Neural basis of autobiographical memory

One line of research seeks to understand the neural network supporting autobiographical memory retrieval. In particular, we examined the role of the hippocampus and how it damage might affect the entire neural network. Lesion studies have suggested this is a key structure involved in the retrieval of autobiographical memories. Neuroimaging data have suggested that, within the autobiographical memory network, the hippocampus is preferentially engaged by autobiographical memory. In one study, we found that the recollective qualities of autobiographical memories predicted the level of hippocampal activity (Addis et al., 2004, Hippocampus). Specifically, autobiographical memories with higher levels of detail, emotionality and personal significance were associated with greater hippocampal activity, even when other attributes, such as recency, were taken into account.

In another study, we examined the networks supporting retrieval of two types of autobiographical memory: (1) specific memories, that is, memories of unique events specific in time and place (e.g., breaking one's leg); and (2) general memories, that is, memories of an event which occurred repeatedly (e.g. walking to school every day; Addis et al., 2004, NeuroImage). Using functional connectivity analyses (spatiotemporal partial least squares), we identified distinct sub-networks supporting these autobiographical memory types. Specific memories were associated with activity in posterior regions supporting visuospatial and imagery processes, while general memories engaged regions supporting conceptual retrieval, such as right temporal pole. Further, these networks differed temporally, consistent with previous cognitive studies, with activity associated with general memories peaking 2 seconds before that associated with specific memories. Despite these differences, those regions which were functionally connected with the hippocampus (i.e., a seed analysis) did not differ according to the type of memory being retrieved, suggesting the hippocampus may be the hub of the autobiographical memory network which is engaged during any form of memory retrieval.

Consequences of hippocampal damage

We have also examined how autobiographical memory retrieval and the associated neural network is altered by left hippocampal damage in temporal lobe epilepsy (TLE) patients (Addis, Moscovitch & McAndrews, 2007, Brain). In this study, we first confirmed the presence of significant left hippocampal atrophy using a linear measure of hippocampal volume (MTL width; Gao et al., 2004). Secondly, we used the Autobiographical Interview (Levine et al., 2002) to probe the integrity of autobiographical memory and found mild deficits in the episodic but not the semantic aspects of autobiographical memory. During the retrieval of residual memories, these patients exhibited significant reductions in activation of the hippocampus, as well as other regions in the autobiographical memory network, even though memory performance was included as a covariate. Further, the effective connectivity of the autobiographical memory network was dramatically different between patients and controls, with an apparent "bypassing" of the left hippocampus.

Remembering the past and imagining the future

A more recent line of research seeks to understand how memories of past events are used in the simulation of future personal experiences. The ability to shift our perspective from the present to the imagining of future scenarios is critical to both personal and social psychological functioning. Evidence from behavioural, neuropsychological and one previous neuroimaging study had suggested there is overlap between the ability to remember the past and imagine the future.

In a recent integration of this literature and the memory distortion literature, Daniel Schacter and I proposed the constructive episodic simulation hypothesis (Schacter and Addis, 2007, Nature; Phil Trans Roy Soc B) in which we outline how the constructive nature of episodic memory is well-suited, and perhaps adapted specifically for, the simulation of future events. Episodic memory is constructive rather than reproductive and is thus prone to various kinds of errors and distortions. However, we argue that it is the constructive character of episodic memory which enables individuals to imagine future episodes, happenings, and scenarios. Because the future is not an exact repetition of the past, simulation of future episodes requires a system that can draw on the past in a manner that flexibly extracts and re-combines elements of previous experiences.

We have been using behavioural and neuroimaging paradigms to test the constructive episodic simulation hypothesis. We have found considerable overlap in the neural networks engaged by past and future events (Addis, Wong & Schacter, 2007, Neuropsychologia). However, future events recruited some regions (e.g., hippocampus, frontal poles) more so than past events, suggesting that future simulation is a more intensive process than remembering. Further analysis of these data found that frontal poles and anterior hippocampus repond to the amount of detail comprising future events, while other regions, such as posterior hippocampus, are responsive to both past and future event detail (Addis & Schacter, 2008, Hippocampus). We are conducting further fMRI studies to better understand the role of the hippocampus in future simulation. For instance, we are exploring hippocampal responses to the novelty of future events, given that novelty can influence hippocampal activity, and future events are novel by definition.

In a recent behavioural study (Addis, Wong & Schacter, 2008, Psychological Science), we examined the episodic specificity of past and future events in young adults, and how this changes with healthy aging. We replicated previous findings that the that number of episodic details in past events was significantly reduced in older adults relative to younger adults. Importantly, this reduction was also evident for future events, and the reduction in episodic details was significantly correlated across past and future events. Finally, the episodic specificity of both past and future events correlated with the integrity of relational memory, consistent with the constructive episodic simulation hypothesis that simulation of future episodes requires a system that can flexibly recombine details from past events into novel scenarios. We are currently using this paradigm to examine the episodic specificity of past and future events in patients with Alzheimer's disease.

Memory and identity

An additional strand of our research considers the role that AM plays in identity. By integrating methods from social psychology and neuropsychology, we have examined the impact of AM loss on the integrity of identity (measured with the Twenty Statements Test and the Tennessee Self Concept Scale) in patients with Alzheimer's disease (Addis & Tippett, 2004, Memory). This revealed that loss of late childhood/early adulthood AMs had the most significant effects on identity, such that identity became weaker (i.e., patients could not generate as much information about themselves) and more abstract in nature, even when controlling for general cognitive decline. We are also interested in how AM contributes to specific aspects of identity such as self-continuity. In a recent chapter (Addis & Tippett, 2008), we reviewed how the episodic and semantic aspects of AM contribute to the content and continuity of identity, and proposed a distinction between narrative and phenomenological forms of self-continuity, the former drawing upon both semantic knowledge and episodic memory while the latter drawing specifically on episodic memories of past events and the experience of the self over time.

Associative encoding

A major role of the hippocampus during memory retrieval appears to be the integration of different recollective aspects of a memory, highlighting its memory for associative information. We have used a semantic-relatedness paradigm to examine associative encoding. This involves presenting triads with varying numbers of semantic relations during encoding. Triads with fewer associations have higher generative load, whereas triads with more associations had higher relational load. We found that hippocampal activity was modulated by the relational load of the encoding task while the left inferior frontal activity was modulated by generative load (Addis and McAndrews, 2006, NeuroImage). Further, there was evidence of strong positive connectivity between these regions which did not differ according to the number of associations provided.

More recently, we have used the same paradigm to examine whether age-related declines in relational encoding reflect dysfunction of inferior frontal gyrus linked with deficient generation of associations, and/or hippocampal dysfunction linked with impoverished binding of associations (Addis, Giovanello et al., in prep). As expected, older adults exhibited decreased levels of activity in both regions relative to young adults. Further, generative load did not modulate the inferior frontal gyrus. However, the hippocampus responded significantly to relational load, suggesting that when provided with associations to bind, hippocampal activity in older adults is comparable to young, consistent with increased recognition accuracy under such conditions. Importantly, this finding suggests that age-related impairments in relational encoding are likely the result of deficits in prefrontal engagement and the generation of associations, rather than the binding of associations by the hippocampus