With recent developments of the digital society, demands involving the scheduling and interleaving of multiple activities have increased significantly for both children and adults (Rideout, Foehr, & Roberts, 2010). The amount of time people spend on engaging in multiple tasks has increased dramatically during the last decades. We have to plan and coordinate multiple goals and intentions in different task contexts and time frames, often delayed and interrupted by other everyday activities. Yet, virtually no prior studies on more complex PM have investigated how we form, monitor, and remember delayed intentions in a single task context versus multiple ongoing task contexts.
Research on PM has been very active and innovative during the last decades, especially since the development of laboratory paradigms, which have been instrumental for conceptual advancement (for overviews, see Brandimonte, Einstein, & McDaniel, 2014; Kliegel, Mackinlay, & Jäger, 2008; McDaniel & Einstein, 2007). In this influential prior work, PM has been considered as a dual-task phenomenon, partly mirroring more traditional research on attention and (retrospective) memory. Indeed, most past studies of PM have investigated how we form and remember a single PM intention (primary task) within a single ongoing task context (secondary task), such as remembering to press the space bar once every minute in the context of a lexical decision task or identifying people wearing glasses in the context of a face recognition task. Note that these PM intentions often need to be repeatedly remembered (cf. taking medication every third hour). However, the task is one-dimensional in the sense that PM intentions refer to one and the same event structure or timeline of the ongoing task (here referred to as single-context PM). In contrast, everyday activities often involve multiple ongoing task contexts or “cognitive threads” (Salvucci & Taatgen, 2008); that is, delayed intentions can be embedded within different task domains (referred to as multiple context PM). Thus, past research on PM might lack generality due to its reliance on overly simplified simulations of everyday activities.
Prominent theories suggest distinct processes to support PM that are typically clustered under the terms of strategic monitoring and spontaneous retrieval. The framework of Preparatory Attentional and Memory Processes (Smith, 2003, 2016) assumes that PM is mainly driven by controlled executive functioning processes in terms of strategic monitoring. The multiprocess framework (Anderson, McDaniel, & Einstein, 2017; Scullin et al., 2018; Shelton & Scullin, 2017), however, assumes that PM performance is additionally guided by automatic, nonstrategic processes and may rather reflect a dynamic interplay of both processes. A general implication of these frameworks is that individuals with reduced executive functions—for example, due to cognitive aging (Schnitzspahn, Stahl, Zeintl, Kaller, & Kliegel, 2013) or frontal lobe damage (Burgess, Gonen-Yaacovi, & Volle, 2011)—have greater difficulties in more complex PM tasks than individuals with more efficient control functions. The aim of this study was not to contrast theories but rather to extend these influential views of PM; we focus on task conditions of PM in which executive control functions might be complemented by other cognitive skills—in the case of multiple-context conditions, by spatial abilities.
In line with prior research, we assume that executive functioning is the main source of individual and developmental differences in PM (Mackinlay, Kliegel, & Mäntylä, 2009; Mäntylä, Carelli, & Forman, 2007). However, this reliance on simplified simulations of PM may have obscured other sources of individual differences in PM. Indeed, most experimental simulations of PM involve dual-task performance in that PM deadlines are embedded along with a single event structure or timeline. However, to better reflect more complexities of PM in everyday life, these experimental simulations should also examine delayed intentions embedded in multiple ongoing tasks. In a similar vein, it should be noted that a majority of the research referred to as “multitasking” is also based on dual-task paradigms (for an overview, see Koch, Poljac, Müller, & Kiesel, 2018), focusing on cognitive bottlenecks (e.g., Pashler, 1994; Schubert, 1999), crosstalk between input and output mechanisms (e.g., Janczyk, Renas, & Durst, 2017), and task order control of two temporally overlapping tasks (e.g., Kübler, Reimer, Strobach, & Schubert, 2018). However, everyday multitasking differs from typical dual-task paradigms, in terms of both number of component tasks and overall duration of the multitasking scenario (see also Burgess, Veitch, de Lacy Costello, & Shallice, 2000; Logie, Trawley, & Law, 2011; Mäntylä, 2013; Redick et al., 2016). Compared with monitoring and scheduling multiple ongoing activities, demands for temporal monitoring are very low in most dual-task conditions, including single-context PM, in which delayed intentions are typically embedded within the temporal structure of the same ongoing task context.
