The influence of interruptions and planning on serial everyday multitasking in older adults

ABSTRACT

Cognitive aging research has studied the influence of healthy aging on the ability to multitask. Yet, little is known about the factors that might improve or impair serial multitasking performance in older adults. Three experiments involving younger and older adults assessed the impact of interruptions and planning on a prop-based test of multitasking. In Experiment 1, 26 younger adults and 25 older adults’ multitasking abilities were assessed; older adults performed significantly more poorly than younger adults. In Experiment 2, 19 younger and 22 older adults were randomly allocated to a group who experienced a one minute unexpected interruption while multitasking or a group with no interruption. The results showed that, when there was an interruption, the age difference disappeared. In Experiment 3, 32 younger and 30 older adults were randomly allocated to a group who were given 3 minutes to write an outline describing how they intended to approach the multitasking task, and another group who were given 3 minutes to label pictures of everyday objects prior to multitasking. Again, when participants were encouraged to plan, no age difference was found. These results highlight the advantage that interruptions and planning might have on serial everyday multitasking performance in older adults.

KEYWORDS:

• Serial multitasking
• executive functions
• aging
• interruptions
• planning

Introduction

Serial or everyday multitasking is the ability to perform a number of tasks, often during a limited time, by effectively switching between the tasks and planning the optimal order to perform them. Such tasks vary in their importance, demands and time and so individuals might not complete one task before starting another (Burgess, 2000a, 2000b; MacPherson, 2022b). Switching between tasks during serial multitasking is self-initiated as there are not salient prompts to signify when to switch tasks. However, there may be cues in the environment that attract one’s attention to engage in another task (e.g., the passing of time, completing several items within a task, completing the task). Individuals might not complete one task before moving to another. The order the tasks are performed in, and the time spent on each task, are self-determined. Typical examples of multitasking in everyday life include preparing a meal, carrying out shopping errands or driving.

One theoretical approach to serial multitasking (Burgess et al., 2000; Logie et al., 2011) proposes that it draws upon a range of cognitive processes including attention, memory and executive abilities but also planning, executing plans, retrospective and prospective memory (Burgess et al., 2000; McAlister & Schmitter-Edgecombe, 2013), as well as working memory (König et al., 2005; Logie et al., 2011). Therefore, serial multitasking requires the coordinated and deliberate employment of several cognitive functions and relies on individuals being able to postpone performing certain tasks while focusing on the performance of others; thus having effective control over one’s attending behavior is important for scheduling and interweaving between tasks (Burgess, 2014; Koechlin, 2013).

While clearly related, serial multitasking is not the same as dual-tasking or concurrent multitasking where two or more tasks can be performed in parallel (Burgess, 2014; Koechlin et al., 1999; MacPherson, 2018, 2022a), or rapid switching between laboratory tasks (e.g., Koch et al., 2010; Monsell, 2003). Dual-tasking or concurrent multitasking is the ability to perform two or more tasks at the same time and typically the tasks administered include digit, word or pattern recall, monitoring tasks (e.g., box crossing or E-checking), tracking a target or continuous choice response time tasks, as well as walking and talking (e.g., Cocchini et al., 2004). Patients with frontal lobe lesions with executive dysfunction have been found to perform more poorly on concurrent dual-tasking than frontal patients without executive dysfunction (Baddeley et al., 1997). The ability to coordinate performance on two tasks is thought to be separate from other executive functions such as shifting or inhibition (Miyake et al., 2000). Standard task-switching paradigms imitate everyday activities that depend on the ability to use contextual cues to influence behavior and to easily switch between different tasks. Individuals are asked to respond to a stimulus according to one of two rules (e.g., press one of two buttons depending on whether a stimulus is odd or even vs. press one of two buttons depending on whether a stimulus is higher or lower than 5). Typically, these rules can alternate or be randomly assigned and switching between the task rules typically incurs additional costs (i.e., switch costs) compared to repeating the same rule (i.e., repeat costs). Switch costs have also been associated with executive functions (e.g., Rubinstein et al., 2001).

Serial multitasking can be assessed in the real-world using tasks such as the Multiple Errands Test (Shallice & Burgess, 1991) where participants are required to complete a number of simple tasks within 15 minutes (e.g., buy a bar of soap) without breaking certain rules (e.g., do not enter a shop without buying something) in a pedestrianized street in London. Shallice and Burgess (1991) also devised the Six Element Test, which forms part of the Behavioural Assessment of the Dysexecutive Syndrome battery (BADS; Wilson et al., 1996), to assess serial multitasking in the clinic. Participants have a time limit of 15 minutes to attempt six subtasks: dictating the route to arrive at the lab and dictating the route to be taken after leaving the lab; two sets of arithmetical problems; and two sets of object naming (Shallice & Burgess, 1991). Similar multitasking paradigms have been devised to assess patients with more severe injuries (e.g., the simplified version of the Multiple Errands Test, MET-SV; Alderman et al., 2003), within a hospital setting (Godbout et al., 2005; Knight et al., 2002) and meal preparation tasks in an occupational therapy department’s kitchen (Chevignard et al., 2008). In addition, Burgess et al. (2000) devised the Greenwich Test, a test similar to the Six Element Test which consists of three complex open-ended tasks to be performed in 10 minutes (i.e., sorting beads, sorting tangled lines on paper, and constructing Meccano models). Participants are assessed on their ability to learn the rules (pre- and post-assessment), plan, task switch in relation to the number of rule breaks, and recollect their own actions.

Neuropsychological research examining multitasking abilities in clinical populations has demonstrated serial multitasking impairments associated with prefrontal lesions (Burgess et al., 2000; Roca et al., 2011; Shallice & Burgess, 1991), traumatic brain injury (Manly et al., 2002), behavioral variant frontotemporal dementia (Torralva et al., 2009), Parkinson’s disease (Roca et al., 2012) and schizophrenia (Laloyaux et al., 2014; Roca et al., 2014). Serial multitasking has also been studied in healthy individuals; in particular, cognitive aging research has studied the influence of healthy aging on the ability to multitask. In everyday life, older adults seem able to carry out day-to-day tasks such as preparing a meal or shopping for food with little difficulty (Phillips et al., 2006). Moreover, research has shown that older adults perform as well as younger adults in terms of rule clarification and recall, and the number of tasks initiated and completed when performing multitasking tasks such as the Six Element Test (Shallice & Burgess, 1991) or the Day Out task (Schmitter-Edgecombe et al., 2012). However, studies also found that older adults perform more poorly than younger adults when forming, retaining and executing plans, and rule following (Kliegel et al., 2000; Levine et al., 2000; McAlister & Schmitter-Edgecombe, 2013; Sanders & Schmitter-Edgecombe, 2012). Using computerized tasks, healthy older adults have been found to make significantly more overall and non-efficiency errors and perform more rule breaks than younger adults when performing a shopping mall virtual environment (VMALL; Rand et al., 2009) and executed plans less effectively when performing a computerized cooking task (the Breakfast task; Craik & Bialystok, 2006). Kosowicz and MacPherson (2017) compared computerized versus prop-based versions of the Breakfast task in the same individuals and found that age-related decrements were only found on the computerized version of the task, not the prop-based one. These studies suggest that older adults can complete as many subtasks as younger adults when serial multitasking but tend to do so less efficiently (Schmitter-Edgecombe et al., 2012) and the age-related deficits may be less when non-computer based tasks are used (Feinkohl et al., 2016; Kosowicz & MacPherson, 2017).

