Abstract

Results of recent functional magnetic resonance imaging (fMRI) studies of memory are not entirely consistent with lesion studies. Furthermore, although imaging probes have identified neural systems associated with processing novel visual episodic information, auditory verbal memory using a novel/familiar paradigm has not yet been examined. To address this gap, fMRI was used to compare the haemodynamic response when listening to recently learned and novel words. Sixteen healthy adults (6 male, 10 female) learned a 10-item word list to 100% criterion, ~1 h before functional scanning. During echo-planar imaging, subjects passively listened to a string of words presented at 6-s intervals. Previously learned words were interspersed pseudo-randomly between novel words. Mean scans corresponding to each word type were analysed with a random-effects model using statistical parametric mapping (SPM96). Familiar (learned) words activated the right prefrontal cortex, posterior left parahippocampal gyrus, left medial parietal cortex and right superior temporal gyrus. Novel words activated the anterior left hippocampal region. The results for the familiar words were similar to those found in other functional imaging studies of recognition and retrieval and implicate the right dorsolateral prefrontal and left posterior medial temporal lobe (MTL) regions. The results for novel words require replication, but are consistent with the substantial lesion and PET literature implicating the anterior MTL as a critical site for processing novel episodic information, presumably to permit encoding. Together, these results provide evidence for an anterior–posterior functional differentiation within the MTL in processing novel and familiar verbal information. The differentiation of MTL functions that was obtained is consistent with a large body of PET activation studies but is unique among fMRI studies, which to date have differed from results with PET. Further, the finding of left MTL lateralization is consistent with lesion-based material-specific models of memory.

Introduction

Functional MRI (fMRI) is well suited for examining the neuroanatomical correlates of human memory and for testing hypotheses about brain regions involved in selective memory processes. Recent blood oxygen level-dependent (BOLD) contrast studies have generally confirmed at a lobar level observations based on lesion studies (Corkin et al., 1997) and PET methods (Squire et al., 1992; Fletcher et al., 1998a, b; Lapage et al., 1998), implicating the frontal lobes and medial temporal/hippocampal system as important substrates of encoding and retrieval (Stern et al., 1996; Gabrieli et al., 1997; Rombouts et al., 1997; Bellgowan et al., 1998; Buckner et al., 1998a, b; Fernandez et al., 1998). However, there has been little attention to functional differentiation within these larger-scale systems, such as selectivity of processing within the anterior versus posterior hippocampal formation. Nonetheless, data using several methodologies have suggested the likelihood of such regional specialization.

The most notable lesion study is the case of H.M., who was severely impaired in encoding new material with relative sparing of retrieval (Scoville and Milner, 1957). Recent MRI studies of H.M. demonstrated that his bilateral resection involved the anterior half of the hippocampus with preservation of the posterior hippocampus and parahippocampal gyrus (Corkin et al., 1997). Similar deficits in encoding were observed in other cases of unilateral hippocampal resection (Milner, 1975). Early reports from Penfield and Mathieson (1974) suggested that encoding for new information occurs in the rostral hippocampal formation (anterior), while remote memories are processed in the caudal portion (posterior). This work has been complemented by the finding of Halgren and colleagues (Halgren et al., 1985) that electrical stimulation of depth electrodes placed in the long axis of the hippocampus in patients with refractory epilepsy disrupts encoding with anterior stimulation and disrupts word retrieval with posterior stimulation. Further, Phelps and colleagues (Phelps et al., 1991) reported retrieval deficits following posterior but not anterior callosal section, which were thought to be due to relative posterior hippocampal formation involvement.

A recent meta-analysis of PET studies involving memory encoding and retrieval and medial temporal activation (Lapage et al., 1998) showed an anatomical dissociation within the hippocampus between encoding and retrieval entirely consistent with the lesion and electrophysiological evidence. Fifty-four activations were found out of 52 studies. Among encoding activations, 17 out of 22 showed anterior medial temporal lobe (MTL) activation (anterior to the posterior commissure), whereas for retrieval 29 out of 32 activations were posterior to the posterior commissure. A reanalysis of the PET literature by Schacter and Wagner (1999) with different study inclusion criteria reached a somewhat different result. For encoding activations in the MTL, 58% (26 out of 45) were anterior, whereas 42% (19 out of 45) were posterior. Thus, the rostrocaudal hippocampal encoding gradient is less clearly defined using the Schacter and Wagner criteria than was originally reported by Lapage and colleagues (Lapage et al., 1998). Both meta-analyses demonstrated that the vast majority of retrieval activations are posterior.

