Trends in Cognitive Sciences
ReviewAssociative memory and the medial temporal lobes
Introduction
Declarative memory comprises memory for personal experiences (episodic memory) and for facts and concepts (semantic memory). Declarative memory is essentially associative, linking together component parts (e.g. words and objects) either directly or via spatial, temporal or other kinds of relationships. These components are represented by neural activity in different parts of the neocortex that project to the medial temporal lobes (MTLs) where they are integrated to create associative memories. At retrieval, one component can cue recall of other components, which reactivates part or all of a ‘memory’. Therefore, recall is inherently associative. Alternatively, previously encoded episodic or semantic information can be recognized. According to the dual-process model 1, 2, recognition memory is mediated by recollection and familiarity. Recollection is cued recall of associated information within a recognition situation; familiarity involves no recall and occurs when exposure to information leads to a ‘feeling’ of memory.
Some researchers believe that familiarity occurs only for item memories and not for inter-item associations (e.g. Ref. [3]). However, others believe that familiarity also occurs for inter-item associations (e.g. Ref. [4]). Items (e.g. cars) comprise components (e.g. lights and windows) that are bound together in such a way that they are, typically, consciously perceived and remembered as single entities. Therefore, item memories are also associative. New items (e.g. unknown words or faces) can be created at encoding through a process of unitization or intra-item association, which binds components together, often using an established spatial framework (e.g. the layout of a face). This creates a representation that is usually perceived and remembered as one entity. However, there are no objective criteria for identifying whether unitization has occurred. The occurrence of unitization needs to be identified by markers other than the presence of familiarity. When two items are encoded together without an obvious ‘object-creating’ framework, it is difficult to prove that a new item has been created when the result does not ‘feel’ unequivocally like an entity. Intuitively, the speed with which new item memories are formed should depend on how easily components fit into a pre-existing template (e.g. a face framework). However, identification of such templates is subjective and, therefore, there might be disagreements. Progress is possible provided that associations that are agreed to be unitized carry measurable costs (e.g. greater difficulty in perceiving and recognizing components) that are not shared by equally well-learned associations that are agreed to be non-unitized.
Despite these difficulties, investigation is needed into whether the functional and neural bases of intra-item associative memory and non-unitized inter-item associative memory differ. Furthermore, inter-item associative memories have distinct characteristics that might have functionally and neurally different bases. Thus, within-domain associative memories are not perceived and remembered as one entity. They form between the same or very similar kinds of items (e.g. two faces; a door and a window) that are likely to be represented by activity in closely adjacent and interacting neocortical neurons. The items in these memories feel like they ‘go together’, but they do not form a single entity. Between-domain associative memories are also not perceived and remembered as one entity. They form between different kinds of items that might come from distinct sensory modalities (e.g. face–voice) or link items spatially, temporally or via another kind of relationship (e.g. A angered B). These items and relationships are likely to be represented by patterned activity in relatively distant and weakly connected neocortical neurons. This processed information feeds into the MTLs where it is integrated to create associative memories and possibly perceptual representations.
Declarative memory comprises intra-item, within-domain and between-domain associations that are bound into memory in the MTLs (Figure 1). How well each of these kinds of associative memories supports familiarity and recollection should be fully explored using appropriate memory-testing procedures (Box 1, Box 2). Current views on how the MTLs support the processes that underlie these associations are discussed below, after which psychological, lesion and functional neuroimaging evidence that discriminates between these views are considered in turn.
Section snippets
The role of the MTLs in associative memory
One view that implies neurofunctional differences among intra-item, within-domain and between-domain memories states that different MTL structures mediate within-domain and between-domain memory in qualitatively distinct ways [4]. This domain dichotomy (DD) view proposes that the perirhinal cortex has a key role in mediating recognition memory for unitized associations and for non-unitized, within-domain associations (e.g. face–face associations) (Figure 2). The perirhinal cortex supports
Psychological evidence from normal subjects
The views described above make different predictions about how well recognition of within- and between-domain associations is supported by familiarity. There is evidence that recognition for rapidly acquired intra-item associations is well supported by familiarity (e.g. Ref. [3]). Most research on within-domain associations has focused on learning to associate unrelated pairs of words. Initially, this research suggested that there was little, if any, familiarity for rapidly learned
Human and animal lesion evidence
The DD view predicts that selective hippocampal lesions will disrupt recognition of between-domain associations but only mildly disrupt within-domain and intra-item associative recognition. The extent of the recognition deficit should be a function of how well familiarity memory can support recognition. The patient YR, who had relatively selective hippocampal lesions, showed the pattern of recognition that is predicted by this view. YR's item familiarity was normal [12] and her performance on
Functional imaging studies in humans
If the MTLs create (encode) and reactivate (retrieve) associative memory representations, encoding leading to subsequent memory should involve the same MTL sites as retrieval, unless the MTL itself produces memory feelings [33] or the information that is encoded and remembered is not the same. Several studies have shown that encoding that leads to item familiarity activates the perirhinal cortex (e.g. Refs 34, 35) and that familiarity itself deactivates this region (e.g. Refs 36, 37, 38), but
Concluding comments
Currently, psychological, lesion and functional imaging evidence is broadly consistent with the DD view, but no studies have directly contrasted the predictions of the different views. A central set of issues concerns whether within-domain associations support associative familiarity more than between-domain associations do, and how much this is influenced by encoding instructions and retrieval test formats. How much familiarity for different kinds of unitized and non-unitized association is
Acknowledgements
We thank Christine Denby for her contributions to the preparation of Figure 1 and Wael El-Deredy, Donna Lloyd, Ellen Poliakoff and Dimitris Tsivilis for their helpful comments on the revised draft.
References (80)
The nature of recollection and familiarity: a review of 30 years of research
J. Mem. Lang.
(2002)The human hippocampus and spatial and episodic memory
Neuron
(2002)The contribution of familiarity to associative memory in amnesia
Neuropsychologia
(2006)Recognition memory for single items and for associations in amnesic patients
Neuropsychologia
(2004)Recognition memory and the human hippocampus
Neuron
(2003)The hippocampus supports both the recollection and the familiarity components of recognition memory
Neuron
(2006)Sparing of the familiarity component of recognition memory in a patient with hippocampal pathology
Neuropsychologia
(2005)Dissociable correlates of recollection and familiarity within the medial temporal lobes
Neuropsychologia
(2004)Memory strength and repetition suppression: multimodal imaging of medial temporal cortical contributions to recognition
Neuron
(2005)- et al.
Encoding activity in anterior medial temporal lobe supports subsequent associative recognition
Neuroimage
(2004)