Prospective memory involves remembering to remember - in other words, successfully executing a planned intention after having completed an unrelated task. If computational models of prospective memory (PM) are to be believed, then PM relies on many of the same mechanisms involved in a huge variety of other tasks. For example, the successful execution of a planned action should rely on the intention being strongly represented in the first place, as well as good retrieval of the intention from memory. In turn, these should both relate to the strength of active maintenance processes in prefrontal cortex, which may be enhanced if the planned action is highly salient.
Successful prospective memory will also tend to rely on the vigilant monitoring and ultimately detection of cues that are relevant to the planned action. This capacity may relate to the strength of prefrontal monitoring processes, which may themselves be enhanced if the prospective memory cues are highly distinct from the planned action or to intervening actions. Similarly, successful prospective memory may be less likely if attentional resources are deployed in intervening tasks.
According to this perspective, individual differences in prospective memory should be related to variance in maintenance/monitoring processes as well as in retrieval from long-term memory. In a 2003 Psychophysiology article, West et al. pursue a similar hypothesis by examining the electrophysiological correlates of prospective memory using event-related potentials (ERPs).
West et al. first reviewed previous research on the electrophysiological correlates of PM, which suggested that an ERP component known as the N300 reflects prospective memory cue detection. In contrast, a late parietal ERP component (the "LPC") has been associated with the detection of PM cues when they are associated with an intention. The authors note that the LPC seems similar in many ways to the P3 ERP component that is often recorded in oddball tasks, where subjects must provide a response to deviant or unusual stimuli. This will be important later.
Methodological details in italics:
West et al. then presented 24 subjects with 1,200 word pairs, each of which subjects classified as related or unrelated. On 95% of trials, either one or both words were presented in a gray-colored font - in either case, subjects were told to perform the relatedness judgment as usual. In the remaining 5% of trials, both words in the pair appeared in cyan, green or yellow, and subjects had to give a different response to those word pairs. For 30 trials subjects merely had to press "C" in response to those colored word pairs, whereas in another 30 trials subjects had to press C for cyan, G for green, or Y for yellow. Scalp electrical evoked potentials were simultaneously recorded from 45 electrodes, while subjects completed this task.
The results showed that the alternate trials (the 5% on which a relatedness judgment was not required) were associated with both a N300 and an LPC wave, relative to the other 95% of trials. However, neither wave was affected by whether subjects responded to all colored-word pairs with the same key, or with color-appropriate keys - therefore, West et al. suggested that these components may reflect the detection of a prospective memory cue. In a second experiment, the authors made PM cue detection more difficult by varying the background color of the screen in each trial - accordingly, the N300 and early components of the LPC were smaller in size than in the first experiment. In contrast, a later component of the LPC was not sensitive to this cue detection manipulation, but actually was sensitive to the number of intentions in both the first and second experiments, implicating it in retrieval processes.
West et al. then repeated the second experiment, but did not require subjects to respond at all to word pairs where either or both words appeared in gray. In this experiment, the early part of the LPC did not differentiate between the two infrequent trial types, whereas the late part of the LPC distinguished between all three trial types similarly across conditions (i.e., it was not affected by changing the background color or the number of intentions).
In summary, these results demonstrate that there are two parts to prospective memory: a cue detection component and an intention-retrieval component. Accordingly, varying the distinctiveness of the cues affected the latency of responding to all trial types, suggesting that this manipulation slowed the "cue-detection" or monitoring component. Simiarly, this manipulation affected the earlier parts of the LPC, thought to correspond to the P3a component of oddball tasks, which is seen in response to novel or infrequent stimuli (and may also index a monitoring process).
Conversely, varying the number of planned intentions had an effect only on the latency of responding to "cue target" trials - possibly by way of increasing the demands on the "intention retrieval" component of prospective memory. Consistent with this interpretation, this manipulation affected the later parts of the LPC, thought to correspond to the P3b component of oddball tasks, which is seen in response to novel but task-relevant items (i.e., only to those infrequent items that require a response).
Donders incarnate in that Method paragraph and two Results paragraphs! Nice experiment and coverage of it!