OUP user menu

Aging affects the engagement of the hippocampus during autobiographical memory retrieval

Eleanor A. Maguire, Christopher D. Frith
DOI: http://dx.doi.org/10.1093/brain/awg157 1511-1523 First published online: 6 May 2003

Summary

Surprisingly little is known about the neural correlates of remembering real life events in the context of normal aging. We therefore asked young and older adults to retrieve real life autobiographical event memories accrued over decades, while their brains were scanned using functional MRI. There were many commonalities between the groups in the wider network of brain areas active during retrieval. Nevertheless, one key difference emerged; while left hippocampal activation was apparent in the young, bilateral hippocampal activation was evident in older adults and direct comparison between the groups confirmed significantly greater right hippocampal activation in older adults. Notably, this difference was specific to autobiographical event memory retrieval, as the groups were comparable in the areas active during semantic memory retrieval. The present findings show that even when stimuli, tasks and performance appear indistinguishable between young and older adults, neural changes occur in aging with specificity in both the memory type and the brain regions affected. In particular, the results reveal that age‐related effects are detectable in the hippocampus. This highlights the need to consider how the dynamic course of normal aging interacts with pathological processes that might also affect the hippocampus. Understanding this relationship may aid prognosis, as well as providing insights into plasticity in the anatomy of memory.

  • Keywords: aging; hippocampus; autobiographical memory; fMRI
  • Abbreviation: fMRI = functional MRI

Introduction

The normal aging process is known to be associated with changes in various cognitive domains including memory (see Light, 1991; Raz, 2000; Cabeza, 2001). There is general agreement that memory in late adulthood is characterized by both spared and impaired abilities. For example, age effects are minimal for implicit memory, priming, memory span and many recognition tasks (see Grady and Craik, 2000). In contrast, age‐related decline is reported for tasks involving cued and free recall, recollection of context and prospective memory (see Burke and Mackay, 2000; Grady and Craik, 2000). Precisely how these effects coalesce with neural changes in the aging brain remains an open question.

Functional neuroimaging is now being used to address this issue but, to date, studies comparing young and older subjects have almost exclusively focussed on the prefrontal cortex, and particularly its response during encoding (Cabeza, 2002). Collectively, the data are somewhat complex, with some reporting less activation of the left prefrontal cortex (e.g. Grady et al., 1995; Cabeza et al., 1997; Anderson et al., 2000; Stebbins et al., 2002), while others observe recruitment of extra frontal regions in older subjects (e.g. Cabeza et al., 1997; Madden et al., 1999; Logan et al., 2002). It has been shown recently that consideration of task demands (Logan et al., 2002) and efficacy of encoding (Morcom et al., 2003) may offer clues into the disparities across encoding studies. To a lesser extent, age‐related effects on memory retrieval have also been examined, again with the emphasis on the prefrontal cortex. A more consistent picture emerges here with a tendency for older subjects to activate bilateral prefrontal areas during retrieval, compared with more right‐lateralized activity in the young (e.g. Backman et al., 1997; Cabeza et al., 1997; Madden et al., 1999). This effect has been characterized as ‘hemispheric asymmetry reduction in older adults’ or HAROLD (Cabeza, 2002), and is now posited to be a task‐independent general aging phenomenon. Why these differences in prefrontal activity are observed is still debated; they may be due to sub‐optimal use of available resources or compensatory strategies to counteract cognitive decline (Grady and Craik, 2000; Cabeza, 2002; Logan et al., 2002).

While it is clear that the prefrontal cortex may hold clues to age‐related memory decline, what about the brain region identified most closely in the neuropsychological literature with memory, namely the hippocampus? It has featured much less reliably in functional neuroimaging studies of aging and memory. It may be that age effects chiefly relate to the prefrontal cortex rather than the hippocampus (Moscovitch and Winocur, 1992). Structural brain imaging has confirmed that volumetric reduction of the frontal cortex with aging is proportionately greater than in many other brain regions including the hippocampus (Raz et al., 1997; Good et al., 2001). Volume, however, is not a direct measure of function and it is notable that many of the memory impairments found in older adults closely parallel those typically observed in patients with lesions involving the hippocampus (see Spiers et al., 2001 for a recent review).

Another possible reason for the inconsistency of hippocampal activation in neuroimaging aging studies may be task‐related. The types of stimuli employed in the studies noted above, such as word lists/pairs, sets of faces (hereafter referred to as laboratory), do not result in consistent hippocampal activations in young subjects in neuroimaging studies, and so cannot provide a reliable springboard for interpreting aging effects in the hippocampus. Since laboratory tasks consistently activate prefrontal regions, it is understandable that this is where the age comparisons have focussed. Why these tasks do not produce reliable hippocampal activations is unclear. Neuropsychological studies have provided widely accepted evidence that the hippocampus is intimately associated with episodic memory (Scoville and Milner, 1957; Squire and Zola‐Morgan, 1991; Cohen and Eichenbaum, 1993; Vargha‐Khadem et al., 1997). This is the memory for the events we experience in the course of our day‐to‐day lives that have a specific temporal and spatial context (Tulving, 1983). Laboratory stimuli in memory scanning experiments, however, are presented close together in time and with much shorter retention intervals (minutes/hours) and in the same spatial context compared with the often decades of time over which real life autobiographical events are experienced. Thus, the temporal and spatial context of a typical autobiographical event is distinct, but this is not usually the case in laboratory tasks. In addition, autobiographical events have more obvious behavioural and social salience, and a clear sense of the ‘mental time travel’ attributed to episodic memory (Tulving, 2002). Diminishment of the emphasis on any or all of these factors may be related to the unreliable engagement of the hippocampus in typical imaging memory experiments.

