www.elsevier.com/locate/ynimg NeuroImage 22 (2004) 1704 – 1714
The hippocampal region is involved in successful recognition of both remote and recent famous faces Frederic A. Bernard, a,* Edward T. Bullmore, a Kim S. Graham, b Sian A. Thompson, c John R. Hodges, b,c and Paul C. Fletcher a a
Brain Mapping Unit and Wolfson Brain Imaging Centre, University of Cambridge, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK MRC Cognition and Brain Sciences Unit, Cambridge CB2 2EF, UK c Department of Clinical Neurosciences, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK b
Received 23 December 2003; revised 16 March 2004; accepted 17 March 2004
There is currently a debate regarding the precise role of medial temporal regions in memory, in particular regarding the time scale of their involvement in conscious recollection of information stored in long-term memory. Using event-related fMRI, we have attempted to contribute to this debate by identifying brain regions associated with the successful recognition of famous faces from two different periods: ‘‘Old’’ faces of people who became famous in the 1960s – 1970s and ‘‘Recent’’ faces of people who became famous in the 1990s. We demonstrate that the hippocampus is involved in the successful recognition of famous faces from both periods and does not appear to distinguish between these two periods. We also highlight a network of brain regions, including the left prefrontal cortex, the retrosplenial cortex, the temporo-parietal junction, the caudate and the right cerebellum, which is activated in association with successful recognition of famous faces. Finally, an analysis of the results obtained during a post hoc episodic recognition task shows the specific involvement of anterior hippocampus in the successful encoding of the unfamiliar faces, which were presented during the fame decision task, suggesting a functional distinction between anterior and posterior parts of the hippocampus, the former being specifically involved in successful episodic encoding and the latter being associated with successful retrieval of semantic information. D 2004 Elsevier Inc. All rights reserved. Keywords: Famous faces; Hippocampal region; Medial temporal lobe
Introduction Understanding the specific contributions of medial temporal regions to long-term memory function remains a major challenge for cognitive neurosciences. For example, the question of whether the hippocampus plays a selective role in episodic
* Corresponding author. Brain Mapping Unit, University of Cambridge, Department of Psychiatry, Addenbrooke’s Hospital, Box 255, Cambridge CB2 2QQ, UK. Fax: +44-1223-336581. E-mail address:
[email protected] (F.A. Bernard). Available online on ScienceDirect (www.sciencedirect.com.) 1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2004.03.036
memory or a more general role in both episodic and semantic memory (together termed declarative memory) is unresolved. Tulving and Markowitsch (1998) have suggested that the hippocampal role is selective to episodic memory with little or no contribution to semantic memory. Squire and Zola (1998) contend that the hippocampal region (including the hippocampus and the adjacent cortical structures) is important for both episodic and semantic memory. Another unresolved question concerns the time scale of hippocampal involvement in conscious recollection of information stored in long-term memory. One view, the classic consolidation model (Squire, 1992), postulates a time-limited involvement of this region in both episodic and semantic memory (together embedded in declarative memory). Another view, which takes into account a finer distinction between episodic and semantic memory, favors a continuous or life-long involvement of the medial temporal regions in recollection of autobiographical episodic memories and a time-limited involvement of these regions in retrieval of semantic information (Nadel and Moscovitch, 1997). However, Cipolotti et al. (2001) have described an amnesic patient, with gross abnormalities in both hippocampi, who has a profound memory impairment for recent and remote famous faces (across four decades). So the apparent discrepancy between the results reported in lesion studies, particularly regarding the time scale of hippocampal involvement in retrieval of semantic information, may be ascribed to several possible factors, among which are the possible neuronal and cognitive reorganizations that may occur as compensatory processes following medial temporal lobe injury. Moreover, while lesion studies can reveal regions that are necessary to perform a task, they do not reveal the distributed brain systems that can be considered sufficient to perform it (Price et al., 1999); in this context, functional neuroimaging studies have a complementary role. As far as we know, only a single study has previously attempted to explore the role of the medial temporal regions in the recollection of semantic information with different levels of historical remoteness or recency. Haist et al. (2001), using a blocked presentation of stimuli with fMRI, showed medial temporal activations only during recognition of famous faces from the 1990s, in comparison to rest. These results suggest
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
that activity in medial temporal regions is historically graded because it was greater during recognition of famous faces from the 1990s relative to recognition of famous faces from more remote decades. However, in a block design, which has a greater sensitivity than an event-related design, the activity associated with a specific kind of event is not examined; instead, activity is averaged over many events. This makes it impossible to distinguish specific stimulus-related effects, e.g., the correctness of stimulus identification, from state-related changes in activity that are tonically maintained across a block of trials, e.g., related to the famousness of the faces (Otten et al., 2002; Rugg, 1998). Thus, transient event-related effects, related to correct identification of remote or recent faces, may be obscured. The purpose of the current study was to replicate and extend the results obtained by Haist et al. (2001) and to reduce ambiguity by using an event-related fMRI paradigm. We were interested in determining the network of brain regions that are specifically activated during successful recognition of famous faces and in establishing, more precisely, the relationship of medial temporal activation to historical recency or remoteness of the famous faces, i.e., either greater activation for more recent faces or equivalent activation by faces from different epochs, by using faces from people who became famous during either the 1960s – 1970s or the 1990s. We addressed this using a two-way factorial design in which both recent (or ‘‘new’’) and remote (or ‘‘old’’) famous and nonfamous faces were presented in the setting of a fame judgement task. The inclusion of the non-famous old condition enables us to establish that any temporal differences between famous old and famous recent faces relates to the longevity of the stored memory rather than to some incidental perceptual difference between modern and older faces. One further area explored in the current study concerns the incidental episodic encoding that would occur in association with face processing. Numerous studies have shown that activity in medial temporal regions was associated with episodic encoding (for reviews, see Lepage et al., 1998; Paller and Wagner, 2002; Schacter and Wagner, 1999). Therefore, as we expect an activation of these regions for the successful recognition of famous faces (Leveroni et al., 2000), it would be of particular interest to determine whether this activation reflects semantic retrieval per se and/or incidental episodic encoding. Thus, following scanning, recognition memory was tested for all of the faces presented during the fame judgement task. This enabled us to estimate the levels of incidental encoding associated with face presentation during scanning, and, furthermore, formed the basis for an analysis of regional activations predictive of subsequent memory for famous and nonfamous faces.
