The Illusion of Multitasking and Its Positive Effect on Performance

Abstract

With technological advancements, the desire, ability, and often necessity to multitask are pervasive. Although multitasking refers to the simultaneous execution of multiple tasks, most activities that require active attention cannot actually be done simultaneously. Therefore, whether a certain activity is considered multitasking is often a matter of perception. This article demonstrates the malleability of what people perceive as multitasking, showing that the same activity may or may not be construed as multitasking. Importantly, although engaging in multiple tasks may diminish performance, we found that, holding the activity constant, the mere perception of multitasking in fact improves performance. Across 32 studies (30 of which had performance-based incentives) containing a total of 8,242 participants, we found that individuals who perceived an activity as multitasking were more engaged and consequently outperformed those who perceived that same activity as single tasking.

Keywords

multitasking performance perception engagement pupil dilation open data open materials

In today’s society, multitasking is an integral part of life (e.g., Ophir, Nass, & Wagner, 2009). People frequently engage in two or more tasks simultaneously, such as switching between tabs on computers and smartphones, checking e-mail and social media, and surfing the Web. A survey of consumers’ mobile habits reported that individuals frequently use their smartphones while watching a movie, during a dinner date, and even at church (Jumio, 2013). Multitasking is also prevalent in the workplace, where most environments necessitate working under time pressure on several tasks simultaneously (Kreckler, Catchpole, Bottomley, Handa, & McCulloch, 2008).

In addition to the prevalence of multitasking, the ability to multitask is seen as a highly desirable trait (Wang & Tchernev, 2012). In a survey we conducted with 434 participants sampled on age, income, and gender to reflect the U.S. population, we found that 84% of participants reported that the ability to multitask is an important trait to have, and 93% said they could multitask as well as or better than the average person (see Table SM 1 in the Supplemental Material available online).

Early research on multitasking (Borger, 1963; Creamer, 1963) found that when working on multiple nonautomatic tasks, individuals cannot actually perform the tasks simultaneously but rather alternate between different activities, engaging in only a single task at a time (Kieras, Meyer, Ballas, & Lauber, 2000; Pashler, 1994). Such switching behavior was found to lower performance because of cognitive-processing limitations and attention residue (e.g., Leroy, 2009; Levy & Pashler, 2001; Pashler, 1994).

If indeed people process only a single task at a time, this means that multitasking is often a matter of subjective perception that, like other perceptions, may be influenced by context. That is, holding the actual activity constant, some situations may cause people to perceive their overall activity as multitasking, whereas other situations may cause people to construe the same activity as single tasking. Accordingly, this article’s central question is how the difference in perception of the same task impacts performance. We propose that the way people mentally construe an activity, as either multitasking or single tasking, can affect their performance on that activity. For example, imagine being asked to watch a video and type everything that is said in the video. If you construe this activity as transcribing, you will probably consider this a single task. However, if you construe the same activity as watching a video and typing at the same time, two distinct tasks done simultaneously, you will be more likely to consider this activity multitasking. This is important because many “single” activities can be broken down into multiple components and construed as multitasking. Inversely, multiple distinct activities can often be collapsed and framed as a single task. Thus, understanding when people perceive an activity as multitasking and how these perceptions might affect performance has important implications across many domains. We set out to explore how simply perceiving a given activity as either multitasking or single tasking affects performance.

Existing research demonstrates that individuals’ motive for investing effort and cognitive control increases with the difficulty of the task (e.g., Kahneman, 1973; Kukla, 1972; Sanders, 1983) as well as with the expectation of task difficulty (e.g., Schrift, Netzer, & Kivetz, 2011). Indeed, several findings support the notion that a more challenging task increases individuals’ attention and ultimately leads to improvement in performance (Hommel, Fischer, Colzato, van den Wildenberg, & Cellini, 2012; Kofman, Meiran, Greenberg, Balas, & Cohen, 2006; Plessow, Fischer, Kirschbaum, & Goschke, 2011). Similarly, neuroscientific observations (Botvinick, Braver, Carter, Barch, & Cohen, 2001; Egner & Hirsch, 2005) suggest that greater difficulty during task performance leads to increased attention. Hommel et al. (2012) summarized that “there is both behavioral and neuroscientific evidence for Hillgruber’s (1912) claim that increasing the challenge of the task spontaneously increases one’s effort to compensate for and to overcome that challenge” (p. 291). Building on these findings, we hypothesized that when individuals construe the impending activity as multitasking, which is perceived to be more challenging, they become more motivated, attentive, and engaged with the task. We hypothesized that this state of increased engagement with the task, which is defined and often operationalized as a state in which individuals are more involved, occupied, interested, and attentive (Bianco, Higgins, & Klem, 2003; Greene & Miller, 1996; Higgins, 2006; Kahneman, 1973), leads to improved performance when people perceive that they are multitasking.

