MSc Brain and Cognitive Sciences
Institute for Interdisciplinary Studies
Literature thesis
Technology-based multitasking
from the lab to the real world
by
Eva Abels
12098396
February, 2020
12 ECTS
Assessor:
Examiner:
Abstract
Modern-day technology is enmeshed in our everyday lives and has become the greatest multitasking enabler. In the laboratory, dual-task and task switching research has demonstrated that task performance worsens when performed in combination with another task. Multitasking studies have also been conducted in “real-world” settings, including study and work environments. However, laboratory and real-life multitasking differ in various ways, and it is unclear whether real-world research has added value. Real-world studies investigating multitasking in- and outside the classroom suggest that using task-irrelevant technologies while studying has detrimental effects on academic performance, and research conducted in the workplace indicates that multitasking behavior is promoted by workflow interruptions, which deteriorate perceived performance and task accomplishment. Although laboratory and real-world studies each have their advantages and disadvantages, both research domains have contributed to our understanding of multitasking. In general, conclusions that were drawn from real-world studies are in accordance with the knowledge that was acquired in the laboratory, with the exception of cumulative interruptions. This aspect of multitasking behavior has been and can be studied in real-life settings only, suggesting that real-world studies have limited added value to multitasking research.
1. Introduction
In our fast-paced, electronic society, doing several things at the same time has become an ordinary phenomenon. Examples abound as teenagers do homework while managing social media accounts, students listen to a lecture while checking incoming messages, and office workers talk on the phone while keeping track of emails. However, performing multiple tasks simultaneously is not a new concept. Humans have been multitasking for as long as they have had competing needs, and psychologists have conducted experiments on the nature and limits of human multitasking since the 1960s. Rapid advancement in information and communication technology (ICT) throughout the last few decades has greatly encouraged multitasking behavior, and technology has come to pervade the way we study and work. In an increasingly digitized world, technology has become the greatest multitasking enabler and amplifier.
Multitasking is commonly defined as performing more than one task at the same time (Pashler, 1994). Consequently, multitasking individuals must decide how to allocate limited resources (e.g., time and attention) across multiple competing tasks to achieve their goals (Adler & Benbunan-Fich, 2012). Although the upside of multitasking is the impression of enhanced productivity, the downside appears to entail a detrimental impact on performance (David, Xu, Srivastava, & Kim, 2013).
The effects of multitasking on task performance have been studied extensively in the laboratory (for a review see Koch, Poljac, Müller, & Kiesel, 2018). These studies have shown that individual task performance generally decreases when performed in combination with another task, compared to when performed in isolation. Laboratory experiments typically study concurrent or sequential multitasking behavior by examining dual-task or task switching performance, respectively (Koch et al., 2018).
In addition to laboratory research, multitasking studies have been conducted in “real-world” settings, such as study and work environments. For students of all ages, technology is enmeshed in their everyday lives, both at school and at home (Carrier, Cheever, Rosen, Benitez, & Chang, 2009). Research has shown that students multitask with ICTs not only while doing homework, but also during class (May & Elder, 2018). Multiple studies have indicated that the off-task use of digital technologies while studying is negatively associated with academic performance (e.g., Downs, Tran, McMenemy, & Abegaze, 2015; Wood et al., 2012).
At work, individuals are frequently interrupted face-to-face and by ICTs, which promotes multitasking behavior. The high daily demand on resources needed to multitask increases psychosomatic stress, diminishes job satisfaction and ultimately reduces performance (Kirchberg, Roe, & Van Eerde, 2015). Research conducted in work environments has demonstrated that multitasking and being interrupted during the workday deteriorate perceived performance (Kirchberg et al., 2015;
Multitasking research has thus been conducted in in the laboratory and in the real world, but these laboratory and real-life forms of multitasking differ in several ways, such that multitasking in reality might not resemble these laboratory experiments at all. For instance, multitasking in the real world is much more complex and might take various forms, and laboratory experiments might not reflect what is actually happening (Carrier, Rosen, Cheever, & Lim, 2015). Moreover, findings from real-world studies may not have contributed anything to the knowledge that was already acquired in the laboratory. In other words, it is unclear whether real-world studies have added value.
Therefore, the aim of this literature review is to synthesize existing research on the effects of multitasking on performance in the laboratory and in the real world, and investigate to what extent real-world studies have added value to multitasking research. Given the pivotal role of ICTs in modern-day multitasking, this paper is mainly focused on technology-based multitasking behavior.
A review of the literature was conducted in the fall of 2019 through PubMed by applying a “snowball” method of using the most recent works to find citations provided in them. Search words and phrases included multitasking, cognition, student, academic performance, work interruptions, job, office and technology. This review analyzed 38 real-world research reports from 2009 to 2019 that primarily investigated the effects of multitasking on performance in study and work environments. Papers were excluded if they did not involve technology-based multitasking behavior, if they were published before 2009 (with the exception of theoretical articles), if they investigated multitasking with music, and if they took place outside a typical study or work setting.
The framework of this literature review is depicted in Figure 1. This paper starts by examining prominent areas of laboratory research in chapter 2, including dual-task and task switching paradigms, the multitasking continuum, interruption research, and training and media multitasking. Then, an overview of real-world papers in study and work environments is provided in chapter 3, each followed by remarks on methodology and (additional) findings, ending with a comparison between study and work multitasking reports. Finally, similarities and (dis)advantages of laboratory and real-world studies are discussed in chapter 4, resulting in a value judgement of real-world research, and afterwards some implications are given for students, office workers and future research.
2. Multitasking in the laboratory
Multitasking has been studied in a wide range of experimental paradigms and multiple theories to characterize multitasking behavior have been established. Prominent examples of such paradigms and theories include dual-task and task switching, the multitasking continuum and interruption research. In addition, recent laboratory studies have focused on a newly emerging variant of multitasking with multiple media sources. These subfields of laboratory multitasking are examined in this chapter, because they all involve some form of performance and technology use or have the potential to do so.
2.1. Dual-task and task-switching paradigms
The majority of laboratory experiments can broadly be divided in dual-task studies, in which tasks are performed concurrently, and task switching experiments, in which task are performed sequentially (for a review see Koch et al., 2018). These different multitasking conditions have differential effects on performance, which can be interpreted as interference or costs and are commonly measured by increased reaction times or reduced accuracy.
Dual-task paradigms involve concurrent performance of two tasks. In single-task blocks, only one task set needs to be kept in working memory, which is a cognitive system with limited capacity to temporarily hold information available for processing (May & Elder, 2018). In dual-task blocks, both task sets require active maintenance, resulting in dual-task costs.
Dual-task performance can be assessed in the psychological refractory period (PRP) paradigm, which consists of two speeded tasks performed in combination (Box 1; Pashler, 1994). Specifically, the stimulus onset asynchrony (SOA) of the stimuli of the tasks is varied, in a way that two stimuli can be presented at the exact same time or with large temporal difference. The typical finding is that variations in SOA do not affect performance on the first task, but performance on the second task decreases with shorter SOA (Koch et al., 2018).
