Keywords

Prospective Memory

Prospective memory (PM) refers to tasks in which one must remember to carry out an intended action at some point in the future. Good PM is vital in everyday life (McDaniel and Einstein 2007), whether remembering an appointment, paying a bill, or taking a prescription. While PM is important for everyone, the consequences of failure can be much greater for older adults. A missed doctor’s appointment or a forgotten pill can have dire repercussions. In addition, older adults complain most about PM failures compared to other memory issues (McDaniel and Einstein), and PM ability declines with age, at least for some types of PM (for a review, see Henry et al. 2004). Given the potential beneficial impact, PM is an ideal target for training, especially in older adults. Yet, very few cognitive training programs in general, or specifically for older adults, have attempted to train PM (see Waldum et al. 2016, for review). Here, we first discuss the theoretical approach—including what to train and how to train it—that has guided our attempts to train PM. We then provide evidence from existing data and current preliminary work supporting and informing this approach.

Theoretical Approach

The fundamental goal in developing a training protocol for PM and a standard goal in cognitive training is to promote transfer or generalization beyond the context of learning. That is, training that learners undergo should be helpful beyond the laboratory and applicable in the real world (see also Guye et al., Karbach and Kray, Könen et al., Rueda et al., Strobach and Schubert, Swaminathan and Schellenberg, this volume). However, transfer following cognitive training has been elusive (see Hertzog et al. 2009; McDaniel and Bugg 2012). With this challenge in mind, our broad approach is to look at existing literature and focus on identifying effective PM strategies that learners can be explicitly taught to apply and generalize more broadly. This is a somewhat innovative approach as other cognitive training protocols have embraced different underlying assumptions. For example, some cognitive training has taken a restorative approach, attempting to enhance the underlying neural physiology to improve cognition (see Lustig et al. 2009, for review; Taatgen, Wenger et al., this volume). Other cognitive training programs include only practice of relevant tasks rather than explicit instruction on how to approach them (e.g., for attentional control: Karbach and Kray 2009; Kramer et al. 1995; Mackay-Brandt 2011; for retrospective-memory control: Jennings and Jacoby 2003; for working memory: Harrison et al. 2013; Redick et al. 2013; see Könen et al., this volume). Even one of the very few training programs aimed at improving PM used only practice and was only somewhat successful in producing transfer (Rose et al. 2015). In contrast, rather than attempting to modify the nervous system or rely on learners gaining spontaneous insights into how to best handle PM tasks through repetitive practice, our approach is to teach effective, efficient strategies with which learners can tackle PM tasks.

We adopted such an approach for several interrelated reasons. First, the PM literature has revealed that dissociated processes underlie different PM tasks (described below), as opposed to perhaps more unitary skills (tasks) that seem to submit to restorative or practice-alone regimens (e.g., attentional control and working memory). Second, PM strategies have been identified that we assume are directly useful in everyday PM tasks (unlike some trained retrospective memory strategies; cf. McDaniel and Bugg 2012). Of note is that PM in the laboratory is quite different than PM in the real world. PM tasks that are encountered in everyday life are widely variable and occur in a myriad of contexts; for example, they include remembering to put a rent check in the mail every month, remembering to pick up a friend at the airport, and remembering to give a housemate a message. By contrast, laboratory PM tasks involve remembering to press a particular key when a given target appears (e.g., the word president or the syllable tor) during an ongoing task (e.g., answering trivia questions; Einstein et al. 1995). Thus, a challenge for a PM training program is creating strong connections between the laboratory training context and the situations learners are faced with in their daily lives (see also Guye et al., this volume). Because practice alone can produce brittle skills that are tightly tied to training (e.g., Healy et al. 2005), we felt that appropriately selected strategies and training could better allow learners to link the laboratory context to everyday PM situations. In fact, Bottiroli et al. (2013) found benefits of a strategy approach for promoting transfer—on retrospective memory tasks—specifically with older adults (see also Wenger et al., this volume; PRIMs Theory in Taatgen, this volume). Third, available evidence suggested that these strategies might help override age-related cognitive limitations that attenuate PM performance for older adults (e.g., Liu and Park 2004). In sum, for PM our aim has been to create and test a cognitive training intervention that is applicable for improving PM in the real world and teaches learners effective practical strategies informed by the basic PM literature.