1705
Experimental task The study used an event-related fMRI design (Fig. 1). The stimuli consisted of 80 black and white pictures of adult faces. Half of the faces were of famous people. Half of the famous faces were of people who became famous in the 1960s – 1970s (old famous faces) and the remaining half were faces of people who became famous in the 1990s (recent famous faces). A full list of the famous persons used in the test is given in Appendix A. Half of the pictures of non-famous faces were adjudged to look old and the remaining half looked recent. The subjects were scanned during a single run while viewing the 40 famous faces, the 40 unfamiliar faces, or a fixation cross (presented 40 times and considered as a baseline), in a randomized order. The stimuli were presented using DMDX (K.I. Forster and J.C. Forster, University of Arizona) on a screen placed comfortably within the subject’s field of view. Each stimulus was shown for 4 s. Volunteers were instructed to press one of two possible buttons of a response box to indicate whether a face was famous or not, as well as for each presentation of the fixation cross (either button). Subjects rested the index and middle fingers of the right hand on the two response buttons. The assignment of the yes response to either the left or right button was alternated across participants. Incorrect answers were excluded from subsequent fMRI analyses. Approximately 30 min after the fMRI session, but without prior warning, subjects were asked to perform a recognition test. During this task, they were presented with the 40 famous and 40 nonfamous faces previously studied among an equal number of foils (40 new famous and 40 new non-famous faces). Subjects were instructed to press one of three possible keys of a keyboard to indicate whether they were recognizing a face with a highconfidence level, with a low-confidence level, or if a face was new. Following this task, subjects were presented again with the famous faces shown during the fame decision task and were asked to recall their name. If the subject was unable to recall the name corresponding to the face, three possible names were presented and a decision was required (33% correct by chance). The two alternative choices were names of famous individuals from the same period as the target face. fMRI data collection Imaging data were collected at the Wolfson Brain Imaging Centre, University of Cambridge, using a Bruker Medspec (Ettlingen, Germany) scanner operating at 3 Tesla. T2*-weighted, echoplanar images, depicting BOLD contrast, were acquired in a single session (TE, 27.5 ms; TR = 1.6 s). Twenty-one 2D slices (4 mm thick; interslice gap, 1 mm; matrix size, 128 128) were acquired at each of 306 time points. The first six EPI images were subsequently discarded to avoid T1 equilibration effects, leaving 300 3D data volumes per session.
Methods fMRI data analysis Subjects Twelve healthy, right-handed volunteers (9 female and 3 male) with a mean age of 58.7 F 6.2 years (FSD) were scanned. The study was approved by the Addenbrooke’s NHS Trust Local Research Ethics Committee and written informed consent was obtained from all subjects.
Preprocessing steps comprised slice acquisition time correction, within-subject image realignment, spatial normalization and spatial smoothing using a Gaussian kernel (8-mm full-width at half-maximum). The time series was high-pass filtered (to a maximum of 1/120 Hz) to remove low-frequency noise. Events were designated as occurring at the presentation of the outcome
1706
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
Fig. 1. Examples of the four types of faces used in the fame decision task: Famous Recent, Famous Old, Non-famous Recent, Non-famous Old.
stimulus. Five event types were modeled: ‘‘Famous Old’’, ‘‘Famous Recent’’, ‘‘Non-famous Old’’, ‘‘Non-famous Recent’’ and ‘‘Incorrect Answers’’. The average hemodynamic response to each event type was modeled using a canonical, synthetic hemodynamic response function (HRF) (Friston et al., 1998). All data analysis was done using statistical parametric mapping in the SPM99 program (Wellcome Department of Cognitive Neurology, London, UK). Parameter estimates for each condition were determined using planned contrasts. The linear combinations of parameter estimates for each contrast were stored as separate images for each subject. These contrast images were entered into a one-sample t test, to permit inferences about condition effects across subjects that generalize to the population (i.e., a ‘random effects’ analysis). These contrasts produced statistical parametric maps (SPMs) of the t statistics at each voxel. Contrasts were thresholded with a P value corresponding to 5% false discovery rate (FDR) to control for multiple comparisons (Genovese et al., 2002). Only activations involving contiguous clusters of at least 15 voxels were interpreted. The locations of maxima of suprathreshold regions were labeled using the nomenclature of Talairach and Tournoux (1988) and anatomical designations of Brodmann (1909).