Nevertheless, it is notable that where close concordance between the neuropsychological literature and functional neuroimaging findings exists, it is in relation to real life memories. While there are acknowledged difficulties in assessing autobiographical event memory (Hodges 1995; Warrington, 1996; Kapur, 1999; Piolino et al., 2002), several studies have reported a network of medial and largely left‐lateralized regions supporting both autobiographical events and general knowledge retrieval when compared with a control task ( Maguire and Mummery, 1999; Maguire et al., 2000, 2001). This network has consistently included the left hippocampus. When these memory types were compared directly, retrieval of autobiographical event memory, in particular, was associated with significantly increased activity in the left hippocampus and medial prefrontal cortex, apparently paralleling neuropsychological findings of preferential involvement of the hippocampus in autobiographical event memory. Interestingly, lateral prefrontal activation is typically not a feature in these experiments. Other neuroimaging studies using different methods to tap autobiographical memory have also reported activation of the hippocampus and often on the left (e.g. Fink et al., 1996; Burgess et al., 2001; Ryan et al., 2001; Gilboa et al., 2002; Piefke et al., 2003). However, to date no study has compared young and older adults, and so nothing is known about the effects of normal aging on the brain circuitry of autobiographical memory. Memory for real life events has been examined in older adults at the cognitive level. Most recently, Piolino et al. (2002) found that autobiographical event memory is more sensitive to the effects of age than semantic memory especially for the remote past (see also Rubin et al., 1986; Nyberg et al., 1996). Nevertheless, they showed that rich autobiographical event memories do persist into old age, enabling older adults to have a sense of reliving personal experiences.

A large proportion of patients who present clinically with memory problems are in late adulthood; this is true also of those with Alzheimer’s disease and even many of the amnesic cases reported in the literature (Spiers et al., 2001, Table 1). Impairment of autobiographical event memory is one of their primary complaints. In order to gain insight into this deficit, we first need to examine the neural correlates of autobiographical event memory in the healthy brain while taking into account the effects of normal aging. This was the aim of the current study. We considered only those memories that were richly recalled, with their distinct spatial and temporal contexts, and personal, behavioural and social salience. Thus, the current experiment aimed to maximize the opportunity for similarities between young and older subjects, allowing us to assess the minimum effects of aging on the neural correlates of autobiographical event memory. In these circumstances, the following questions were asked:

(i) Would any differences (such as under‐ or extra‐ recruitment) be evident between young and older adults in the network of brain regions supporting semantic memory on the one hand and autobiographical event memory on the other?

(ii) Would the hippocampus be more engaged by autobiographical memory retrieval compared with other types of memory, and would there be differences between young and older subjects in this regard?

(iii) Finally, would differences be present in prefrontal areas as in previous neuroimaging studies of memory and aging?

Materials and methods

Subjects

There were 24 participants in total. Twelve subjects were young (mean age 32.42 years; SD 4.21; range 23–39; six males, six females). Twelve subjects were older (mean age 74.75 years; SD 4.92; range 67–80; six males, six females). There was one left‐handed male in each group. All subjects were healthy and had been screened to exclude neurological, psychiatric or systemic conditions or medication that might affect brain function. In addition, the structural MRI scan of each subject was examined by a neuroradiologist to rule out pathology. The mean number of years of education beyond the age of 16 years was 5.5 (SD 3.43) for the young and 4.58 (SD 2.61) for the older subjects (the difference was not significant; P = 0.47). All subjects were high functioning (mostly university educated), autonomous community‐dwellers. The older subjects, whilst retired, were all active in cultural pursuits, continuing education or with responsibilities in various associations. Given the auditory nature of the experiment, the hearing of each subject was checked prior to scanning to ensure the tasks could be performed in the scanning context. All subjects gave informed written consent in the study which was approved by The National Hospital for Neurology and Neurosurgery and The Institute of Neurology Joint research ethics committee.

Image acquisition and data analysis

Data were acquired using a 2 Tesla Magnetom VISION (Siemens GmbH, Erlangen, Germany) MRI system. Contiguous multislice T2*‐weighted functional MRI (fMRI) images were obtained [TE (echo time) = 40ms] with echo‐planar imaging (EPI), whole head (32 slices, each 3 mm thick, 3.1 s per volume). A structural MRI scan using a standard 3D T1 weighted sequence was acquired from each subject. Functional images were processed and analysed using Statistical Parametric Mapping (SPM99; Wellcome Department of Imaging Neuroscience, London, UK; Friston et al., 1995). Briefly, the first five volumes were discarded to allow for T1 equilibration effects. Images were realigned to correct for interscan movement, normalized to a standard EPI template based on the Montreal Neurological Institute reference brain and resampled to 3 × 3 × 3 mm3. The T1 structural volume was co‐registered with the mean realigned EPI volume and normalized with the same deformation parameters. The normalized images were smoothed with an isotropic 8 mm full‐width half‐maximum (FWHM) Gaussian kernel to accommodate intersubject anatomical variability. As this was an event‐related study, the haemodynamic response to each stimulus event (statement plus response) was modelled by a canonical haemodynamic response function (HRF) and its first‐order temporal derivative. The six head movement parameters were included as confounds. First level linear contrasts of parameter estimates for each subject were taken to the second‐level and a random effects analysis was performed. A threshold of P < 0.001 uncorrected for multiple comparisons was employed throughout for the fMRI analyses. Activations involving clusters of at least three contiguous voxels were interpreted. Structural images were analysed using an optimized method of whole brain voxel‐based morphometry (see Ashburner and Friston, 2000) implemented in SPM99, employing a smoothing kernel of 10 mm FWHM. Regionally specific differences in grey matter density between subject groups were assessed. The significance threshold was set at P < 0.05 corrected for multiple comparisons (see Results for description of additional analyses).