Despite little work on training PM, the broader PM literature indicates a number of strategies that learners could use to improve their PM. As just noted, there are different types of PM that rely on different processes (McDaniel and Einstein 2007), and accordingly are associated with different effective strategies. Focal PM tasks involve cues that are presented in the focus of attention and thus are easy to recognize as a cue for performing the related task. For example, seeing a coworker in the hallway can act as a focal cue to give that person a message. In other words, simply seeing that coworker might automatically bring to mind the PM task of relaying the message. Because PM intentions like this are associated with focal cues that can stimulate spontaneous retrieval of the intention, they can be performed without actively looking for the cue. Previous research indicates that creating a strong association between the anticipated cue and the PM intention (an implementation intention strategy taking the general form, “When X occurs, I will remember to perform Y”) can improve performance on focal tasks (e.g., McDaniel and Scullin 2010). This strategy has been explored more broadly and shows effectiveness beyond healthy aging: In a subsequent section, we report recent research with Parkinson’s disease (PD) patients that train an implementation-intention strategy.

In contrast, nonfocal tasks involve cues that occur outside the focus of attention and are therefore more difficult to notice. For instance, one may need to stop at the grocery store after work, but the store itself is not easy to notice in the midst of a routine drive home where one must pay attention to traffic, etc. Here, actively monitoring for the cue is needed in order to successfully notice (Einstein et al. 2005), otherwise one might drive right by the store. The implementation intention strategy that is effective for focal tasks would not be as helpful in nonfocal PM tasks since the key is to notice the cue in the first place (Breneiser 2007). Thus, the best strategy for nonfocal tasks may be to simply check for the cue frequently and actively attend to that intention (an event monitoring strategy; see also Wenger et al., this volume).

Similarly, time-based PM tasks, wherein an intended action must occur at a particular time, require this type of active monitoring. Furthermore, the only cue is the time itself, whereas in focal and nonfocal tasks, events are the cues. This type of task is especially challenging for older adults (Einstein et al. 1995). Prior work indicates that learners who check the clock more often as the target time nears perform intended actions more frequently (Einstein et al.). Consistent with this finding, older adults are less likely than younger adults to ramp up their monitoring as the target time approaches (Einstein et al.; Park et al. 1997). Teaching older adults to use this strategic clock-checking may be the most effective strategy for improving their performance on time-based tasks.

Beyond the specific strategies to teach older adults, an important question is how to implement the training. In what form should these strategies be taught such that older adults learn them well and learn to apply them outside the context of learning? Several key factors may be critical for designing the most beneficial training program.

Key Factors for Training: The EXACT Study

As part of a larger cognitive training and aerobic exercise program (EXACT; McDaniel et al. 2014), McDaniel and colleagues developed a protocol specifically aimed at improving PM through strategy use (Waldum et al. 2016 describe this protocol in detail; see also Pothier and Bherer, this volume). Five main components were implemented in an 8-week intensive intervention. First, learners were given explicit instructions about effective strategies to use in PM tasks, specifically tailored for each type of task. Second, both to increase the generalizability of training and capitalize on previous memory research, the training context varied greatly. In terms of generalizability, as mentioned above, PM tasks are widely variable, both in task type (focal, non-focal, and time-based) and in context. Accordingly, learners were trained using several on-going tasks that tapped different types of PM. Encountering various scenarios during training might make learners’ approach more flexible and resilient in the face of new PM challenges. Additionally, learners may start to be able to identify the different types of PM tasks and then transfer the appropriate strategies accordingly. This line of reasoning is also consistent with memory research on encoding variability wherein multiple contexts at the learning stage can improve later memory for the to-be-remembered material (Hintzman and Stern 1978).