An example of a dual-task experiment involved simultaneous performance of two choice-reaction time tasks (Box 1) containing real-world elements, resembling a crosswalk situation (Allen, Lien, Ruthruff, & Voss, 2014). Participants had to discriminate between colored signals (green = “go” and red = “wait”; similar to pedestrian signals) and tones (white noise and a honking horn; similar to traffic sound), with varying SOA between the tasks. As hypothesized, dual-task interference was observed, with deteriorated performance in the second task (Allen et al., 2014).
Task switching paradigms include sequential performance of two tasks, which can be presented in two types of blocks (Strobach, Liepelt, Schubert, & Kiesel, 2012). In single-task blocks, either the first or the second task is presented exclusively. In mixed-task blocks, participants are required to perform both tasks, as indicated by a predefined sequence (Rogers & Monsell, 1995) or a
Manzey, 2016). Both task sets must be maintained in working memory, because they will be needed again soon (Koch et al., 2018). Mixed-task blocks contain task switches and task repetitions between trials, which allows for assessment of two types of performance costs: mixing and switch costs (Box 1; Koch, Prinz, & Allport, 2005; Rogers & Monsell, 1995; Strobach et al., 2012).
Examples of task switching experiments include the number-letter task (Alzahabi & Becker, 2013; Minear, Brasher, McCurdy, Lewis, & Younggren, 2013; Ophir, Nass, & Wagner, 2009) and the dots-triangles task (Box 1; Baumgartner, Weeda, van der Heijden, & Huizinga, 2014). In such experiments, participants were required to respond to the same set of stimuli with varying task instructions. As expected, subjects showed longer reaction times when repeating tasks in mixed-task blocks than in single-task blocks, and responded slower when switching tasks compared to when repeating a task, indicative of mixing and switch costs, respectively.
Box 1
Laboratory tasks and measures as used in dual-task and task switching paradigms
• Choice-reaction time task = multiple stimuli each require a different response, and participants are required to react as fast as possible.
• Dots-triangles task = varying numbers of red dots and green triangles are presented, and participants are required to indicate whether there are more dots on the left or right side, or more triangles on the top or bottom part.
• Mixing costs = difference between performance in repetition trials within mixed blocks and performance in single-task blocks.
• Number-letter task = participants are required to categorize a number into even or odd while categorizing a letter into vowel or consonant.
• Switch costs = difference between performance in switch trials and performance in repetition trials within the mixed-task blocks.
As described above, numerous laboratory studies have identified performance decrements when two or more tasks have to be accomplished within a limited time period. Such findings have often been explained by referring to the bottleneck theory (Koch et al., 2018; Maslovat et al., 2013; Pashler, 1994). Bottleneck models theorize that the processing of multiple stimuli eventually reaches a processing filter, which allows the only one item at a time. The processing of the second task is then postponed or “queued” until the first task is completed, delaying the response. In other words, the limited-capacity processing bottleneck is simply unable to process two tasks in parallel and is thus serial (Koch et al., 2018; Pashler, 1994), questioning if concurrent multitasking is actually possible.
To summarize, dual-task experiments typically assess concurrent task performance in PRP-paradigms, while task switching studies investigate mixing and switch costs as indicators of sequential task performance. Such laboratory studies have also shown that performing multiple tasks is subject to limited-capacity bottlenecks, in which processing occurs strictly serial.
2.2. The multitasking continuum
Instead of an explicit separation, multitasking has also been described in terms of a continuum, as a function of time spent on one task before switching to another (Salvucci & Taatgen, 2011; Salvucci, Taatgen, & Borst, 2009). At one end of the continuum, there are tasks that involve constant switching (Fig. 2, left). This is similar to concurrent multitasking as tasks are, in essence, performed simultaneously. At the other end, tasks involve longer time spans between switches (Fig. 2, right), akin to sequential multitasking. According to the multitasking continuum, concurrent and sequential multitasking are not totally separate concepts, but can be represented on the same spectrum.
Figure 2. The multitasking continuum. Adapted from Salvucci and Taatgen (2009).
In addition, the multitasking continuum can overcome a major limitation of dual-task and task switching research. Although these study domains are characterized by high experimental control, they lack the freedom to self-organize how and when tasks are processed.On the basis of the multitasking continuum, different strategies with varying degrees of multitasking are feasible.
There are a few laboratory studies that investigated such different multitasking strategies. Two papers examined task performance in three multitasking conditions: discretionary, where participants were free to decide when and how frequently to switch tasks; mandatory, where they were forced to switch tasks at certain timepoints; and sequential, where they had to perform tasks in succession (Adler & Benbunan-Fich, 2015; Buser & Peter, 2012). Buser and Peter (2012) reported that subjects who executed the tasks sequentially outperformed those in the other two conditions. Adler and Benbunan-Fich (2015), however, found that mandatory multitasking led to lowest performance, but only when the task was considered difficult, because when the task was deemed easy, mandatory multitasking resulted in the best performance. It was argued that the perceived difficulty of the task might mediate the relationship between the degree of multitasking and task performance (Adler & Benbunan-Fich, 2015).
Another laboratory study investigated different multitasking strategies by analyzing individual response behavior (Reissland & Manzey, 2016). This resulted in the identification of three subgroups based on their individual multitasking strategies, but no performance differences were reported
between groups. Nevertheless, this study showed that there are differences in individual preference for multitasking strategies.
In conclusion, the multitasking continuum allows more freedom to apply strategies with different degrees of multitasking, although laboratory studies that investigated such varying multitasking strategies have found inconclusive results with respect to task performance.
2.3. Interruptions
Other than differences in multitasking strategies, there are various incentivesto interrupt one task to perform another. Task interruptions have been studied extensively in the laboratory and are defined as events that cause interference with the primary task (Fletcher, Potter, & Telford, 2018). Interruptions typically occur in the following order: engagement in the primary task, which is suspended temporarily by the secondary task, and after completion of the secondary task, the primary task is resumed (Fig. 3; Couffe & Michael, 2017; Katidioti, Borst, & Taatgen, 2014). Being interrupted while performing a task generally results in longer task completion time, which can be explained by interruption and resumption lags (Fig. 3; Lee & Duffy, 2015). For example, longer resumption lags following interruptions were suggested to reduce the chance of making mistakes (Brumby, Cox, Back, & Gould, 2013).
Interruptions studies are particularly relevant with regard to ICTs, since technological developments have expanded the opportunities for task interleaving, resulting in more interruptions and more choices for users (Duggan, Johnson, & Sørli, 2013). An example of a laboratory study using technology-based interruptions involved carrying out a visual pattern-matching task on a computer while communicating with a partner via instant messaging (IM) or online voice chat (Wang et al., 2012). In both communication conditions, interruptions decreased pattern-matching performance.