Study materials and procedure

Using a similar paradigm to that reported previously (Maguire et al., 2000, 2001) several weeks prior to scanning, young and older subjects were interviewed at length to check general and personal factual knowledge, and to ascertain details of their personal event memories and knowledge of public events from the time they were 4–5 years old to the present day. Only events that involved precise recall of time and place of occurrence, and were definitely recollected by the subject and not told to them by a third party, and were recalled with the most detail, were considered for inclusion in the scanning experiment. The richness of details was scored in a similar manner to that described by Nadel et al. (2000), that is, by counting the number of details provided for each event. The memories selected for inclusion were those that were recalled with most details (but excluding those events where the details were of one particular type, e.g. for a memory involving a car, where the details pertained only to the visual/perceptual aspects of the car, but with few details about the overall event involving the car). This approach was used to exclude event memories that had become ‘semanticised’ (Moscovitch et al., 2000; Rosenbaum et al., 2001). There was no difference in the amount or richness of detail of recent and remote memories included in the scanning experiment. We were also careful to exclude public events that had clear autobiographical associations. The interview information was used to construct stimuli specific to each individual subject. Four memory types were considered: autobiographical events, public events, autobiographical facts and general knowledge.

Autobiographical event memory was our primary interest. As the older subjects have older memories compared with young subjects, the effect of aging would be confounded by the age of memories. We explicitly controlled for this by including the public events memory condition. For each autobiographical memory event, a public event matched for age (i.e. that occurred around the same time) was included. Thus, comparison of autobiographical events with public events controls for the age of memories, and the further comparison of this contrast between the two subject groups (see Table 4) is not confounded by the differential effect of memory age. Note that we do not directly compare the autobiographical event (AE) memory conditions of the two subject groups (i.e. AEolder versus AEyoung), as this would be confounded by memory age. Inclusion of the public events condition also controlled for additional factors pertinent to autobiographical event memory such as orienting to the past, and retrieval of episodes with specific spatio‐temporal contexts. Autobiographical events are also characterized by being self‐relevant. We included the autobiographical fact condition to examine the effect of self‐relevance in the absence of any association with a particular episode or experience. In order to look at basic semantic knowledge, we had the general knowledge condition that featured facts that were without either self‐relevance or episodic associations. In order to appreciate the extended networks of brain regions involved in retrieving these memory types, a lower level control condition was included that controlled for basic auditory, motor, attention and basic word recognition.

The stimuli comprised sentences presented auditorily during fMRI scanning via headphones that also dampened the scanner noise. The task was to listen to a statement and verify if it was true or false via a key press response. This task has been found in several previous studies to be easily accomplished by subjects and to maintain attention throughout the scanning session and, for episodic memories, causes subjects to recall the original events ( Maguire et al., 2001). There were 24 statements (scanning events) for each memory type and 24 control sets of words. In the control case, subjects heard sets of function words/prepositions and had to decide if the last word of the set had one or more syllables. The ratio of true: false (and 1:2 syllables for the control task) was 3:1, with the false statements comprising obvious adjustments to genuine memories. As described previously (see Maguire et al., 2001), it has been found that subjects remember the original event in the case of ‘false’ as well as true statements. Each statement was between 3–4 s long and subjects responded up to a maximum of 8 s from statement onset. Across all memory conditions, statements were controlled for length of presentation, number of syllables and frequency of proper nouns. For autobiographical events and public events, the stimuli were matched for the age of the memories, with an even spread throughout the lifetime of a subject. Subjects also rated each autobiographical and public event for emotional valence using a 7 point scale ranging from –3 (very unhappy) to +3 (very happy). They also rated the intensity of their feelings in relation to each memory event on a seven point scale raging from 1 (low intensity) to 7 (high intensity). Autobiographical and public event memories were matched overall for valence and intensity within groups and between the older and young groups. Order of conditions was randomized across subjects. Examples of stimuli include: (i) autobiographical events: You were the Christmas star in the school nativity play; (ii) autobiographical facts: Winkey and Frawley were friends at school; (iii) public events: Phil Collins sang in the UK and US for Band Aid; (iv) general knowledge: Cox’s orange pippin is a type of apple; and (v) control: He ago otherwise this a off therefore.

Results

Behavioural performance

Scores from the tasks performed during scanning show that, overall, the performance of the young and elderly subjects was well matched (see Table 1 ). The only difference was for the autobiographical fact retrieval task where young subjects scored slightly better, but in the context of high scoring throughout. There were no significant reaction time differences across conditions or between groups. Subjects were debriefed after scanning using a simple procedure. They were asked to state what they had been thinking about when they heard the stimuli. They were also asked to say if they thought about the prior interview during scanning or when hearing the stimuli. Subjects confirmed that, for event memories, the task had caused them to recall the original event, typically with a subjective sense of re‐experiencing. None of the subjects reported thinking about the prior interview during scanning, and none reported any of the tasks to be more difficult than another.

View this table:
Table 1

Mean behavioural performance

Condition Young Older P
Autobiographical events 23.33 (0.78) 22.92 (0.51) ns
General knowledge 23.67 (0.49) 23.42 (0.67) ns
Public events 23.17 (0.94) 22.58 (0.67) ns
Autobiographical facts 23.75 (0.62) 22.83 (0.94) <0.01
Control 23.58 (0.51) 23.17 (0.94) ns

Scores were out of a possible 24. ns = no statistically significant differences when compared. SDs in brackets.

fMRI findings

Memory networks

In the first instance, we compared each of the memory tasks with the control task in order to appreciate the extended networks of brain regions supporting the memory types.

Young. For the young group, retrieval of each of the memory types was associated with activation of a medial and largely left lateralized network of brain regions (detailed in Table 2 A–D). This included the medial frontal cortex, left temporal pole, left hippocampus, left middle temporal cortex, left parahippocampal gyrus, retrosplenial cortex, left temporo‐parietal junction and right posterior cerebellum. Examples of this network are also shown in Fig. 1 A,C during general knowledge and autobiographical event retrieval, respectively.