Third, combined with the wide variety of laboratory tasks, homework was added to the program. That is, learners were given assignments to complete outside the laboratory regarding PM situations they faced in daily life. Explicit practice applying the training they received in the lab to their regular lives is likely to be beneficial for later transfer (e.g., Schmidt et al. 2001). Fourth, as the training program went on, the difficulty of the tasks increased. Learners were asked to keep in mind more PM objectives, and the nature of the tasks also became more challenging. Simultaneously, the trainer’s involvement decreased from initially providing explicit strategy instruction prior to each training task to later expecting the learners to use the relevant strategies without prompting. This idea of increasing the difficulty across the training program is consistent with the broader literature on cognitive training. In the restorative approach, the demands of the task are incrementally increased to push the ultimate level of acquisition of the trained skill (e.g., retrospective memory training: Jennings and Jacoby 2003; attentional training: Mackay-Brandt 2011). Additionally, in the occupational therapy domain, strategies are trained such that learners are required to initiate and apply the strategies across activities that systematically differ in physical similarity and context but remain at the same level of complexity. In this sideways approach, task difficulty is only increased after strategy transfer has been observed (Toglia 2011). Again, intervention is designed to encourage transfer and generalize the training to learners’ everyday lives.

Fifth, a key component of the EXACT project was to evaluate the training effects with computer simulations of cognitively challenging real-world tasks (e.g., cooking breakfast, Craik and Bialystok 2006; remembering health-related information and the sources of that information). To evaluate the PM training effects, older adults completed (pre- and posttraining) a simulation of going through the course of a day for three successive days (the Virtual Week task; Rose et al. 2010). During the course of each day, the older adults have to remember a number of prospective tasks, such as “remember to drop off dry cleaning when you go shopping” and “remember to take asthma medication at 11 a.m. and 9 p.m.” (in the game, a person’s token passes squares that indicate the virtual time for the day).

The results of the EXACT study were especially encouraging with regard to training PM (see McDaniel et al. 2014). Eight weeks of cognitive training on laboratory PM tasks with the components discussed above produced significant gains (from pre to posttests) in remembering to perform the real-world Virtual Week PM tasks relative to a control that did not receive PM training or an aerobic exercise control (a real clock, time-based task did not show training effects). By contrast, cognitive training did not produce significant gains for cooking breakfast or memory for health information tasks. However, the EXACT study was not designed to isolate the impact of particular training components to the success of the training protocol for improving PM; accordingly, many basic issues remain unanswered (see Waldum et al. 2016, for detailed discussion).

Briefly, the cognitive training included attentional control training tasks and retrospective training tasks in addition to PM training; thus, though plausible, it remains uncertain that the PM training alone would be sufficient to produce transfer to the ecologically valid VW tasks. Also, the PM training protocol included a number of components—including using a different laboratory task each week (variable training) and explicit strategy instruction—either or both theoretically could have been instrumental in promoting transfer. Initial support for the value of these components comes from noting that in the EXACT protocol, the attentional control training, and the retrospective memory training, following the precedent from the literature, generally did not include explicit strategy instruction and repeatedly used the same training task over the course of 8 weeks. As just mentioned, there was no significant transfer of training to the real-world attentional control task (cooking breakfast) or to the real-world retrospective memory task (memory for health information). Clearly, experiments that directly compare variable training (varying the parameters of the practice task, rather than keeping it constant; e.g., Kerr and Booth 1978; Goode et al. 2008) to single-task training and directly compare explicit strategy instruction with a typical practice-only procedure (e.g., Kramer et al. 1995; Jennings and Jacoby 2003) would provide valuable insights as to the importance of these factors in promoting the generalizability of cognitive training.

Finally, a feature of the EXACT project that poses practical limitations is that the cognitive training was a huge undertaking, requiring a great deal of commitment and investment from the trainers and the learners. A major practical issue is whether a more efficient training program focusing on PM per se and restricting training to one session (rather than multiple sessions as in EXACT) could support transfer of strategies to real-world PM tasks. Initial studies have reported significant improvements with older adults in everyday-like PM tasks using a brief implementation-intention instruction for the target PM task (Liu and Park 2004, with healthy adults; Shelton et al. 2016, with older adults with mild Alzheimer’s disease; see also, Lee et al. 2016, for effective implementation-intention use with AD patients). Accordingly, it seemed possible that a single PM strategy training session could support transfer, and if so, then an efficient and nondemanding training protocol could be provided to older adults to improve their everyday PM success. We tested this possibility in a new experiment, reported next.