Results Behavioral data Fig. 2A presents the overall accuracy for recognition of famous faces and rejection of non-famous faces. Subjects correctly recognized 80.4% of the old famous faces and 69.6% of the recent famous faces; percentage correct rejection of non-famous faces was 91.7% for both old and new nonfamous faces. Subjects were significantly more accurate at rejecting non-famous faces than recognizing famous faces [ F(1,11) = 9.51, P = 0.01], but their reaction time (RT) for rejecting a non-famous face was significantly slower than for recognizing famous faces [ F(1,11) = 12.48, P = 0.005]. Analysis of RTs also yielded a fame by recency interaction [ F(1,11) = 14.15, P = 0.003] (see Fig. 2B). Performance data for the test of subsequent recognition memory, for those faces that had been correctly judged as famous or non-famous, are displayed in Fig. 3. Mean percent name recall and recognition scores regarding the famous faces recognized during the scanning test were 71.56% and 98.16%, respectively. For non-famous faces only, there was a significant relationship between subsequent memory outcome and encoding
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
1707
the fusiform face area), the medial temporal regions, the anterior cingulate cortex, the cerebellum, and frontal regions (especially the inferior and medial frontal gyri; see Fig. 4). These activations, mainly bilateral, were more widespread and generally associated with higher t scores for Famous than for Non-famous faces. The effect of Famous faces [Famous faces—Baseline] was associated with additional activations in the medial parietooccipital region (cuneus/precuneus), the posterior cingulate, and middle and anterior frontal regions; whereas the effect of Nonfamous faces [Non-famous faces—Baseline] highlighted an additional activation in the left postcentral gyrus. Effects of fame The comparison between Famous and Non-famous face processing [Famous – Non-famous] yielded significant activations in a network of brain regions including the caudate, the thalamus, the left frontal cortex (in middle and inferior frontal gyri), the right cerebellum, the temporo-parietal junction (mainly lateralized on the left), the precuneus, the posterior cingulate, and the medial temporal regions (Fig. 5). The reverse comparison [Non-famous – Famous] did not reveal any significant regions of activation. Effects of recency No effects of recency [Recent – Old] were significant at the pre-set threshold (FDR, P < 0.05) for either the Famous or the Non-famous faces. To enhance the sensitivity of these contrasts, without producing an unacceptable level of Type I error, we adopted a region-of-interest approach using small volume correction for Gaussian random field theory to control the family-wise Type I error at P < 0.05. Six regions of interest (each 5-mm radius) were designed bilaterally in anterior hip-
Fig. 2. (A) Percentage of correct recognition of Famous Old (FO) and Famous Recent (FR) faces and correct rejection of Non-famous Old (NFO) and Non-famous Recent (NFR) faces. (B) Mean of reaction times to faces. Error bars indicate SEM.
response latency [ F(1, 95) = 8.191, P = 0.005]; that is, RTs were significantly longer for the fame decision on non-famous faces later remembered with high confidence (1265 ms) than for the decision on non-famous faces that were later forgotten (1085 ms). Imaging data Face processing The fMRI results are summarized in Table 1 and Figs. 4 – 7. The main effects of Non-famous [Non-famous faces—Baseline] and Famous faces [Famous faces—Baseline] both revealed significant activations in the occipito-temporal regions (including
Fig. 3. Percentage of Famous faces recognized with a High (FH) or Low (FL) confidence level or Forgotten (FF), and percentage of Non-famous faces recognized with a High (NFH) or Low (NFL) confidence level or Forgotten (NFF). Black bars indicate percentage of False Alarms for Famous (FHFA) and Non-famous (NFHFA) foils recognized with a high confidence level. Error bars indicate SEM.