Consistent with this hypothesis, multiple studies involving different tasks in which participants were incentivized on the basis of their performance have shown that merely perceiving an activity as multitasking increases motivation and improves performance. Using physiological measures, we found that individuals who construe the activity as multitasking are more attentive and engaged with the task. Next, we report 5 studies that tested our hypothesis (6 additional studies are fully reported in the Supplemental Material, and the other 21 are summarized in Table SM 4 in the Supplemental Material) and conclude with a meta-analysis.

Studies 1a and 1b: Multitasking Perceptions and Their Impact on Performance

The goal of Studies 1a and 1b was to test whether framing a certain activity as multitasking, as opposed to single tasking, improves performance. In Study 1a, we asked participants to watch and transcribe an educational video. In Study 1b, we asked participants to summarize a virtual lecture.

Study 1a

Method

One hundred sixty-two participants (62% female; mean age = 21.04 years) were recruited from a northeastern university to participate in an hour-long lab session in exchange for a base payment of $10. Participants in this and all subsequent studies provided informed consent and were aware that participation was voluntary. In addition, all research assistants were blind to the hypothesis. The sample size for this and all subsequent studies was determined on the basis of an expected medium effect size (Cohen’s d ≈ 0.5) that was observed in an initial study (see Study SM 5 in the Supplemental Material). The sample size for studies conducted in the university lab varied on the basis of the number of participants recruited for a standard lab session occurring over a week (normally 150–200 participants). In this and all subsequent studies, all participants were included in the analysis unless otherwise specified.

Participants were asked to watch an educational video and transcribe what was said in it. Each participant was then randomly assigned to either the multitasking or single-tasking conditions. Participants assigned to the multitasking condition were told that they would be working on two tasks concurrently and, therefore, would need to multitask. The first task, entitled the learning task, was described as a test of individuals’ learning abilities and required participants to watch an educational video from Animal Planet’s Shark Week. The second task, entitled the transcribing task, was described as a test of individuals’ writing skills that required participants to transcribe exactly what was said in the video. Thus, through framing, we intended to make participants in this condition feel as if they were working on two separate tasks concurrently.

Participants assigned to the single-tasking condition were asked to perform the exact same activity, merely framed differently. Specifically, participants in this condition were told that they would be working on a learning task meant to test individuals’ learning and writing abilities. The task was described as watching and transcribing an educational video from Animal Planet’s Shark Week. Thus, in both conditions, participants performed the exact same activity and were told that their learning and writing abilities would be tested. However, in the single-tasking condition, the activity was framed as a single task, and in the multitasking condition, it was framed as two tasks done concurrently. We validated that this framing manipulation worked as intended in a separate pretest using the same population (for details, see Section SM 1a in the Supplemental Material).

All participants learned that they would receive an additional $0.02 for each word they correctly transcribed and that they could work for as long as they liked until the video ended after 6 min. The first measure of performance was how many words were transcribed and how accurately participants transcribed them. The second measure of performance tested participants’ comprehension of the information provided in the video. Specifically, at the end of the study, we administered a 10-question multiple-choice pop quiz about the video’s content.

Results

The first measure of performance that we analyzed was the number of words transcribed in each condition. Participants assigned to the multitasking condition transcribed significantly more words (M = 274.13, 95% confidence interval, or CI = [247.02, 301.24]) than participants assigned to the single-tasking condition (M = 229.60, 95% CI = [199.53, 259.67]), F(1, 160) = 4.63, p = .033, d = 0.34. Analyzing how accurate participants’ transcriptions were using a text match application, we found that participants assigned to the multitasking condition wrote more words accurately (M = 223.77, 95% CI = [200.20, 247.34]) than participants assigned to the single-tasking condition (M = 177.20, 95% CI = [152.88, 201.52]), F(1, 160) = 7.27, p = .008, d = 0.42.