An Initial Experiment with Healthy Older Adults

The focus in our new experiment was to evaluate the success of PM training for a single 60–90 minute training session that compared the success of explicit prospective-memory strategy training relative to a practice-only condition and a test-retest control condition. We also included a new prospective-memory strategy training condition that we developed: Guided metacognitive training, described in the next section.

Metacognitive Strategy Training

It may be that neither explicit strategy instruction nor practice alone is most optimal. Instead, guided use of effective strategies that integrates metacognitive components may extend benefits of strategy training by helping a person recognize when and why a particular strategy is applicable and thus increase the probability of generalization (see Schaeffner et al., this volume). Metacognitive strategy training focuses on the general process of how to go about a task, including analyzing task demands, strategy generation and selection, and self-monitoring and self-evaluation of performance (Toglia 2018). A learner-centered approach that actively engages the participant in a collaborative process of planning or choosing strategies and evaluating effectiveness can be integrated with metacognitive strategy training by using systematic questions and guided prompts to facilitate self-generation of strategies (McEwen et al. 2018; Toglia 2018). Learner-centered approaches, such as guided discovery, are rooted in constructivism theories of learning that suggest that learning is enhanced when the learner is actively engaged in the process of discovering solutions themselves (e.g., McDaniel and Schlager 1990).

Preliminary evidence supporting the use of guided metacognitive strategy techniques in enhancing transfer of learning or generalization has been reported for older adults (Bottiroli et al. 2013; Dawson et al. 2014) as well as for cognitive rehabilitation of executive functions in individuals with stroke or brain injury (Skidmore et al. 2014; Toglia et al. 2010). For example, Bottiroli et al. (2013) found that transfer of learning was facilitated in older adults by encouraging active involvement in analyzing memory tasks involving lists, stories, locations, or paired-associates and adapting strategies to meet task demands. Guided metacognitive strategy training, however, has not been applied to PM training. Another important question, therefore, is whether PM strategies are best learned through explicit instruction or through guided metacognitive methods.

In the following experiment, we adapted a guided metacognitive strategy framework described by Toglia (2018) to the training of PM. The framework is outlined in Table 1 and consists of three components: (1) preactivity discussion on analyzing task demands, identifying similarities with meaningful activities, and self-generation of strategies; (2) mediation during the task to facilitate self-monitoring and the use of alternative strategies when needed; and (3) after-task questioning aimed at promoting self-evaluation of performance and strategy use.

Table 1 Guided metacognitive strategy framework for prospective memory training
Full size table

Healthy older adults were assigned to one of four separate experimental groups: metacognitive strategy training, explicit strategy training, practice only, and a no-training control. Approximately 20 participants were assigned to each group (tested at both Washington University in St. Louis and Mercy College). To give some indication of the sample characteristics, participants’ ages ranged from 60 to 90 with a mean age of 69.65, and all participants were living independently in the community. Montreal Cognitive Assessment (MoCA) scores ranged from 18 to 30, with a mean of 26.85 (for 61 out of 81 participants); 20 participants came from a subject pool with preexisting archival data (Knight Alzheimer’s Disease Research Center at Washington University in St. Louis) and did not have MoCA scores, but were all screened as cognitively normal.

To assess PM, the Virtual Week task (VW task, previously described) was administered approximately 1 week before and after a single strategy training session. After completing the pretraining VW task, participants returned 1 week later for the training session (the retest control did not return to the lab at this point). This session included three different computerized PM games, previously described by Waldum et al. (2016), with increasing difficulty across the tasks (focal + time-based, nonfocal + time-based, a combination of focal + nonfocal + time-based).