1708
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
Table 1 Principal brain regions showing significant signal increases for the following comparisons: Non-famous faces vs. baseline; Famous faces vs. baseline; and Famous faces vs. Non-famous faces Cluster size (voxels)
Brain regions
Non-famous vs. Baseline 14,823 L lingual gyrus R lingual gyrus L inferior occipital gyrus L hippocampus L fusiform gyrus R fusiform gyrus R inferior occipital gyrus R cerebellum R hippocampus L colliculus R parahippocampal region R cerebellum 729 L anterior cingulate R medial frontal gyrus 481 L postcentral gyrus 164 R putamen 207 R orbital frontal gyrus R inferior frontal gyrus 76 L orbital frontal gyrus L inferior frontal gyrus 52 L amygdala Famous vs. Baseline 24,108 L cuneus L hippocampus Lingual gyrus L fusiform gyrus R cerebellum R posterior cingulate L hippocampus R fusiform gyrus L colliculus L amygdala R parahippocampal region L cerebellum L inferior occipital gyrus R inferior occipital gyrus R precuneus L posterior cingulate 2067 L medial frontal gyrus
BA
Coordinates x
y
t score
z
17/18 17/18 18
10 10 36
90 80 88
4 3 6
13.92 11.71 11.38
37 37 18
24 42 44 44
31 55 47 80
3 16 16 1
11.20 10.38 9.66 9.14
30
2 20 6 10
60 33 29 35
32 2 7 0
7.39 7.37 7.34 6.50
32 8
6 4 2
66 23 40
30 38 42
5.40 11.55 6.52
3
38
21
49
8.19
11
22 26
6 40
6 17
5.53 5.17
47
34
40
17
4.21
11
28
33
10
4.48
47
36
27
3
4.14
32
10
11
3.71
17/18 37
4 26 0 42
66 29 82 53
7 4 1 16
15.86 15.12 13.23 10.97
23/31
2 4
56 56
33 12
9.85 9.45
24 44
14 48
13 18
8.58 8.34
10 20 14
29 8 37
5 11 0
7.62 7.46 7.35
18
6 36
60 88
29 11
7.16 6.81
18
44
82
4
5.39
7 23/31
2 12
68 55
29 18
4.43 3.68
2
22
50
8.18
31
37
8
Table 1 (continued) Cluster size (voxels)
Brain regions
Famous vs. Baseline 2067 L anterior cingulate R anterior cingulate L medial frontal gyrus 523 R inferior frontal gyrus R inferior frontal gyrus R middle frontal gyrus 383 R middle frontal gyrus R inferior frontal gyrus R middle frontal gyrus 130 L inferior frontal gyrus L inferior frontal gyrus 191 R medial frontal gyrus L superior frontal gyrus 634 L inferior frontal gyrus L inferior frontal gyrus 106 L middle frontal gyrus 83 L medial frontal gyrus 76 L medial frontal gyrus Famous vs. Non-famous 1851 R head of caudate L head of caudate R thalamus L thalamus 402 R cerebellum 902 L middle frontal gyrus L middle frontal gyrus L middle frontal gyrus L middle frontal gyrus L inferior frontal gyrus 47 R superior frontal gyrus 121 R lingual gyrus 1381 L temporo-parietal junction L middle temporal gyrus
BA
Coordinates x
y
t score
z
32 32 9
8 20 10
23 27 48
41 28 22
7.50 5.40 3.82
47
32
40
12
5.18
11
36
42
14
5.11
11
28
52
16
4.67
46
44
36
18
5.00
45
42
28
13
4.44
9/46
55
29
28
4.43
11
26
29
12
5.00
47
40
23
10
4.78
10
2
65
24
4.96
9
8
60
30
4.11
44
50
13
23
4.77
45
46
18
5
4.17
6
34
36
28
4.30
9/10
6
55
10
4.28
10
4
66
3
3.15
14
8
7
10.94
9
10 2 6 48 42
8 5 15 68 21
9 9 14 30 34
8.59 6.93 5.85 10.19 9.48
8
30
18
45
8.75
46
46
26
24
7.70
6
34
12
51
5.01
45
53
32
13
4.55
8
10
41
40
8.56
17 39
24 48
73 65
15 24
8.15 8.03
21
63
51
4
6.96
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714 Table 1 (continued) Cluster size (voxels)
Brain regions
Famous vs. Non-famous 1381 L temporo-parietal junction 558 L medial frontal gyrus L medial frontal gyrus L superior frontal gyrus 2798 R precuneus R posterior cingulate L posterior cingulate L precuneus 50 Anterior cingulate 574 R temporo-parietal junction R temporo-parietal junction 49 L hippocampus L hippocampus 101 L medial frontal gyrus R medial frontal gyrus 110 R inferior parietal lobe 79 L inferior frontal gyrus 130 L body of caudate L hippocampus L thalamus 58 L hippocampus 49 R inferior frontal gyrus 16 R parahippocampal region
BA
Coordinates x
y
t score
z
1709
fame decision task were considered. The most significant activation was in the left anterior hippocampus (x = 20, y = 9, z = 15; t = 5.40; P = 0.0001, uncorrected). The Dm effect for non-famous faces was also significant by a more tightly controlled regional analysis of voxels in a 7-mm radius centered on anterior hippocampus (x = 22, y = 16, z = 14; coordinates obtained from the contrast ‘‘Famous vs. Non-famous’’), using a small volume correction to maintain familywise error at 5% (see Fig. 7).
40
44
68
40
6.17
8
16
39
42
7.90
9
14
38
26
7.23
10
26
55
17
5.65
Discussion
7 23/31
10 8
54 57
43 19
7.30 6.75
Face processing
31
16
47
41
6.32
7 24 40
2 6 48
60 2 51
38 28 23
6.20 7.24 7.22
39
53
59
27
5.36
10
22 28 6
16 20 62
14 19 8
6.89 4.16 6.77
Face processing, in the setting of a fame-decision task, was associated with activation predominantly in occipito-temporal regions, including the fusiform face area (FFA) (Kanwisher et al., 1997). Our result suggests that this region seems to be primarily involved in the processing of faces (familiar or unfamiliar) and is not sufficient to result in successful recognition of famous faces. Bilateral activations of the medial temporal regions suggest a contribution of these structures in the attempt to match perceived faces with pre-existing semantic representations stored in longterm memory.
