each word in the presentation sequence during encoding. Subjects were instructed to first indicate whether or not a presented triplet was “old” (had been seen during the preceding study block) or “new.” If subjects responded old, they were then asked to reorder the nouns in the triplet to match the order of presentation of the nouns during the study phase using a magnet-compatible button box. Three buttons on the box corresponded to the left, middle, or right position of the test triplet words on the display, and participants were instructed to reorder the triplets by sequentially pressing the buttons for the display positions according to their memory for the original presentation order for that triplet. For example, if the noun triplet (PRINCE → FALCON → SALMON) was presented during encoding and displayed as SALMON–PRINCE–FALCON at test, subjects would correctly reorder this triplet by sequentially pressing buttons 2-3-1, indicating that the original sequence was the word in the middle position (PRINCE) followed by the word in the right position (FALCON) and then the word in the left position (SALMON). Each triplet remained on the screen until subjects completed reconstruction of the order after which time the next triplet appeared.

fMRI Procedures

Scanning was performed on a 3T Siemens Allegra scanner at the Center for Brain Imaging at New York University. Functional data were acquired using a gradient-echo echo-planar sequence (time repetition [TR] = 2500 ms, time echo = 30 ms). Each volume contained 40 slices (3mm × 3mm × 3mm voxel size) oriented perpendicular to the long axis of the hippocampus with 0.6 mm interslice gap. During each run, 324 volumes were collected, with 1944 volumes collected across all 6 runs. High-resolution T1-weighted (magnetization prepared rapid gradient echo) images were collected for anatomical localization. Head motion was minimized with an inflatable pillow and foam padding. Visual stimuli were projected onto a screen viewed with a mirror, and participant responses were collected with a magnet-compatible button box.

Data were preprocessed using SPM5 (Wellcome Department of Cognitive Neurology, London). Images were corrected for differences in slice acquisition timing followed by motion correction across all runs. Each subject's functional data were then coregistered to their own T1 anatomical image. Anatomical images were spatially normalized to Montreal Neurological Institute (MNI) space represented by the T1 template image in SPM5 using a 12-parameter affine transformation and a nonlinear transformation using cosine basis functions. Functional images were spatially normalized by applying the normalization parameters estimated during the anatomical normalization for each subject to the coregistered functional images. Functional images were resampled into 3-mm cubic voxels and then spatially smoothed with an 8-mm full-width at half-maximum isotropic Gaussian kernel.

Statistical analysis was performed using the random effects general linear model in Brain Voyager QX 1.9 (Brain Innovation). Encoding trials were sorted according to individual participants’ responses during encoding and their subsequent memory performance. Specifically, encoding trials for which participants reported successful imagery (see Stimuli and Behavioral Procedures above) were sorted according to whether the triplet was subsequently recognized and reordered (ordered), recognized and misordered (misordered), or not recognized (forgotten). Individual trials were modeled as a boxcar spanning the length of each trial (7.5 s, 3 TRs) convolved with a canonical hemodynamic response function. Encoding trials during which subjects reported no imagery or failed to respond were modeled as events of no interest, along with low-frequency components (<3 cycles per scan) of the signal modeled using a Fourier basis set as implemented in Brain Voyager, and subject- and session-specific effects. In addition to the subsequent memory model, a separate task performance model was constructed in which encoding trials given successful imagery responses were collapsed across subsequent memory performance in a single “task” regressor. Other aspects of the task model (encoding trials of no interest, low-frequency regressors, subject-, and session-specific effects) were as in the subsequent memory model.

Effects were estimated using a subject-specific fixed-effects model, with session-specific effects and low-frequency signal components treated as confounds. Linear contrasts were used to obtain subject-specific estimates for each effect. The subject-specific estimates derived from each of these contrasts were entered into a second-level group analysis treating subjects as a random effect. For hypothesis-driven analyses in the MTL, regions were considered reliable if they consisted of at least 16 contiguous voxels that exceeded an uncorrected threshold of P < 0.01. This threshold and spatial extent were obtained from a Monte Carlo simulation using an anatomical mask comprising the MTL in the mean brain across subjects and the spatial smoothness of the data as implemented in the Analysis of Functional Neuroimages package (AFNI; http://afni.nimh.nig.gov/afni/doc/manual/AlphaSim). The resulting significance/extent threshold was estimated to result in a mapwise false-positive rate of P < 0.05. Identification of active regions in the task model (i.e., task > baseline) was carried out at P < 0.0001 with a minimum of 5 contiguous voxels.

MTL regions contributing to the encoding of order information were identified in the group-level contrast between “ordered” and “misordered” trials. This contrast enabled us to identify regions whose activations differentiated according to differences in subsequent order memory while controlling for task processing and other aspects of memory. As the trials comprising the ordered and misordered bins were matched for stimulus display, encoding task, response requirements, and subsequent recognition memory for occurrence of 3 words together, differing in the subsequently measured ability to correctly reconstruct the triplet order, regions identified in this contrast are likely to contribute to the formation of temporal order memory rather than contributing to more general differences between contrasted trials. In addition, the role of these regions in supporting subsequent recognition memory was explored by extracting estimates of activation in the ordered, misordered, and forgotten conditions in each identified MTL region for a subset of participants and comparing responses across conditions. Specifically, these planned comparisons were carried out using the 12 participants who had at least 10 subsequently forgotten trials across the experimental session. For these participants, activation estimates in the ordered, misordered, and forgotten conditions were compared using 2-tailed paired t-tests.