For the metacognitive strategy group, after a general introduction to types of PM (i.e., time-based, focal, and nonfocal tasks), participants were then presented with PM tasks and asked to identify the type of PM required by the task. Next, guided questioning was used to help the person identify how the PM training task was similar to everyday activities or situations, and the person was given the opportunity to try the PM games using their own methods. During the activity, the examiner stopped and mediated performance as errors occurred and guided the person to reassess the effectiveness of their method. If needed, the person was encouraged to adjust or generate alternative strategies.

For the explicit strategy group, participants were instructed on different strategies depending on task demands (i.e., focal + time-based, nonfocal + time-based, or focal + nonfocal + time-based). The strategy training for focal tasks was to use implementation intention encoding (e.g., “When the focal target X occurs, I will remember to perform Y”) repeated aloud and visualized (see McDaniel and Scullin 2010). The strategy training for time-based tasks encouraged participants to ramp up clock monitoring behavior when approaching the appropriate time (see Einstein et al. 1995). Finally, the strategy trained for nonfocal tasks was active monitoring, which involved trying to maintain a state of active cue-searching (Einstein et al., 2005).

The practice-only condition, after receiving a general introduction to types of PM, received no strategy instructions, and simply practiced the PM tasks during the training session. The control condition received no training. One week after completing the training session, participants completed the VW assessment again. The control completed the pre- and posttest VW assessments separated by 2 weeks.

The proportions of correctly detected PM targets as a function of assessment time (pre and post) and training condition (control, explicit, practice-only, and metacognitive) are shown in Fig. 1. There was a significant increase in scores from pretest (M = 0.49, SE = 0.03) to posttest (M = 0.63, SE = 0.03). However, there was no effect of training approach, nor was there any interaction between the two variables. The explicit and practice-only conditions obtained modest gains from pre- to posttest (0.08 and 0.10, respectively) and the metacognitive group obtained the greatest increase (0.18).

Fig. 1
A grouped column chart with error bars plot proportion correct versus condition. The columns are the tallest for the post-test than the pretest.

Proportion of PM targets detected on Virtual Week from pre- to posttest as a function of training condition

Full size image

This pattern is initially encouraging regarding the benefits of metacognitive training; however, the control group performed surprisingly well, too, also increasing by 0.18 from pre- to posttest. One interpretation is that, due to low sample size, random assignment did not adequately balance individual differences across groups, such that participants in the control group were by chance more able learners compared to those in the other groups. Another interpretation rests on the following feature of the experiment: The pre- and posttest VW versions were identical to one another. Accordingly, it is possible that the increases in performance on VW, for at least the control group, reflected practice of the specific PM tasks encountered during both pre- and posttesting, rather than acquisition of more general PM skills and strategies. We had not expected this improvement on VW in a no-trained control given previous research with repeated administration of VW (e.g., McDaniel et al. 2014); however, that research used intervals of 6 months between pre- and posttesting, not the 2 weeks used here. In retrospect, the experiment could have been more sensitive had we used different versions of VW at pre- and posttesting that incorporated different particular PM tasks.

Nevertheless, two speculative conclusions might be offered. First, the metacognitive strategy training seems more promising for training PM transfer than does practice alone or even explicit strategy training. The second conclusion follows from the observation that the improvement from pre- to posttest in the training groups was not more robust than that displayed in the control group. It may be that a brief one-session training is not sufficient to adequately train PM skills and strategies that significantly transfer. Clearly, these possibilities merit further research.

PM Training in Pathological Aging Older Adults: Evidence from Parkinson’s Disease Patients

Effective training of PM also has important applications beyond healthy aging. Some work has extended findings in healthy aging to attempts to improve PM in pathological aging. Here, we mainly focus on our findings regarding Parkinson’s disease, though work has also been done on very mild Alzheimer’s disease and other forms of dementia (e.g., Burkhard et al. 2014). For example, prior work on older adults with very mild Alzheimer’s disease indicated that a simple implementation intention encoding intervention can improve focal PM performance in both laboratory tasks (Lee et al. 2016) and simulated real-world tasks (the VW task; Lee et al. (2016)). Similar work has been done for those with Parkinson’s disease (PD) because this disease seems to cause PM impairments in forming and remembering intentions (Kliegel et al. 2011).