Fig. 1 Memory retrieval networks. Transverse views of glass brains permitting observation of all activations simultaneously. See Table 2A,B and Table 3A,B for details of activations. (A) Brain regions more active during the retrieval of general knowledge compared with the control task in the young group. (B) Brain regions more active during the retrieval of general knowledge compared with the control task in the older group. The results are very similar for the two groups. (C) Brain regions more active during the retrieval of autobiographical events compared with the control task in the young group. (D) Brain regions more active during the retrieval of autobiographical events compared with the control task in the older group. While overall there are similarities between the two groups, the older group activates the right hippocampus to a greater extent than the young group.

View this table:
Table 2

Main fMRI comparisons for the young group

Brain region Coordinates (x, y, z) Peak Z
A (AE–con)*
Medial frontal cortex (BA 10) –3, 54, –3 5.10
Left temporal pole (BA 21/38) –48, 9, –30 4.89
Right temporal pole (BA 38) 48, 18, –33 4.54
Left hippocampus –28, –13, –18 4.91
Right hippocampus 21, –18, –15 4.29
Left middle temporal cortex (BA 21) –54, –6, –24 4.95
Left parahippocampal gyrus (BA 20/36) –21, –39, –15 4.39
Retrosplenial cortex (BA 31) –6, –63, 24 6.02
Left temporo‐parietal junction (BA 39) –45, –63, 21 4.26
Right posterior cerebellum 30, –81, –36 4.60
Right ventral frontal cortex (BA 44/45) 54, 36, 0 4.75
B (GK–con)*
Medial frontal cortex (BA 10) –3, 54, 0 3.96
Left temporal pole (BA 20/38) –36, 15, –42 4.65
Left hippocampus –21, –11, –20 4.41
Left middle temporal cortex (BA 21) –60, –3, –27 3.85
Left parahippocampal gyrus (BA 36) –30, –33, –21 4.60
Retrosplenial cortex (BA 30) –3, –51, 12 5.07
Left temporo‐parietal junction (BA 39) –51, –69, 24 4.43
Right posterior cerebellum 21, –90, –36 4.95
Left ventral frontal cortex (BA 11) –42, 39, –21 3.77
C (PE–con)
Medial frontal cortex (BA 10) –3, 63, 9 4.72
Left temporal pole (BA 38) –42, 9, –36 4.50
Left hippocampus –22, –15, –18 5.30
Left middle temporal cortex (BA 20/21) –60, –12, –21 5.57
Left parahippocampal gyrus (BA 36) –33, –24, –24 4.80
Retrosplenial cortex (BA 23/31) –3, –57, 12 6.58
Left temporo‐parietal junction (BA 39) –48, –60, 18 5.74
Right posterior cerebellum 24, –84, –39 5.90
Left ventral frontal cortex (BA 11) –48, 36, –15 4.25
D (AF–con)
Medial frontal cortex (BA 10) –6, 57, 0 4.76
Left temporal pole (BA 38) –39, 12, –39 5.35
Left hippocampus –22, –15, –19 4.19
Left middle temporal cortex (BA 21) –60, –3, –27 4.51
Left parahippocampal gyrus (BA 36) –2,7 –24, –24 3.98
Retrosplenial cortex (BA 23/30) –3, –54, 6 4.75
Left temporo‐parietal junction (BA 39) –48, –63, 18 4.73
Right posterior cerebellum 18, –84, –39 5.14
Left ventral frontal cortex (BA 11/47) –39, 30, –21 4.08
E (AE–GK) (GK‐AE = ns)
Medial frontal cortex (BA 10) –3, 54, 0 5.58
Left hippocampus –33, –18, –15 5.31
Retrosplenial cortex (BA 31) –3, –60, 27 6.00
Dorsal medial nucleus of the thalamus 0, –9, 0 3.99
F (AE–PE) (PE–AE = ns)
Medial frontal cortex (BA 10) –3, 51, –3 4.76
Left hippocampus –27, –12, –24 3.99
Retrosplenial cortex (BA 31) –9, –60, 24 5.83
G (AE–AF) (AF–AE = ns)
Medial frontal cortex (BA 10) –6, 48, 0 3.85
Left hippocampus –24, –15, –15 3.75
Retrosplenial cortex (BA 23/31) –6, –57, 18 5.30

*See also Fig. 1. AE = autobiographical event memory; AF = autobiographical facts; BA = Brodmann area; con = control task; GK = general knowledge; ns = no statistically significant differences when compared; PE = public event memory.

Older. For the older group too, retrieval of each of the memory types was associated with activation of a medial and largely left lateralized network of brain regions (detailed in Table 3A–D) very similar to those activated in the young group. Examples of this network are also shown in Fig. 1B,D during general knowledge and autobiographical event retrieval respectively.