Yet, little is known about the factors that might improve or impair the type of serial multitasking performance assessed in the neuropsychological literature. One likely feature of everyday multitasking is interruptions, which are part of daily life and may cause disruption to on-going task performance (Burgess, 2000a), and the timing of these events cannot be controlled by the individual (Speier et al., 1999). For example, while cooking a meal, an individual may be disrupted by a ringing telephone, or somebody may ask them a question that occupies their attention. Interruptions have been studied in occupational settings such as emergency medicine (Chisholm et al., 2000) and aviation (Latorella, 1999), and the results seem to suggest that interruptions can have a detrimental impact on tasks with similarities to multitasking situations (Edwards & Gronlund, 1998; Gillie & Broadbent, 1989). However, interruptions do not inevitably disrupt performance and can even be advantageous in certain circumstances. For example, Manly et al. (2002) found that interrupting “alerts” (auditory tones) improved the multitasking performance of patients with traumatic brain injuries who were told to use the alerts as prompts to think about what they were currently doing and what their overall goals were. Moreover, while Law et al. (2004) demonstrated serial multitasking impairments in their patients with dysexecutive syndrome compared to healthy controls, they found that both groups were resistant to the effects of interruptions. This suggests that the ability to manage interruptions is separate from the ability to multitask and the effect of interruptions might depend on the interaction of factors associated with the multitasking paradigm. Therefore, like patients with dysexecutive syndrome, older adults may not show any particular sensitivity to the effects of interruptions despite age-related declines in their executive abilities (Daigneault et al., 1992; de Frias et al., 2006; Lamar & Resnick, 2004; MacPherson et al., 2002; Mittenberg et al., 1989).

Another factor that might influence serial multitasking performance in healthy aging is the ability to plan and to adhere to the results of a plan. Models of everyday multitasking originating from studies of frontal patients (Burgess et al., 2000) and healthy younger individuals (Logie et al., 2011) examine how different components of the cognitive system work in concert rather than isolation. Burgess et al. (2000) proposed that planning might be one of the cognitive concepts that maintains the ability to multitask, together with retrospective and prospective memory. Subsequently, Logie et al. (2011) suggested that planning can be fractionated into task-ordering and goal-setting processes that occur before performing the task (preplanning) and task reordering and goal adjustment processes that ensue while performing the task (online plan modifications). Law et al. (2013) asked participants to memorize a list of errands presented to them as a preplanned sequence: one group received an optimally ordered plan (good plan) while the other group received a suboptimal plan (poor plan). However, participants were allowed to rearrange the errands while they performed the task. Law et al. (2013) found that individuals who began with a good self-generated plan performed better than those who began with a poor plan. Nonetheless, participants who stuck with their original plan (regardless of whether it was good or bad) tended to perform better than participants who changed their plans while performing the task. Similar findings were found for prospective memory, an important aspect of multitasking. Trawley et al. (2011) found that following a plan, rather than changing the plan online, resulted in better overall multitasking performance. The authors concluded that changing the errands’ order during the task disrupted the performance of the errands that had yet to be completed, canceling out any potential advantage from changing a suboptimal plan. These findings suggest that individuals who are encouraged to plan prior to serial multitasking, regardless of the quality of the plan, perform better than individuals who do not plan.

The focus of the present study was to investigate the impact of interruptions and planning on serial multitasking abilities in healthy adult aging. The multitasking paradigm adopted for the three experiments was that used by Law et al. (2004), which was based on the Greenwich Test described by Burgess et al. (2000). Participants have four subtasks to attempt within a limited time, and have to apply an effective strategy to collect red-colored items in each subtask in order to obtain a high score. In Experiment 1, we investigated whether younger and older adults differed in their performance on the Law et al. (2004) paradigm. Given that age differences are not found in terms of the number of tasks initiated and completed when using prop-based multitasking tasks, we predicted that older adults would not perform more poorly than younger adults. In Experiment 2, we examined the influence of interruptions on multitasking in older adults. Based on the results from the studies of dysexecutive patients, we might expect that older adults do not show any particular sensitivity to the effects of interruptions. Finally, in Experiment 3, we investigated whether older adults would benefit from preplanning prior to performing multitasking. We predicted that those who engaged in preplanning would perform more efficiently. To our knowledge, the effects of interruptions and planning on serial multitasking abilities in healthy aging have not been directly assessed.

Experiment 1

Methods

Participants

Twenty-six younger adults (16 women) aged between 19 and 39 years (M = 24.65, SD = 4.89) and 25 older adults (17 women) aged between 60 and 80 years (M = 69.92, SD = 5.92) were tested. The younger age group had significantly higher years of full-time education than the older age group (M = 16.62, SD = 2.17, range = 12–21; M = 14.28, SD = 3.02, range = 9–19 for younger and older adults respectively), t(49) = 3.18, p < .005. Participants were either students at University of Edinburgh or members of the Department of Psychology, University of Edinburgh volunteer panel, were recruited via word-of-mouth, or were known to the experimenter. The older group had a mean of 29.40 out of 30 (SD = 1.15, range = 26–30) on the Mini-Mental State Examination (MMSE; Folstein et al., 1975). Inclusion criteria included: aged between 18 and 40 years or 60 and 80 years, normal or corrected to normal hearing and vision, living independently in the community, and no self-reported history of the neurological or psychiatric disorders listed on the Wechsler Adult Intelligence Scale-Third Edition (WAIS-III UK; Wechsler, 1997a) and the Wechsler Memory Scale-Third Edition (WMS-III UK; Wechsler, 1997b) selection criteria. All participants were native English speakers. All participants gave written, informed consent in accordance with the Department of Psychology Research Ethics Committee guidelines.

Materials

Multitasking task

The multitasking test used by Law et al. (2004) was administered. It consisted of four subtasks: Telephone task; Bricks Construction task; Envelopes task and Beads task. Participants were instructed to attempt all four subtasks in 10 minutes but were told it was not possible to complete them all. Participants were able to attempt the tasks in any order and could switch between them as often as they liked. In each subtask, the red items were worth extra points (i.e., 10 points) compared to other colored items which were worth one point. The aim was to achieve as many points as possible, but participants would lose all their points for a subtask if they violated a rule, and they would lose 100 points if they failed to attempt a subtask. The subtasks were:

Beads task

Participants were provided with an example string of beads approximately 55 cm long. The string was made up of 26 sections of colored beads with three beads in each section. There was a large round bead at one end of the beads to indicate the start of the pattern. The sections of red beads were the 2^nd, 6^th, 12^th, and 19^th sections. The thread used was 0.2 cm thick and the beads were 1 cm long, and 0.9 cm in diameter, with a 0.5 cm hole in the middle for threading the string. Participants were presented with a 105 cm long piece of string with a large bead at the end to show the start point. An open box contained the beads necessary to replicate the pattern on the example string of beads. Participants were instructed to take only one bead out of the box at a time; therefore, the strategic approach to this task was to complete the string of beads as far as a red section before moving onto another subtask.

