In the current experiment, the contribution of recollection and familiarity were examined by measuring their closely associated neural correlates – i.e., the Left Parietal and Mid-Frontal old/new effects respectively. Here, we discuss each correlate in turn, beginning with the Mid-Frontal old/new effect. Although the experiment was a test of recollection, it was possible that the behavioural expression of recollection may actually reflect the contribution of familiarity – widely believed to be a continuous process (Murdock, 1974; Gillund & Shiffrin, 1984; Mandler, 1980; Yonelinas, 1994, 2001a, 2001b). Although the source task was designed by Harlow and Donaldson (2013) to limit the contribution of familiarity, they did not provide any direct test of this assumption, relying simply on the requirement to make source discrimination to isolate recollection. In the current experiment however, the contribution of familiarity was likely given the presence of an old/new item recognition task. Analysis of the data, however, revealed that although Mid-Frontal old/new effects were present, the magnitude of the effect did not significantly differ between response categories.
Although caution should always be taken when interpreting a null result, the pattern of ERP data taken together with an experimental paradigm that discourages familiarity for source information supports the assumption made by Harlow and Donaldson (2013) that source accuracy mainly reflected recollection.
168 6.4.2. Overview of the Left Parietal old/new effect
The primary aim of the current experiment was to test between the Graded and All-or-None accounts of the Left Parietal old/new effect. As expected, analysis of the ERP data revealed significant Left Parietal old/new effects for all three response categories within the 500-800ms time window, reflecting the contribution of recollection. The pattern of data revealed a graded pattern, with the magnitude of the Left Parietal old/new effect being larger for High precision responses, reducing in size for Low precision responses and smallest for Guessed responses. Statistical analysis of the data partially confirmed that a graded pattern was present. To be clear, the Left Parietal old/new effect was statistically larger for High precision responses compared to Guessed responses, but not between High and Low, or Low and Guessed, responses.
The pattern of Left Parietal old/new effects is difficult to interpret. In one instance, the current pattern of Left Parietal old/new effects may reflect the nature of the underlying neural signal supporting recollection. Perhaps the recollection signal is only significantly different from guessing when trials are recollected with fairly high precision – resulting in a significantly large Left Parietal old/new effect. Alternatively, recollection associated with lower precision trials may be more variable than high precision trials resulting in a signal that is both indistinguishable from activity elicited by either high and guess responses. How this pattern of data fits with the existing models of recollection is however, unclear. For example, it is possible that the neural signal supporting recollection operates in an all-or-none fashion. By this account, no differences were observed between High and Low responses because both exceed a recollection threshold. The absence of any significant differences in the size of the Left
169 Parietal old/new effect between Low and Guessed responses, however, is difficult to reconcile with an all-or-none account – i.e., the magnitude of the Left Parietal old/new effect was not statistically different for Low responses that reflect above-thresholded responses, and Guessed responses that reflect sub-thresholded responses.
Alternatively, it is possible that the Left Parietal old/new effect is in fact graded, but the experimental paradigm was not sensitive enough to detect the differences between High, Low, and Guessed response categories. The inclusion of an old/new task, for example, meant that potential changes in recollection due to source accuracy were confounded with recollection for ‘old’ items. Although the old/new task facilitates comparisons with previous ERP source experiments, it more than likely introduces additional non-criterial recollection of information not related to the source task.
Further, the nature of the adapted source recollection task also meant that participants did not provide an equal number of source judgements (as a result of the initial old/new recognition task), resulting in large variability of response frequencies. More specifically, the difference among individual thresholds among participants, for example, may have resulted in a significant proportion of recollected trials being incorporated into the guess bin – particularly for those participants with low thresholds (i.e., those thresholds above 90°). These recollected trials may have resulted in a larger left parietal old/new effect than would be expected if the bin only reflected non-criterial recollection associated with the old/new task. In short, the ERP results reveal that the underlying neural signal associated with recollection was present and was larger for High compared to Guessed responses but the predicted graded pattern did not reach statistical significance.
170 6.4.3. Summary
The aim of the current experiment was to test between the Graded and All-or-None accounts of the Left Parietal old/new effect during a novel source task. The behavioural data was entirely consistent with the finding that the behavioural expression of recollection is both variable and thresholded. The ERP data, by contrast, failed to differentiate between a Graded and All-or-None account of the underlying neural mechanism supporting recollection. Although the graded Left Parietal old/new effect observed in the current study did not reach significance, the pattern of data suggests that the underlying neural mechanism may be sensitive to positional response accuracy.
Given the possible insensitivity of the current task to detect a graded pattern in the ERP data, it was decided to remove the old/new task and provide a more direct measurement of retrieval success. The results of this experiment are reported in the next chapter.
171