10
8
60
10
4.26
Effects of fame
40
36
43
32
6.42
44
50
3
20
6.24
47
20 24 22 32 32
32 27 25 24 20
16 5 5 16 18
5.63 4.97 4.36 5.46 4.44
30
26
19
4.37
The successful recognition of famous faces (by comparison with the successful rejection of non-famous faces) activates anterior (x = 22, y = 16, z = 14) and posterior (x = 24, y = 27, z = 5) parts of the left hippocampus (see Fig. 5Ba). This result suggests that hippocampus is involved in mediating effective access to the memory trace or, in other words, conscious recollection of preexisting semantic representations stored in long-term memory (engram). Comparable data have been reported by Kapur et al. (1995) and Leveroni et al. (2000), apart from the fact that the right hippocampus was activated in the latter study. In addition to the left hippocampus, we found several other regions were significantly more activated during the successful recognition of famous faces than during the successful rejection of non-famous faces, including the left prefrontal cortex, the right cerebellum, the posterior cingulate, the precuneus, the caudate and the temporo-parietal junction. The left prefrontal cortex is almost invariably activated during tasks involving semantic processing (Fletcher and Henson, 2001) and several theories of its contribution to such processing have been put forward. For example, it may be involved in maintenance of semantic information in working memory (Gabrieli et al., 1998), and/or in selection of task-relevant semantic attributes (ThompsonSchill et al., 1997). It may possibly reflect covert naming of some of the famous faces, although Haist et al. (2001) did not observe any activation of this region during a covert-naming task. While the current study does not distinguish between these possibilities, it is theoretically predicted to see a higher level of left frontal activation in association with famous faces compared to unfamiliar ones because the former have greater meaning for subjects. Coactivation of posterior cingulate/medial parietal cortex in association with famous faces is also in keeping with given previous studies by Leveroni et al. (2000) and Shah et al. (2001) showing activation of this region during the recognition of familiar people (faces only in the former study and faces + voices in the latter study, in which an
Location is with respect to the system of Talairach and Tournoux (1988); probability threshold for significance was controlled at FDR = 5%. L = left; R = right.
pocampus (x = F22, y = 16, z = 14) + posterior hippocampus (x = F24, y = 27, z = 5) + entorhinal cortex (x = F24, y = 10, z = 32). However, this more focused analysis also failed to show significant effects of recency on medial temporal lobe activation (Fig. 6). Regions predictive of subsequent memory performance Characterization of a ‘‘Dm’’ effect (differential neural activity based on memory) depends on analysis of a sufficient number of subsequently remembered and subsequently forgotten items from the encoding phase. Subsequent confident recognition of famous faces was uniformly high with an insufficient number of forgotten items to carry out a Dm analysis for famous faces. We therefore present only the results of a Dm analysis confined to non-famous faces. That is, we explored the neuronal signature of non-famous face processing when those faces were subsequently recognized with confidence compared to those that were not subsequently recognized. Only non-famous faces recognized as such during the
1710
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
Fig. 4. Orthogonal ‘‘glass brain’’ views for the comparisons Non-famous faces vs. Baseline (A) and Famous faces vs. Baseline (B) are shown sagittally from the right (upper left of each figure), coronally from behind (right), and transversely from above (lower left). These images were thresholded at P < 0.05 (FDR corrected, cluster size >15).
activation restricted to retrosplenial cortex was reported). Moreover, Valenstein et al. (1987) and Rudge and Warrington (1991) have described patients with impaired recognition memory for faces after a lesion to this region. It is probably relevant that retrosplenial cingulate cortex forms a critical point in fronto-hippocampal connectivity (Goldman-Rakic et al., 1984). Thus, the coactivation of frontal cortex, hippocampus and retrosplenial cingulate cortex in association with famous faces (which show higher levels of incidental episodic encoding) is consistent with the anatomical connections and can plausibly be regarded as a correlate of successful semantic retrieval. This, however, is speculative because patients with lesions of the retroplenial cingulate cortex are characterized by episodic rather than semantic memory problems, although, to our knowledge, memory retrieval for famous faces has not yet been assessed with such patients. The temporo-parietal junction has also been previously associated with semantic processing (Gorno-Tempini et al., 1998; Price et
al., 1997; Vandenberghe et al., 1996), and the retrieval of ‘‘realworld’’ memories (Maguire and Mummery, 1999). It remains difficult, however, to understand the precise role of this region. Further studies will be required to address this issue. The same conclusion can be drawn regarding the right cerebellum, which has been repeatedly involved in various cognitive tasks (Cabeza and Nyberg, 2000). The activation of the caudate was unexpected. This region has been associated with procedural memory tasks (Peigneux et al., 2000; Rauch et al., 1997). One possible explanation for its activation in these data relates to the mean age of our subjects, which was approximately 60 years. Several functional neuroimaging studies have observed age-related brain activation changes associated with various cognitive tasks (Cabeza, 2000). These changes may reflect age-related degenerative or compensatory processes. Therefore, our subjects may possibly have spontaneously or automatically recruited these regions, which are included in fronto-striatal loops, to compensate for inevitable
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
1711
Fig. 5. (A) Orthogonal ‘‘glass brain’’ views for the comparison Famous faces vs. Non-famous faces are shown sagittally from the right (upper left), coronally from behind (right), and transversely from above (lower left). These images were thresholded at P < 0.05 (FDR corrected, cluster size > 15). (B) Regions of significant activation superimposed on a T1-weighted anatomical image in standard space, showing increased activity in (a) the left hippocampus, (b) the left frontal cortex, (c) the bilateral temporo-parietal junction and the right cerebellum, (d) a region including the precuneus/cuneus and the retrosplenial cortex, and (e) the head of caudate bilaterally for the comparison Famous faces vs. Non-famous faces. These images were thresholded at P < 0.05 (FDR corrected, cluster size >15).
