Characterising and measuring human episodic memory
Harlow, Iain Malcolm
MetadataShow full item record
Episodic memory, the ability to store and retrieve information from our past, is at the very heart of human experience, underpinning our identity and relationship with the world. Episodic memory is not a unitary phenomenon: in dual-process theory, researchers draw a distinction between familiarity, a rapid and automatic sense of oldness to a previously encountered stimulus ("I know that face"), and recollection, the reactivation of additional context from a particular episode ("We met at the York conference"). A fundamental objective in the study of human memory is to ground recollection and familiarity in neural terms. This requires accurately measuring the contribution of each from behavioural data, which in turn relies on an accurate characterisation of recollection. This thesis introduces a novel source retrieval task to demonstrate that recollection has two critical, and fiercely contested, properties: it is thresholded, i.e. it can fail completely, and successful recollection is graded, i.e. it varies in strength. The consequences of this characterisation are explored. Firstly, familiarity and recollection are functionally separable retrieval mechanisms. Secondly, the models currently used to measure the contribution of each are generally flawed, and a corrected model is described which better fits, and explains, the extant data. Finally, the frequency of recollection is shown to be dissociable from its strength, a result which links behavioural data more strongly than before to a neurocomputational account of episodic memory, and which suggests a relationship between the representational overlap of memory traces and their retrieval. This thesis necessitates a change in the way behavioural memory data is modelled, and consequently the interpretation of evidence underpinning neuroanatomical accounts of memory experience. Significantly, however, it also moves the field beyond a long-running debate and provides a deeper dual-process framework with which to address outstanding questions about the relationship between, and neural basis of, episodic memory processes.