As an additional measure of performance, we checked how well participants performed on the unannounced quiz administered at the end of the study. As predicted, participants assigned to the multitasking condition performed significantly better on the quiz (M = 6.60, 95% CI = [6.21, 6.99]) than participants assigned to the single-tasking condition (M = 5.81, 95% CI = [5.30, 6.32]), F(1, 160) = 5.82, p = .017, d = 0.38.

We next analyzed how much time participants spent transcribing the video using log-transformed time spent on the task. Because the video ended for all participants after 6 min, the data were right censored. Thus, we used a Cox regression survival analysis and found that there was not a significant difference in persistence (multitasking: M = 2.43, 95% CI = [2.38, 2.48]; single tasking: M = 2.32, 95% CI = [2.24, 2.40]); b = 0.32, 95% CI = [−0.15, 0.79]; Wald χ²(1, N = 162) = 1.82, p = .178, d = 0.23. This analysis method was employed for all subsequent persistence analyses.

Moreover, controlling for time differences, we found that participants assigned to the multitasking condition wrote more words per second (M = 0.91, 95% CI = [0.86, 0.96]) than participants assigned to the single-tasking condition (M = 0.84, 95% CI = [0.78, 0.90]), F(1, 160) = 3.49, p = .064, d = 0.29.

In this study, we triggered the perception of multitasking by explicitly telling participants that the activity involved multitasking. Accordingly, one concern may be that the effect is driven by explicitly mentioning the notion of multitasking and that it will not persist when individuals spontaneously construe their activity as either multitasking or single tasking. Although this is not concerning from a policy perspective (using an explicit manipulation to trigger multitasking perceptions is easy to implement), it is important to examine whether the effect persists without explicitly mentioning multitasking. We addressed this in Study 1b.

Study 1b

Method

Two hundred ninety participants (50% female; mean age = 36.53 years) were recruited from Amazon’s Mechanical Turk (MTurk). Participants were paid a base payment of $0.90 and were told that they could earn additional compensation on the basis of their performance. All participants worked on the exact same task. The task required participants to watch an online lecture from an educational platform about Pangaea and Earth’s geographical history and take notes about its content. Participants were told that they would receive additional compensation on the basis of how much they remembered about the lecture and how detailed their notes were. To ensure that the computer audio was working, prior to engaging in the study, we had participants watch a video clip that instructed them to write a test word. Three participants did not write the correct test word and were excluded from the analysis.

Each participant was randomly assigned to one of two framing manipulations, which were validated in a pretest (for details, see Section SM 1b in the Supplemental Material). In the multitasking condition, watching the lecture and taking notes were described as two distinct tasks that would be performed simultaneously. Of importance, the word multitasking was not explicitly mentioned in this study. Rather, the instructions emphasized to participants, “you will be both watching the lecture and taking notes concurrently.” The single-tasking condition described watching the lecture and taking notes as a single task. Participants in both conditions were allowed to quit at any time until the lecture ended (6 min). Following the task, participants indicated how boring they found the activity to be (measured on a scale from 1 to 7). No significant differences were observed on this measure (p = .52), and it will not be discussed further.

Two coders, who were blind to the hypothesis and conditions, coded the quality of the notes. The coders evaluated participants’ notes on (a) how detailed, thorough, and comprehensive the notes were; (b) their overall clarity; and (c) how much effort they thought the participant put into writing the notes. All items were measured on a scale from 1 to 7 and were highly correlated (α = .97). Therefore, we collapsed these measures to form a single measure of quality. Further, the two coders’ ratings were highly correlated (r = .91, p < .001) and were therefore averaged, resulting in a single quality score for each participant.

Results

Seventeen participants (6%) who spent less than 2 standard deviations of time (log transformed) on reading the manipulation (i.e., < 1.92 s) were excluded from the analyses. The same pattern of results held when these participants were included.

Participants’ performance was analyzed on the basis of the coders’ quality measure. As predicted, the average note quality was higher in the multitasking condition (M = 3.76, 95% CI = [3.48, 4.04]) compared with the single-tasking condition (M = 3.32, 95% CI = [3.03, 3.61]), F(1, 271) = 4.55, p = .034, d = 0.26. In addition, participants assigned to the multitasking condition wrote, on average, more words (M = 64.16, 95% CI = [56.45, 71.87]) compared with participants assigned to the single-tasking condition (M = 51.20, 95% CI = [41.76, 60.64]), F(1, 271) = 5.30, p = .022, d = 0.28. No difference was observed on how long participants worked on the task (p = .477) because the majority of participants (86%) viewed the entire lecture. Taken together, the results of Studies 1a and 1b support the hypothesis that performance on a given activity improves when it is perceived as a multitasking (as opposed to a single-tasking) activity.