age-related brain changes and hence to perform the task efficiently. In support of this interpretation, Jacome (1986) reported a 59year-old patient who developed prosopagnosia after a lesion of the caudate. The hypothesis of age-related change in frontostriatal activation by declarative memory processes could be
experimentally tested by future studies of young and older participants scanned using encoding and retrieval paradigms, such as the event-related design used here, which appropriately control for differential task performance (number of successful trials).
1712
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
mined not only by measuring its activity while healthy subjects perform a cognitive task but more particularly by taking into account the functional connectivity between this region and other regions or, in other words, its neural context (McIntosh et al., 2003). Therefore, one possible extension of our study would be to check whether the connectivity between the hippocampus and other brain regions is the same or not when subjects are successfully recognizing old and recent famous faces. Another further extension would be to take into account the possible autobiographical significance that some famous people may have (Westmacott and Moscovitch, 2003; Westmacott et al., 2004) and check whether this factor may have an effect on the hippocampal activation associated with the successful recognition of famous faces from different periods. It would be an interesting dimension to take into account in future imaging studies involving recognition of famous faces. Regions predictive of subsequent memory performance Fig. 6. Box-plots of activity in the left anterior hippocampus (x = 22, y = 16, z = 14), the right anterior hippocampus (x = 22, y = 16, z = 14), the left posterior hippocampus (x = 24, y = 27, z = 5), the right posterior hippocampus (x = 24, y = 27, z = 5), the left entorhinal cortex (x = 24, y = 10, z = 32), and the right entorhinal cortex (x = 24, y = 10, z = 32) for both Old and Recent Famous faces. The level of activity between these two conditions is indistinguishable.
Effects of recency One of the most important observations in this study is the indistinguishable level of activity of the hippocampus during the successful recognition of both old and recent famous faces. Of course, the absence of a statistical difference between hippocampal activation by old and recent faces must be treated cautiously in the light of our modest sample size: Type II error is a risk. With this in mind, we explored the comparison by a region-of-interest analysis focused on hippocampus and entorhinal cortex with a less-conservative statistical threshold. Even under these circumstances, no significant differences were seen in hippocampal and entorhinal activation by old and recent faces. The apparent discrepancy between our results and those obtained by Haist et al. (2001) can be explained by the fact that their use of a block design did not allow a disambiguation of activation by recognized and unrecognized famous faces. Also, in this study, participants were required to indicate with a movement of their left index finger whether they could recall the name of the person in the photograph. Of course, it is possible to recognize a famous person facially, and hence to have access to semantic associations stored in long-term memory, without being able to name this person. It is probably also germane to note that Haist et al. (2001) studied a small number of subjects (eight) and used a fixed-effects analysis to demonstrate a hippocampal recency effect whereas we have used a more conservative but more generalizable random-effects analysis. Although our results regarding hippocampal activation tend to support the view of a long-lasting involvement of this structure in the retrieval of information stored in semantic memory (Cipolotti et al., 2001), they do not entirely refute the idea of a differential involvement of this structure in recollection of semantic information with different degrees of historical remoteness. Current views conceive that the role of a specific brain region can be deter-
Finally, an additional analysis of the results obtained during the post hoc recognition task revealed the specific involvement of left anterior hippocampus in the successful encoding of the nonfamous faces, which were presented during the fame decision task. This involvement of a medial temporal region in Dm effects is consistent with previous findings (for review, see Paller and Wagner, 2002) and confirms the role of this structure in the binding and storage of memory traces associated with successful encoding. Interestingly, an activation of this left anterior part of the hippocampus was also found for successful recognition of famous faces (by comparison with successful rejection of nonfamous faces). Although there was an insufficient number of forgotten items to carry out a Dm analysis with famous faces, it seems reasonable to think that an activation of the left anterior hippocampus, common to successfully encoded non-famous faces and successfully recognized famous faces, corresponds to a specific component of the medial temporal regions that might be involved in incidental episodic encoding of faces presented during the experiment, regardless of prior exposure or fame. Thus, the results of the present study are compatible with suggestions of
Fig. 7. Regions of significant activation superimposed on a T1-weighted anatomical image in standard space, showing increased activity in the left anterior hippocampus (yellow + green) for the comparisons of Non-famous faces recognized vs. Non-famous faces forgotten in a post hoc recognition test. Activity in this region survives a small volume correction (7-mm radius) for multiple comparisons (FWE and FDR, P < 0.05). The left anterior and posterior hippocampal activations associated with [Famous faces – Non-famous faces] are shown in green.