Moreover, the difference between human and digital interruptions has been explored as well, and it was reported that human interruptions resulted in much shorter interruption lags than interruptions by ICTs (Nees & Fortna, 2015). Apparently, technological interruptions offer more temporal flexibility in how and when to manage the interruption, while human interruptions seem to require a more acute response.
Figure 3. The time course of a task interruption. Task A = the primary task; task B = the secondary task; interruption lag = the
time between the interruption and the start of the secondary task, resumption lag = the time between completing the secondary task and returning to the primary task.
Previous multitasking research has identified two types of interruptions that lead to the suspension of the current task to attend to another: internal and external interruptions (Jett & George, 2003). The former is defined as voluntary decisions due to personal choices or thought processes, while the latter refers to external signals, notifications or environmental triggers. Interestingly, it was reported that reaction times were longer after internal interruptions compared to external interruptions (Katidioti et al., 2014).
Laboratory research has mainly focused on studying external interruptions by using fixed switches, while multitasking in real life is often caused by voluntary self-interruptions (Courage, Bakhtiar, Fitzpatrick, Kenny, & Brandeau, 2015). Nonetheless, there exist some studies that did investigate internal interruptions in the laboratory. For instance, in the studies by Adler and Benbunan-Finch (2012, 2013) participants were presented with six problem-solving tasks (Box 2). They were required to solve these problems with a fixed duration for each task, but were free to switch whenever they wanted, similar to the above-mentioned discretionary multitasking condition. It was reported that accuracy declined when more switches were made, such that high multitaskers scored lowest compared to other multitaskers and non-multitaskers (Adler & Benbunan-Fich, 2012, 2013). Self-interruptions due to negative triggers were associated with increased switching behavior, and it was suggested that negative self-interruptions may unleash a downward spiral that ultimately degrades performance (Adler & Benbunan-Fich, 2013). Remarkably, from the subjects who reported not having interrupted themselves, only half actually had no self-interruptions, indicating that subjects cannot adequately assess their own behavior (Adler & Benbunan-Fich, 2013).
To sum up, interruptions generally delay task completion and laboratory studies have mainly focused on external interruptions, but insights into internal interruptions have been provided as well. Box 2
Description of the six problem-solving tasks in the studies of Adler and Benbunan-Fich (2012, 2013)
The primary task was a Sudoku puzzle of medium difficulty level. The goal of a Sudoku problem is to insert the numbers 1 to 9 into all the boxes in a 9-by-9 grid, in a way that each row, column, and 3-by-3 box contains each number only once.
There were five additional tasks of shorter duration: one textual, two numeric series and two visual. The textual task was a word production exercise in which participants were required to create 20 new words by unscrambling the letters of a given word. The numeric series tasks were number series problems in which participants were required to fill in the missing number in the sequence. The visual tasks were odd one out challenges, where subjects were required to find the shape that did not match the pattern presented in a set of four other shapes. For the numeric series and visual tasks, there were two tasks sets, each consisting of ten problems.
2.4. Training and media multitasking
As extensive practice generally leads to performance improvement, laboratory studies have investigated potential training effects of multitasking. For example, Strobach and colleagues (2012) reported reduced mixing and switch costs after eight practice sessions. Despite such indications that extensive training can decrease the associated costs, it is unclear whether this reflects a general development of multitasking skills or a specific learning effect that applies only to the trained task (Enriquez-Geppert, Huster, & Herrmann, 2013). Overall, studies have indicated that laboratory training could help overcome some cognitive bottlenecks for the relevant tasks, but results in limited transfer when other multitasking activities are involved (Cardoso-Leite, Green, & Bavelier, 2015).
Another form of multitasking training is related to the habit of concurrently using multiple media sources, referred to as media multitasking, which has been of great interest in recent laboratory research. A diary study revealed that the average time spent on media is nearly seven hours per day (Voorveld & van der Goot, 2013), and more than four switches per minute were observed for concurrent computer and TV use (Brasel & Gips, 2011). Modern-day technology enables multitasking thus not only in the form of an off-task activity, but also with multiple media streams, such as scrolling through Facebook, checking incoming messages and watching television at the same time.
With regard to practice effects of media multitasking, two opposing hypotheses have been proposed: the scattered attention and trained attention hypothesis (Van Der Schuur, Baumgartner, Sumter, & Valkenburg, 2015). According to the former, media multitasking accelerates the depletion of the attentional resource, which decreases performance of the primary task, whereas the latter states that frequent media multitasking positively impacts performance via training effects and enhancement of cognitive processes.
Investigating the effects of media multitasking on task performance has become increasingly popular in multitasking research in the laboratory. In the pioneering study of Ophir and colleagues (2009), the relationship between chronic media multitasking and cognitive control was investigated. To discriminate between light and heavy media multitaskers (LMMs and HMMs, respectively), the questionnaire-based media multitasking index (MMI) was developed. It should be noted, however, that the MMI has been modified and updated in later research, due to the ongoing development in everyday technology use (Elbe, Sörman, Mellqvist, Brändström, & Ljungberg, 2019).
As HMMs perform more multitasking behavior on a daily basis, they were intuitively assumed to be better multitaskers than LMMs. Nevertheless, results showed that HMMs had worse cognitive control abilities than LMMs, which was attributed to HMMs’ possibly greater susceptibility to interference from irrelevant environmental stimuli and memory representations. Moreover, it was argued that media multitasking might adjust cognitive control strategies, with HMMs focusing more on breadth and LMMs more on depth (Ophir et al., 2009).
Replication and follow-up laboratory studies have yielded inconsistent results (for reviews see Cardoso-Leite et al., 2015; May & Elder, 2018; Van Der Schuur et al., 2015). Some studies supported the findings of Ophir et al. (2009), and found that HMMs exhibited lower working memory performance (Uncapher, Thieu, & Wagner, 2016) and deficits in implicit learning (Edwards & Shin, 2017), as compared to LMMs and a third group of medium media multitaskers (MMMs). Another study reported decreased cognitive performance of HMMs, though MMMs occasionally outperformed both LMMs and HMMs, indicating that the effects of increasing media multitasking behavior might be nonlinear (Cardoso-Leite et al., 2016). Together, these findings suggest that long-term media multitasking is associated with a wider attentional scope, which may permit task-irrelevant information to compete with task-relevant information, thereby advocating for the scattered attention hypothesis.
Other laboratory studies failed to provide evidence for cognitive decrements in HMMs, and instead presented no between-group differences or even increased performance of HMMs. In a replication study by Minear et al. (2013), no differences were reported between HMMs and LMMs, and another report found comparable dual-task scores but better task switching performance of HMMs (Alzahabi & Becker, 2013). HMMs reported having more problems in executive domains in their everyday lives, but were better at ignoring irrelevant distractions (Baumgartner et al., 2014). In addition, higher levels of media multitasking were related to increased multisensory integration (Lui & Wong, 2012) and lower switch costs in attention-shifting tasks (Elbe et al., 2019). The fact that HMMs are not deficient in all cognitive tasks suggests that media multitasking might have some beneficial effects, which supports the trained attention hypothesis (Van Der Schuur et al., 2015).