Studies 2a and 2b: Measuring and Manipulating Multitasking Perceptions

Unlike Studies 1a and 1b, Study 2a measured participants’ perceptions of multitasking instead of manipulating them. This allowed us to test whether the instructions trigger a behavior that would not occur naturally. Using this natural variation in how people construed the activity, we examined how such perceptions correlate with performance. To support causal claims, we used the same task as in Study 2a but manipulated multitasking perceptions.

Study 2a

Method

Eighty participants (43% female; mean age = 36.66 years) were recruited from MTurk to take part in this study. Participants were asked to work on an assignment involving two distinct puzzles. The first puzzle was a word puzzle in which participants observed a 15 × 15 matrix of letters and were asked to find as many words as possible in a horizontal, vertical, or diagonal pattern in the matrix. The second puzzle was an anagram task in which participants observed a 10-letter string and were asked to construct as many words as possible using the letters in the string. Participants were paid a base payment of $0.40 and were told that each correct answer (i.e., a word of 4 or more letters) would earn them an additional $0.01. The two puzzles appeared on the same screen side by side. Participants worked on the tasks concurrently for 4 min and could submit as many words as they could find (see Fig. 1).

Fig. 1.

Word-puzzle task (left) and anagram task (right) used in Study 2.

After participants finished working on the task, we measured the extent to which they perceived their activity as either multitasking or single tasking using two types of measurements. First, to prevent participants from stating post hoc that they perceived their activity as multitasking and to increase the validity of their responses, we incentivized participants to answer truthfully by employing a particular response format. Specifically, after participants finished working on the task, we told them that they would be matched with a partner and that they and their partner would need to indicate whether they perceived working on the activity as a multitasking or single-tasking activity. If their partner and they both responded to the question in the same way, they would each receive an additional $0.05 bonus payment. Four additional binary items were used to measure perceptions of multitasking (for details, see Section SM 2a in the Supplemental Material). Responses to all five binary measures were combined into a single measure of multitasking perception (α = .61).

Results

One participant was more than 2 standard deviations from the mean performance and was therefore excluded from the analyses. Results held when this participant was included.

We regressed the number of words found in the puzzles on the measure of multitasking perception. We examined both the overall number of words submitted as well as the number of correct words found (these two measures were highly correlated), r(77) = .98, p < .001. As predicted, we found a significant positive relationship between the perception of multitasking and number of words attempted (b = 4.33, 95% CI = [0.83, 7.82]), t(77) = 2.47, p = .016, d = 0.56, as well as the number of correct words (b = 3.83, 95% CI = [0.64, 7.02]), t(77) = 2.39, p = .019, d = 0.55. That is, the more participants felt that they were multitasking, the better they performed. Obviously, one cannot make any causal claims on the basis of this study alone because reverse causality could drive the observed correlation. To test causality, we used the same activity in Study 2b, but instead of measuring the perception of multitasking, we manipulated it.

Study 2b

Method

Two hundred thirty-seven paid online participants were recruited from MTurk (55% female; mean age = 36.33 years) to take part in this study, which employed the same puzzles and same incentives described in Study 2a. Each participant was randomly assigned to one of two conditions. In the multitasking condition, the two puzzles were described as relating to two different studies (perceptual study and identification study), were separated on the screen by a vertical line, and had different background colors. In the single-tasking condition, both puzzles were described as being part of the same study (perceptual-identification study) and were not visually distinguished by different background colors or separated by a line. Unlike in Study 2a, participants were allowed to quit the tasks at any time until 4 min, thus enabling us to examine persistence as another indication of performance.

Because the framing manipulation was relatively subtle, we also included an additional factor intended to further strengthen the manipulation. Specifically, we also manipulated (between subjects) whether we disclosed to participants that in this study some participants would work on a single study whereas others would work on two studies at the same time. The disclosure manipulation did not produce any main effects or interactions on any of the dependent variables or with the framing manipulation. We therefore collapsed the analyses and do not discuss this factor further. This does not meaningfully change the pattern of results.

As a manipulation check, we asked participants to indicate to what extent they felt they were multitasking (1 = not at all, 4 = somewhat, 7 = totally) and whether they felt like they were completing two different tasks, a single task with two components, or a single task (they selected one of the three descriptions that best matched their experience).