In contrast to the lesion and PET studies, recent fMRI reports have primarily detected activation associated with encoding in the posterior MTL for both verbal (Rombouts et al., 1997; Fernandez et al., 1998; Kelley et al., 1998; Wagner et al., 1998b) and visual–spatial information (e.g. Stern et al., 1996; Gabrieli et al., 1997; Bellgowan et al., 1998). Stern and colleagues (Stern et al., 1996) scanned eight healthy subjects while presenting novel scenes alternating with a single repeating control scene. The posterior parahippocampal area was activated bilaterally, right more than left. In a similar paradigm, Gabrieli and colleagues (Gabrieli et al., 1997) found similar posterior activation in the hippocampal region during presentation of novel scenes to six subjects. In a separate experiment, Gabrieli and colleagues (Gabrieli et al., 1997) also studied recognition of previously viewed line drawings and found anterior hippocampal activation. Other fMRI studies of encoding (Rombouts et al., 1997; Bellgowan et al., 1998; Brewer et al., 1998; Kelley et al., 1998; Wagner et al., 1998b) also primarily report posterior MTL activation (usually parahippocampal and fusiform gyri). Importantly, there are three fMRI studies of retrieval that are consistent with the PET data showing posterior MTL activation for retrieval of previously studied words (Schacter et al., 1997) and of topographical visual stimuli (Aguirre et al., 1996; Aguirre and D'Esposito, 1997).

In view of the differences between recent fMRI reports and prior lesion and PET studies, and because verbal memory processes have been under-represented in recent fMRI studies, we designed an event-related verbal memory task targeting the temporal–hippocampal system. It has recently been demonstrated in several reports that the excellent temporal resolution of fMRI permits examination of the BOLD response to intermixed classes of individual stimuli or events (Buckner et al., 1996a, 1998; Zarahn et al., 1997; Rosen et al., 1998). Indeed, several recent studies have successfully examined human memory processes with event-related fMRI paradigms (Schacter et al., 1997; Buckner et al., 1998a; Friston et al., 1998). Rather than assuming a steady state, such as is required in `blocked' or `box-car' designs, event-related fMRI allows the analysis of activation patterns in response to discrete events within a scanning run. The goal of this study was to examine the activation patterns in response to two event classes (novel and familiar words) obtained from the same experimental trial. We designed this event-related memory task after the P300 oddball event-related potential paradigm, which shows novelty-related changes in the hippocampus (Knight, 1996; Knight and Nakada, 1998) without requiring executive/motor responses. We hypothesized that by grouping fMRI scans by event class we would be able to detect differential responses to novel and familiar words in medial temporal and prefrontal regions. Based on the lesion and PET literature we further hypothesized that novel words would activate the anterior MTL region associated with encoding processes, and familiar words would activate the right prefrontal and medial temporal regions associated with the hemispheric encoding/retrieval asymmetry model (Buckner et al., 1996b; Tulving et al., 1996; Lapage et al., 1998).

Methods

Participants

Sixteen (6 male, 10 female) healthy adult volunteers participated (mean age = 31 ± 11 years; mean education = 14.6 ± 2.6 years). Subjects were administered a handedness examination and those with left or mixed dominance were excluded. Personal and familial medical history and symptoms, including neurological, psychiatric and substance abuse areas, were reviewed using a semi-structured interview protocol. All subjects reporting past or present symptoms of a major psychiatric (axis I) disorder or neurological disorder (including head injury with loss of consciousness for >5 min) were excluded. Structural MRI scans were acquired for all subjects and reviewed by a board-certified neuroradiologist (A.M.) blind to clinical status (intermixed with other clinical scans). Potential subjects who were taking psychoactive or vascular-related medications were excluded. Thirteen subjects underwent a brief cognitive evaluation, and these results, along with other demographics, are presented in Table 1. All subjects gave written informed consent under an Institutional Review Board-approved protocol.

Activation task

During scanning, subjects were asked to listen carefully to a 48-item word list of concrete nouns presented at a rate of one word every 6 s. Ten of the words had been memorized 1 h before scanning (familiar words); the remaining 38 words were novel. The familiar words were pseudo-randomly positioned throughout the task and these were always separated by at least one novel word. Old and new words were equated on imagery and frequency values in the English language (Francis and Kucera, 1982). Subjects were tested for free recall of the word list immediately prior to scanning and all were able to recite the 10 words.

Scan procedure

Scans were acquired using a 1.5 T GE Signa scanner and a multi-axial local gradient head coil system (Medical Advances, Inc., Milwaukee, Wis., USA). A single-shot gradient echo, echo-planar sequence was used to provide whole-brain coverage [TR (repetition time) = 3000 ms; TE (echo time) = 40 ms; field of view = 24 cm (20 or 23 sagittal slices 6 mm thick)], yielding a 64 × 64 matrix with 3.75 mm in-plane resolution. Prior to scanning, linear shims were optimized. A time series of 98 T2*-weighted volumes was acquired during the task.