A recent meta-analysis found limited evidence for the association between media multitasking and information processing tasks in the laboratory, questioning whether this relationship exists at all (Wiradhany & Nieuwenstein, 2017). It was suggested that media multitasking might be related to problems of inattention rather than general cognitive errors (Ralph, Thomson, Cheyne, & Smilek, 2014). Alternatively, preexisting individual differences in personality traits or underlying neurocognitive profiles may lead to distinct patterns of media use (Uncapher & Wagner, 2018).
In summary, the newly emerging form of multitasking with multiple media streams has recently received much attention, but laboratory research has provided mixed support for the scattered attention and trained attention hypotheses. Longer follow-up studies are needed to further investigate whether repeated practice of media multitasking will improve performance over time.
3. Multitasking in the real world
In addition to in the laboratory, multitasking has also been investigated in real world settings, such as study and work environments. On a daily basis, students and office workers commonly engage in multitasking behavior with ICTs (Kirchberg et al., 2015; May & Elder, 2018; Voorveld & van der Goot, 2013). Although electronic devices such as laptops and smartphones facilitate collaboration and provide access to (online) resources, they serve as a major source of distraction as well, with the potential to affect task performance.
This chapter provides an overview of real-world multitasking research, divided into study and work environments. For study settings, in-class and homework studies are described separately. Lastly, study and work multitasking are compared.
3.1. Multitasking in study environments
At school and in college, long spans of focused attention are required for studying and learning (Carrier et al., 2015). However, students are surrounded by real-time digital information that constantly places demands on their attention and interferes with their focus. It was reported that students work on a task on average for less than six minutes before switching to another, which was mostly due to distractions from social media and texting (Rosen, Carrier, & Cheever, 2013). Studying while also engaging in ICTs for nonacademic purposes is believed to have a detrimental impact on learning, as multiple tasks must compete for the limited cognitive capacity of the learner (Zhang, 2015).
Students were reported to engage in multitasking behavior in the classroom as well as while doing homework, both of which affect academic performance, although in a different manner. While doing homework, students can compensate for deficiencies in their understanding by re-doing a task, but this is not possible inside the classroom (May & Elder, 2018). This was explained by in-class time constraints, which are particularly prevalent in situations where the lecturer determines the pace and amount of instruction.
Twenty-five real-world reports investigating multitasking in study environments met the inclusion criteria of this paper, which were categorized based on their study designs. In-class and homework studies are examined separately.
3.1.1. Multitasking studies in the classroom
An overview of in-class multitasking studies is presented in Table 1, including one observational, six cross-sectional and ten experimental reports.
The observational study reported that most students were distracted by the internet during a lecture, though there were no differences in post-lecture test performance between distracted and non-distracted students (Nalliah & Allareddy, 2014).
Results from cross-sectional studies indicated that in-class multitasking was negatively associated with grade point average (GPA; Burak, 2012), and engaging in internet activities for nonacademic purposes was related to lower grades (Ravizza, Hambrick, & Fenn, 2014; Zhang, 2015; Zureick, Burk-Rafel, Purkiss, & Hortsch, 2018). In-class texting frequency was negatively associated with class grade, but not with GPA (Clayson & Haley, 2012), whereas another study reported a negative relationship with GPA for both text messaging and Facebook use (Junco, 2012).
Experimental studies reported that students who performed additional non-lecture activities in the classroom scored lower on post-lecture tests than students who only took notes (Sana, Weston, & Cepeda, 2013), which was also observed in conditions with phone ringing (End, Worthman, Mathews, & Wetterau, 2010), distraction by Facebook (Downs et al., 2015; Marone, Thakkar, Suliman, O’Neill, & Doubleday, 2018; Wood et al., 2012) and texting (Ellis, Daniels, & Jauregui, 2010; Gingerich & Lineweaver, 2014; Kuznekoff & Titsworth, 2013; Rosen, Lim, Carrier, & Cheever, 2011). A study that investigated the context of texting and Twitter messages found that sending and receiving irrelevant messages negatively impacted performance, while messages related to lecture content had no effect (Kuznekoff, Munz, & Titsworth, 2015). Moreover, students who received and sent more words in their texts scored lower, though this was moderated by the elapsed time, with longer delays resulting in increased performance (Rosen et al., 2011). This suggests that message content and elapsed time may play a role in the relationship between multitasking with texting and academic performance.
In summary, most studies reported that in-class multitasking is adversely related to measures of academic performance and reduces performance on post-lecture tests.
3.1.2. Multitasking studies while doing homework
Research reports investigating multitasking while doing homework are shown in Table 2, including one observational, four cross-sectional, two experimental and one longitudinal study.
The observational study reported that accessing Facebook at least once during a 15-minute study session at home was related to lower GPA (Rosen, Carrier, et al., 2013).
Cross-sectional research indicated that using Facebook (Gabre & Kumar, 2012) and/or texting (Junco & Cotten, 2012) while doing schoolwork was negatively associated with GPA, and students who frequently use instant messaging (IM) while studying reported lower perceived academic outcomes (Junco & Cotten, 2011). Besides, students that performed heavy media multitasking while doing homework had lower class grades than students who did not perform heavy media multitasking (Martín-Perpiñá, Poch, & Cerrato, 2019).
Experimental studies found that students who were IMing while reading took longer to finish the passage compared to students who were not simultaneously reading and IMing, but no differences
Crawford, 2009). This indicates that students can compensate for IMing by re-reading (parts of) the passage, in accordance with the reasoning of May and Elder (2018).
Results from the longitudinal study showed that some forms of media consumption in the first semester, including cell phone use, social networking, television and movie viewing, were negatively associated with GPA in the second semester (Walsh, Fielder, Carey, & Carey, 2013). This suggests that specific forms of media use might decrease later study performance.
To conclude, multitasking while doing homework was mostly negatively related to (long-term) academic performance, although homework settings may provide opportunities to compensate for knowledge deficits.
3.1.3. Remarks on real-world multitasking in study environments
Altogether, real-world multitasking studies that were conducted in- and outside the classroom indicate that using ICTs for nonacademic purposes while studying has detrimental effects on academic performance. However, it is important to note a few methodological issues and (additional) findings with reference to self-report measures, surroundings, lecture and sample variations, and attitudes.
Most studies used self-report measures to examine ICT use, but these can lead to overestimates and can be inadequate. Although significant correlations were found between students’ reported and actual internet time (Moreno et al., 2012) and Facebook use (Junco, 2013), self-reported time was estimated much higher than the actual time, with a respective difference of 2.5 and 2 hours. This implies that students cannot accurately assess their own ICT use, leading to an overestimation of their true behavior and possibly inflates effect sizes. In addition, not all self-report measures could adequately assess multitasking, as some surveys did not permit occurrence of more than two activities at the same time (Burak, 2012; Walsh et al., 2013) while this is feasible in multitasking scenarios. Therefore, a reliable, universal self-report measure for multitasking while studying should be developed, perhaps separately for in- and outside the classroom.