Results

Behavioral

Subjects’ old/new recognition memory for the noun triplets was high. Of the triplets for which subjects reported successful imagery during encoding, the mean percentage of old noun triplets recognized as old was 86.32% (standard deviation [SD] 10.15%). On average, participants recognized significantly more triplets that were reported as “imaged in order” during the encoding session (90%; SD 8.1%) than those that were reported as “imaged but not in order” (82%; SD 12.2%) (t16 = 2.89; P < 0.05). Mean false alarm rate across participants was 4.41% (SD 6.57%). This high level of recognition accuracy was likely the result of the short retention intervals used in this paradigm. Of the noun triplets that were correctly recognized as old, subjects were able to correctly reorder 45.20% (SD 15%), which was significantly above the chance level of 16% (t16 = 12.03, P < 0.0001). Among recognized triplets, there was no significant difference in the percentage of correctly reordered triplets between those encoding trials rated as imaged in order versus not in order (mean percent correct 49% and 42%, respectively; t16 = 1.5, P = 0.15). For the subset of 12 subjects included in the imaging analyses comparing recognized to forgotten triplets (see fMRI procedures above), the mean percentage of triplets recognized at test was 82.65% (SD 7.5%) and the mean percentage of recognized triplets correctly reordered was 41.04% (SD 7.1%), which, as in the full set of subjects, was significantly better than chance performance (t11 = 12.27, P < 0.0001). The false alarm rate for this set of subjects was 5.74% (SD 7.46%).

The present study was designed to explore MTL regions distinguishing the associative encoding of word triplets with and without subsequent order memory. However, it is possible that some of the triplets may have been recognized and/or reordered on the basis of familiarity, particularly those that were presented in the same order as when studied. To assess whether familiarity might have been differentially contributing to any of our behavioral conditions, we compared recognition and subsequent order memory as a function of the left to right screen position of triplets at test. Because presentation order at test was randomized (see Stimuli and Behavioral Procedures), we were able to examine recognition and order memory performance for those triplets whose words’ left to right screen position at test matched the presentation order of words during encoding, with the first word presented during encoding appearing in the left screen position at test, the second word during encoding in the middle screen position at test, and the third word during encoding in the right screen position at test. We reasoned that if familiarity-based processes were making significant contributions to behavioral performance, this should be most evident for those test trials when the screen position of words at test essentially provides a copy cue by matching the temporal order at encoding to the left to right position at test. Comparison of recognition rates for test triplets with left to right positions matching the encoding presentation order against triplets with other test presentation arrangements indicated no difference in recognition memory (t16 = 0.11, P = 0.9). Examination of the proportion of recognized triplets that were subsequently correctly reordered also yielded no effect of matching test order compared with other test orders, with recognized triplets whose screen order at test matched presentation order during encoding reordered at the same rate as other test arrangements (t16 = 1.57, P = 0.14). These results suggest that familiarity-based processes did not dominate performance in this study, as such an effect would likely be evident in a difference in performance for those test trials providing the most direct mapping from encoding order to test presentation.

Functional Magnetic Resonance Imaging

Task Regions

A comparison of all encoding trials compared with baseline fixation trials will reveal regions related to cognitive, perceptual, and motor demands of the task. Accordingly, this contrast revealed activation in numerous areas including the left inferior frontal gyrus (IFG; Brodmann areas [BAs] 44, 45, 47), right dorsolateral prefrontal cortex, bilateral parahippocampal cortex, fusiform and lateral occipitotemporal cortices, bilateral parietal cortex, cerebellum, and basal ganglia (full list of regions available upon request). Activation in both anterior and posterior left IFG, in combination with posterior parietal and lateral temporal cortices, has often been noted in tasks that make demands on phonological and semantic retrieval and/or selection, whether they require imagery (Davachi et al. 2003) or not (Thompson-Schill et al. 1997; Badre and Wagner 2007). In addition, as expected in tasks that require associative processing, strong engagement of the MTL was seen.

Subsequent Memory Analyses

Ordered > misordered: MTL regions.

In order to reveal MTL encoding activation predictive of subsequent memory for the sequence of items in each triplet, trials for which the triplet was later recognized and correctly reordered (ordered) were contrasted against those that were later recognized without sequence memory (misordered). These 2 conditions were equivalent in terms of trial structure and encoding task demands and for memory for the occurrence of a triplet. Hence, the contrast between them will likely identify regions contributing to order memory rather than differences in encoding task demands or other forms of memory. This contrast revealed significant clusters in bilateral MTL. Within the hippocampus, significant subsequent order memory effects were seen in right anterior and mid-hippocampus (peak MNI coordinates 32, −10, −13 and 33, −19, −10) and left hippocampus/subiculum (−18, −24, −9) (Fig. 2A). In addition, regions in bilateral parahippocampal cortex were observed to exhibit greater activation during subsequently ordered compared with subsequently misordered trials (MNI coordinates: 35, −27, −15 and −31, −37, −8) (Fig. 3A).