View this table:
Table 3

Main fMRI comparisons for the older group

Brain region Coordinates (x, y, z) Peak Z
A (AE–con)*
Medial frontal cortex (BA 10) –3. 54. 0 5.64
Left temporal pole (BA 21/38) –48, 9, –24 5.10
Right temporal pole (BA 38) 42, 15, –36 4.59
Left hippocampus –18, –18, –21 5.53
Right hippocampus 21, –15, –12 5.29
Left middle temporal cortex (BA 21) –57, –3, –24 5.18
Right superior temporal cortex (BA 21) 57, 3, –15 4.59
Left parahippocampal gyrus (BA 20/36) –26, –28, –18 4.23
Retrosplenial cortex (BA 31) –3, –63, 15 6.36
Left temporo‐parietal junction (BA 39) –42, –63, 18 5.39
Right posterior cerebellum 30, –78, –36 5.98
Dorsal medial nucleus of the thalamus 3, –6, –3 5.03
Right ventral frontal cortex (BA 47) 48, 30, –9 4.60
B (GK–con)*
Medial frontal cortex (BA 10) –6, 54, 0 4.01
Left temporal pole (BA 38) –45, 9, –39 4.90
Left hippocampus –21, –9, –21 4.73
Left middle temporal cortex (BA 21) –60, –9, –24 4.11
Left parahippocampal gyrus (BA 36) –24, –33, –21 4.13
Retrosplenial cortex (BA 23/30) –6, –57, 6 4.19
Left temporo‐parietal junction (BA 39) –45, –63, 21 5.13
Right posterior cerebellum 24, –84, –39 5.51
Left ventral frontal cortex (BA 11/47) –39, 33, –18 3.80
C (PE–con)
Medial frontal cortex (BA 10) –18, 60, 9 3.89
Left temporal pole (BA 20/38) –42, 15, –42 4.59
Left hippocampus –26, –17, –17 4.17
Left middle temporal cortex (BA 21) –60, –12, –24 4.21
Left parahippocampal gyrus (BA 28/36) –27, –21, –21 3.97
Retrosplenial cortex (BA 23/30) 3, –60, 9 4.63
Left temporo‐parietal junction (BA 39) –39, –66, 24 3.99
Right posterior cerebellum 15, –84, –39 5.78
D (AF–con)
Medial frontal cortex (BA 10) –3, 60, 9 4.54
Left temporal pole (BA 20/38) –39, 9, –45 4.48
Left hippocampus –22, –21, –15 4.01
Left middle temporal cortex (BA 21) –60, –6, –24 4.48
Left parahippocampal gyrus (BA 36) –24, –21, –24 3.78
Retrosplenial cortex (BA 23/31) –6, –63, 9 4.46
Left temporo‐parietal junction (BA 39) –45, –60, 18 4.12
Right posterior cerebellum 24, –84, –39 3.80
Left ventral frontal cortex (BA 47) –48, 36, –12 4.59
E (AE–GK) (GK–AE = ns)
Medial frontal cortex (BA 10) –6, 51, 0 3.63
Left hippocampus –21, –15, –15 4.04
Right hippocampus 21, –15, –18 3.85
Left middle temporal cortex (BA 21) –51, –9, –12 5.36
Right middle temporal cortex (BA 21) 57, –3, –18 3.85
F (AE–PE) (PE–AE = ns)
Medial frontal cortex (BA 10) –3, –54, 3 3.85
Left hippocampus –21, –12, –18 4.28
Right hippocampus 24, –12, –18 4.10
Left middle temporal cortex (BA 21) –60, –3, –24 4.05
Right middle temporal cortex (BA 21) 57, –12, –24 3.63
Retrosplenial cortex (BA 31) –6, –57, 24 5.21
G (AE–AF) (AF–AE = ns)
Medial frontal cortex (BA 10) –3, 54, 0 5.64
Left hippocampus –21, –15, –15 5.51
Right hippocampus 21, –15, –15 5.29
Left middle temporal cortex (BA 21) –57, –3, –24 4.01
Right middle temporal cortex (BA 21) 60, –12, –21 4.02
Retrosplenial cortex (BA 23/31) –6, –63, 15 6.36

*See also Fig. 1. AE = autobiographical event memory; AF = autobiographical facts; BA = Brodmann area; con = control task; GK = general knowledge; ns = no statistically significant differences when compared; PE = public event memory.

Young versus older. The striking similarity between the young and older groups is clearly demonstrated when the memory versus control contrasts are formally compared between the two groups (see top panel of Table 4; also Fig. 1). For general knowledge, public events and autobiographical facts, there were no statistically significant differences between the groups. There is only one instance when the two groups differed; the older group activated the right hippocampus more than the young group only during the retrieval of autobiographical events. No brain areas were more active in the younger compared with the older group.

View this table:
Table 4

Direct fMRI comparisons between young and older groups

Comparison (brain regions) Coordinates (x, y, z) Peak Z
[(AEy–cony)–(AEo–con o)] ns
[(AEo–cono)–(AEy–con y)]
 Right hippocampus 21, –15, –12 3.80
[(GKy–cony)–(GKo–con o)] ns
[(GKo–cono)–(GKy–con y)] ns
[(PEy–cony) –(PEo–con o)] ns
[(PEo–cono) –(PEy–con y)] ns
[(AFy–cony)–(AFo–con o)] ns
[(AFo–cono)–(AFy–con y)] ns
[(AEy–GKy)*–(AEo–GK o)*] ns
[(AEo–GKo)–(AEy–GKy )]*
 Right hippocampus 24, –15, –12 4.14
[(AEy–PEy)–(AEo–PEo )] ns
[(AEo–PEo)–(AEy–PEy )]
 Right hippocampus 21, –18, –15 3.79
 Right middle temporal     cortex (BA 20/21) 54, –15, –24 3.43
[(AEy–AFy)–(AEo–AFo )] ns
[(AEo–AFo)–(AEy–AFy )]
 Right hippocampus 21, –6, –21 3.85

*See also Fig. 2. o  = older subjects;y = young subjects; AE = autobiographical event memory; AF = autobiographical facts; BA = Brodmann area; con = control task; GK = general knowledge; ns = no statistically significant differences when compared; PE = public event memory.

Autobiographical events versus the other memory types

In order to assess the brain regions that are particularly active during autobiographical event memory retrieval, we compared it to each of the other memory types.

Young. The comparison of autobiographical event memory retrieval with retrieval of any of the other memory types (general knowledge, public events and autobiographical facts) in each case revealed increased activity in the left hippocampus, medial frontal and retrosplenial cortices for autobiographical events (see Table 2E–G; see also Fig. 2). No brain regions were more active for any of the other memory types compared with autobiographical events.