Within a classroom, not only students’ own multitasking behavior, but also their surroundings are believed to affect academic performance. In an additional experiment of Sana et al. (2013), it was reported that students who were in direct view of a multitasking peer performed worse compared to those who were not. However, another study found that performance for students who took notes on paper did not differ between laptop-free and laptop-permitted zones, indicating that laptop use did not impair the achievement of surrounding students (Aguilar-Roca, Williams, & O’Dowd, 2012). Thus, findings with regard to proximity to multitasking behavior remain inconclusive and should be further explored in future research.
An important methodological issue among in-class reports is variation in lecture duration and modality. The duration ranged from one lecture of 12 minutes to three lectures of 20 minutes or was
in a few cases even unknown, and lectures were presented live in some studies while video podcasts were used in others. Since performance decrements were observed in most multitasking conditions, these differences may have limited impact. However, increased multitasking behavior was observed in online lectures compared to live presentations (Burak, 2012), and students who always attended live lectures outperformed students who always watched videos (Zureick et al., 2018). This suggests that lecture modality could mediate the relationship between multitasking and academic achievement, such that video lectures promote multitasking behavior, which in turn has a stronger negative impact on learning.
Another source of variation between studies were differences in sample size and composition. Importantly, sample characteristics such as sex and mean age were in a number of cases not provided at all. Since students were used as the sample in all studies and findings are fairly consistent, this matter may be of little concern.
An apparent inconsistency was reported with respect to students’ attitudes and beliefs. Students claimed to be aware of the detrimental influence on performance and agreed that they should not use technologies for nonacademic purposes while studying; yet, they still continued to engage in the behavior (Clayson & Haley, 2012; Junco & Cotten, 2011; Rosen et al., 2011). Perceived multitasking abilities were found to decrease from before to after an in-class multitasking experiment (Downs et al., 2015), whereas a weeklong intervention specifically designed to raise awareness for the detrimental effects of using ICTs while learning failed to changes students’ attitudes (Terry, Mishra, & Roseth, 2016). Despite the fact that students seem to recognize that multitasking degrades academic performance, it remains difficult, if not impossible, to change their behavior. Hence, multitasking while studying appears to be a persistent problem.
In short, variations in lecture duration and study sample may be of minor importance, but lecture modality and students’ attitudes appear to be greater issues. Future directions include the development of universal self-report measures for student multitasking and further unravelling the impact of being surrounded by multitaskers.
3.2. Multitasking in work environments
Besides study environments, the workplace is another real-world setting in which technology-based multitasking is highly prevalent. The degree of multitasking behavior at work is promoted by multiple factors, with workflow interruptions being the most prominent. On a daily basis, office workers experience frequent interruptions, with an interruption rate of 12.5 per hour (Cades, Werner, Boehm-Davis, & Arshad, 2010) and a regular working episode duration of less than three minutes (Wajcman & Rose, 2011). The high demands on cognitive resources necessary for interruption management can
Moreover, it was found that interruptions at work are predominantly mediated through technology rather than face-to-face (Wajcman & Rose, 2011). Although ICTs offer flexibility and control to employees, they also experience being electronically “tied” to work, for instance via email (Schlachter, McDowall, Cropley, & Inceoglu, 2018). In fact, it was reported that per day, employees spent nearly one and a half hours on email and checked their inboxes seventy-seven times (Mark, Iqbal, Czerwinski, Johns, & Sano, 2016). Such behavior may increase technostress, referring to feelings of stress due inability to cope with ICTs, which can negatively affect job performance (Yin, Davison, Bian, Wu, & Liang, 2014).
3.2.1. Multitasking studies while working
An overview of multitasking research in the workplace is presented in Table 3. Thirteen real-world reports investigating multitasking in a work environment were included, of which seven were observational, five experimental and one longitudinal.
Observational studies reported that multitasking during the workday was negatively related to performance (Kirchberg et al., 2015) and that interruptions were perceived as detrimental to task accomplishment (Sonnentag et al., 2018) and task performance (Pachler et al., 2018). Specifically, resumption lags following external interruptions were longer compared to when the interruption was self-initiated, suggesting that it takes longer to resume a task after an external alert (Cades et al., 2010). Time spent on email was negatively related to productivity, and employees who used self-interrupting and batching strategies had higher productivity scores with longer email duration compared to those who relied on notifications (Mark et al., 2016). In contrast, a positive correlation between multitasking ability and performance was observed as well (Sanderson, Bruk-Lee, Viswesvaran, Gutierrez, & Kantrowitz, 2013), and another study reported that office workers experienced interruptions as part of their job and were not negatively affected by them (Zoupanou, 2015).
In experimental conditions without email, employees switched less frequently and spent more time on each window before switching, compared to their baseline email management style (Mark, Voida, & Cardello, 2012). When email was checked only once a day, office workers spent less time on email compared to when email was checked regularly (Bradley, Brumby, Cox, & Bird, 2013). However, another study reported no differences in concurrent task performance between batching and continuous email strategies, though batching email was related to reduced email time (Akbar et al., 2019). In study with fixed window switches, analogous to the aforementioned mandatory multitasking condition, different window switching conditions did not result in performance differences, but the use of virtual desktops was found to reduce resumption lags (Jeuris & Bardram, 2016). A follow-up study with one window switching sequence reported that half of the switches contained one or more errors, indicating that task switching decreases performance (Jeuris, Tell, Houben, & Bardram, 2018).
The longitudinal study followed one sample for five years and another for eight months, and in both groups, increases in interruptions over time reduced job satisfaction (Keller, Meier, Elfering, & Semmer, 2019), which may eventually impair work performance (Kirchberg et al., 2015).
To be brief, multitasking at work is promoted by workflow interruptions, commonly via email, and most studies reported a negative impact of multitasking and interruptions on (perceived) performance.
3.2.2. Remarks on real-world multitasking in work environments
Overall, real-world multitasking research that was conducted in the workplace indicates that multitasking and workflow interruptions affect performance negatively. However, some methodological matters and (additional) findings with respect to sample compositions, interruption sources, performance measures, personal variables and email strategies must be discussed.
The samples used in the workplace studies samples consisted of working individuals, but the exact sample compositions are disputable. The mean age showed large variability between studies or was in a few cases not provided at all, and the samples mostly consisted of people with white collar jobs or were knowledge workers. Besides, some samples were as small as four participants or were all female, restricting the use of proper statistics. Such issues reduce generalizability of the findings to the total working population.