Envelopes task

Participants were presented with 25 sheets of A4 (21 cm × 29.6 cm) colored paper in three piles: one with 10 sheets of blue paper; one with 10 sheets of yellow paper; and one with five sheets of red paper. There was also a pile of 25 white letter (11.4 cm × 23.4 cm) envelopes. Participants were instructed to put one sheet of paper, folded into thirds, into each envelope but not to seal the envelopes. The envelopes could be filled with the colored paper in any order. Therefore, the best strategy in this subtask was to fill the envelopes with red paper before using any of the other colored paper.

Bricks construction task

Participants were presented with a hollow square structure (8 cm × 8 cm x 12.5 cm) built from Lego™ bricks. The structure was made up of 13 layers of eight 2 × 4 bricks (or the appropriate number of 2 × 2 bricks). Each layer contained bricks of the same color, but the layers varied in color and no two consecutive layers were the same color. The red brick layers were in the 2nd, 6^th and 11^th layers from the base. Participants were given the exact number of bricks to replicate the structure in the same shape and size. Points were awarded for every complete layer rather than every individual brick; therefore, the best strategy was to complete a red layer before shifting to another subtask.

Telephone task

Participants were presented with a telephone directory and a list of 20 names from the residential section of the directory with five of the names printed in red ink. Participants were instructed to look up the names in the telephone directory and write down their corresponding telephone numbers. They could look up the names in any order; therefore, the best strategy was to look up the names printed in red ink first.

Additional materials

A silent digital clock was evident to the participants to allow them to monitor the time.

Procedure3

An instruction sheet provided participants with the task instructions and task rules. After reading the instruction sheet, participants were asked a series of 12 cued recall questions to ensure they understood and remembered the instructions. If participants did not know the answers, they were explained to them. Participants then performed the multitasking task for 10 minutes.

Data analysis

Data were analyzed using R (R Core Development Team, 2018). To compare the performance of younger and older participants in terms of multitasking performance, independent samples t-tests were conducted. The data were collected as part of a student project with the stopping rule being the end of the data collection period. G*Power 3 (Faul et al., 2007) was used to compute the effect size post-hoc for the difference between the younger group (n = 26) and the older group (n = 25) with 80% power at the 5% alpha level. The analysis revealed a large effect size with Cohen’s d = 0.80 (Cohen, 1988; Faul et al., 2007).

Results

Table 1 shows the means and standard deviations for the multitasking performance of the younger and older age groups.

Table 1. The means and standard deviations in parentheses for the younger and older adults performing the multitasking task in Experiment 1.

	Younger Group (n = 26)	Older Group (n = 25)
Efficiency Total No. Red Items	13.38 (2.17)	9.48 (3.23)
Overall Total No. Points	150.19 (25.81)	105.32 (33.97)
Beads	51.08 (19.70)	31.12 (19.66)
Letters	44.54 (11.99)	34.16 (18.13)
Bricks Construction	14.19 (9.22)	11.20 (8.56)
Telephone	40.38 (12.16)	28.84 (17.07)

Multitasking efficiency

Multitasking efficiency was measured by the number of completed red items. An independent samples t-test demonstrated that the mean multitasking efficiency score was significantly higher for the younger compared to the older group, t(49) = 5.08, p < .001.

Overall multitasking performance

Overall performance was taken as the number of available items achieved by the participant, irrespective of their color. An independent samples t-test revealed that the mean overall multitasking performance score was significantly higher for the younger group compared to the older group, t(49) = 5.32, p < .001.

Individual subtask performance

An independent samples t-test revealed that the mean Beads score was significantly higher for the younger group compared to the older group, t(49) = 3.62, p < .001. Mann Whitney U Tests demonstrated that the younger group (Mdn = 50) performed better than the older group (Mdn = 40) on the Letters task, U = 484, p < .005, and the younger group (Mdn = 45) performed better than the older group (Mdn = 22) on the Telephone task, U = 450, p < .05. There was not a significant difference between the two groups on the Bricks construction task (p = .46).

Discussion

Experiment 1 examined whether there are age-related differences in performance on the Law et al. (2004) serial multitasking paradigm, which is a prop-based task. The previous literature has shown that older adults perform as well as younger adults in terms of the number of tasks initiated and completed when prop-based assessments of serial multitasking are used (Kliegel et al., 2000; Levine et al., 2000; McAlister & Schmitter-Edgecombe, 2013; Sanders & Schmitter-Edgecombe, 2012). However, in contrast to these findings, our older adults completed significantly fewer red items and had a significantly lower overall performance score than younger adults. They also performed more poorly on the Beads, Letters and Telephone tasks compared to younger adults.

The Law et al. (2004) paradigm adopted in the current study involves subtasks that include folding paper into envelopes, building a Lego™ structure, searching for names in a phone directory and threading beads onto a string. Therefore, some of these subtasks rely on an individual’s motor abilities to perform the subtasks successfully. In contrast, other serial multitasking paradigms that have not shown age effects tend to involve subtasks that rely less on the coordination of bimanual movements and more on cognitive abilities, such as solving arithmetical problems, naming pictures or copying objects (Kliegel et al., 2000; Levine et al., 2000). Other paradigms involved more naturalistic subtasks, which older adults are more likely to be familiar with given they perform them in everyday life, such as retrieving items from another room or using the microwave (McAlister & Schmitter-Edgecombe, 2013; Sanders & Schmitter-Edgecombe, 2012). The age differences in this experiment might be explained by older adults’ slower coordination of bimanual movements (Stelmach et al., 1988; Wishart et al., 2000). While efforts were made to ensure the components of the subtasks were not too small for older adults to manipulate, the beads may have been more difficult for older adults to thread onto the string and older adults may have been slower at searching through the pile of Lego™ bricks. However, as age differences were not found for the Bricks construction task, but were for all other subtasks, older adults may have preferred to perform this task and focused on it at the expense of the other ones. Unfortunately, we do not have completion time data for the individual subtasks to allow us to examine whether this might be the case or not.

While dual-tasking or concurrent multitasking (Burgess, 2014; Koechlin et al., 1999; MacPherson, 2018), and rapid switching (e.g., Koch et al., 2010; Monsell, 2003) do not focus on the broader demands of everyday multitasking, there is likely some overlap between these task paradigms and serial multitasking. In terms of aging, dual-task or concurrent multitasking provide conflicting findings where some studies report poorer performance in healthy older adults compared to younger adults, especially when the demands of the simultaneous tasks are high (Craik, 1977; McDowd & Craik, 1988; Salthouse et al., 1984), while others report no age-related differences (Baddeley et al., 2001; Anderson et al., 2011; Argiris et al., 2020; Kilb & Naveh Benjamin, 2014). Importantly, these latter studies are inclined to equate single-task performance across the age groups by adjusting the difficulty of each task for individual ability. With rapid task switching paradigms, there are again inconsistencies in terms of age effects where some studies have shown that older adults have greater switch costs compared to younger adults (e.g., Cepeda et al., 2001; Hillman et al., 2006; Kray et al., 2002) but other studies do not (e.g., Kray et al., 2005; Reimers & Maylor, 2005; Salthouse et al., 1998). Several meta-analyses have reported that, while older adults show general slowing, there are not age deficits in terms of switch costs (Chen & Hsieh, 2023; Verhaeghen, 2011; Wasylyshyn et al., 2011). These inconsistencies may be due to the different paradigms adopted, the amount of practice, and the similarity between tasks in terms of stimuli presentation and response modality; factors which have the potential to also contribute to age differences in serial multitasking.