Overall statistical approach

Functional MRI analyses were performed with statistical parametric mapping, on a voxel-by-voxel basis, using a general linear model approach (Friston et al., 1995; Worsley and Friston, 1995; Worsley et al., 1996) as implemented in SPM96. The task was initially analysed for each subject as an individual time series before inclusion in multi-subject analyses. Theoretical issues and practical implementation of multi-subject analyses of fMRI data have been the subject of much recent work and debate. The main analyses reported here used the Random Effects procedure recently developed by Holmes and Friston (1998) as described below. The principal advantage of this method is the elimination of highly discrepant variances between and within subjects in constructing an appropriate error term for hypothesis-testing to permit generalizability to the population.

Preprocessing steps

All scans were cropped to eliminate most non-brain voxels. Spatial realignment using the SPM96 six-parameter model was performed on all raw scan data prior to further analysis to remove any minor (subvoxel) motion-related signal change.

Random effects procedure

For the multi-subject group analyses, the procedure of Holmes and Friston (Holmes and Friston, 1998) assumes input of one scan per subject for each condition and then performs a mixed-model analysis to account for both random effects (scan) and fixed effects (task condition). Scans for novel and familiar trials were segregated prior to analysis. The scans representing the familiar words were matched with an equal number of randomly selected scans from the novel condition and these scans were converted to mean condition images. The mean input images for the random effects analysis for each subject were obtained by calculating the mean image for the familiar and novel word presentations. Data analysis adjusted for the haemodynamic response function (HRF) by offsetting the design vector by 1 TR (i.e. 4.5-s lag on average). The 4.5-s average offset was designed to capture the leading edge of the HRF with minimal sensitivity to venous influences. A general linear model analysis was performed on a voxel-by-voxel basis with two contrasts of interest identifying voxels with higher activation during the familiar condition and novel condition.

Spatial normalization and smoothing

Prior to multi-subject analyses, mean images were spatially normalized to the Montreal standardized atlas space using a 12-parameter affine approach and a T2*-weighted template image. The optional use of non-linear warping by spatial basis functions was limited to 2 × 2 × 2 and eight iterations. During normalization all scans were resampled to isotropic voxels (2 mm3). Spatial smoothing to a full width half maximum (FWHM) of 15 mm3 was then performed to help ensure the validity of analysis across subjects. To reduce the possibility that highly localized `sharp' foci in hypothesized regions might be attenuated by spatial smoothing, analyses were repeated at 8 mm3 FWHM.

Probability thresholds

For a priori hypothesis testing, critical probability thresholds were uncorrected and set at P ≤ 0.05 for assessing predefined search regions such as the MTLs and prefrontal cortex. Otherwise the height threshold value was set to P ≤ 0.01 with an extent threshold of 70 voxels for exploration of additional regions. In view of our neuroanatomically constrained hypotheses regarding expected MTL and frontal regions of activation based on prior functional imaging, electrophysiological and lesion studies, a multiple comparison correction strategy designed for exploratory searches of the entire brain volume would have been overly conservative.

Results

Results of multi-subject analysis of familiar words (Fig. 1) indicated significant activations for word recognition in the two principal hypothesized regions: the posterior parahippocampal gyrus (x, y, z; –24, –46, –8; Z = 2.94, P = 0.002) and the right prefrontal cortex (54, 14, 32; Z = 3.49, P < 0.0001). Other significant areas of activation shown in Fig. 1 included the left posterior parietal cortex, right superior temporal gyrus and anterior right cingulate.

It was predicted that novel words would activate the anterior hippocampal region but not the prefrontal cortex. Results for the novel words (Fig. 2) showed activation in the left anterior hippocampal region (–20, –24, –14; Z = 2.15, P = 0.016) and this activation was anterior to the MTL cluster activated for familiar words. The locations of peak MTL activations for each condition are presented on a standardized atlas brain in Fig. 3.

The activation by event-type difference within the left MTL region is illustrated in Fig. 4. The region × condition interaction indicated a dissociable functional difference within the left MTL for novel versus familiar words where novel words were associated with relative anterior activation (12 out of 16 subjects) and familiar words with posterior MTL activation (13 out of 16 subjects). This effect was fairly consistent in that ≥75% of subjects showed the dissociation. Individual time series analyses for scans of the three or four subjects not showing the effect observed for the group as a whole did not reveal any consistent secondary pattern.