It is also crucial to take the origin of the interruption into account, as workflow disturbances can arise from internal or external sources (Jett & George, 2003). It appeared that the greatest interrupters are employees themselves (Courage et al., 2015; Wajcman & Rose, 2011) and self-initiated email checking was the most popular strategy (Mark et al., 2016). Regardless, multiple studies did not allow for the occurrence of internal interruptions by design (Jeuris & Bardram, 2016; Jeuris et al., 2018; Keller et al., 2019; Sanderson et al., 2013; Sonnentag et al., 2018), likely resulting in an underrepresentation of the true effect. However, a few reports did investigate internal interruptions (Cades et al., 2010; Mark et al., 2016, 2012), which is crucial for an accurate representation of interruption behavior.
Another central question involves the diverse operationalization of performance. Particularly self-report ratings were applied, but several other measures were used as well, which complicates interpretations and comparisons of results. Recently, two measures for work interruptions were developed: the work interruptions resiliency (WIR) measure (Zide, Mills, Shahani-Denning, & Sweetapple, 2017) and the workplace interruptions measure (WIM) (Wilkes, Barber, & Rogers, 2018). The use of such measures in future research should be encouraged to improve the integration of findings between different studies.
Other than performance, some studies investigated additional personal variables as polychronicity, stress and well-being. Polychronicity, defined as the preference to multitask, was described as an important predictor for the extent of multitasking behavior at work (König, Oberacher, & Kleinmann, 2010). Polychronicity was found to moderate the relationship between interruptions and performance ratings, such that polychronic individuals’ performance was less affected on days with frequent multitasking (Kirchberg et al., 2015; Pachler et al., 2018; Sanderson et al., 2013). Moreover, job performance was suggested to be optimal for employees with both the ability to multitask and the preference to do so (Sanderson et al., 2013). Apparently, individuals differ in their multitasking preferences, which may influence their performance.
Stress and well-being were also examined in a few studies, with methods such as thermal imaging, heart rate monitors and surveys. Workflow interruptions were reported as the most commonly experienced stressors in the workplace (Pachler et al., 2018), day-level multitasking was negatively associated with well-being in the evening (Kirchberg et al., 2015), and increases in interruptions over longer time periods were related to more psychosomatic complaints (Keller et al., 2019). This implies that there exists a cumulative effect of interruptions. However, findings regarding email-related stress were inconsistent: more time spent on email was related to elevated stress levels (Mark et al., 2016, 2012) and highly stressed office workers answered emails faster and with more anger (Akbar et al., 2019), but no differences in stress levels between email conditions were reported as well (Bradley et al., 2013). Thus, findings related to stress and well-being have shown that interruptions may have cumulative effects, though the influence of email is still ambiguous.
With reference to email, there were a couple of studies that examined different email management styles. It was found that limiting email access by batching or daily checks resulted in less time spent on email without decreases in performance (Akbar et al., 2019; Bradley et al., 2013; Mark et al., 2016). Also, employees showed less task switching behavior without email, and their colleagues did not report detrimental effects when the participants were off email (Mark et al., 2012). Evidently, the batching approach seemed to be the most beneficial for managing email.
In sum, multitasking studies in the workplace may be less generalizable, but have provided insights into the role of personal variables as polychronicity, stress and well-being and revealed that batching email is the most efficient strategy. Future research should use uniform performance measures, include internal interruptions and further explore cumulative effects of interruptions.
3.3. Comparison of study and work multitasking
Although findings from real-world studies in study and work environments are mostly in agreement, such that technology-based multitasking decreases performance, there are three important
differences between these settings, including the operationalization of performance, relevance of distractive ICTs, and motive for multitasking behavior.
The operationalization of performance varied not only within, but also between research domains. In study settings, performance generally referred to clear-cut assessments as GPA and post-lecture test performance, whereas at work, equivocal and more subjective estimates as self-rated task accomplishment and performance were used. This implies that comparing results between student populations is more straightforward, while this is not the case for work scenarios. Therefore, the use of validated, uniform measures as the WIR (Zide et al., 2017) or WIM (Wilkes et al., 2018) in future workplace studies is advised.
Study and work environments also differ in how distractive technologies relate to the primary task. Students multitasked with ICTs, especially social media, whereas work studies were focused on interruptions, mostly via email. This makes it seem as if students were distracted by technologies solely for nonacademic purposes, while multitasking at work was caused by only job-relevant interruptions It could thus be assumed that office workers do not experience non-work-related distractions, while in reality, there is a lack of studies investigating multitasking in the workplace with work-irrelevant ICTs. Nonetheless, it was argued that all types of interrupting technologies increase multitasking behavior, in spite of the (ir)relevance, as they all disrupt the primary task (J. van den Eerenbeemt, personal communication, January 22, 2020). Hence, the relevance of the distraction to the primary task may be unrelated to task performance.
Another difference between study and work multitasking is related to the underlying motive to engage in the behavior. Students tend to multitask with ICTs because of feelings of anxiety and fear of missing out (FOMO; Clayson & Haley, 2012; Przybylski, Murayama, DeHaan, & Gladwell, 2013; Rosen, Whaling, Rab, Carrier, & Cheever, 2013). In contrast, multitasking in the workplace is promoted by open-plan office spaces, which are considered beneficial for collaboration and flexibility, but appear to degrade concentration and performance (Di Blasio, Shtrepi, Puglisi, & Astolfi, 2019). Alternatively, the fear of missing something important has been reported as a prominent motivation for technology-based multitasking, which could be applicable to both students and office workers (J. van den Eerenbeemt, personal communication, January 22, 2020). This implies that students and employees can show multitasking behavior for different reasons, but also for the same motives.
Briefly, differences in performance operationalization between study and work multitasking research appear to be the greatest issue. The relevance of the distraction to the primary task may not be of any influence and motives to engage in multitasking behavior can differ but also correspond between students and office workers.
Table 1
Research reports investigating multitasking in the classroom
Authors (year) Study sample Multitasking measure Method Performance measure Main result
Observational
Nalliah & Allareddy (2014) 26 (8 F, Mage unknown) Distractions Lecture of unknown duration. Post-lecture test No differences between distracted and non-distracted students.
Cross-sectional
Burak (2012) 774 (67.1% F, Mage = 20.75) Multitasking behavior Survey GPA Multitasking scores were negatively associated with GPA.
Clayson & Haley (2012) 198 (159 F, Mage = 23.3) Texting Survey Class grade and GPA Texting frequency was negatively associated with class grade, but
not associated with GPA.
Junco (2012) 1839 (64% F, Mage = 21) ICT use Survey GPA Facebook and text messaging were negatively associated with GPA.
Ravizza et al. (2014) 170 (F and Mage unknown) ICT use Survey Exam grade Internet use was negatively associated with exam grade.
Zhang (2015) 176 (127 F, Mage = 21.55) Laptop multitasking Survey Class grade Laptop multitasking was negatively associated with class grade.
Zureick et al. (2017) 888 (F and Mage unknown) Multitasking behavior Survey Class grade Multitasking was negatively associated with class grade.
Experimental
Downs et al. (2015) 204 (119 F, Mage = 19.55) Facebook 6 conditions: Facebook distracted, notes on paper,
no media use, mixed distraction, notes on laptop, distracted combo. Video lecture of 25 minutes.