Figure 2.

Figure 2.

Open in a new tab

Hippocampal regions showing greater BOLD activation during subsequently ordered relative to subsequently misordered trials. (A) Right anterior and mid-hippocampus and left hippocampus/subiculum displayed on mean anatomical brain. (B) Estimates of mean activation across the subset of participants included in the old/new analyses (see Materials and methods) in each hippocampal region for subsequently forgotten, recognized but misordered, and recognized and ordered trials. Lines indicate paired t-test results in the pairwise comparisons (* denotes 2-tailed P < 0.05).

Figure 3.

Figure 3.

Open in a new tab

Parahippocampal regions showing greater BOLD activation during subsequently ordered relative to subsequently misordered trials. (A) Left and right parahippocampal regions of interest displayed on mean anatomical brain. (B) Estimates of mean activation across the subset of participants included in the old/new analyses (see Materials and methods) in each parahippocampal region for subsequently forgotten, recognized but misordered, and recognized and ordered trials. Lines indicate paired t-test results in the pairwise comparisons (* denotes 2-tailed P < 0.05).

To explore whether MTL regions identified in the ordered versus misordered contrast also show subsequent recognition memory effects distinguishing recognized from missed trials, we extracted beta estimates from these regions for ordered, misordered, and forgotten trials for the subset of subjects who had enough misses (see fMRI procedures). These estimates were then compared to see whether these regions showed activation patterns consistent with a role in supporting the subsequent discrimination of old and new triplets. In all regions, as in the full group of subjects, activation during subsequently correctly ordered trials elicited greater activation than subsequently misordered trials (follow-up t-tests in all MTL regions: t16 values > 3.5; P values < 0.005) (Figs 2B and 3B), and each of the hippocampal regions, along with the left parahippocampal cortex, showed greater activation during subsequently ordered trials than subsequently forgotten trials (all t16 values > 2.5; P values < 0.05). However, none of the regions showed activation patterns consistent with contributions to subsequent recognition memory, as there were no differences between subsequently recognized but misordered trials and subsequently forgotten trials. Although the mean beta estimates across subjects for misordered and forgotten trials show numerical differences in some regions, follow-up t-tests in all hippocampal and parahippocampal regions showed no significant effects or trends toward differences between misordered and forgotten (all t16 values < 1; P values > 0.3) (Figs 2B and 3B). Hence, none of the MTL regions that emerged in the statistical map of ordered > misordered also showed activation consistent with overall event memory, as encoding activations in these regions did not differentiate between trials that were subsequently forgotten and those that were recognized but without order memory. Further exploratory contrasts of misordered versus forgotten trials at a relaxed threshold of P < 0.005 with 5 contiguous voxels did not identify any of the MTL regions identified in the ordered versus misordered contrast. This suggests that, in our paradigm at least, activation in these MTL regions does not differentiate according to the general encoding of episodic details, and the close temporal proximity of each within-triplet item at encoding makes it unlikely that order judgments at test were based on relative memory strength for individual items. Instead, the observed MTL activations appear to be related specifically to the encoding of information supporting subsequent memory for the order in which words were encountered.

Interestingly, although the pattern of activation across conditions was similar in each of the 3 hippocampal regions, with ordered trials eliciting significantly greater activation than both misordered and forgotten trials, the parahippocampal regions showed divergent patterns. The left parahippocampal region showed a pattern similar to that in hippocampus, with ordered trials marked by greater activation than both the misordered and forgotten trials, which were statistically equivalent to each other. In contrast, the right parahippocampal region shows evidence of a u-shape (Fig. 3B), with ordered activations significantly greater than misordered but not forgotten trials. Although the low trial numbers in the forgotten condition in this study warrant interpretive caution, this pattern parallels findings in other subsequent memory studies of the parahippocampal cortex (Shrager et al. 2008; Staresina and Davachi 2008). These authors have interpreted elevated levels of MTL activation in missed or low-memory strength conditions as reflective of encoding processes related to information other than that which is tested directly. As experimental tests of subsequent memory tend to be constrained in terms of the criteria constituting successful memory performance, it is possible that elevated activation in MTL during missed trials reflects encoding of details not related to the task or the subsequent memory test. The right parahippocampal results presented here might be explained in terms of such an account, although strong conclusions of about these effects, as well as the differences between left and right parahippocampal activations, will need to wait for studies more directly concerned with subsequently forgotten trials, as the low trial numbers in that condition in this study may be yielding unreliable beta estimates for subsequently forgotten trials.

Ordered > misordered: other regions.

In addition to the targeted exploration of MTL regions showing an ordered > misordered effect, we performed contrasts to identify other regions showing this effect. At conventional brainwise significance levels (P < 0.0001, minimum of 5 contiguous voxels), no regions showed this pattern. Exploratory analyses at relaxed thresholds (P < 0.005, uncorrected, minimum 5 contiguous voxels) identified regions in left striatum, left anterior temporal lobe, left ventrolateral prefrontal cortex (VLPFC; see Frontal Contributions), right lateral temporal cortex, and cerebellum. However, the thresholds used to identify these regions are fairly liberal, and further examination of these regions’ contributions to order memory formation will need to be addressed in studies with targeted hypotheses and analyses.