The poorer overall score in the older group suggests they were more likely to break the rules than the younger group. Examples of rule breaks included constructing the tower of bricks around the entire edge of the base platform so the structure was too big, threading only the red beads on the string in the Beads task, and writing addresses instead of phone numbers in the Telephone task. However, it is unlikely that the older participants did not understand the rules of the multitasking test, as all participants were able to answer the cued recall questions, which assessed understanding and memory for the rules immediately prior to testing. If participants did not respond correctly to the questions, the answers were provided again. A more likely explanation is that the older participants were less able to later remember the task rules when performing the multitasking paradigm, and as a result, they made more rule breaks. Indeed, older adults may have attempted less red items due to forgetting that the red items were worth more. However, since they attempted fewer items overall, it is more likely that older adults simply had less opportunity to use the red items. To be able to conclude that poorer memory is the reason why older adults attempted significantly fewer red items and achieved a significantly lower overall performance score, future work is required, as one limitation of this experiment was that participants were not asked the cued recall questions again at the end of the study.

After determining that older adults showed poorer multitasking performance compared to younger adults on the Law et al. (2004) multitasking paradigm, the aim of Experiment 2 was to examine whether older adults are sensitive to interruptions, a common occurrence when serial multitasking in everyday life (e.g., the doorbell ringing while you are cooking dinner). Interruptions were defined as stimuli that required attention (Clapp & Gazzaley, 2012) during the performance of the multitasking paradigm. One group was interrupted 7 minutes into the multitasking paradigm (i.e., interruption condition) to perform an unrelated task for one minute before returning to multitasking, whereas the other group was not interrupted (i.e., no interruption condition).

Experiment 2

Methods

Participants

Nineteen younger adults (9 women) aged between 20 and 39 years (M = 26.95, SD = 5.19) and 22 older adults (18 women) aged between 60 and 80 years (M = 66.68, SD = 5.25) took part in Experiment 2. The younger age group had significantly more years of full-time education than the older age group (M = 16.37, SD = 2.01, range = 12–21; M = 14.38, SD = 3.25, range = 9–21 for younger and older adults respectively), t(38) = 2.30, p < .05. The older group had a mean of 29.73 out of 30 (SD = 0.46, range = 29–30) on the Mini-Mental State Examination (MMSE; Folstein et al., 1975). The recruitment and inclusion criteria were the same as Experiment 1. All participants were native English speakers and gave written, informed consent in accordance with the Department of Psychology Research Ethics Committee guidelines.

Materials

Multitasking task. The same multitasking task as in Experiment 1 was administered.

Procedure

The same procedure as in Experiment 1 was adopted except that participants in each age group were randomly allocated to one of two independent groups to perform the multitasking task for 10 minutes. One group was interrupted 7 minutes into the test (i.e., interruption condition) and the other group was not interrupted (i.e., no interruption condition).

Participants were aware that they would be interrupted and were explained the instructions for the task beforehand. The interruption task was unrelated to the on-going situation; those who were interrupted were asked to complete the Silly Sentences task (Baddeley et al., 1985) for one minute. Participants indicated on an A4 sheet of paper whether the presented sentences were “true” or “false” (e.g., “Lions are living creatures, Buses are made from apples”). The 10-minute timer was paused during the interruption and restarted one minute later.

Data analysis

Data were analyzed using R (R Core Development Team, 2018). To investigate the effects of age group and interruptions on multitasking performance, the data were analyzed using a 2 (age group: young vs. old) x 2 (condition: interruption vs. no interruption) analysis of variance (ANOVA). For data that were not normally distributed, a Box-Cox power transformation (Box & Cox, 1964) was applied. Other transformations such as log and square root transformations were not used as the assumptions were still violated when they were applied. Post-hoc comparisons were calculated with the Tukey HSD test with a .05 level of significance. If after a Box-Cox power transformation, the data still did not meet the assumptions for ANOVA, separate Mann Whitney U tests were conducted to compare the younger and older adults’ performance on the interruption condition and then the no interruption condition. The data were collected as part of a student project with the stopping rule being the end of the data collection period. G*Power 3 (Faul et al., 2007) was used to compute the effect size post-hoc for the two-way interaction in the ANOVAs using a sample size of 41 participants and 80% power at the 5% alpha level. The analysis revealed a large effect size of Cohen’s f = 0.45 (Cohen, 1988; Faul et al., 2007).

Results

The demographic data for the younger and older participants in the interruption and no interruption conditions are summarized in Table 2. A 2 × 2 ANOVA showed no significant main effects of age group (p = .45) or interruption (p = .17) or an age group by condition interaction (p = .22) on years of education. In terms of MMSE, a 2 × 2 ANOVA revealed no significant main effect of age group or condition or an age group x condition interaction (p > .05).

Table 2. Participant demographic information for Experiment 2 means are presented with standard deviations in parentheses.

Age Group	Age (years)	Full-time education (years)	MMSE (out of 30)
Younger
Interruption (n = 9)	26.33 (4.64)	16.11 (1.45)	29.78 (0.44)
No interruption (n = 10)	27.50 (5.84)	16.60 (2.46)	29.20 (0.79)
Older
Interruption (n = 12)	66.83 (4.93)	15.18 (3.28)	29.67 (0.49)
No interruption (n = 10)	66.50 (5.87)	13.50 (3.14)	29.80 (0.42)

Note: MMSE = Mini Mental State Examination.

Multitasking efficiency

Figure 1 demonstrates the mean multitasking efficiency scores (i.e., the number of completed red items) for each age group. As the data were not normally distributed, a Box-Cox power transformation was used to achieve normality. Participants’ multitasking efficiency scores were raised to the power of λ = 1.68. A 2 (age group: young vs. old) x 2 (condition: interruption vs. no interruption) ANOVA showed a significant main effect of condition, F(1, 37) = 5.59, p < .05, η_p² = .02, where the interruption group had significantly higher efficiency scores. There was also a significant age group by condition interaction, F(1, 37) = 6.14, p < .05, η_p² = .14. Post-hoc Tukey HSD tests revealed that younger adults achieved significantly higher efficiency scores than older adults in the no interruption condition (p < .01). The two age groups did not significantly differ in the interruption condition (p = .99). There was no significant difference in younger adults’ efficiency scores (p = .58) or older adults’ efficiency scores (p = .08) between conditions. There was no significant main effect of age group (p = .81).

Figure 1. Multitasking performance for younger and older adults in the interruption and no interruption conditions. Top: Multitasking efficiency as measured by the total number of red items completed; Bottom: Overall multitasking performance as measured by the total number of points obtained. Error bars represent standard error of the mean.