Post-lecture test Lowest mean scores in the Facebook distracted condition. In the mixed distraction condition, distracted students scored lower than non-distracted students.
Ellis et al. (2010) 62 (36 F, Mage unknown) Texting 2 conditions: texting and controls. Lecture of
unknown duration. Post-lecture test Students who texted scored lower than controls. End et al. (2010) 71 (48 F, Mage = 20.21) Phone ringing 2 conditions: ringing cellphone and controls. Video
lecture of unknown duration. Post-lecture test Students in the ringing condition scored lower than controls. Gingerich & Lineweaver
(2014) 67 (F and M
age unknown);
56 (40 F, Mage unknown)
Texting Two samples with 2 conditions: texting and
controls. Lecture of 12 minutes. Post-lecture test Students who texted scored lower than controls in both samples. Kuznekoff & Titsworth
(2013) 47 (F unknown, Mage = 18) Texting 3 conditions: control, low-distraction, high-distraction. Video lecture of 12 minutes. Post-lecture tests Students who were highly distracted scored lower than students who were lowly distracted and controls. Kuznekoff et al. (2015) 145 (F unknown, Mage = 18) Texting and Twitter 9 conditions: control and high- or low distracting
sending/receiving (ir)relevant messages. Video lecture of 12 minutes.
Post-lecture tests Sending and receiving irrelevant messages negatively affected performance, while relevant messages had no effect. Marone et al. (2018) 20 (17 F, Mage = 20.1) Facebook 2 conditions: visual Facebook distraction and
controls. Lecture of 42 minutes. Post-lecture test Students who were visually distracted by Facebook scored lower than controls. Rosen et al. (2011) 185 (80% F, Mage = 25) Texting 3 conditions: no/low, moderate and high
distraction. Video lecture of 30 minutes. Post-lecture test Students who were highly distracted scored lower than students who were not/lowly distracted. Sana et al. (2013) 44 (25 F, Mage = 18.9) Online tasks 2 conditions: laptop users and laptop users with
online tasks. Lecture of 45 minutes. Post-lecture test Students who performed additional online tasks scored lower than students who only took notes on their laptops. Wood et al. (2012) 145 (116 F (Mage = 20.67);
29 M (Mage = 19.56))
Texting, email, MSN
and Facebook 7 conditions: texting, emailing, MSN, Facebook, paper-and-pencil, word processing, natural use of technology. Three lectures of 20 minutes.
Post-lecture test Students using Facebook scored lower than students in the paper-and-pencil condition. Students who did not use any technology scored higher than students who used some form of technology.
Abbreviations: F = females, Mage = mean age; GPA = grade point average; ICT = information and communication technologies; M = males; MSN = the Microsoft network.
Table 2
Research reports investigating multitasking while doing homework
Authors (year) Study sample Multitasking measure Method Performance measure Main result
Observational
Rosen et al. (2013) 263 (146 F, Mage unknown) Computer windows Behavior assessment of a 15-minute study session. GPA Accessing Facebook at least once was associated with lower GPA.
Cross-sectional
Junco & Cotten (2012) 1774 (64% F, Mage = 21) ICT use Survey GPA Facebook use and texting were negatively associated with GPA.
Martín-Perpiñá et al.
(2019) 977 (51.9% F, M
age = 14.37) MMI Survey Class grades High media multitaskers had lower class grades than non-high
media multitaskers.
Experimental
Bowman et al. (2010) 89 (43 F, Mage = 20.17) IM 3 conditions: IM before reading, IM during reading,
no IM. Reading time, post-reading test Reading time was longer in the IM during reading condition, but performance scores did not differ between conditions. Fox et al. (2009) 69 (F and Mage unknown) IM 2 conditions: IM before reading and IM during
reading. Reading time, post-reading tests Reading time was longer in the IM during reading condition, but performance scores did not differ between conditions.
Longitudinal
Walsh et al. (2013) 483 (483 F, Mage = 18.1) Media use Surveys in the first and second semester. GPA Cell phone use, social networking and television and movie viewing
were negatively associated with later GPA.
Abbreviations: F = females, Mage = mean age; GPA = grade point average; IM = instant messaging; ICT = information and communication technologies; MMI = media multitasking index.
Table 3
Research reports investigating multitasking in the workplace
Authors (year) Study sample Multitasking measure Method Performance measure Main result
Observational
Cades et al. (2010) 4 (4 F, Mage unknown) Interruptions Multiple screen recording sessions of 30- or
60-minutes. Resumption lag External interruptions resulted in longer resumption lags than internal interruptions. Kirchberg et al. (2015) 93 (41 F, Mage unknown) Multitasking behavior Diary study during mornings and evenings of 5
consecutive workdays. Self-rated performance Day-level multitasking was negatively related to performance. Mark et al. (2016) 40 (20 F, Mage unknown) Email Computer logging and daily surveys for 12
workdays. Self-rated productivity Email duration was negatively related to productivity. Self-interrupting and batching checkers had higher productivity scores with longer email duration than those who relied on notifications. Pachler et al. (2018) 149 (40.94% F, Mage =
42.99) Interruptions Diary study during evenings of 5 consecutive work days. Self-rated performance Interruptions were negatively related to performance. Sanderson et al. (2013) 119 (34% F, Mage unknown) Email Split-screen tasks: problem-solving and email. Supervisor ratings Multitasking ability was positively correlated to performance.
Sonnentag et al. (2018) 174 (54% F, Mage = 42.7) Interruptions Daily surveys for 5 consecutive work days. Self-rated task
accomplishment Interruptions were negatively associated with perceived task accomplishment. Zoupanou (2015) 9 (4 F, Mage = 39.75) Interruptions Interviews Subjective experiences All participants perceived interruptions as part of their job, and
were not negatively affected by them.
Experimental
Akbar et al. (2019) 63 (45 F, Mage = 23.75) Email 2 conditions: batch or continual email, with
concurrent task. Task performance, time on email Concurrent task performance did not differ between conditions. Batching email was associated with less time on email. Bradley et al. (2013) 7 (3 F, Mage = 28) Email Within-subjects: once-a-day or frequent email
strategy, for 5 consecutive work days. Time on email In the once-a-day strategy, less time was spent on email than in the frequent strategy. Jeuris & Bardram (2016) 16 (4 F, Mage unknown) Window switching 2 x 2: Windows or virtual desktops, and task set A
or B. Same sequence as Jeuris et al. (2018). Resumption lag, productivity, accuracy Virtual desktops were related to shorter resumption lags. There were no differences in productivity and accuracy scores. Jeuris et al. (2018) 7 (1 F, Mage unknown) Window switching 12 task switches during 50 minutes. Four tasks,
each requiring 2-4 application windows. Accuracy Half of the task switches contained one or more errors. Mark et al. (2012) 13 (6 F, Mage = 46) Email Within-subjects: 3-day baseline and 5-day
no-email condition. Behavior assessment of two days (one in each condition) and computer logs.