As a number of regions were also identified as contributing to task performance in the task > baseline contrast (see above), we interrogated these regions to determine whether any showed a subsequent order memory effect. With exception of a marginal effect in VLPFC, none of the regions identified in the task contrast showed differential activation according to subsequent order memory.

Frontal Contributions to Order Memory Encoding

Previous neuropsychological observations have suggested that damage to frontal regions impairs performance on tasks requiring retrieval or reconstruction of episodic temporal information (Shimamura et al. 1990; McAndrews and Milner 1991; Mangels 1997). As bilateral frontal regions were identified in the task versus baseline contrast in the present study, we extracted beta estimates for subsequent memory conditions from these regions to assess whether they showed activation patterns consistent with preferential encoding of temporal or order information. With the exception of a trend in VLPFC, none of the left or right PFC regions from the task contrast showed such a pattern, with subsequently misordered and subsequently ordered trials eliciting comparable levels of frontal activations during encoding in these regions. The left VLPFC region (−40, 33, 8) showed marginally greater activation during ordered than misordered trials (t16 = 1.79, P = 0.09), consistent with findings from primate neurophysiology demonstrating PFC neuronal selectivity for the rank order of items in a to-be-remembered sequence (Ninokura et al. 2003, 2004).

However, the more general absence of an order encoding effect in frontal cortex in the present study may be congruent with the neuropsychological reports. There is evidence that patients with frontal damage are particularly impaired in the deployment of strategic processes during retrieval (Mangels 1997). As the BOLD measurements in the present study were obtained during encoding, we might not expect to see frontal contributions to accurate order encoding because these contributions might be most pronounced during the retrieval of such information.

Similarly, one might expect differences in activation during subsequently ordered and misordered trials in frontal regions contributing to imagery, as the encoding task involved the construction incorporating a mental image of the 3 words in a triplet. However, the ordered and misordered trials included in these analyses were all drawn from test trials with triplets for which participants reported successful imagery during encoding. As such, the current results support a temporal encoding process that does not depend on differences in imagery but instead seem to derive from mnemonic processes indexed by MTL activation during encoding.

MTL and Recognition Memory Effects

The results presented above provide evidence for MTL contributions to encoding processes enabling subsequent recovery of order information, but these none of those regions show a differentiation according to recognition memory performance (i.e., correctly reporting triplets as old, regardless of subsequent order memory). Similarly, the direct contrast of misordered > forgotten trials yielded no MTL regions. Although the present study was designed to examine the differences between encoding processes during trials yielding later recognition with and without order memory, there is abundant evidence of a role for MTL cortical encoding activations in supporting subsequent item memory (e.g., Davachi et al. 2003; Ranganath et al. 2004), and the lack of such findings here should be addressed. Recognition performance in this study was high, resulting in few trials in the forgotten condition, and analysis of this condition was possible for only a subset of participants. Even within this subset, trial numbers in this condition were low, perhaps leading to unreliable estimates of activation levels and obscuring recognition effects. Interestingly, however, a parahippocampal cortex region (−39, −20, −25) identified in the task > baseline contrast exhibited a trend toward a recognition memory effect, with ordered and misordered trials eliciting marginally greater activation than forgotten trials (t11 = 2.21, P = 0.049 for ordered vs. forgotten; t11 = 1.75, P = 0.11 for misordered vs. forgotten). This region was more anterior than the parahippocampal regions coming out of the ordered > misordered and did not show an order memory effect (t16 = 0.04, P = 0.97).

It should be noted that although the parahippocampal regions exhibiting an order effect here are interpreted as supporting subsequent order memory, this is consistent with data showing a role for MTL cortex in the encoding of item information. Indeed, we speculate that the combined parahippocampal and hippocampal order regions are supporting a process whereby item information is processed and maintained in MTL cortex, enabling relational binding across these items to be carried out in the hippocampus.

Discussion

Memory for episodes unfolding across time requires mechanisms for bridging temporal gaps between the constituent elements of an episode in order to encode the relationships between these elements. The ability to remember when things occurred with respect to each other points to encoding processes that can preserve temporal order information. Computational models have previously proposed that the MTL can support such processes through both the maintenance of item-based sequence information in parahippocampal cortex and across-item binding in the hippocampus (Wallenstein et al. 1998; Jensen and Lisman 2005). Addressing the hypothesis that MTL regions support encoding of temporal order, we found activation in bilateral MTL consistent with a role in temporal encoding. Specifically, a contrast of encoding period BOLD activation for triplets that were subsequently correctly reordered compared with those that were subsequently misordered revealed extensive activation in bilateral hippocampus and parahippocampal cortex, despite intact recognition memory in both conditions. These results are consistent with a hypothesized role for MTL in the maintenance and encoding of temporal sequence information (Wallenstein et al. 1998; Jensen and Lisman 2005) and recent observations of hippocampal and parahippocampal contributions to temporal order retrieval (Lehn et al. 2009).