Fig. 2 Direct comparison of autobiographical and semantic memory. Results of the direct comparison of autobiographical events and general knowledge are shown on coronal views at the level of the hippocampus on mean structural MRI images of the relevant group. See Table 2E, Table 3E and Table 4 for full details of activations. (A) For young subjects, there was increased activation of the left hippocampus for autobiographical events. (B) For older subjects, there was increased activation of both left and right hippocampi for autobiographical events. (C) Direct comparison of the young and older subjects for this contrast confirmed significantly more activation of the right hippocampus in older subjects. Two other activations are evident, but were less than the minimum extent threshold – see Material and methods. (D) The plot shows the percentage blood oxygenation level dependent (BOLD) signal change averaged across subjects (relative to grand mean over voxels and scans) for autobiographical event memory retrieval. This is relative to the general knowledge condition. Error bars represent the between‐subjects SE. This is derived from the maximum voxel in the hippocampus as identified in the random effects analysis of each subject group. LHy = left hippocampus young; LHo = left hippocampus older; RHo = right hippocampus older.

Older. As with the young group, comparison of autobiographical event memory retrieval with retrieval of any of the other memory types in each case showed a similar set of brain regions more active for autobiographical events (see Table 3 E–G and Fig. 2B). In addition to the left hippocampus, medial frontal and retrosplenial cortices, the right hippocampus and bilateral middle temporal cortices were also more active in the older group when autobiographical event memory was compared with the other memory types.

Young versus older. Direct statistical comparison of the two groups revealed that, for each contrast involving autobiographical event memory compared with one of the other memory types, the older subjects consistently activated the right hippocampus more than the younger subjects (see lower panel of Table 4 and Fig 2 C,D). No brain regions were more active in the younger compared with the older group.

Additional comparisons

Several other comparisons were of interest. In order to examine the effect of self‐relevance in the absence of any association with a particular episode or experience, we compared autobiographical fact retrieval with general knowledge retrieval. Thus, both comprised semantic knowledge, but one was self‐relevant. For both young (Y) and older (O) groups, two areas were more active for autobiographical facts compared with general facts, medial frontal cortex [(Y: x, y, z coordinates = –3, 63, 9; Z  = 4.99) (O: –3, 63, 9; Z  = 3.85)] and retrosplenial cortex [(Y: –3, –45, 21; Z = 5.80) (O: 0, –48, 21; Z = 4.85)]. No areas were more active for general knowledge compared with autobiographical facts.

We also compared the two ‘non‐self’ conditions, namely public events and general knowledge. In the young group, left middle frontal gyrus (–48, 21, 45; Z = 3.74) and left middle temporal cortex (–57, –3, –21; Z  = 4.10) were more active for public events than general knowledge. For the older group, bilateral middle frontal gyri (–42, 15, 48; Z = 3.61 / 48, 24, 45; Z  = 3.94) were more active for public events than general knowledge. No brain regions were more active for general knowledge compared with public events.

Given that the group of participants included 50% males and 50% females, we examined the main effect of gender for the principal comparisons, as well as age × gender interactions (although the study was not specifically or optimally designed to examine this). On formal statistical comparison, there were no significant main effects of gender, and no significant interaction of age and gender (P  < 0.001 uncorrected).

Structural brain imaging findings

Given the significant age difference between the subject groups, it is possible that group difference observed in the fMRI data may be influenced by structural brain differences. We therefore compared the structural MRI images of the young and older groups using voxel‐based morphometry. The results are detailed in Table 5 (see also Fig. 3 ). In line with previous reports (e.g. Good et al., 2001), greater grey matter volume in the young group was apparent in several cortical areas, but importantly not in any brain regions that were active in the functional comparisons. Areas of volume difference included pre and post central regions, the insulae and intraparietal sulci. There were no areas of greater volume in the older group compared with the young group. There were no volume differences evident in the medial temporal lobes or the hippocampi. To further clarify the latter finding, we examined the data using a more liberal statistical threshold (P < 0.001 uncorrected for multiple comparisons) and did not observe any medial temporal or hippocampal group differences.

Fig. 3 Voxel‐based morphometry between young and older groups. Areas of greater grey matter volume in the young compared with the older group are shown on the mean structural MRI images of all subjects. (A) Coronal view showing the intraparietal sulci. (B) Transverse view showing pre/post central and medal frontal differences. (C) Sagittal view showing the insula, occipital and pre/post central differences.

View this table:
Table 5

Voxel‐based morphometry structural brain comparisons between groups

Brain region Coordinates (x, y, z) Peak Z
Y > O*
Left post central gyrus (BA 1,2,3) –50, –15, 33 5.80
Left precentral gyrus (BA 4) –45, –12, 60 5.53
Right post central gyrus (BA 1,2,3) 59, –24, 54 4.99
Right precentral gyrus (BA 4,6) 36, –20, 68 5.41
 (BA 3,4) 45, –17, 50 4.89
Right inferior occipital gyrus (BA 19) 47, –87, –12 5.36
Medial frontal cortex (BA 8) 0, 38, 36 5.04
Left intraparietal sulcus (BA 40) –48, –35, 47 4.98
Right intraparietal sulcus (BA 40) 44, –35, 53 4.97
Left insula –44, –14, 2 4.73
Right insula 44, –14, 11 5.27
Left posterior cerebellum –47, –77, –44 4.97
O > Y
ns

*See also Fig. 3. O = older subjects; ns=no statistically significant differences when compared; Y = young subjects.

Discussion

The current study examined how the neural correlates of autobiographical event memory retrieval are affected by normal aging. Specifically, we set out to address three questions:

(i) Are there differences (such as under‐ or extra‐ recruitment) between young and older adults in the network of brain regions supporting semantic memory on the one hand, and autobiographical event memory on the other?

(ii) Is the hippocampus more engaged by autobiographical memory retrieval compared with other types of memory, and are there differences between young and older subjects in this regard?

(iii) Are there prefrontal activation differences as in previous neuroimaging studies of memory and aging?