Overall multitasking performance

Figure 1 also shows the mean overall multitasking performance scores for each age group across each condition. A Box-Cox power transformation was applied where the overall multitasking scores were raised to the power of λ = 1.48. A 2 × 2 between-subjects ANOVA revealed no significant main effect of age group (p = .90) or condition (p = .07) on performance. However, there was a significant age group by condition interaction, F(1, 37) = 4.21, p < .05, η_p² = .10. Post-hoc Tukey HSD tests revealed that the younger adults who received no interruption scored significantly higher than the older adults who received no interruption (p < .05). There was no other significant differences between the groups (p > .21).

Individual subtask performance

Table 3 demonstrates the means and standard deviations for the younger and older groups performing the subtasks under the interruption and no interruption conditions. As the data were not normally distributed for the Beads task, a Box-Cox power transformation was used to achieve normality. Participants’ Beads scores were raised to the power of λ = 0.70. A 2 × 2 between-subjects ANOVA revealed no significant main effect of age group (p = .25), or condition (p = .73) or age group x condition (p = .97) interaction. For the Letters task, Mann Whitney U Tests revealed no significant difference between younger and older adults in the interruption condition (p = .34) or the no interruption condition (p = .38). A 2 × 2 between-subjects ANOVA conducted on the Bricks construction task revealed a significant main effect of age group, F(1, 37) = 8.90, p < .01, η_p² = .19, where the younger adults scored significantly higher than the older adults. The main effect of condition (p = .49) and the age group x condition (p = .22) interaction were not significant. Finally, a 2-way between-subjects ANOVA conducted on the Telephone task score revealed a significant main effect of condition, F(1, 37) = 9.63, p < .005, η_p² = .04, where the interruption condition was performed better than the no interruption condition. The age group x condition interaction was also significant, F(1, 37) = 9.63, p < .005, η_p² = .22. Post-hoc Tukey HSD tests revealed that younger adults achieved significantly higher Telephone task scores than older adults in the no interruption condition (p < .05) but not the interruption condition (p = .37). The main effect of age group was not significant (p = .11).

Table 3. The means and standard deviations in parentheses for the younger and older adults performing the multitasking subtasks in Experiment 2.

	Younger Interruption	Younger No interruption	Older Interruption	Older No Interruption
Beads	38.56 (18.12)	36.10 (17.47)	29.33 (18.42)	26.80 (18.09)
Letters	39.22 (21.40)	51.40 (2.95)	42.42 (20.34)	37.00 (23.35)
Bricks Construction	11.00 (7.37)	8.00 (5.56)	3.00 (5.44)	4.80 (6.05)
Telephone	27.89 (17.68)	39.00 (15.95)	39.67 (15.80)	18.00 (16.01)

Silly sentences task

The younger and older participants in the interruption condition did not significantly differ in terms of the number of correctly completed sentences (M = 22.11, SD = 6.43; M = 27.83, SD = 11.28 respectively), t(19) = 1.36, p = .19.

Discussion

Experiment 2 examined the influence of interruptions on serial multitasking performance in aging. In terms of efficiency scores, age differences were only found in the no interruption condition where older adults were less efficient than younger adults, completing fewer red items. However, when there was an interruption, the two age groups did not significantly differ in the amount of red items completed. The same pattern of performance was found for the overall score. These findings suggest that interruptions benefit rather than disrupt older adults’ multitasking performance. Similarly, Manly et al. (2002) found that the multitasking performance of patients with traumatic brain injuries significantly improved and did not significantly differ from the control group when patients were exposed to interrupting tones. One explanation for this finding may be that the interruption introduced a planned break from the demands of the multitasking scenario, enabling older adults to regroup before returning to the unstructured and more demanding multitasking paradigm. During the interruption, older adults were able to refocus on the tasks and reflect on what they were doing and what their overall goals were. To our knowledge, this is the first published study to examine the influence of interruptions on older adults’ serial multitasking abilities. Indeed, Law et al. (2004) found no effect of interruption on the same multitasking paradigm in younger adults or patients with executive dysfunction. Moreover, Shum et al. (2013) examined the influence of unexpected interruptions on prospective memory, a component of multitasking, in older adults. They found that interruptions only impaired time-based prospective memory but not event- or activity-based prospective memory and there was no interaction with age. Therefore, older adults did not find the interruption more challenging or disruptive than younger adults did. Together, these findings suggest that interruptions do not disrupt serial multitasking performance in healthy aging.

Studies investigating externally imposed interruptions have tended to be conducted in other multitasking contexts such as emergency medicine (Chisholm et al., 2000) or aviation (Latorella, 1999). Unlike the current findings, these studies suggest that interruptions have a negative impact on tasks with similarities to multitasking situations (Edwards & Gronlund, 1998; Gillie & Broadbent, 1989). However, often the disruption is observed in terms of increased response times rather than accuracy measures (Eyrolle & Cellier, 2000; Gillie & Broadbent, 1989; Trafton et al., 2003) or memory for the items after the interruption (Edwards & Gronlund, 1998). Therefore, future studies should examine whether timing measures are more sensitive to interruptions compared to accuracy in healthy older adults.

The number of silly sentences that participants answered during the one-minute interruption task did not significantly differ between the younger and older groups. This suggests both age groups put the same effort into performing the interruption task; older adults did not perform similarly to younger adults on the multitasking paradigm due to a trade-off in performance between the two tasks. It may be that the Silly Sentences task was not demanding enough to have a disruptive effect. However, given that the older adults were poorer at serial multitasking than younger adults in the no interruption condition, but showed improved performance in the interruption condition, this is unlikely to be the case. Studies have shown that ongoing task performance is more likely to be disrupted if the interruption is complex, or comparable to the on-going task (Edwards & Gronlund, 1998; Gillie & Broadbent, 1989; Li et al., 2011; Speier et al., 1999). Therefore, future research might examine the effect of different types of interruption on serial multitasking performance in younger and older adults.

Individuals in the one-minute interruption condition were told beforehand that they would be presented with an unrelated interruption while multitasking. This was to avoid spending time explaining the Silly Sentences task instructions during the actual interruption. Given that the interruption was expected, it may not have been as disruptive as an unexpected one; therefore, older adults might not experience benefits of an unexpected interruption. Another explanation might be that, because the participants in the interruption condition knew they would be interrupted, they hurried to complete more tasks prior to the interruption compared to the group who were not under such pressure, and this could have affected their performance irrespective of the interruption itself. However, these explanations are unlikely, given that Law et al. (2004) did not inform their participants that they would be interrupted and they found that patients with dysexecutive syndrome were also resistant to the effects of unexpected interruptions, regardless of their impaired multitasking performance, using the same serial multitasking paradigm.

In our study, those in the no interruption condition were simply asked to perform the multitasking paradigm without stopping. One might argue that a better control condition would be for all participants to be told that they would be interrupted, but then only the interruption group perform the Silly Sentences task, or the no interruption group to spend the same amount of time away from the multitasking paradigm. While we selected the same control condition Law et al. (2004), future work might adopt one of these other control conditions to determine whether similar results are found. Our current findings suggest that interruptions do not inevitably disrupt multitasking performance and can even be beneficial to older adults in certain serial multitasking situations.