As all the activations in this contrast came from trials that yielded successful subsequent recognition memory for the old/new status of word triplets and were matched for stimulus display and task demands during encoding, these regions appear to be contributing specifically to the encoding of temporal order memory rather than memory in general. Nonetheless, it is possible that these regions might be contributing to something like the quantity or strength of memory encoding rather than specifically to information related to temporal aspects of the event. In other words, it is possible that these encoding activations correlate with the amount of encoded information, with the difference between subsequently ordered and misordered trials lying in the amount, rather than kind, of remembered details (Staresina and Davachi 2008). To explore this possibility, we further examined activation in these regions during encoding trials for those triplets that were subsequently not recognized. In each of the hippocampal and parahippocampal regions identified in the ordered versus misordered contrast, the activation for subsequently missed trials did not differ from subsequently recognized and misordered trials. If the contributions of these regions to encoding was related more generally to the amount or strength of the resulting memory, we would expect encoding period activation for subsequently missed trials to fall below that for subsequently remembered and misordered trials, as the misordered trials, unlike the forgotten trials, were encoded with enough information to enable recognition but not order memory, whereas the ordered trials were encoded in a manner enabling recognition and order reconstruction. Instead, the observed pattern suggests that the differentiation between subsequently ordered and misordered trials in the identified hippocampal and parahippocampal regions reflects processing related to the encoding of temporal information. This interpretation is based on the observation that there were no differences between encoding activations in the misordered and forgotten differences in any of the MTL regions showing a subsequent order memory effect, and exploratory contrasts of misordered versus forgotten trials at a reduced threshold failed to identify any of the MTL regions in question. However, we should use caution when interpreting our null findings, particularly because triplet recognition, in the present paradigm, was high enough that we are working with relatively few trials in the subsequently forgotten condition. Although we attempted to ameliorate this issue by restricting the misordered versus forgotten analyses to the subset of subjects with higher number of forgotten trials, it still may be that we were unable to observe any effects due to a lack of power.

Hippocampal Contributions to Order Encoding

The results reported here are consistent with and extend the extant literature on hippocampal contributions to episodic encoding. It is well established that the hippocampus plays an important role in the binding of episodic associations (Davachi et al. 2003; Ranganath et al. 2003; Kirwan and Stark 2004; Uncapher et al. 2006; Diana et al. 2007; Johnson and Rugg 2007), and evidence for a role for the hippocampus in binding elements across time comes from a recent study showing that the successful binding of event details occurring discontiguously in time elicits elevated hippocampal activation relative to the successful binding of such elements when they are temporally overlapping (Staresina and Davachi 2009). Further evidence for a role for the hippocampus in temporal aspects of memory comes from lesion work in rats (Fortin et al. 2002; Kesner et al. 2002) and fMRI investigation of temporal order retrieval (Lehn et al. 2009). The novel hippocampal results reported here complement this body of work by showing that hippocampal processing during encoding contributes to binding of information across time in a manner supporting the subsequent ability to correctly reconstruct the order in which a sequence of episodic elements were previously encountered.

Parahippocampal Contributions to Order Encoding

Theoretical frameworks hypothesize that extra-hippocampal cortical regions might serve to buffer information in a manner enabling longer term associative encoding of discontiguous events by the hippocampus (Jensen and Lisman 2005). It has been proposed that oscillatory dynamics in such buffers enables the compression of sequences of events unfolding on supra-LTP timescales in a manner that preserves the relative temporal ordering across these events, and evidence for oscillatory coupling between hippocampus and surrounding cortex during successful encoding has been reported in humans (Fell et al. 2001). Further support for a role for MTL cortex in processing temporal aspects of events comes from a recent study showing that medial entorhinal cortex (MEC) neurons reliably distinguish the temporal context within which overlapping sequences are experienced (Lipton et al. 2007). This has been interpreted as evidence that information flowing into the hippocampus via the parahippocampal–MEC pathway (Witter et al. 2000) contains temporal information that may be used to link events when memory for the order of their occurrence has consequences for behavior (Eichenbaum and Lipton 2008). In humans, retrieval of temporal order information recruits both the hippocampus and parahippocampal cortex (Lehn et al. 2009), with engagement of the MTL showing sensitivity to the resolution of order memory required (St Jacques et al. 2008). St Jacques et al. (2008) reported increased engagement of the parahippocampal cortex during order discriminations between events occurring close, relative to further, in time. The results reported here complement these findings by showing activation in parahippocampal cortex that is consistent with a role for this region in the encoding of within-event memory information supporting order judgments that are unlikely to be based on relative memory strength for individual items.