The results confirm that there were many commonalities between young and older adults in the activated brain areas. The overall pattern of results corresponds closely with previous studies of this kind in young subjects where a largely medial and left‐sided network of brain regions was found to support memory retrieval, but with the left hippocampus being most active for autobiographical event memory (Maguire and Mummery, 1999; Maguire et al., 2000, 2001). Nevertheless, one key difference between the two groups emerged; while left hippocampal activation was indeed apparent in the young, bilateral hippocampal activation was evident in older adults and direct comparison between the groups confirmed greater right hippocampal activation in the older adults. Notably, this difference was specific to autobiographical event memory retrieval, as young and older adults were indistinguishable in terms of areas active during semantic memory retrieval. There were no prefrontal differences evident between the two groups.

The particular vulnerability to aging of autobiographical event memory but not semantic memory has been reported behaviourally (e.g. Nyberg et al., 1996; Piolino et al., 2002). Typically, such studies measured the total number of memories, amount of detail recalled and the distribution of memories across the lifespan of subjects of different ages (Moscovitch et al., 2000; Piolino et al., 2002). While the current fMRI findings of age effects for autobiographical events but not for semantic memory appear to fit with the general pattern of behavioural reports, our memory measure was purposely biased towards memories that were richly recalled—thus one might have expected no differences between the memory types. However, the present results show that even when stimuli, tasks and performance appear indistinguishable, and where there is no volume reduction in the medial temporal region, fMRI reveals ‘hidden’ and quite specific consequences of aging.

Functional significance of the age effect

To interpret the age‐related change, one needs first to consider whether this difference is explicable merely in terms of an overall disparity in fMRI signal response between young and older subjects. Reduced amplitude of the blood oxygenation level dependent (BOLD) response and increased variability have occasionally been reported in older adults (D’Esposito et al., 1999; Huettel et al., 2001). However, this is unlikely to be the cause of the current effect, as findings were anatomically focal rather than global and specific to one memory type, while in all other respects the groups were similar. We also rule out simple performance/task effects, given that performance was matched on the explicit tasks during scanning, and all memories that were included in the experiment were salient and richly described by the subjects.

The age effects were not due to under‐recruitment of areas in older subjects due to absence or non‐utilization of resources as in other studies (e.g. Grady et al., 1995; Cabeza et al., 1997; Logan et al., 2002). There was no difference in the activation of the left hippocampus between groups, merely additional recruitment of the right by the older subjects. The present data are more consistent with the HAROLD model (Cabeza, 2002) where extra regions were recruited during task performance. This model has been primarily related to prefrontal age effects. The current data show that this model may also characterize age effects in the hippocampus during retrieval. As observed in the Introduction, many neuroimaging aging studies do not result in activation of the hippocampus in either young or older subjects, making the ‘generalizability’ of the HAROLD model difficult to evaluate. In studies that have found hippocampal activation, one reported no difference between young and older subjects in terms of hippocampal activation during recollection of words (Schacter et al., 1996), while another found increased co‐activation of hippocampus and prefrontal cortex in the young during retrieval of object locations (Schiavetto et al., 2002). In the latter study, however, young subjects performed better than older subjects, presenting a potential confound. The only other studies to report hippocampal findings were at encoding/perception and not retrieval as in our study; thus they are not directly comparable. The results of these studies are very mixed, with more hippocampal activation reported in the young (Grady et al., 1995, 1999; Mitchell et al., 2000), more activation in older subjects (Grady et al., 2000) or no differences between young and old (Morcom et al., 2003) reported.

Several previous fMRI studies involving autobiographical memory retrieval also reported bilateral activation of the hippocampus. Interestingly, the subjects in Ryan et al. (2001) had a mean age of 60 years and five out of seven of them showed bilateral hippocampal activation during autobiographical memory retrieval, in line with the current findings, although they did not directly compare young and older subjects, and did not include other memory types for comparison with autobiographical events. Using a similar paradigm to the present study, Maguire et al. (2001) reported bilateral hippocampal activation in a young patient during memory retrieval. However, he showed bilateral hippocampal activity for all memory types including general knowledge (when compared with the control task). Given that his pathology was of perinatal origin, the ubiquitous bilateral activations in his case may have more to do with reactive plasticity in the developing brain than a similar process to natural aging. Piefke et al. (2003), using a region of interest approach focussed on the hippocampus, reported bilateral hippocampal responses in young subjects during the retrieval of recent compared with more remote autobiographical memories that were selected to have a strong emotional valence. It may be that the emotional valence was related to the increased right hippocampal activity in their young subjects (e.g. see Fink et al., 1996). In addition, they report significant differences between their recent and remote memories in terms of ratings of picture‐likeness, emotionality, richness of detail and sense of re‐experiencing, which may also relate to the bilaterality of the findings. By contrast, the autobiographical memories of our young and older subjects were matched on these factors, both between groups and within groups for memory age (see Material and methods).

Several alternative mechanisms have been posited to account for lateralization reductions in normal aging (see Grady and Craik, 2000; Cabeza, 2002 for a review; Logan et al., 2002). Changes may be compensatory and reflect either neuroanatomical re‐organization of memory or differences in the cognitive processes engaged during a task. Alternatively, it has been suggested that the lack of lateralization may be associated with the failure by older subjects to recruit specialized brain regions during a particular task, in a reversal of the process of functional differentiation during childhood (Cabeza, 2002). A third possibility is that the additional activations in the older subjects have no functional significance. Functional neuroimaging reveals brain regions that are involved, but not necessarily crucial for task performance (Price et al., 1999). Overall, the present data do not favour one over another of these possibilities. The anatomical and task specificity in our view speak against the effects having no functional significance. As with other neuroimaging memory studies of aging, we accept the difference between young and older subjects has functional relevance, and now consider in more detail what the basis of the age effect might be.