Discussion

These results demonstrate separable neural systems for processing novel and familiar words. We observed selective lateralized MTL activation to the left hemisphere for both novel and familiar stimuli. Within the left MTL, an anterior–posterior dissociation was found; novel words activated an anterior area in the left hippocampus and familiar words activated an area in the left posterior parahippocampal region. Additionally, the right dorsolateral prefrontal, right superior frontal and left medial parietal regions showed activation to the familiar words. These results show greater left MTL involvement for auditory–verbal word presentation in contrast to right or bilateral activation for visual stimuli seen in prior studies (Stern et al., 1996; Aguirre and D'Esposito, 1997; Gabrieli et al., 1997; Bellgowan et al., 1998; Brewer et al., 1998; Kelley et al., 1998).

Medial temporal lobe

This study demonstrates that MTL activation can be evoked using auditory–verbal stimuli with BOLD contrast fMRI. We were able to detect predicted differential MTL activation for novel versus familiar word presentations in the same experimental scanning run. This provides new support for models positing dissociable circuitry subserving verbal encoding and recognition processes within the left MTL. While no fMRI studies have looked at novelty/familiarity differences for auditory verbal material, Dolan and Fletcher (1997) studied auditory–verbal novel versus familiar word pairs using PET and found increased anterior left hippocampal and parahippocampal gyrus activation to novel word pairs in six subjects.

Medial temporal lobe activation during memory tasks has been observed inconsistently in functional imaging studies. The aggregate MTL PET findings demonstrate a dissociation along the anterior–posterior direction of the MTL similar to that reported here. However, recent fMRI studies using novel compared with familiar scenes have shown differing results. Although Gabrieli and colleagues (Gabrieli et al., 1997) found dissociable activation for encoding and recognition, the pattern was the reverse of our data and most PET studies. In that study, novel information (encoding) activated the posterior parahippocampal gyrus while the recognition condition activated the hippocampal formation more anteriorly. However, there are important methodological differences between the study of Gabrieli and colleagues (Gabrieli et al., 1997) and ours that might account for this difference. The current study used whole-brain echo-planar imaging with sagittal acquisition, and novel and familiar verbal stimuli were intermixed pseudo-randomly within the same scanning trial in an event-related design. In contrast, Gabrieli and colleagues (Gabrieli et al., 1997) imaged in an oblique coronal plane through the hippocampus, they used differing stimulus types to study each memory process (photographs of scenes for the novelty/encoding trial and line drawings for the recognition trial) and they presented stimuli visually in blocked epochs. Our findings, using verbal stimuli, suggest there may be important differences in medial temporal activation patterns for novel and familiar verbal and visual stimuli. Further, our findings support the accumulating PET and prior lesion data suggesting that the anterior hippocampus subserves the formation of new memories for verbal stimuli.

In contrast to PET studies, many fMRI studies of memory have failed to detect MTL activation, and when MTL activation is reported it has most often involved the posterior region. The explanation for this may be methodological. Using echoplanar imaging, signal loss within the medial temporal region is a common susceptibility artefact due to different magnetic susceptibilities of the brain and surrounding bone and air sinuses (Ojemann et al., 1997). This artefact is often overlooked when statistical maps obtained from echo-planar images are displayed on high-resolution anatomical images, as is customary in fMRI studies. It is also possible that task differences in the PET and fMRI studies have led to differing results (see discussion by Schacter and Wagner, 1999). Some studies implicate posterior MTL in novelty detection (Elliott and Dolan, 1998). This may account for the posterior MTL activation observed using fMRI but is not consistent with our findings. Another possibility is that the use of complex visual scenes may recruit posterior temporal regions involved in visual processing (Ungerleider and Haxby, 1994); this seems less likely in that many PET studies have also used visual stimuli. These issues are likely to become clearer as the body of literature accumulates and as the signal-to-noise ratio of fMRI within the anterior MTL improves.

