Empirical Support For The Pam Theory

Much of the supporting evidence for the PAM theory comes from examining performance on the ongoing activity. A second line of support comes from the application of multinomial modeling techniques. With respect to the former, consider that the PAM theory proposes that successful prospective memory performance requires resource-demanding preparatory attentional processes. If we assume that an individual has a limited pool of conscious capacity, and if the individual assigns some of these resources to engaging in preparatory, attentional processes, this leaves fewer resources for the ongoing task. If the ongoing task is sufficiently resource demanding, it should be affected negatively by the addition of the prospective memory task. We can evaluate this specific prediction by comparing performance on the ongoing task when the ongoing task is performed alone with performance on the ongoing task when the prospective memory task is embedded in the ongoing task. The preparatory attentional processes must be engaged prior to the retrospective recognition processes, so the preparatory attentional processes will be engaged on nontarget trials as well as target trials. Thus, evidence for preparatory attentional processes should be found on nontarget trials.

A number of studies have demonstrated a cost to ongoing activities. Smith (2003) found that participants who had to perform a prospective memory task that was embedded in an ongoing lexical decision task were 200 to 300 ms slower on non-prospective-memory-target trials than were participants in a control condition who performed the lexical decision task alone. A cost to ongoing activities has been shown in a number of different laboratories using different ongoing tasks, different target events, and different age groups (Einstein et al., 2005; Loft & Yeo, in press; Marsh et al., 2005, 2006; Marsh, Hicks, Cook, Hansen, & Pallos, 2003; Smith, 2003; Smith & Bayen, 2004, 2006; Smith et al., 2007, under review; West, Krompinger, & Bowry, 2005). West et al. (2005) demonstrated that the cost was found on trials that preceded successful prospective memory hits, but not prospective memory misses, indicating that the cost associated with the prospective memory task is caused by functionally important processes.

Although many of these demonstrations do not distinguish between the multiprocess framework and the PAM theory, recent work by Smith et al., 2007; see also Smith, 2001) does provide a direct comparison of these two explanations. As mentioned earlier, the multiprocess view proposes that when prospective memory tasks involve salient target events, a strong association between the target and the action, a target that is processed in the course of completing the ongoing task, and a simple action, the intention will be retrieved automatically. Smith et al. found evidence of a cost to ongoing activities when using a task that meets these criteria. For instance, when participants are asked to respond to their own name (a very salient target), their performance on an ongoing lexical decision task was poorer in comparison to participants in a control condition who performed only the lexical decision task. In a different experiment conducted by Smith et al., participants performed an ongoing color-matching task (Smith & Bayen, 2004, 2006). All of the participants performed two blocks of 62 trials of a color-matching task that involved seeing four colored rectangles followed by a string of Xs. Their task was to decide if the color in which the Xs were displayed matched one of the four colored rectangles presented on that trial. The colors used were red, green, blue, yellow, and white rectangles and X strings on a black background. Half of the participants, those in the control condition, performed the color-matching task only, and the other half performed a prospective memory task in the second block of trials. The prospective memory task was to press the P key if they saw a pink string during the color-matching task. The occurrence of the pink string was salient: The color change was spontaneously detected by participants in the control condition, despite not having been told about the appearance of an unusual color. This task simultaneously meets all of the criteria outlined by the multiprocess framework for a prospective memory task to be considered automatic: The action is simple (press a key), the action and target are well associated (press the P key for a pink string), the target is salient, and the dimension along which the target was defined (color) was processed as part of the focal processing engaged by the ongoing task. Thus, the multiprocess view would not predict a cost to ongoing task performance in this case. In contrast, the PAM theory would predict a cost to the ongoing task.

Performance on the ongoing task was evaluated for the five trials preceding each trial (the six prospective memory target trials and the four trials following each target were excluded from the analysis). There was no difference in baseline (block 1 before the prospective memory instructions were given) reaction time or accuracy on the color-matching task between the two conditions. Each individual's baseline reaction time was subtracted from his or her reaction time in block 2. This difference score was evaluated to determine whether the addition of the prospective memory task affected the ongoing task. The control condition showed a significant practice effect; on average this group was over 100 ms faster in the second block than in the first block. In contrast, the prospective memory group failed to show a practice effect and importantly, the difference scores for the prospective memory group were significantly related to performance on the prospective memory task, with larger difference scores (i.e., a greater cost) associated with better performance. In other words, resource-demanding processes were being engaged even on non-prospective-memory-target trials, and these processes are functionally related to success on the prospective memory task.

Why did we find a cost in a task that simultaneously met all of the criteria for automaticity, whereas in some cases others have not found a significant cost using single targets? Marsh et al. (2003) failed to find a significant cost in one condition, but the difference was in the direction of a cost. Einstein et al. (2005) did demonstrate a significant cost in all experiments in which a control condition was compared with a prospective memory condition, but they failed to find a significant cost in three out of nine comparisons of control and prospective memory conditions. In two of these three nonsignificant cases, the direction was consistent with a cost. It is possible that a lack of power contributed to the nonsignificant findings. As demonstrated in the Smith, et al. (2007) study, the cost appeared as a lack of a practice effect for the prospective memory condition, whereas a practice effect was demonstrated in the control condition. This is a cost nonetheless, but clearly indicates that the resource demands of preparatory attentional processing can be very subtle. In the Einstein et al. (2005) study, the power to detect small effects was less than .13 in all cases, and never better than .6 for detecting even medium-sized effects. Highlighting power as a potential factor, a comparison in Experiment 3 of Einstein et al. that failed to reach significance with 32 participants was significant in Experiment 4 using the identical procedures, but including 104 participants.