Basis of the age effect

Does the recruitment of the right hippocampus help support the same cognitive operations as in young subjects, or different ones? This requires consideration of the nature of human hippocampal involvement in memory. Although the precise hippocampal contribution to memory is still open for debate, this issue was recently examined in a review of neuropsychological, structural and functional neuroimaging findings (see Burgess et al., 2002). Briefly, it was concluded that the right hippocampus has a role directly related to that in non‐humans, namely spatial memory and, in particular, allocentric representations (O’Keefe and Nadel, 1978; Burgess et al., 2002). In the left hippocampus, this allocentric role may have evolved to be useful for long‐term memory in general, perhaps providing an index‐like code (or traces, see also Nadel and Moscovitch, 1997) that interacts with neocortical storage regions to retrieve contextually specific autobiographical memories.

The additional activation of the right hippocampus may directly reflect the increased use or salience of spatial context in older subjects. The lack of age effects for general knowledge retrieval, where spatial context is not an issue, supports this view. Examination of the memories included in the scanning experiment and the original interviews did not reveal any obvious spatial bias in older subjects. However, this may not be manifested in subject’s overt reports. Why is it that the location of events might become more salient or is more enduring than other aspects of life events? Spatial memory is a fundamental cross‐species function; perhaps late adulthood is associated with a reversion to that which is most fundamental, a form of dedifferentiation (Cabeza, 2002). Alternatively, it may be that with a greater number of memories accrued over a more years by older compared with the young subjects, additional resources are needed to distinguish the relevant memories, and the environments in which they occurred is one way to achieve this.

It is possible that the right hippocampus has the capacity to perform exactly the same cognitive processes as the left hippocampus, such that if the left hippocampus is in physiological decline then the right can compensate. A related more radial view might be because humans in their current form have only quite recently begun to have such long life expectancies that the left hippocampus reaches its capacity (of index codes, traces). The right hippocampus is then called upon for any spare capacity it can offer. If this is the case, one might expect to see a bias towards the support of recent autobiographical memories by the right hippocampus, (see Maguire and Frith, 2003; Piefke et al., 2003).

As people and their memories age, it has been posited that autobiographical event memories become ‘semanticised’ and that this accounts for the apparent preservation of remote memories compared with recent memories in patients and older subjects (Moscovitch et al., 2000; Rosenbaum et al., 2001). Moscovitch et al. (2000) have shown a deficit in recall of details with time interval in older subjects. It may be that in the present study the autobiographical memories of the young and older subjects were qualitatively different, with the latter’s being more like semantic memories. By including only those memories that were recalled in rich detail with a full sense of ‘mental time travel’ (Tulving, 2002), we believe we circumvented such a possibility. If the autobiographical event memories of the older subjects were more like semantic knowledge, then one might predict that comparison of autobiographical events with autobiographical facts would reveal little difference. However, retrieval of autobiographical events resulted in more activity in several areas including the hippocampi. In addition, it is interesting that autobiographical facts did not activate the hippocampus more than non‐self relevant facts. This is in line with the view that self‐relevant episodes are crucial for engaging the hippocampus (Vargha‐Khadem et al., 1997). That autobiographical fact retrieval was associated with activity in medial frontal cortex is concordant with recent findings of activation in this region for general self‐referential tasks (Gusnard et al. 2001; Kelley et al. 2002).

Memory decline in the elderly has also been attributed to cognitive factors such as a reduction in processing speed (Salthouse, 1996) and attentional resources (Craik and Byrd, 1982). If activation or reaction time differences across all of the current tasks had been observed between the two age groups, then such explanations might have been plausible. We believe the most parsimonious explanation for the recruitment of the right hippocampus during autobiographical memory retrieval in older adults relates to the use of its spatial processing or reserve capacity, and this will be investigated in future studies.

Overall, the possible origins of the age effect seem to be more in line with a compensation view (Cabeza, 2002), although little is known about the functional circuitry of autobiographical development in childhood to, as yet, properly evaluate differentiation and then dedifferentiation by comparison. Logan et al. (2002) found that the supply of cognitive strategies reversed some age effects during encoding, while others were less malleable. They attributed the latter to a neurogenic origin (Cabeza, 2002), namely physiological changes in the brain. Given that the young and older adults in the present study performed comparably, we believe our hippocampal age effects have a similar physiological origin.

Conclusions

The current results show that, even in the context of preserved memory retrieval, age‐related effects are detectable in the hippocampus. This underlines the need to consider how normal aging interacts with pathological processes that might also affect the hippocampus. The effects of such pathology may differ depending on the age of the individual. As observed in the Introduction, a large proportion of patients who present clinically with memory problems are in late adulthood, as are many of the amnesic cases reported in the literature (Spiers et al., 2001, Table 1). Making inferences about the general structure of memory without considering age may be misleading. The subjects in the present study were in two groups, separated in age by 40 years. It will also be important to know exactly when or over what timescale this additional hippocampal recruitment occurs.

As with previous neuroimaging studies of this kind, lateral prefrontal activations were minimal and the young and older subjects did not differ in this regard. This obviously contrasts with the majority of imaging studies of aging and memory where the prefrontal region was the focus of activation and change. As alluded to at the outset, it may be the nature of the salient and distinct stimuli that shifts the focus to the hippocampus in the case of more real life stimuli. In the current study, we ensured a high level of memory retrieval; however, it may be that, in normal circumstances where recall is more variable, prefrontal regions would then be engaged, with concomitant effects such as under‐recruitment (Logan et al., 2002). In addition, the focus here was on retrieval; it may be that at encoding frontal activity and age‐related changes, perhaps even for general knowledge, may emerge.

While this is the first neuroimaging aging study to focus on real life events and we have identified neural changes that occur in aging with specificity in both the memory type and the brain region affected, clearly much remains to be understood. With increasing longevity of the population, understanding the mechanisms of these age‐related changes has important implications for the expectations of older adults in society.

Acknowledgements

We wish to thank Catriona Good for examination of structural scans and Alexa Morcom for assistance with subject recruitment. The authors are supported by the Wellcome Trust.

References

View Abstract