Right prefrontal cortex

The right dorsolateral prefrontal cortex was activated for familiar but not novel words in our study. While right prefrontal activation during retrieval paradigms has been widely observed in functional neuroimaging studies, our findings regarding word recognition are particularly interesting in view of recent theoretical interpretations of right prefrontal activation. Several authors have suggested that the right prefrontal cortex is involved in the process of retrieval mode or retrieval attempts (Kapur et al., 1995; Nyberg et al., 1995; Cabeza et al., 1997) or retrieval context (Wagner et al., 1998a). This essentially reflects a cognitive `set' in which incoming information is specifically monitored with reference to previously learned information. Fletcher and colleagues (Fletcher et al., 1998b) also emphasized the role of the prefrontal cortex in retrieval as an executive process. In that study, right dorsolateral prefrontal activation during retrieval was evident for an internally controlled retrieval process and the ventral lateral prefrontal cortex was more activated when the experimenter guided retrieval with external cues. Further, in a blocked design fMRI experiment of visually presented words with two levels of encoding (shallow and deep), Buckner and colleagues (Buckner et al., 1998b) demonstrated greater right prefrontal activation when words were deeply encoded; those results were attributed to subject-initiated monitoring strategies. The task we used differed from those in most other studies because it involved passive rather than active participation during the word presentation specifically in order to minimize executive and motor task components. Nonetheless, subjects showed peak activation in the right dorsolateral prefrontal cortex to familiar words, consistent with other studies that did require responses to stimuli. This raised a question as to whether our subjects also engaged in monitoring or decision strategies despite the fact that they were not given an explicit instructional set to monitor the word list for prior learned words, and were not asked to make any overt response. When subjects were debriefed after the experiment, most did in fact report self-initiated monitoring strategies. Therefore, our data are consistent with the hypothesis that the right prefrontal cortex is important for the strategic monitoring of memory stimuli (Kapur et al., 1995; Fletcher et al., 1998b).

Limitations

Tests of neuroanatomical hypotheses in predefined regions of interest using P values not corrected for brain volume have a potentially significant risk of yielding false positive results (i.e. type 1 statistical error). The recently developed threshold adjustment algorithm for small volumes (Worsley et al., 1996) was employed in a supplementary analysis. All reported activations except the anterior hippocampal activation in the novel > familiar contrast exceeded the adjusted threshold criteria. However, the latter activation was predicted and the adjustment may be overly conservative when used in a random effects design, leading to a type 2 error. fMRI usually involves relatively small changes in signal between conditions that are normally offset by a high number of degrees of freedom when analysed as a time series. However, in a random-effects analysis, perhaps the most valid approach for multi-subject fMRI, degrees of freedom are dramatically reduced. Given these factors, a random effects analysis of fMRI data, while having the advantage of greater generalizability, will be less sensitive to signal changes between conditions than a time series analysis (Friston et al., 1999). Additionally, our mean condition images were selectively averaged, having only 10 time points for each event type, further reducing the available signal for these contrasts. Nonetheless, Fig. 2 clearly demonstrates the hypothesized anterior left MTL activation. In fact, it is the only activation in any region usually associated with the memory system and it does not appear to be due to noise in general or artefact in the MTL region. The fact that this activation cluster was consistent with our a priori hypothesis and was observed in near isolation increases our confidence in its validity. Clearly, replication of these findings will be important. Another limitation is that our task design used a constant offset between the stimuli and scans. Thus, we were unable to assess regional differences in the onset and duration of the haemodynamic response or assess the effect of possible differential onset on the magnitude of the response. Future studies should take advantage of recent advances in image processing software, such as the SPM event-related kit, that permits non-constant stimulus–scan offsets and the randomization of interstimulus intervals. For individual time series analysis, such an approach has the advantage of being able to model the onset, amplitude and dispersion of the HRF on a voxel-by-voxel basis. A remaining challenge is the incorporation of random effects models in this type of temporal processing so that group data can be appropriately analysed.

Conclusions

Our results lend support to functional neuroanatomical hypotheses regarding differential anterior/posterior MTL and left/right frontal system involvement in encoding/retrieval comparisons and material-specific memory lateralization (i.e. left MTL for verbal material). MTL activation using auditory stimuli was demonstrated, confirming that our posterior MTL retrieval results are not due to visual stimulation as used in most fMRI studies. This initial experiment did not separate encoding and retrieval processes from novelty/familiarity detection, processes that are often linked. However, our findings are consistent with lesion, PET and electrophysiological studies of encoding/retrieval and novelty/familiarity. It is noteworthy that our passive listening task was sufficient to evoke regionally specific memory changes using BOLD contrast, consistent with previous studies that actively manipulated these task dimensions.

Table 1

Descriptive demographic and cognitive data

 Mean SD Range 
WAIS = Wechsler Adult Intelligence Scale; CVLT = California Verbal Learning Test. *Psychometric measurements were available only on 13 subjects. 
Age 30.9 11.3 19–53 
Education 14.6  2.6 12–20 
Vocabulary WAIS-IIIa* 13.0  3.2  8–17 
Matrix Reasoning WAIS-III (n = 13) 12.2  2.1  8–15 
CVLT Acquisition (n = 13) 59.1  7.6 42–70 
CVLT Recognition Discrimination (n = 13) 96.2  8.9 89–100 
 Mean SD Range 
WAIS = Wechsler Adult Intelligence Scale; CVLT = California Verbal Learning Test. *Psychometric measurements were available only on 13 subjects. 
Age 30.9 11.3 19–53 
Education 14.6  2.6 12–20 
Vocabulary WAIS-IIIa* 13.0  3.2  8–17 
Matrix Reasoning WAIS-III (n = 13) 12.2  2.1  8–15 
CVLT Acquisition (n = 13) 59.1  7.6 42–70 
CVLT Recognition Discrimination (n = 13) 96.2  8.9 89–100 
Fig. 1