The focus of the third and final experiment was to further examine the factors that influence multitasking performance in healthy aging. Previous studies of serial everyday multitasking involving patients with frontal lobe damage (Burgess et al., 2000) and healthy younger individuals (Logie et al., 2011) have proposed that planning might be one of the cognitive concepts that maintains the ability to serially multitask. Studies have shown that participants who are encouraged to plan tend to perform better than participants who do not plan (Law et al., 2004; Trawley et al., 2011). Here, we investigated whether encouraging older adults to consider the task order and set goals prior before performing the multitasking paradigm (i.e., preplanning) might improve performance. Participants in the plan condition were given 3 minutes to consider how they would approach the multitasking paradigm prior to performing it. Participants in the no plan condition were given 3 minutes to label pictures of everyday objects.

Experiment 3

Methods

Participants

Thirty-two younger adults (20 women) between the ages of 19 to 33 years were recruited through an online advertisement. Thirty older adults (21 women) aged 65 to 79 years were recruited via the Psychology Department volunteer panel at the University of Edinburgh. All participants were native English speakers and had normal or corrected to normal hearing and vision. Participants had no history of drug or alcohol dependence, had never experienced a serious head injury, and did not report having any psychiatric or neurological conditions.

The younger group received significantly more years of full time education than the older group, t(60) = 3.23, p < .005. All older adults scored above the cutoff scores of 24 out of 30 on the MMSE. The study was approved by the Department of Psychology Research Ethics Committee.

Materials

Background neuropsychological measures

Participants also completed some background neuropsychological tests. These included the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999) to assess full scale IQ and the Wechsler Test of Adult Reading (WTAR; Wechsler, 2001) as a measure of verbal abilities. To assess executive abilities, the Tower Test from the Delis-Kaplan Executive Function System (D-KEFS) and the D-KEFS Color-word Interference Test (Delis et al., 2001) were administered. The Coding subtest from the Wechsler Adult Intelligence Scale – Fourth Edition (WAIS-IV; Wechsler, 2008) was also administered to assess speed of processing.

Multitasking task

The same multitasking task as Experiments 1 and 2 was administered.

Procedure

Participants in each age group were randomly allocated to either the plan or no plan conditions. Participants in the plan condition were given 3 minutes to write an outline describing how they intended to approach the multitasking task. They were permitted to refer back to their plans throughout the session. Participants in the no plan condition were given 3 minutes to label pictures of everyday objects, which consisted of the first 200 pictures from Snodgrass and Vanderwart (1980). After the 3 minutes had elapsed, all participants were given 10 minutes to complete the multitasking test.

Data analysis

Data were analyzed using R (R Core Development Team, 2018). In instances where the data were not normally distributed, Box-Cox power transformations were used as the assumptions for ANOVA were still violated using log and square root transformations. The data were analyzed using a 2 (age group: young vs. old) x 2 (condition: plan vs. no plan) analysis of variance (ANOVA). Post-hoc analyses were conducted using Tukey’s HSD post-hoc comparisons. If the data still did not meet the assumptions for ANOVA after Box-Cox power transformation, separate Mann Whitney U tests were conducted to compare the younger and older adults’ performance on the plan condition and then the no plan condition. Pearson’s or Spearman’s correlational analyses were also performed to evaluate the relationships between the background neuropsychological measures and multitasking performance depending on whether the data were normally distributed or not. Finally, 2 (age group: young vs. old) x 2 (condition: plan vs. no plan) analyses of covariance were conducted on a subset of the data entering Coding as a covariate to examine whether the effects remained when controlling for speed of processing. Again, the data were collected as part of a student project and data collection concluded at the end of the data collection period. G*Power 3 was used to compute the effect post-hoc for the two-way interaction in the multitasking ANOVAs using a sample size of 62 participants and 80% power at the 5% alpha level. The analysis revealed a medium to large effect size with Cohen’s f = 0.36 (Cohen, 1988; Faul et al., 2007).

Results

Background neuropsychological measures

Table 4 demonstrates the demographics and background neuropsychological test performance of the younger and older adults in the plan and no plan conditions. A two-way ANOVA revealed a main effect of age group, F(1, 58) = 3.99, p = .05, η_p² = .15, where the older group had a significantly fewer years of education than the younger group. However, the age group by condition interaction was not significant (p > .67); therefore, the younger adults in the plan and no plan conditions and the older adults in the plan and no plan conditions did not significantly differ in their years of education.

Table 4. Participant demographic information and background neuropsychological test performance for younger and older groups in the plan and no plan conditions in Experiment 3. Means are presented with standard deviations in parentheses.

		Age (years)	Full-time education (years)	WASI IQ	WTAR IQ	Coding Number completed in 2 mins	Stroop Inhibition Completion time	Tower Total score
Younger	Plan (n = 16)	25.44 (4.16)	16.69 (2.50)	118.13 (10.34)	111.44 (4.68)	84.31 (16.52)	16.44 (7.22)	19.31 (4.16)
	No Plan (n = 16)	22.88 (3.26)	16.63 (2.06)	117.25 (11.92)	112.31 (4.60)	86.81 (12.75)	15.75 (4.23)	17.38 (3.93)
Older	Plan (n = 15)	70.67 (4.50)	14.40 (2.64)	127.33 (8.98)	116.33 (3.90)	67.33 (17.06)	28.93 (8.28)	17.33 (3.33)
	No Plan (n = 15)	72.20 (4.90)	14.80 (2.93)	122.40 (9.69)	114.60 (3.40)	68.33 (14.73)	27.67 (12.94)	16.47 (3.18)

Note: WASI = Wechsler Abbreviated Scale of Intelligence; WTAR = Wechsler Test of Adult Reading.

A 2 × 2 ANOVA conducted on the WASI scores revealed no significant main effect of age group or condition or two-way interaction (p > .17). For the WTAR, a Box-Cox power transformation raised the scores to the power of λ = 3.00. Neither the main effects of age group or condition nor the age group x condition interaction were significant (p > .13). For Coding, there was a main effect of age group, F(1, 58) = 11.24, p < .01, η_p² = .26, where the older adults had significantly lower scores than the younger adults. There was not a main effect of condition or age group by condition interaction (p > .85). For the executive measures, the Stroop data were raised by a Box-Cox power transformation to the power of λ = 0.27. The ANOVA revealed a main effect of age group, F(1, 58) = 11.24, p < .01, η_p² = .26, where the older adults performed significantly more poorly than the younger adults. However, there was no significant effect of condition or 2-way interaction (p > .53). Finally, for the Tower Test, there were no significant main effects or age group x condition interaction (p > .50).

Multitasking efficiency

Figure 2 shows the mean multitasking efficiency scores for each age group across each condition. The ANOVA revealed a significant main effect of age group, F(1, 58) = 15.04, p < .001, η_p² = .15, and condition, F(1, 58) = 8.89, p < .005, η_p² = .06. There was also a significant age group by condition interaction, F(1, 58) = 5.27, p < .05, η_p² = .08. Post-hoc Tukey HSD tests revealed that younger adults achieved significantly higher efficiency scores than older adults in the no plan condition (p < .005). However, the two age groups did not significantly differ in the plan condition (p = .92). The older adults had significantly higher total scores in the plan condition than the no plan condition (p < .05). The younger adults did not significantly differ in their efficiency scores in the plan compared to the no plan condition (p = .99).