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
a functional distinction within the hippocampus between an anterior part, hypothetically specialized for incidental episodic encoding, and a posterior part, associated with explicit semantic retrieval regardless of the historical remoteness of retrieved information. This formulation corresponds well with results obtained by Strange et al. (1999) who have highlighted a functional dissociation within the hippocampus. In this study, the anterior part of the hippocampus was activated by generic novelty whereas a more posterior part was activated by familiarity of stimuli. These considerations emphasize the importance of caution when medial temporal activations are found during semantic retrieval tasks, because they can be associated with episodic encoding. In conclusion, the striking findings from this event-related fMRI study using famous faces divided into two different epochs of historical recency strongly support the hypothesis of a continuous or life-long involvement of the hippocampus in the retrieval of information stored in long-term memory. In addition to this, our results demonstrate a role in successful semantic retrieval for several other regions including left prefrontal cortex, retrosplenial cortex and temporo-parietal junction. Finally, we have highlighted a functional distinction between anterior and posterior parts of the hippocampus, the former being specifically involved in incidental episodic encoding and the latter being associated with explicit retrieval of semantic information.
Acknowledgments F.A.B. is supported by a Merck Sharp and Dohme Post-doctoral Fellowship award and was supported by INSERM. P.C.F. is supported by the Wellcome Trust. The Wolfson Brain Imaging Centre is supported by an MRC Co-operative Group Grant. The authors are indebted to R. Bisbrown-Chippendale for her help in this study.
Appendix A
1960s – 1970s
1990s
Woody Allen George Best Bjorn Borg Karen Carpenter Fidel Castro Julie Christie Alec Douglas Home Bob Dylan John F. Kennedy John Lennon Sophia Loren Steve McQueen Bobby Moore John Profumo Angela Rippon Peter Sellers Jackie Stewart Barbara Streisand Twiggy Harold Wilson
David Beckham Tony Blair George W. Bush Bill Clinton George Clooney Jill Dando Leonardo Di Caprio Celine Dion Hugh Grant Geri Halliwell Tom Hanks Tim Henman Nicole Kidman Ellen MacArthur Sophie Rhys-Jones Julia Roberts Michael Schumacher Britney Spears Robbie Williams Boris Yeltsin
1713
References Brodmann, K., 1909. Vergleichende lokalisationslehre der grosshirnrinde in ihren prinzipien dargestellt auf grund des zellenbaues Barth, Leipzig. Cabeza, R., 2000. Cognitive aging. In: Cabeza, R., Kingstone, A. (Eds.), Handbook of Functional Neuroimaging of Cognition. The MIT Press, Cambridge. Cabeza, R., Nyberg, L., 2000. Imaging cognition II: an empirical review of 275 PET and fMRI studies. J. Cogn. Neurosci. 12, 1 – 47. Cipolotti, L., Shallice, T., Chan, D., Fox, N., Scahill, R., Harrison, G., Stevens, J., Rudge, P., 2001. Long-term retrograde amnesia. . .the crucial role of the hippocampus. Neuropsychologia 39, 151 – 172. Fletcher, P.C., Henson, R.N.A., 2001. Frontal lobes and human memory: insights from functional neuroimaging. Brain 124, 849 – 881. Friston, K.J., Fletcher, P., Josephs, O., Holmes, A., Rugg, M.D., Turner, R., 1998. Event-related fMRI: characterizing differential responses. NeuroImage 7, 30 – 40. Gabrieli, J.D., Poldrack, R.A., Desmond, J.E., 1998. The role of left prefrontal cortex in language and memory. Proc. Natl. Acad. Sci. U. S. A. 95, 906 – 913. Genovese, C.R., Lazar, N.A., Nichols, T., 2002. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. NeuroImage 15, 870 – 878. Goldman-Rakic, P.S., Selemon, L.D., Schwartz, M.L., 1984. Dual pathways connecting the dorsolateral prefrontal cortex with the hippocampal formation and parahippocampal cortex in the rhesus monkey. Neuroscience 12, 719 – 743. Gorno-Tempini, M.L., Price, C.J., Josephs, O., Vandenberghe, R., Cappa, S.F., Kapur, N., Frackowiak, R.S., 1998. The neural systems sustaining face and proper-name processing. Brain 121, 2103 – 2118. Haist, F., Bowden Gore, J., Mao, H., 2001. Consolidation of human memory over decades revealed by functional magnetic resonance imaging. Nat. Neurosci. 4, 1139 – 1145. Jacome, D.E., 1986. Subcortical prosopagnosia and anosognosia. Am. J. Med. Sci. 292, 386 – 388. Kanwisher, N., McDermott, J., Chun, M.M., 1997. The fusiform face area: a module in human extrastriate cortex specialized for face perception. J. Neurosci. 