Orthogonal maximum intensity projections of SPM for familiar words showing peak clusters in the left posterior parahippocampal gyrus (P = 0.002; x, y, z, –24, –46, –8), right dorsolateral prefrontal cortex (P < 0.001; 54, 14, 32), right superior temporal gyrus (P < 0.001; 56, –24, 14), right anterior cingulate/medial superior frontal gyrus (P = 0.001; 18, 44, 30), right mid-cingulate (P = 0.003; 6, –6, 50) and left medial posterior parietal lobe (P < 0.001; –14, –76, 44). Height threshold P = 0.01, extent threshold = 70 voxels.

Fig. 1

Orthogonal maximum intensity projections of SPM for familiar words showing peak clusters in the left posterior parahippocampal gyrus (P = 0.002; x, y, z, –24, –46, –8), right dorsolateral prefrontal cortex (P < 0.001; 54, 14, 32), right superior temporal gyrus (P < 0.001; 56, –24, 14), right anterior cingulate/medial superior frontal gyrus (P = 0.001; 18, 44, 30), right mid-cingulate (P = 0.003; 6, –6, 50) and left medial posterior parietal lobe (P < 0.001; –14, –76, 44). Height threshold P = 0.01, extent threshold = 70 voxels.

Fig. 2

SPM maximum intensity projection showing areas of activation for novel words. Peak clusters are seen in the left hippocampal formation (P = 0.016; x, y, z, –20, –24, –14) and inferior posterior regions.

Fig. 2

SPM maximum intensity projection showing areas of activation for novel words. Peak clusters are seen in the left hippocampal formation (P = 0.016; x, y, z, –20, –24, –14) and inferior posterior regions.

Fig. 3

Orthogonal views of hypothesized regions of activation in the left hippocampus/parahippocampal gyrus for the familiar (top) and novel (bottom) word types. Both conditions activated medial temporal/hippocampal regions; the familiar words activated posterior MTL while novel words activated a more anterior hippocampal region. A liberal statistical threshold (height threshold P = 0.05, no extent threshold, uncorrected) was employed to enhance visualization of activation.

Fig. 3

Orthogonal views of hypothesized regions of activation in the left hippocampus/parahippocampal gyrus for the familiar (top) and novel (bottom) word types. Both conditions activated medial temporal/hippocampal regions; the familiar words activated posterior MTL while novel words activated a more anterior hippocampal region. A liberal statistical threshold (height threshold P = 0.05, no extent threshold, uncorrected) was employed to enhance visualization of activation.

Fig. 4

Signal difference (familiar–novel) in left anterior and posterior medial temporal lobe regions. The anterior region in the body of the left hippocampus (–20, –24, –14) was more activated to novel words (12 out of 16 subjects). The posterior region in the parahippocampal gyrus (–24, –46, –8) showed activation to familiar words (13 out of 16 subjects).

Fig. 4

Signal difference (familiar–novel) in left anterior and posterior medial temporal lobe regions. The anterior region in the body of the left hippocampus (–20, –24, –14) was more activated to novel words (12 out of 16 subjects). The posterior region in the parahippocampal gyrus (–24, –46, –8) showed activation to familiar words (13 out of 16 subjects).

The authors wish to thank Judy R. O'Jile, Cheryl Brown, Heather A. Wishart, Leslie C. Baxter, Jennifer D. Schoenfeld, James Ford, Charles B. Owen and Fillia Makedon for their assistance and Anthony Wagner for providing helpful information and comments. We also wish to thank Karl Friston, Andrew Holmes and Matthew Brett for helpful advice and for the software needed for statistical image processing. Support for this research was provided by the Ira DeCamp Foundation, the National Institute of Disability and Rehabilitation Research (H-133670031), the Alzheimer's Association (IIRG 94–133), the National Institutes of Health (NS-10563) and New Hampshire Hospital.