Figure 2. Multitasking performance for younger and older adults in the plan and no plan conditions. Top: Multitasking efficiency as measured by the total number of red items completed; Bottom: Overall multitasking performance as measured by the total number of points obtained. Error bars represent standard error of the mean.

Overall multitasking performance

Figure 2 also demonstrates the overall multitasking performance scores for each age group. A Box-Cox power transformation raised the scores to the power of λ = 1.24. A 2 (age: young vs. old) x 2 (condition: plan vs. no plan) ANOVA showed a significant main effect of age group, F(1, 58) = 16.85, p < .0005, η_p² = .20, and condition, F(1, 58) = 7.08, p < .05, η_p² = .05. There was also a significant age group by condition interaction, F(1, 58) = 3.95, p = .05, η_p² = .06. Post-hoc Tukey HSD tests revealed that younger adults achieved significantly higher total scores than older adults in the no plan condition (p < .001). However, the two age groups did not significantly differ in the plan condition (p = .57). The older adults had significantly higher total scores in the plan condition than the no plan condition (p < .05). The younger adults did not significantly differ in their total scores in the plan compared to the no plan condition (p = .99).

Individual subtask performance

Table 5 shows the means and standard deviations for the younger and older groups performing the subtasks during the plan and no plan conditions. As the data were not normally distributed for any of the subtasks even after Box-Cox power transformations, separate Mann Whitney U Tests were conducted. For the Beads task, older adults (Mdn = 38) performed significantly more poorly than younger adults (Mdn = 65) in the no plan condition, U = 206, p < .001. However, there was not a significant difference between the younger and older adults in the plan condition (p = .12). For the Letters task, separate Mann Whitney U Tests revealed that older adults (Mdn = 30) had a significantly lower Letters score than younger adults (Mdn = 42) in the no plan condition, U = 168.5, p = .05. In the plan condition, the younger and older adults did not significantly differ (p = .39). In the Bricks construction task, Mann Whitney U Tests demonstrated that our younger and older adults did not significantly differ on the plan (p = .90) or no plan conditions (p = .69). Finally, in the Telephone task, Mann Whitney U Tests showed that older adults (Mdn = 50) performed better in the plan condition than younger adults (Mdn = 40), U = 57, p < .001, but there was no difference between the groups in the no plan condition (p = .36).

Table 5. The means and standard deviations in parentheses for the younger and older adults performing the multitasking subtasks in Experiment 3.

	Younger Plan	Younger No plan	Older Plan	Older No plan
Beads	58.44 (23.30)	60.44 (14.86)	42.87 (21.70)	34.00 (20.74)
Letters	45.06 (13.38)	40.25 (13.54)	41.87 (12.55)	32.27 (14.26)
Bricks Construction	16.50 (8.16)	18.06 (9.04)	14.53 (8.45)	15.47 (7.31)
Telephone	38.25 (12.48)	41.25 (9.57)	46.73 (9.15)	36.07 (13.96)

Correlational analyses

Correlations between performance on the background tests and multitasking are reported in Table 6. Both multitasking efficiency and total scores significantly correlated with Coding, where the faster the speed of processing, the better the multitasking performance. None of the other background measures correlated with multitasking performance. Multitasking efficiency and total scores significantly correlated with one another. WTAR and WASI IQ significantly correlated with one another, and both executive measures (i.e., Stroop and Tower tests) correlated with Coding where the higher the executive performance, the faster the speed of processing.

Table 6. Correlation matrix of the cognitive test scores (n = 62).

	Measure	1.	2.	3.	4.	5.	6.	7.
1	WTAR		.58^a	−.16^a	.21^a	−.08	−.06^a	−.09^a
2.	WASI	<.0001		.08	−.05^a	.02	.23	.17^a
3.	Coding	.22	.53		−.50^a	.26^a	.40	.44^a
4.	Stroop	.10	.69	<.001		−.20^a	−.17^a	−.23^a
5.	Tower	.87	.87	<.05	.12		.23	.19^a
6.	Multitasking Efficiency	.66	.07	<.01	.20	.07		.97^a
7.	Multitasking Total Score	.50	.19	<.001	.08	.15	<.0001

Note: Pearson’s r or Spearman’s rho^a (upper diagonal) and ps (lower diagonal) are reported. WTAR = Wechsler Test of Adult Reading; WASI = Wechsler Abbreviated Scale of Intelligence.

Analysis of covariance

While Coding was found to be a significant covariate (p = .02), the ANCOVA result indicated that the significant main effects of age group, F(1, 57) = 7.47, p < .01, η_p² = .03, condition, F(1, 57) = 9.93, p < .01, η_p² = .05 and the age group by condition interaction, F(1, 57) = 5.41, p < .01, η_p² = .06, remained. The same result was found for the multitasking total scores where Coding was a significant covariate (p = .005) but the main effects of age group, F(1, 57) = 8.33, p < .01, η_p² = .04, condition, F(1, 57) = 8.90, p < .01, η_p² = .05 and the age group by condition interaction, F(1, 57) = 4.62, p < .05, η_p² = .05, remained.

Discussion

Experiment 3 examined whether encouraging older adults to plan prior to serial multitasking might improve their performance. The results found that, while older adults performed significantly more poorly than younger adults in the no plan condition, they performed as well as younger adults in the plan condition. The same pattern was found for both multitasking scores. Additional analysis of performance on the subtasks, age differences were not found on any subtask in the plan condition, except the Telephone task where older adults performed better than younger adults. In the no plan condition, older adults performed more poorly on the Beads and Letters tasks than younger adults. These results suggest that planning improves multitasking performance for older adults, and if given the opportunity to plan ahead, older adults are as efficient as younger adults on a test of serial multitasking.

While the current study found that preplanning improved older adults’ serial multitasking performance, we did not examine the efficiency of the plan or whether participants adhered to their plan while multitasking. Law et al. (2013) found that younger adults who began with a good plan performed better than those who began with a poor plan; however, participants who stuck with their original plan (regardless of whether it was good or bad) tended to perform better than participants who changed their plans while performing the task. Therefore, while preplanning did improve multitasking performance in older adults, future work might examine whether simply sticking to any plan, regardless of its quality, is sufficient to help performance.

Our younger and older adults did not significantly differ in terms of IQ based on their WASI or WTAR scores, and yet an age-related difference in multitasking (in the no plan condition) was still found. This suggests that intellectual abilities do not entirely explain the multitasking impairments in our older adults. Indeed, previous research has also reported multitasking impairments in frontal patients compared to healthy controls even after controlling for fluid intelligence (Roca et al., 2010). Therefore, our current study provides further evidence to support the claim that everyday serial multitasking situations may not be associated with individual differences in fluid intelligence (Kievit et al., 2014).