Given the assumption that temporal complexity is a characteristic feature of time-based PM (and multitasking), remembering delayed intentions across independent task contexts is expected to require a higher degree of monitoring and coordination of deadlines along separate timelines and cognitive threads. Following this line of reasoning, remembering future intentions may involve higher cognitive processes other than executive functioning. Indeed, prior research showed that we understand and handle rather basic aspects of time (e.g., duration, sequence) by translating them in a spatial reference frame (Bender & Beller, 2014; Bonato, Zorzi, & Umiltà, 2012; Boroditsky, 2000; Casasanto & Boroditsky, 2008; Dehaene & Brannon, 2011; Núñez & Cooperrider, 2013). On the basis of this notion, we propose a spatiotemporal hypothesis of more complex forms of PM. We suggest that representing complex temporal patterns of deadlines as spatial relations is a basic and computationally efficient cognitive strategy because it alleviates cognitive control demands and possibly provides complementary processing advantages. That is, individuals with more effective spatial abilities can better meet the high temporal demands when handling immediate and delayed intentions in multiple contexts by relying on these additional spatial-relational processes. Consistent with earlier work in different areas of cognitive sciences, support for this “time in space” relational process has been observed in the field of multitasking. Several studies provided evidence that spatial ability, as measured by the mental rotation task (MRT), and executive functioning, as measured by the critical component of working memory (WM) updating, make independent contributions to multitasking performance (Mäntylä, 2013; Mäntylä, Coni, Kubik, Todorov, & Del Missier, 2017; Todorov, Del Missier, Konke, & Mäntylä, 2015; Todorov, Del Missier, & Mäntylä, 2014; Todorov, Kubik, Carelli, Del Missier, & Mäntylä, 2018).
Because these studies involved monitoring and coordination of multiple tasks with minimal demands on both prospective and retrospective memory, we aimed here to test the primary predictions of the spatiotemporal hypothesis within a complex PM paradigm by manipulating the complexity of the PM context. A methodological challenge in investigating PM in multiple contexts is that PM intentions should refer to different (vs. same) subtasks while eliminating differences in other task-related factors, including task complexity and expertise. Furthermore, for a fair contrast between PM involving a single task versus multiple task contexts, the ongoing task should be the same in both conditions. Also, the subtasks should be virtually identical and independent in that monitoring one task should not provide information about the state of the other tasks. As far as we know, no earlier studies have contrasted complex PM performance in single versus multiple contexts while minimizing task-specific differences and eliminating cross-task alignment.
In the present individual-differences study, we used a time-based PM paradigm in which delayed intentions were embedded in an ongoing task. The latter involved multiple simultaneous tasks rather than a single ongoing task as in most laboratory simulations of PM. We increased the complexity of the ongoing task for three reasons. First, as noted earlier, many everyday activities involve multiple ongoing activities with a mix of actions and intentions, rather than an isolated task (such as making lexical decisions), in which delayed intentions are embedded. Second, because PM performance is typically related to ongoing task costs (McDaniel & Einstein, 2007; Smith, 2003), the background activities should be virtually identical between the single-context and multiple-context conditions to allow a fair comparison. Third, we intended to examine PM performance of varying context complexity (PM intentions related to a single context vs. multiple contexts) in a setting of high cognitive load in which attentional resources for task management and temporal processing are shared with the ongoing task. If participants need to temporally monitor, update, and switch among different ongoing tasks, executive control and temporal processing are already taxed and should interfere more with scheduling and tracking of multiple intentions in complex PM performance (cf. McNerney & West, 2007; Occhionero, Esposito, Cicogna, & Nigro, 2010). Thus, potential differences in recruiting these limited cognitive resources should become more accentuated when participants attempt to additionally meet the increased temporal demands of PM tasks in a multiple (compared to single) task context. Finally, this task arrangement provided an opportunity to test the predictions of the spatiotemporal hypothesis in the context of multitasking performance (possibly replicating our earlier work) and of delayed intentions (possibly extending the hypothesis to more complex PM).
In this study, participants monitored four digital clocks, or more specifically counters, and needed to respond whenever one of the differently colored counters displayed readings that fulfilled a specific rule (e.g., multiples of 20; see also the Methods section for details). To prevent cross-task monitoring, the clocks ran at different rates (see also Mäntylä, 2013). Embedded in these ongoing monitoring activities, participants were instructed to remember and execute PM intentions (i.e., to press the space button) at certain deadlines (represented by the specific readings of the counters). In the single-context condition, participants needed to remember the PM intentions referring to the deadlines within one and the same counter; for example, they needed to respond when the counter showed “110,” “210,” and “310.” In the multiple-context condition (manipulated within subjects), participants needed respond whenever any of three different counters reached any of the deadlines. That is, participants had to respond when, for example, the red, blue, or yellow counters would show the time reading of “104.” Specifically, participants needed first to monitor the (occluded) counters, then to activate the counter with an approaching deadline, and finally to respond in time when the target reading appeared in the counter. We used four different sets of PM deadlines, and their temporal distribution was the same in both conditions. Participants also completed separate tasks of spatial ability (as measured by MRT) and WM updating (as measured by the matrix-monitoring task).
Following the spatiotemporal hypothesis and our earlier work, we expected that individual differences in both the matrix-monitoring performance and MRT performance would predict PM performance and that these contributions would be selective. More specifically, we expected that individual differences in matrix-monitoring performance would predict PM performance and, more importantly, that MRT performance would incrementally explain variability only in PM performance in the multiple-context condition in which remembering delayed intentions becomes more complex in terms of temporal monitoring and coordination. However, we expected that individual differences in MRT performance would not, or at least would to a lesser degree, contribute to single-context PM performance. We assumed that transforming within-context temporal relations (i.e., before vs. after along the same timeline) to spatial relations would not bring additional computational benefits for remembering PM intentions.