Foster et al. (2017) studied individuals with mild to moderate PD without dementia on the VW PM task described above. First, participants completed the VW task without any special instructions. Then, a week later participants again performed the VW task. Prior to doing so, half were instructed to form implementation intentions. That is, they were told to create a “When X, I will do Y” statement, repeat it out loud three times, and then visualize performing the task at the correct time in the game. The other half simply repeated the PM tasks out loud three times. Regardless of the instructions, participants improved compared to their initial performance. This was especially true for event-based compared to the time-based tasks. More importantly, the implementation intention strategy training led to better performance than the verbal repetition task when participants completed nonrepeated tasks—tasks that were only presented once during the overall VM task—compared to the ones that were repeated.

These strategies were then extended to self-reports of naturalistic PM experiences. Goedeken et al. (2018) examined PD patients’ experience of PM via the Prospective and Retrospective Memory Questionnaire Prospective Scale (PRMQ-Pro) 1 week before and 1 month after the same two training techniques: implementation intention strategy training and verbal repetition. The training occurred within the context of the VW task, but participants were then instructed to use the strategies as much as possible in their daily lives. Those in the verbal repetition actually showed a decline on the PRMQ-Pro, whereas those in the implementation intention group showed no change. Here, the effectiveness of the implementation intention training seemed to be in preventing decline rather than in improving PM. Of course, a limitation of this work is that it is based on patients’ self-reports rather than actual performance on naturalistic PM tasks. Still, taken together, the findings are heartening in that training strategies can not only be taught and implemented by PD patients but can lead to maintenance of PM, if not even improvements. As progress is made in understanding the mechanisms and strategies for effective improvement of PM for healthy older adults, it appears fruitful to then test these techniques for those with clinical issues.

Conclusions

A unique aspect of our research is the appreciation of different types of PM tasks, with training oriented toward informing learners of these differences and highlighting particular strategies targeted at the different types of tasks. It seems that a parallel approach for retrospective memory training might be considered to improve outcomes for assisting older adults with their everyday retrospective memory challenges (cf. McDaniel and Bugg 2012). However, our new, though preliminary, results suggest that a relatively brief training session may not be enough to produce transfer of learned PM strategies to at least a simulation of real-world PM tasks. At this point, we remain optimistic that the present training approach, with training extended beyond one session, might benefit older (and younger) adults in improving their everyday prospective remembering. Clearly, however, a definitive conclusion awaits more complete experimental findings.

More generally, our research is attempting to examine and identify essential ingredients of cognitive training that enhance successful outcomes and generalization. There are many choices to be made in developing cognitive training, and as researchers, we need to be confident that those decisions will provide the greatest improvement (Cochrane and Green, this volume). Fundamentally of interest is what we are trying to train. Many programs have targeted cognitive capacities themselves (see Guye et al., Könen et al., Rueda et al., Wenger et al., this volume). Instead, our approach is to focus on teaching effective strategies that older adults can then use to tackle the PM situations they face.

One concern is how to implement this kind of strategy training, starting with how extensive the training ought to be. Though several sessions may be beneficial, the right kind of single training session may help older adults, which is a more practical proposition. In such a single session, the variability of the tasks that participants are exposed to in training is likely to be critical to later generalizability; experiencing a few different tasks may allow for more robust and flexible strategy development and application. In strategy training, it seems that guided metacognitive strategy training might be the best (see Schaeffner et al., this volume). Having such support in instruction has promise for older adults in comparison to allowing them to try and develop their own approach to PM tasks on their own.

Finally, the ecological validity of the training and the assessments of learning and transfer are critical. PM looks quite different inside and outside the laboratory. Thus, it is an important goal to foster the transfer of effective strategy use from training to the real world. As such, training programs must consider the balance and inclusion of laboratory training, homework, and simulated real-world activities during training such as the VW task. As these different considerations are explored, we are confident that an effective and efficient PM training for older adults will emerge, one that promotes transfer and generalizability to the real-world PM challenges.