Time on task, window
switching In the no-email condition, employees switched less and spent longer on each window before switching, than in the baseline condition.
Longitudinal
4. Discussion
In modern-day culture, it is commonly thought that doing multiple things within the same timeframe increases productivity. As technology has become omnipresent in contemporary society, this belief is further reinforced by an increasing number of options for multitasking with ICTs. Nevertheless, scientific research has offered a less optimistic outlook on the effectiveness of multitasking behavior. This literature review has shown that detrimental effects of multitasking on performance were observed not only in the laboratory, but also in real-world experiments taking place in- and outside the classroom and in the workplace. Despite this analogous findings, it is still obscure to what extent real-world studies have added value to multitasking research.
Before answering this question, similarities and (dis)advantages of laboratory and real-world research are debated. Based on that assessment, a conclusion regarding the added value of real-world research is drawn. Finally, some implications for students, office workers and future research are provided.
4.1. Comparison between laboratory and real-world multitasking
Based on the aforementioned studies, laboratory and real-world multitasking research are compared by analyzing advantages and disadvantages of both study domains, as depicted in Table 4. Several similarities are discussed as well, which include the investigation of multiple strategies, long-term effects, internal and external interruptions, differences in individual preference and a common limitation of lack of smartphone multitasking research.
Table 4
Advantages and disadvantages of laboratory and real-world multitasking research.
Advantages Disadvantages
Laboratory research • Clear measures • Controllable setting • Timeless
• Lacks ecological validity • Isolated interruptions
Real-world research • Cumulative interruptions • Equivocal measures • Quickly outdated
Laboratory studies have investigated multitasking in dual-task and task switching paradigms, while real-life multitasking might take a variety of forms, only sometimes resembling these classical experiments. In reality, there are often more than two tasks, and a task interleaving strategy is regularly applied by switching back and forth before finishing any one task (Duggan et al., 2013). In other words, there is considerably more variation in how people choose to allocate their resources and priorities during tasks in real-life situations (Carrier et al., 2015). However, the multitasking continuum enabled the application of different strategies with varying degrees of multitasking behavior in the laboratory. Hence, the effects of varying forms of multitasking on performance can be and have been investigated in both research fields.
In both laboratory and real-world studies, performance measures typically involve a component of time and/or accuracy. Nevertheless, measurements in the laboratory are clear and unambiguous, while real-world studies strongly vary in the operationalization of performance, resulting in an equivocal measure. This is an obvious advantage of laboratory studies and disadvantage of real-world research, and findings between these research domains should be translated with caution.
Another characteristic of laboratory studies is high experimental control, as researchers are in charge of what task the participants work on and for how long. The use of controllable laboratory settings is favored by indications that individuals have restricted insight into their own behavior, such as overestimations of media time (Junco, 2013; Moreno et al., 2012) and underestimations of switching frequency (Adler & Benbunan-Fich, 2013; Brasel & Gips, 2011). Though, neither highly controlled scenarios nor limited interruption lengths necessarily reflect how individuals are affected by and cope with interruptions in the real world (Cades et al., 2010). People perform many scheduled and spontaneous tasks as part of their everyday lives (Courage et al., 2015) and real work can last for hours (Kirchberg et al., 2015). Although the real world can be chaotic and is less controllable, the lack of ecological validity is a substantial disadvantage of laboratory research.
Both laboratory and real-world studies have examined long-term effects of multitasking behavior. Several laboratory studies have investigated potential training effects, such as via media multitasking, with ambiguous results (Wiradhany & Nieuwenstein, 2017). Longitudinal studies have also been performed in the real world, in which multitasking was found to have negative long-term effects on performance (Keller et al., 2019; Walsh et al., 2013). Thus, both research fields have investigated some form of longitudinal effects, but longer follow-up studies are required (May & Elder, 2018).
In addition, both research fields allowed for assessment of both internal and external interruptions. Laboratory research has mainly focused on studying external interruptions by using fixed switches, yet there exist a number of laboratory studies in which participants were free to decide when to switch tasks and interrupt themselves (Adler & Benbunan-Fich, 2012, 2013, 2015; Buser & Peter, 2012; Reissland & Manzey, 2016). Multiple real-world reports did only permit the occurrence of external interruptions by design (Bowman et al., 2010; End et al., 2010; Jeuris & Bardram, 2016; Jeuris et al., 2018; Keller et al., 2019; Sanderson et al., 2013; Sonnentag et al., 2018), while others did include internal interruptions (Cades et al., 2010; Mark et al., 2016, 2012). It is recommended that future studies allow for performance of self-interruptions, as they form a major part of interruption behavior. Moreover, laboratory studies have presented useful insights into isolated interruptions, but have not sufficiently provided an understanding of the cumulative effects that are characteristic in
coping processes become increasingly complex and cognitively demanding (Baethge, Rigotti, & Roe, 2015). Therefore, the effects of isolated interruptions reported in the laboratory do not simply generalize to cumulative effects that happen while studying or working. The possibility to study cumulative effects is thus an advantage of real-world research.
Another similarity between laboratory and real-world studies is that differences in individual multitasking preferences have been recognized. Laboratory research has identified subgroups based on their individual multitasking strategies (Reissland & Manzey, 2016) and real-world studies have investigated the moderating role of polychronicity personality traits (Kirchberg et al., 2015; Pachler et al., 2018; Sanderson et al., 2013). These individual differences in multitasking preference may be an interesting subject for future research or should at least be included as a possible confounder in the relationship between multitasking and task performance.
Furthermore, new technologies are being created and used at such a quick pace that it is difficult for researchers to capture the effects of these rapidly changing technologies. As a result, real-world studies become outdated quickly. Reports that only involved computers are no longer relevant, since studying and working with laptops, tablets and smartphones has become inherent in modern-day society. However, laboratory research is less susceptible to obsolescence due to technological development, because the experiments and measures are, in general, of a more timeless nature. As a side effect, the fact that real-world research is quickly outdated could be considered a counterargument for the ecological validity of such studies, as they seem to be relevant for only a limited period of time. Anyhow, the declining relevance of real-world research is an important disadvantage, and conversely, the more timeless character of laboratory settings can be regarded as beneficial.
A major limitation that laboratory and real-world studies have in common is that multitasking with smartphones is relatively under researched, since the majority of the reports only involved computer-based multitasking. The lack of empirical research on the cognitive impacts of smartphone technology is understandable, given that the relevant technology itself is still nascent and constantly evolving (Wilmer, Sherman, & Chein, 2017). However, research in this area will soon be applicable to most of the world’s population, making it of vital importance to get a deeper understanding of the effects of multitasking with smartphone technology.
4.2. Conclusion: the added value of real-world research
Both laboratory and real-world research have contributed to our understanding of multitasking and have offered useful insights into varying multitasking strategies, long-term effects, internal and external interruptions and differences in individual multitasking preferences. Besides, laboratory and real-world studies each have their advantages and disadvantages. Laboratory studies are clear,