Importantly, the control and prospective memory conditions in Einstein et al. (2005) were always manipulated within subjects in a counterbalanced design. The choice to use a within-subjects control condition in a counterbalanced design, as opposed to using both a within-subjects baseline block that always preceded the prospective memory block combined with a true control condition as used in Smith, et al. (2007), could have contributed to the different outcomes. When using the within-subjects counterbalanced control condition, for half of the participants the control block comes after the prospective memory condition. This could interfere with their performance on the control block, despite the fact that they are told that they do not have to perform the prospective memory task during the control block. Guynn (2005) demonstrated just such a side effect: Participants who were given prospective memory instructions prior to the control block were slower on the control trials than participants who received the prospective memory instructions after the control block, even though the participants knew that they did not need to perform the prospective memory task during the control block. Thus, the particular procedures used could have contributed to the lack of a significant cost in Einstein et al. (2005).

Although there have been four failures to find a significant cost (one case in Marsh et al., 2003; three cases in Einstein et al., 2005), on the whole, the results of the effect of a prospective memory task on ongoing task performance support the proposal made by the PAM theory that resource-demanding preparatory atten-tional processes are involved in successful prospective memory performance. In particular, when avoiding potential problems associated with a counterbalanced design for control blocks and while using a task that meets all of the multiprocess framework's criteria for automaticity, performance on a prospective memory task is associated with a cost to ongoing task performance (Smith, et al., 2007).

Evidence regarding a cost to the ongoing task is encouraging but limited in a number of different ways. First, the cost measures provide only an indirect indicator that participants are engaging in preparatory attentional processes. Cost to ongoing tasks could be caused in part by factors such as rehearsal in addition to the basic preparatory attentional processes. Second, given that the cost is most often found on reaction time measures (often with relatively few observations per participant), issues such as noise, baseline differences, and insufficient power can complicate interpretations of these findings, particularly in the case of small differences. Fortunately, many prospective memory tasks lend themselves to multinomial modeling, which provides converging evidence for evaluating the PAM theory.

Smith and Bayen (2004, 2005, 2006) introduced a multinomial model of prospective memory. The model provides estimates of the contributions of underlying cognitive processes, specifically preparatory attentional processes and retrospective recognition processes, to observable prospective memory performance. The model has been shown to be identifiable and has been validated (Smith & Bayen, 2004). With respect to what sorts of variables would be expected to influence the estimates of the underlying cognitive processes, the retrospective recognition processes should be influenced by variables that influence recognition, such as the discriminability of targets and nontargets and the amount of time allowed for target encoding (Smith & Bayen, 2004). Other variables that influence recognition performance would also be expected to influence the parameter estimates of retrospective memory in many cases. However, the small number of target events generally used in prospective memory makes predictions based on findings in the typical retrospective recognition tasks, employing many more target events, complicated.

Factors that influence the availability of cognitive resources or the allocation of resources would also be expected to influence the likelihood of engaging in preparatory attentional processes. Motivation, task importance, expectations about when preparatory attentional processes should be engaged or how difficult a task will be, and reminders could all potentially affect preparatory attentional processes. Task importance and task context have both been shown to affect estimates of preparatory attentional processes (Smith & Bayen, 2004). Using the model, Smith and Bayen (2005) showed that preparatory attentional processing is sensitive to differences in working memory capacity and working memory load, as would be expected if preparatory attentional processes are resource demanding. In addition, Smith and Bayen (2006) found that adult age differences in prospective memory performance were driven by age differences in preparatory atten-tional processes, also as would be expected if preparatory attentional processes are resource demanding, given evidence of reduced resources for older adults (Zacks, Hasher, & Li, 2000). As a brief aside, one might wonder about the demonstration of age differences in the resource-demanding preparatory attentional processes in light of demonstrations of no adult age differences on naturalistic tasks. It could well be that in noncontrolled settings, the older adults are devoting a larger component of their cognitive resources to engaging in preparatory attentional processing and thereby performing as well as the younger adults. We do not know exactly how it is that the tasks are being accomplished, nor how this might affect other ongoing activities, in these naturalistic tasks. The fact that we find age differences in preparatory processes in a relatively demanding laboratory task does not rule out the possibilities that under the right circumstances older adults could choose or be encouraged to redirect sufficient resources to the preparatory attentional processes to increase their performance to a level matching that seen with younger adults. The results of the Smith and Bayen (2006) study are consistent with other work showing no age differences in recall of the action and failure to maintain intentions (Cohen, Dixon, Lindsay, & Masson, 2003; Cohen, West, & Craik, 2001; West & Craik, 1999; West, Murphy, Armilio, Craik, & Stuss, 2002; see Smith & Bayen, 2006, for discussion). The results of Smith and Bayen also converge with Kliegel et al. (2000), in which no age differences were found in plan retention, albeit Kliegel et al. employed a much more complicated task context.

+1 -1

Post a comment