In addition, performance on the multitasking paradigm did not correlate with performance on the traditional tests of executive functions (i.e., Stroop or Tower Test). The lack of correlations among different executive tests supports previous findings in the literature (e.g., Argiris et al., 2020; Duncan et al., 1997; Miyake et al., 2000). However, executive tests may not correlate for other reasons such as low reliability, different strategy use, and task impurity (Burgess, 1997; Shallice & Burgess, 1996; Stuss & Alexander, 2000). Yet, studies have also shown that patients with impaired executive functioning in everyday life often perform normally on executive function tests but are impaired in serial multitasking (Eslinger & Damasio, 1985; Goldstein et al., 1993; Chevignard et al., 2000; Shallice & Burgess, 1991). Our results support the notion that tests that try to simulate real-life executive abilities may provide unique information about everyday functioning not obtained by traditional executive tests.

Slower processing speed was associated with poorer multitasking performance, which is likely due to certain subtasks being dependent on an individual’s motor abilities to perform them effectively. As deterioration in motor control occurs in healthy adult aging (Smith et al., 1999), including coordination difficulties (Seidler et al., 2002), more erratic movements (Darling et al., 1989), and slowing of movement (Diggles-Buckles, 1993), these motor deficits might explain the age differences in multitasking. However, as controlling for processing speed did not remove the effects of age and condition, age-related slowing in mental processing or motor abilities cannot be the only underlying cause for the multitasking impairment in our older adults.

General discussion

To our knowledge, this is the first study to examine the influence of interruptions and planning on serial everyday multitasking in healthy adult aging using paradigms typically used in the neuropsychological literature. The results of the above three experiments demonstrated that, while older adults performed more poorly than younger adults on a prop-based multitasking paradigm, significant age-related differences were not found when participants experienced interruptions (Experiment 2) or were encouraged to preplan prior to multitasking (Experiment 3). In the real world, older adults appear able to carry out everyday tasks that rely on multitasking such as preparing a meal or shopping for food with minimal difficulty (Phillips et al., 2006). Our study suggests that both preplanning and experiencing interruptions support older adults in their ability to successfully multitask in the real world compared to the laboratory.

In all three experiments, in conditions where there was no intervention, age-related differences on the Law et al. (2004) multitasking paradigm were consistently found. This finding is in contrast to previous work, including our own, where age differences on certain aspects of multitasking are not found when using more naturalistic or prop-based tasks (Garden et al., 2001; Kosowicz & MacPherson, 2017; Levine et al., 1998; Schmitter-Edgecombe et al., 2012). The age differences in the current study might be explained by the types of subtasks adopted. For example, tasks such as threading beads onto a string require the manipulation of small objects, which is more difficult for older adults due to their slower coordination of bimanual movements (Stelmach et al., 1988; Wishart et al., 2000). Older adults may also be less familiar with certain tasks as they do not typically perform them in everyday life such as constructing a tower from Lego™ bricks. Future work might explore whether these same benefits of interruptions and planning are found using other serial multitasking paradigms.

Serial multitasking tends to be studied within the field of neuropsychology and few such paradigms currently exist in the literature to assess serial multitasking. In the current study, we adopted the paradigm used by Law et al. (2004), as it was designed to be more challenging for healthy individuals. In Experiment 2, analysis of performance on the subtasks suggests that the benefit of an interruption on older adults’ multitasking performance is driven by the Telephone task. One might argue that younger adults are likely to have less experience of looking up numbers in a telephone book compared to older adults and so this subtask biases older adults. Yet, we did still report age differences on the Telephone task in Experiment 1 and Experiment 2 in the no interruption condition, where younger adults performed significantly better than older adults. Moreover, Experiment 2 was only adequately powered to reliably reveal large-sized effects; our sample size prevented the reliable estimation of small and medium effects. Therefore, we may not have had sufficient power to identify more subtle effects on the other subtasks. Indeed in Experiment 3, which was better powered, a benefit of planning was found for all subtasks except the Bricks construction task. Future work should examine performance on the subtasks in a larger sample, as well as recording completion times for the different subtasks to examine age differences in the amount of time that it takes to complete each subtask and the amount of time spent on each subtask.

While the current study has focused on interruptions (Experiment 2) and planning (Experiment 3), there are likely other factors which might support older adults’ ability to multitask in everyday life. For example, in real-life, older adults are likely to rely on external cues such as timers to remind them when to perform certain tasks. Moreover, while our prop-based task is considered more naturalistic than computerized multitasking assessments such as the Breakfast task (Craik & Bialystok, 2006), we cannot ultimately determine the ecological validity of our paradigm without comparing performance against more real-world multitasking assessments (e.g., Garden et al., 2001; Shallice & Burgess, 1991). The ecological validity of assessments is an important factor when assessing older adults (Phillips et al., 2012).

Moreover, while benefits of interruptions and planning were found using our prop-based multitasking paradigm, these same benefits may not be found when using computer-based tasks. Previous studies examining multitasking in older adults using computerized tasks versus prop-based (Kosowicz & MacPherson, 2017) or virtual (Feinkohl et al., 2016) tasks have shown that older adults perform more poorly when assessed using computer-based tasks. More recently, some instruments to assess cognitive abilities have been replaced by computerized tasks (e.g., Deary et al., 2011; Kush et al., 2012; Logie et al., 2011) to improve the accuracy and ease of data collection, the reliability of use, and to avoid biases, yet poor performance in older adults on computer-based tasks may not always reflect true deficits and so should be interpreted with caution by clinicians and researchers (Feinkohl et al., 2016). Understanding whether these same advantages are experienced by older adults when using computerized tasks would be an important question for future work.

One limitation of the present study was that we did not include any background measures in Experiments 1 and 2 due to time limitations. However, we were able to include measures of executive function, IQ and processing speed in Experiment 3. Here, we found that multitasking performance was not correlated with IQ or executive abilities, which is in line with previous studies in the literature that suggest that multitasking and executive abilities (Eslinger & Damasio, 1985; Goldstein et al., 1993; Chevignard et al., 2000; Shallice & Burgess, 1991) and IQ (Kievit et al., 2014; Roca et al., 2010) are separate cognitive abilities. There was however an association between multitasking and processing speed but again, when we controlled for processing speed, the age differences in multitasking were still found, suggesting that speed cannot account entirely for the poorer performance of our older adults.

It should be noted that our study included mainly female participants who were recruited via convenience sampling. Moreover, our participants were highly educated with the lowest level of education being 10 years and a mean of around 14 years. This may limit the generalizability of our results to males and lower educated individuals. Our younger adults had significantly more years of education than our older adults, due to our younger group tending to be students who were coming to the end of their undergraduate degree or who were completing a postgraduate degree. While we did not have background neuropsychological data for our participants in Experiments 1 and 2, our older adults did show age differences on our coding and Stroop measures in Experiment 3 suggesting that our older adults are typical of the older population. Despite these limitations, our results indicate that interruptions and planning play an important role in improving serial multitasking abilities in older adults and highlight the need for future research in this area.

Taken together, we have highlighted the advantage of interruptions and planning on serial multitasking performance in older adults. While older adults performed more poorly than younger adults on a prop-based multitasking paradigm, both interruptions and preplanning improved performance so age differences were no longer found. These findings have real-life applicability in terms of offering ways that older adults’ multitasking abilities might be improved in everyday life. However, future work should examine whether these same benefits are found using other prop-based and computerized multitasking paradigms.

Acknowledgments

Thanks to Brandon Gautier for his valuable help with data collection.

Disclosure statement

We have no known conflict of interest to disclose.