Mechanisms of Order Memory

Prior investigations of the retrieval of temporal information have made a distinction between memory judgments that can be performed on the basis of a strength or familiarity signal and those that require or make use of information directly relating the to-be-discriminated items, perhaps through direct retrieval of temporal relational information or recollection of episodic details (Suzuki et al. 2002; Konishi et al. 2006; Dudukovic and Wagner 2007; St Jacques et al. 2008). Such studies have found increased MTL activation at the time of retrieval when participants make recency versus item recognition decisions (Dudukovic and Wagner 2007), when recency decisions are carried out between items encoded relationally and between those that are not (Konishi et al. 2006), and when order discriminations are between items experienced closely, as opposed to separated, in time (St Jacques et al. 2008). Together, these results indicate that the hippocampus and surrounding cortex play a preferential role in order memory retrieval when inter-item relational information is particularly salient or when a relative strength signal between 2 items cannot resolve differences in presentation order. The encoding results reported here complement these findings by showing that such differences are also present at the time of encoding. Such a finding might be expected, as engagement of MTL during retrieval of order information implies the attempt to retrieve information that was encoded in the first place, and influential theories of hippocampal function posit an overlap between the circuits engaged during encoding and those providing mnemonic information at the time of retrieval (Alvarez and Squire 1994; O'Reilly and Norman 2002). The present results, showing significant subsequent order memory effects in MTL, provide novel evidence that the differences previously observed during order retrieval tasks might reflect differences in the encoding of information upon which such decisions are made.

Although the task design in this study argues against an account of order memory performance based on relative item strength within triplets, as the 3 items to be reordered are presented within a 5-s window with the memory test occurring approximately 10 min later, it may still be the case that the observed MTL cortex activations correlating with order memory reflect processes other than the encoding of temporal order information per se. Another possibility is that order memory in this paradigm is dependent on the replay of an imagined scenario constructed during encoding. During encoding, subjects were encouraged to form a vivid mental image of the referents of the 3 nouns interacting in the order of presentation. Thus, it is possible that MTL activation during encoding reflected better imagery of the referents in order. This would be consistent with the role of the hippocampus and parahippocampal gyrus in scene imagery (Davachi et al. 2003). However, we do not think this can fully explain our findings because among recognized triplets, there was no performance benefit for trials reported as imaged in order during encoding compared with those reported as imaged but not in order. This implies that mechanisms other than “reading out” the original presentation order from an ordered scenario constructed during encoding account, at least in part, for the effects observed here. Consequently, although participants may have relied in part on a replay of the imagined scenario during performance of the order task, there seems to be other information upon which such memory decisions were made. Although the preceding discussion argues against an account of the current findings in terms of replay of participants’ imagined scenarios, we do not mean to imply that episodic order memory does not rely on other reconstructive or inferential processes. It may very well be the case that participants in the present study were in fact recollecting the original encoding episode and basing decisions on order information falling out of that recollection, whether through imagery or based on other aspects of the recollection.

Finally, it is unlikely that the memory task in this study is process pure (Jacoby 1991), drawing from a single kind of mnemonic information. Instead, it is likely that the memory decisions were based on whatever information was available and diagnostic at the time of retrieval. This may have included the application of familiarity, or strength, signals as well as more elaborative recollective processes. Although the behavioral results presented here argue against a strong contribution of familiarity as the recognition and accurate reordering rates were comparable for those test items whose left to right screen position matched their temporal order at encoding and those which did not, an important goal for future work will be specification of the kinds of information underlying episodic memory for order information.

Conclusion

In summary, the present work examined MTL contributions to episodic encoding of sequence information. The results reported here are consistent with a set of hypotheses regarding the role of the MTL in the maintenance and storage of episodic sequence memory (Wallenstein et al. 1998; Lisman 1999; Jensen and Lisman 2005) and extend knowledge of the hippocampal and parahippocampal cortical contributions to episodic memory formation. Although previous work has identified a role for hippocampus and parahippocampal cortex in the episodic encoding and retrieval of associations (Davachi et al. 2003; Aminoff et al. 2007), the pattern of results described in this study constitute the first report of a role for these regions in supporting the encoding of information enabling subsequent recovery of episodic sequence information.

Funding

National Institute of Mental Health R01 (grant MH-074692 to L.D.).

Acknowledgments

Conflict of Interest: None declared.

