An overarching aim of this thesis is to explore the question: Can event-related potential (ERP) data refine theories concerning the operation(s) supporting episodic recognition memory? More specifically: is one, or are two, memory access process(es) integral to recognition? Global Matching and Bayesian models posit that a single, context-dependent memory access process can support single-item episodic recognition. Global Matching models assume that memory access involves a matching operation (between memory cue[s] and trace[s]), the decision being based on the strength of that operation. Dual-process theorists assume that two processes are fundamentally involved: a rapid, automatic familiarity assessment, and recollection, a strategic, context-dependent retrieval process akin to recall (although distinct versions differ in important ways). Can ERP data help to discriminate between these alternatives?
ERP 'old/new' effects comprise a relative amplitude positivity in ERPs evoked by items that have been presented within an experiment (old), relative to those evoked by new items; They evidently incorporate at least two ERP components: the N400 and the LPC (late positive complex). These effects have been interpreted within the dual-process framework: An N400 effect has been proposed to index familiarity, and the LPC effect, recollection. The former has been largest at frontal electrodes, and termed the 'FN400' However, those interpretations should be considered speculative for a number of reasons, including: marked differences between various authors' notions of familiarity or recollection; a reliance on assumptions which have subsequently been shown not to hold, and; in some cases, a lack of sufficient discriminatory power. Hence the single-process assumption that the recognition decision is based on the strength of a context-dependent memory access process, warrants further consideration in the interpretation of ERP old/new effects.
In the two experiments reported in Chapter Two, strength was manipulated at study: Test lists comprised equal numbers of words that had been presented thrice (strong), or once (weak), and of new, unstudied words. In both experiments, N400 effect amplitude indexed strength in a graded manner (i.e., strong>weak>new): This effect was largest at parietal electrodes (left>right; in ear-referenced ERPs); And no corresponding FN400 effect was obtained. ERPs were also computed according to both strength and the recognition decision: LPC effect amplitude indexed recognition decision accuracy (i.e., correct>incorrect), and exhibited a centro-parietal scalp distribution (left>right).
On the basis of data reported in Chapter Two, it was not possible to detemine whether or not the N400 effect reflects the outcome of a context-dependent (versus -independent) memory access process. Hence the principal aim of the experiments reported in Chapter Three was to address this issue, and also to provide further converging evidence for the above proposals. Hence normative frequency of words' occurrence in the language - which is known to reliably affect performance on various tasks, including recognition - and strength, were both manipulated, and recognition was tested in the first experiment. The N400 strength effect was obtained in low, but not high, frequency words' ERPs: This may constitute an ERP correlate of the mirror effect for word frequency. Across six strength-by-frequency conditions, LPC amplitude closely resembled the recognition performance data: This is consistent with the view that this effect indexes decisional factor(s) including accuracy.
The design of the next experiment was identical, except that normative frequency judgments were required at test. N400 amplitudes were positive in high versus low, frequency words' ERPs; Whereas this difference was in the opposite direction (although non-significant) in recognition ERPs. Considering the different contextual cue(s) applied to memory between the two tasks, the expected consequent strength levels (for a given level of normative frequency, in a given task) correspond to these N400 patterns. Hence these outcomes are consistent with the view that the N400 strength effect indexes the output of a context-dependent memory access process; Such as that posited by Global Matching and Bayesian models. In recognition ERPs, LPC amplitudes were greater for low versus high frequency words, but this difference was reversed (although non-significant) in frequency judgment ERPs. Given corresponding differences in the performance data, these results are consistent with the view that the LPC effect is sensitive to decisional factor(s) including accuracy. Hence, the proposals reported in Chapter Two were supported and extended, by the data reported in Chapter Three.
The conclusions summarised above indicate that the N400 and LPC effects - like the relevant behavioural effects - can be explained within the framework of single-process theories. Hence the current research has provided electrophysiological data that converge with existing empirical and modelling approaches, which posit that a single, context-dependent memory access process is integral to single-item episodic recognition. More generally. this research exemplifies the utility of ERP data to effectively constrain memory theory.