References

Aguirre GK, D'Esposito M. Environmental knowledge is subserved by separable dorsal/ventral neural areas.
J Neurosci
 
1997
;
17
:
2512
–8.
Aguirre GK, Detre JA, Alsop DC, D'Esposito M. The parahippocampus subserves topographical learning in man.
Cereb Cortex
 
1996
;
6
:
823
–9.
Bellgowan PSF, Binder JR, Hammeke TA, Frost JA, Possing ET. Configural associations and novelty are necessary for hippocampal-dependent encoding.
Neuroimage
 
1998
;
7 (4 Pt 2)
:
S51
.
Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD. Making memories: brain activity that predicts how well visual experience will be remembered [see comments].
Science
 
1998
;
281
:
1185
–7. Comment in: Science 1998; 281: 1151–2.
Buckner RL. Event-related fMRI and the hemodynamic response.
Hum Brain Mapp
 
1998
;
6
:
373
–7.
Buckner RL, Bandettini PA, O'Craven KM, Savoy RL, Petersen SE, Raichle ME, et al. Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging [see comments].
Proc Natl Acad Sci USA
 
1996
;
93
:
14878
–83. Comment in: Proc Natl Acad Sci USA 1996; 93: 14302–3.
Buckner RL, Raichle ME, Miezin FM, Petersen SE. Functional anatomic studies of memory retrieval for auditory words and visual pictures.
J Neurosci
 
1996
;
16
:
6219
–35.
Buckner RL, Koutstaal W, Schacter DL, Dale AM, Rotte M, Rosen BR. Functional-anatomic study of episodic retrieval. II. Selective averaging of event-related fMRI trials to test the retrieval success hypothesis.
Neuroimage
 
1998
;
7
:
163
–75.
Buckner RL, Koutstaal W, Schacter DL, Wagner AD, Rosen BR. Functional-anatomic study of episodic retrieval using fMRI. I. Retrieval effort versus retrieval success.
Neuroimage
 
1998
;
7
:
151
–62.
Corkin S, Amaral DG, Gonzalez RG, Johnson KA, Hyman BT. H.M.'s medial temporal lobe lesion: findings from magnetic resonance imaging.
J Neurosci
 
1997
;
17
:
3964
–79.
Cabeza R, Kapur S, Craik FIM, McIntosh AR, Houle S, Tulving E. Functional neuroanatomy of recall and recognition: PET study of episodic memory.
J Cogn Neurosci
 
1997
;
9
:
254
–65.
Dolan RJ, Fletcher PC. Dissociating prefrontal and hippocampal function in episodic memory encoding.
Nature
 
1997
;
388
:
582
–5.
Elliott R, Dolan RJ. Neural response during preference and memory judgments for subliminally presented stimuli: a functional neuroimaging study.
J Neurosci
 
1998
;
18
:
4697
–704.
Fernandez G, Weyerts H, Schrader-Bolsche M, Tendolkar I, Smid HG, Tempelmann C, et al. Successful verbal encoding into episodic memory engages the posterior hippocampus: a parametrically analyzed functional magnetic resonance imaging study.
J Neurosci
 
1998
;
18
:
1841
–7.
Fletcher PC, Shallice T, Dolan RJ. The functional roles of prefrontal cortex in episodic memory. I. Encoding.
Brain
 
1998
;
121
:
1239
–48.
Fletcher PC, Shallice T, Frith CD, Frackowiak RS, Dolan RJ. The functional roles of prefrontal cortex in episodic memory. II. Retrieval.
Brain
 
1998
;
121
:
1249
–56.
Francis WN, Kucera H. Frequency analysis of English usage: lexicon and grammar. Boston: Houghton Mifflin; 1982.
Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: a general linear approach.
Hum Brain Mapp
 
1995
;
2
:
189
–210.
Friston KJ, Fletcher P, Josephs O, Holmes A, Rugg MD, Turner R. Event-related fMRI: characterizing differential responses.
Neuroimage
 
1998
;
7
:
30
–40.
Friston KJ, Holmes AP, Price CJ, Buchel C, Worsley KJ. Multi-subject fMRI studies and conjunction analyses. Neuroimage. In press 1999.
Gabrieli JDE, Brewer JB, Desmond JE, Glover GH. Separate neural bases of two fundamental memory processes in the human medial temporal lobe.
Science
 
1997
;
276
:
264
–6.
Halgren E, Wilson CL, Stapleton JM. Human medial temporal lobe stimulation disrupts both formation and retrieval of recent memories.
Brain Cognition
 
1985
;
4
:
287
–95.
Holmes AP, Friston KJ. Generalisability, random effects and population inference.
Neuroimage
 
1998
;
7 (4 Pt 2)
:
S754
.
Kapur S, Craik FI, Jones C, Brown GM, Houle S, Tulving E. Functional role of the prefrontal cortex in retrieval of memories: a PET study.
Neuroreport
 
1995
;
6
:
1880
–4.
Kelley WM, Miezin FM, McDermott KB, Buckner R, Raichle ME, Cohen NJ, et al. hemispheric specialization in human dorsal frontal cortex and medial temporal lobe for verbal and nonverbal memory encoding.
Neuron
 