References

  1. Alvarez P, Squire LR. Memory consolidation and the medial temporal lobe: a simple network model. Proc Natl Acad Sci USA. 1994;91:7041–7045. doi: 10.1073/pnas.91.15.7041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aminoff E, Gronau N, Bar M. The parahippocampal cortex mediates spatial and nonspatial associations. Cereb Cortex. 2007;17:1493–1503. doi: 10.1093/cercor/bhl078. [DOI] [PubMed] [Google Scholar]
  3. Badre D, Wagner AD. Left ventrolateral prefrontal cortex and the cognitive control of memory. Neuropsychologia. 2007;45:2883–2901. doi: 10.1016/j.neuropsychologia.2007.06.015. [DOI] [PubMed] [Google Scholar]
  4. Clayton NS, Dickinson A. Episodic-like memory during cache recovery by scrub jays. Nature. 1998;395:272–274. doi: 10.1038/26216. [DOI] [PubMed] [Google Scholar]
  5. Cohen NJ, Eichenbaum H. Memory, amnesia, and the hippocampal system. Cambridge (MA): MIT Press; 1993. [Google Scholar]
  6. Dale AM. Optimal experimental design for event-related fMRI. Hum Brain Mapp. 1999;8:109–114. doi: 10.1002/(SICI)1097-0193(1999)8:2/3<109::AID-HBM7>3.0.CO;2-W. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Davachi L, Mitchell JP, Wagner AD. Multiple routes to memory: distinct medial temporal lobe processes build item and source memories. Proc Natl Acad Sci USA. 2003;100:2157–2162. doi: 10.1073/pnas.0337195100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Diana RA, Yonelinas AP, Ranganath C. Imaging recollection and familiarity in the medial temporal lobe: a three-component model. Trends Cogn Sci. 2007;11:379–386. doi: 10.1016/j.tics.2007.08.001. [DOI] [PubMed] [Google Scholar]
  9. Dudukovic NM, Wagner AD. Goal-dependent modulation of declarative memory: neural correlates of temporal recency decisions and novelty detection. Neuropsychologia. 2007;45:2608–2620. doi: 10.1016/j.neuropsychologia.2007.02.025. [DOI] [PubMed] [Google Scholar]
  10. Eichenbaum H, Lipton PA. Towards a functional organization of the medial temporal lobe memory system: role of the parahippocampal and medial entorhinal cortical areas. Hippocampus. 2008;18:1314–1324. doi: 10.1002/hipo.20500. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Ergorul C, Eichenbaum H. The hippocampus and memory for “what,” “where,” and “when”. Learn Mem. 2004;11:397–405. doi: 10.1101/lm.73304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fell J, Klaver P, Lehnertz K, Grunwald T, Schaller C, Elger CE, Fernandez G. Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nat Neurosci. 2001;4:1259–1264. doi: 10.1038/nn759. [DOI] [PubMed] [Google Scholar]
  13. Fortin NJ, Agster KL, Eichenbaum HB. Critical role of the hippocampus in memory for sequences of events. Nat Neurosci. 2002;5:458–462. doi: 10.1038/nn834. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Jacoby L. A process dissociation framework: separating automatic from intentional uses of memory. J Mem Lang. 1991;30:513–541. [Google Scholar]
  15. Jensen O, Lisman JE. Hippocampal sequence-encoding driven by a cortical multi-item working memory buffer. Trends Neurosci. 2005;28:67–72. doi: 10.1016/j.tins.2004.12.001. [DOI] [PubMed] [Google Scholar]
  16. Johnson JD, Rugg MD. Recollection and the reinstatement of encoding-related cortical activity. Cereb Cortex. 2007;17:2507–2515. doi: 10.1093/cercor/bhl156. [DOI] [PubMed] [Google Scholar]
  17. Kesner RP, Gilbert PE, Barua LA. The role of the hippocampus in memory for the temporal order of a sequence of odors. Behav Neurosci. 2002;116:286–290. doi: 10.1037//0735-7044.116.2.286. [DOI] [PubMed] [Google Scholar]
  18. Kirwan CB, Stark CE. Medial temporal lobe activation during encoding and retrieval of novel face-name pairs. Hippocampus. 2004;14:919–930. doi: 10.1002/hipo.20014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Konishi S, Asari T, Jimura K, Chikazoe J, Miyashita Y. Activation shift from medial to lateral temporal cortex associated with recency judgements following impoverished encoding. Cereb Cortex. 2006;16:469–474. doi: 10.1093/cercor/bhi126. [DOI] [PubMed] [Google Scholar]
  20. Kumaran D, Maguire EA. The dynamics of hippocampal activation during encoding of overlapping sequences. Neuron. 2006;49:617–629. doi: 10.1016/j.neuron.2005.12.024. [DOI] [PubMed] [Google Scholar]
  21. Lehn H, Steffenach HA, van Strien NM, Veltman DJ, Witter MP, Haberg AK. A specific role of the human hippocampus in recall of temporal sequences. J Neurosci. 2009;29:3475–3484. doi: 10.1523/JNEUROSCI.5370-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Levy WB, Steward O. Temporal contiguity requirements for long-term associative potentiation/depression in the hippocampus. Neuroscience. 1983;8:791–797. doi: 10.1016/0306-4522(83)90010-6. [DOI] [PubMed] [Google Scholar]
  23. Lipton PA, White JA, Eichenbaum H. Disambiguation of overlapping experiences by neurons in the medial entorhinal cortex. J Neurosci. 2007;27:5787–5795. doi: 10.1523/JNEUROSCI.1063-07.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lisman JE. Relating hippocampal circuitry to function: recall of memory sequences by reciprocal dentate-CA3 interactions. Neuron. 1999;22:233–242. doi: 10.1016/s0896-6273(00)81085-5. [DOI] [PubMed] [Google Scholar]