17, 4302 – 4311. Kapur, N., Friston, K.J., Young, A., Frith, C.D., Frackowiak, R.S., 1995. Activation of human hippocampal formation during memory for faces: a PET study. Cortex 31, 99 – 108. Lepage, M., Habib, R., Tulving, E., 1998. Hippocampal PET activations of memory encoding and retrieval: the HIPER model. Hippocampus 8, 313 – 322. Leveroni, C.L., Seidenberg, M., Mayer, A.R., Mead, L.A., Binder, J.R., Rao, S.M., 2000. Neural systems underlying the recognition of familiar and newly learned faces. J. Neurosci. 20, 878 – 886. Maguire, E.A., Mummery, C.J., 1999. Differential modulation of a common memory retrieval network revealed by PET. Hippocampus 9, 54 – 61. McIntosh, A.R., Rajah, M.N., Lobaugh, N.J., 2003. Functional connectivity of the medial temporal lobe relates to learning and awareness. J. Neurosci. 23, 6520 – 6528. Nadel, L., Moscovitch, M., 1997. Memory consolidation, retrograde amnesia and the hippocampal complex. Curr. Opin. Neurobiol. 7, 217 – 227. Otten, L.J., Henson, R.N., Rugg, M.D., 2002. State-related and item-related neural correlates of successful memory encoding. Nat. Neurosci. 5, 1339 – 1344. Paller, K.A., Wagner, A.D., 2002. Observing the transformation of experience into memory. Trends Cogn. Sci. 6, 93 – 102. Peigneux, P., Maquet, P., Meulemans, T., Destrebecqz, A., Laureys, S., Degueldre, C., Delfiore, G., Aerts, J., Luxen, A., Franck, G., Van der Linden, M., Cleeremans, A., 2000. Striatum forever, despite sequence
1714
F.A. Bernard et al. / NeuroImage 22 (2004) 1704–1714
learning variability: a random effect analysis of PET data. Hum. Brain Mapp. 10, 179 – 194. Price, C.J., Moore, C.J., Humphreys, G.W., Wise, R.J., 1997. Segregating semantic from phonological processing during reading. J. Cogn. Neurosci. 9, 727 – 733. Price, C.J., Mummery, C.J., Moore, C.J., Frackowiak, R.S., Friston, K.J., 1999. Delineating necessary and sufficient neural systems with functional imaging studies of neuropsychological patients. J. Cogn. Neurosci. 11, 371 – 382. Rauch, S.L., Whalen, P.J., Savage, C.R., Curran, T., Kendrick, A., Brown, H.D., Bush, G., Breiter, H.C., Rosen, B.R., 1997. Striatal recruitment during an implicit sequence learning task as measured by functional magnetic resonance imaging. Hum. Brain Mapp. 5, 124 – 132. Rudge, P., Warrington, E.K., 1991. Selective impairment of memory and visual perception in splenial tumours. Brain 114, 349 – 360. Rugg, M.D., 1998. Convergent approaches to electrophysiological and hemodynamic investigations of memory. Hum. Brain Mapp. 6, 394 – 398. Schacter, D.L., Wagner, A.D., 1999. Medial temporal lobe activations in fMRI and PET studies of episodic encoding and retrieval. Hippocampus 9, 7 – 24. Shah, N.J., Marshall, J.C., Zafiris, O., Schwab, A., Zilles, K., Markovitsch, H., Fink, G.R., 2001. The neural correlates of person familiarity: a functional magnetic resonance imaging study with clinical implications. Brain 124, 804 – 815.
Squire, L.R., 1992. Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans. Psychol. Rev. 99, 195 – 231. Squire, L.R., Zola, S.M., 1998. Episodic memory, semantic memory, and amnesia. Hippocampus 8, 205 – 211. Strange, B.A., Fletcher, P.C., Henson, R.N.A., Friston, K.J., Dolan, R.J., 1999. Segregating the functions of human hippocampus. Proc. Natl. Acad. Sci. U. S. A. 96, 4034 – 4039. Talairach, J., Tournoux, P., 1988. Co-planar Stereotaxic Atlas of the Human Brain. Thieme, Stuttgart. Thompson-Schill, S.L., D’Esposito, M., Aguirre, G.K., Farah, M.J., 1997. Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation. Proc. Natl. Acad. Sci. U. S. A. 94, 14792 – 14797. Tulving, E., Markowitsch, H.J., 1998. Episodic and declarative memory: role of the hippocampus. Hippocampus 8, 198 – 204. Valenstein, E., Bowers, D., Verfaellie, M., Heilman, K.M., Day, A., Watson, R.T., 1987. Retrosplenial amnesia. Brain 110, 1631 – 1646. Vandenberghe, R., Price, C., Wise, R., Josephs, O., Frackowiak, R.S., 1996. Functional anatomy of a common semantic system for words and pictures. Nature 383, 254 – 256. Westmacott, R., Moscovitch, M., 2003. The contribution of autobiographical significance to semantic memory. Mem. Cogn. 31, 761 – 774. Westmacott, R., Black, S.E., Freedman, M., Moscovitch, M., 2004. The contribution of autobiographical experience to semantic memory: evidence from Alzheimer’s disease, semantic dementia and amnesia. Neuropsychologia 42, 25 – 48.