1998
;
20
:
927
–36.
Knight R. Contribution of human hippocampal region to novelty detection.
Nature
 
1996
;
383
:
256
–9.
Knight RT, Nakada T. Cortico-limbic circuits and novelty: a review of EEG and blood flow data. [Review].
Rev Neurosci
 
1998
;
9
:
57
–70.
Lapage M, Habib R, Tulving E. Hippocampal PET activations of memory encoding and retrieval: the HIPER model.
Hippocampus
 
1998
;
8
:
313
–22.
Milner B. Psychological aspects of focal epilepsy and its neurosurgical management.
Adv Neurol
 
1975
;
8
:
299
–321.
Nyberg L, Tulving E, Habib R, Nilsson LG, Kapur S, Houle S, et al. Functional brain maps of retrieval mode and recovery of episodic information.
Neuroreport
 
1995
;
7
:
249
–52.
Ojemann JG, Akbudak E, Snyder AZ, McKinstry RC, Raichle ME, Conturo TE. Anatomic localization and quantitative analysis of gradient refocused echo-planar fMRI susceptibility artifacts. [Review].
Neuroimage
 
1997
;
6
:
156
–67.
Penfield W, Mathieson G. Memory. Autopsy findings and comments on the role of hippocampus in experiential recall.
Arch Neurol
 
1974
;
31
:
145
–54.
Phelps EA, Hirst W, Gazzaniga MS. Deficits in recall following partial and complete commissurotomy.
Cereb Cortex
 
1991
;
1
:
492
–8.
Rombouts SARB, Machielsen WC, Witter MP, Barkhof F, Lindeboom J, Scheltens P. Visual association encoding activates the medial temporal lobe: a functional magnetic resonance imaging study.
Hippocampus
 
1997
;
7
:
594
–601.
Rosen BR, Buckner RL, Dale AM. Event-related functional MRI: past, present, and future. [Review].
Proc Natl Acad Sci USA
 
1998
;
95
:
773
–80.
Schacter DL, Wagner AD. Medial temporal lobe activations in fMRI and PET studies of episodic encoding and retrieval. Hippocampus. In press 1999.
Schacter DL, Buckner RL, Koutstaal W, Dale AM, Rosen BR. Late onset of anterior prefrontal activity during true and false recognition: an event-related fMRI study.
Neuroimage
 
1997
;
6
:
259
–69.
Scoville WB, Milner B. Loss of recent memory after bilateral hippocampal lesions.
J Neurol Neurosurg Psychiatry
 
1957
;
20
:
11
–21.
Squire LR, Ojemann JG, Miezin FM, Petersen SE, Videen TO, Raichle ME. Activation of the hippocampus in normal humans: a functional anatomical study of memory.
Proc Natl Acad Sci USA
 
1992
;
89
:
1837
–41.
Stern CE, Corkin S, Gonzalez RG, Guimaraes AR, Baker JR, Jennings PJ, et al. The hippocampal formation participates in novel picture encoding: evidence from functional magnetic resonance imaging.
Proc Natl Acad Sci USA
 
1996
;
93
:
8660
–5.
Tulving E, Markowitsch HJ, Craik FE, Habib R, Houle S. Novelty and familiarity activations in PET studies of memory encoding and retrieval.
Cereb Cortex
 
1996
;
6
:
71
–9.
Ungerleider LG, Haxby JV. `What' and `where' in the human brain. [Review].
Curr Opin Neurobiol
 
1994
;
4
:
157
–65.
Wagner AD, Desmond JE, Glover GH, Gabrieli JDE. Prefrontal cortex and recognition memory: Functional MRI evidence of context-dependent retrieval processes.
Brain
 
1998
;
121
:
1985
–2002.
Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Dale AM, et al. Building memories: remembering and forgetting of verbal experiences as predicted by brain activity [see comments].
Science
 
1998
;
281
:
1188
–91. Comment in: Science 1998; 281: 1151–2.
Worsley KJ, Friston KJ. Analysis of fMRI time-series revisited—again [comment].
Neuroimage
 
1995
;
2
:
173
–81. Comment on: Neuroimage 1995; 2: 45–53.
Worsley KJ, Marrett S, Neelin P, Vandal AC, Friston KJ, Evans AC. A unified statistical approach for determining significant signals in images of cerebral activation.
Hum Brain Mapp
 
1996
;
4
:
58
–73.
Zarahn E, Aguirre G, D'Esposito M. A trial-based experimental design for fMRI.
Neuroimage
 
1997
;
6
:
122
–38.