  25. Mangels JA. Strategic processing and memory for temporal order in patients with frontal lobe lesions. Neuropsychology. 1997;11:207–221. doi: 10.1037//0894-4105.11.2.207. [DOI] [PubMed] [Google Scholar]
  26. McAndrews MP, Milner B. The frontal cortex and memory for temporal order. Neuropsychologia. 1991;29:849–859. doi: 10.1016/0028-3932(91)90051-9. [DOI] [PubMed] [Google Scholar]
  27. Ninokura Y, Mushiake H, Tanji J. Representation of the temporal order of visual objects in the primate lateral prefrontal cortex. J Neurophysiol. 2003;89:2868–2873. doi: 10.1152/jn.00647.2002. [DOI] [PubMed] [Google Scholar]
  28. Ninokura Y, Mushiake H, Tanji J. Integration of temporal order and object information in the monkey lateral prefrontal cortex. J Neurophysiol. 2004;91:555–560. doi: 10.1152/jn.00694.2003. [DOI] [PubMed] [Google Scholar]
  29. O'Reilly RC, Norman KA. Hippocampal and neocortical contributions to memory: advances in the complementary learning systems framework. Trends Cogn Sci. 2002;6:505–510. doi: 10.1016/s1364-6613(02)02005-3. [DOI] [PubMed] [Google Scholar]
  30. Ranganath C, Johnson MK, D'Esposito M. Prefrontal activity associated with working memory and episodic long-term memory. Neuropsychologia. 2003;41:378–389. doi: 10.1016/s0028-3932(02)00169-0. [DOI] [PubMed] [Google Scholar]
  31. Ranganath C, Yonelinas AP, Cohen MX, Dy CJ, Tom SM, D'Esposito M. Dissociable correlates of recollection and familiarity within the medial temporal lobes. Neuropsychologia. 2004;42:2–13. doi: 10.1016/j.neuropsychologia.2003.07.006. [DOI] [PubMed] [Google Scholar]
  32. Schendan HE, Searl MM, Melrose RJ, Stern CE. An FMRI study of the role of the medial temporal lobe in implicit and explicit sequence learning. Neuron. 2003;37:1013–1025. doi: 10.1016/s0896-6273(03)00123-5. [DOI] [PubMed] [Google Scholar]
  33. Scoville WB, Milner B. Loss of recent memory after bilateral hippocampal lesions. J Neurol Neurosurg Psychiatry. 1957;20:11–21. doi: 10.1136/jnnp.20.1.11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Shimamura AP, Janowsky JS, Squire LR. Memory for the temporal order of events in patients with frontal lobe lesions and amnesic patients. Neuropsychologia. 1990;28:803–813. doi: 10.1016/0028-3932(90)90004-8. [DOI] [PubMed] [Google Scholar]
  35. Shrager Y, Kirwan CB, Squire LR. Activity in both hippocampus and perirhinal cortex predicts the memory strength of subsequently remembered information. Neuron. 2008;59:547–553. doi: 10.1016/j.neuron.2008.07.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Squire LR, Stark CE, Clark RE. The medial temporal lobe. Annu Rev Neurosci. 2004;27:279–306. doi: 10.1146/annurev.neuro.27.070203.144130. [DOI] [PubMed] [Google Scholar]
  37. St Jacques P, Rubin DC, LaBar KS, Cabeza R. The short and long of it: neural correlates of temporal-order memory for autobiographical events. J Cogn Neurosci. 2008;20:1327–1341. doi: 10.1162/jocn.2008.20091. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Staresina BP, Davachi L. Selective and shared contributions of the hippocampus and perirhinal cortex to episodic item and associative encoding. J Cogn Neurosci. 2008;20:1478–1489. doi: 10.1162/jocn.2008.20104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Staresina BP, Davachi L. Mind the gap: binding experiences across space and time in the human hippocampus. Neuron. 2009;63:267–276. doi: 10.1016/j.neuron.2009.06.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Suzuki M, Fujii T, Tsukiura T, Okuda J, Umetsu A, Nagasaka T, Mugikura S, Yanagawa I, Takahashi S, Yamadori A. Neural basis of temporal context memory: a functional MRI study. Neuroimage. 2002;17:1790–1796. doi: 10.1006/nimg.2002.1303. [DOI] [PubMed] [Google Scholar]
  41. Thompson-Schill SL, D'Esposito M, Aguirre GK, Farah MJ. Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation. Proc Natl Acad Sci USA. 1997;94:14792–14797. doi: 10.1073/pnas.94.26.14792. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Tulving E. Episodic memory: from mind to brain. Annu Rev Psychol. 2002;53:1–25. doi: 10.1146/annurev.psych.53.100901.135114. [DOI] [PubMed] [Google Scholar]
  43. Uncapher MR, Otten LJ, Rugg MD. Episodic encoding is more than the sum of its parts: an fMRI investigation of multifeatural contextual encoding. Neuron. 2006;52:547–556. doi: 10.1016/j.neuron.2006.08.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Wallenstein GV, Eichenbaum H, Hasselmo ME. The hippocampus as an associator of discontiguous events. Trends Neurosci. 1998;21:317–323. doi: 10.1016/s0166-2236(97)01220-4. [DOI] [PubMed] [Google Scholar]
  45. Witter MP, Naber PA, van Haeften T, Machielsen WC, Rombouts SA, Barkhof F, Scheltens P, Lopes da Silva FH. Cortico-hippocampal communication by way of parallel parahippocampal-subicular pathways. Hippocampus. 2000;10:398–410. doi: 10.1002/1098-1063(2000)10:4<398::AID-HIPO6>3.0.CO;2-K. [DOI] [PubMed] [Google Scholar]

Articles from Cerebral Cortex (New York, NY) are provided here courtesy of Oxford University Press

RESOURCES