For exploratory purposes, participants in all conditions were asked to answer several questions about their multitasking habits and feelings of productivity after finishing their assignment (reported in the Section SM 2b in the Supplemental Material). None of these measures moderated the effect; therefore, we do not discuss these further.

Results

Manipulation checks

Participants assigned to the multitasking condition indicated that they perceived their activity as multitasking (M = 4.40, 95% CI = [4.08, 4.72]) to a greater extent than those assigned to the single-tasking condition (M = 2.52, 95% CI = [2.24, 2.80]), F(1, 235) = 76.06, p < .001, d = 1.14. Furthermore, of the participants who were assigned to the multitasking condition, 46% indicated perceiving their activity as working on two separate tasks (vs. 5% in the single-tasking condition), Pearson χ²(1, N = 237) = 51.35, p < .001, r = .47, and 11% indicated perceiving their activity as working on a single task (vs. 71% in the single-tasking condition), Pearson χ²(1, N = 237) = 88.81, p < .001, r = .61. Thus, the manipulation worked as intended.

Performance

An analysis of variance revealed that participants assigned to the multitasking condition submitted, on average, more words (M = 14.42, 95% CI = [13.07, 15.77]) than those assigned to the single-tasking condition (M = 8.08, 95% CI = [7.17, 8.99]), F(1, 235) = 57.86, p < .001, d = 0.99.¹ Again, the number of words submitted and the number of correct words were highly correlated, r(235) = .98, p < .001. Analyzing the number of correct words revealed the same pattern (multitasking: M = 13.65, 95% CI = [12.31, 14.99]; single tasking: M = 7.50, 95% CI = [6.64, 8.36]), F(1, 235) = 56.34, p < .001, d = 0.98.

Persistence

Participants worked on the puzzles for 4 min, at most, but could quit at any point prior to that. Using survival analysis of log (time), we found that participants who were assigned to the multitasking condition persisted longer (M = 2.35, 95% CI = [2.08, 2.62]) than those assigned to the single-tasking condition (M = 2.26, 95% CI = [1.73, 2.79]), b = 1.17, 95% CI = [1.75, 5.96]; Wald χ²(1, N = 237) = 14.02, p < .001, d = 0.50.

Moreover, even after controlling for the time participants spent on the tasks, we found that participants in the multitasking condition still submitted more words, F(1, 234) = 46.49, p < .001, d = 0.89, and performed better, F(1, 234) = 44.60, p < .001, d = 0.87, suggesting that the quality of work, and not only the overall time spent on the task (i.e., persistence), drove the improvement in performance.

Discussion

By both manipulating and measuring multitasking perceptions, Studies 2a and 2b further support the notion that performance improves when individuals construe their activity as multitasking. Across different incentivized activities, participants performed better and earned more when they merely perceived the same activity as multitasking.

Admittedly, although we kept the actual activity fixed in Studies 2a and 2b, we did not control or restrict participants’ work sequence on the word puzzles. That is, some participants may have switched more often than others, and such variation in work sequence may have, at least partially, driven the results. Although such an account is less plausible for the tasks employed in Studies 1a and 1b, which involved transcribing and summarizing an educational video, we directly addressed this account using two additional studies (see Studies SM 5 and SM 6 in the Supplemental Material, where this is fully reported). In these studies, we used two distinct paradigms that experimentally controlled the work sequence. That is, we externally imposed a specific pattern by which participants switched back and forth between the tasks. We found that, even when we controlled for switching, perceptions of multitasking still improved performance, demonstrating that differential switching alone cannot account for the effect.

Study 3: Physiological Measures of Engagement

Study 3 employed the same paradigm used in Studies 2a and 2b but used eye-tracking technology to measure participants’ pupil dilation while working on the tasks. Pupil dilation has been used to measure individuals’ attentional and mental effort, processing load, and arousal (e.g., Beatty & Lucero-Wagoner, 2000; Bradley, Miccoli, Escrig, & Lang, 2008; Hoeks & Levelt, 1993; Kahneman & Beatty, 1966). Therefore, we used this validated physiological measure to examine whether the increase in performance is driven by greater engagement, which is often defined by greater attention and effort (Higgins, 2006).

Method

One hundred fifteen participants were recruited from a behavioral lab at a northeastern university (60% female; mean age = 20.46 years) in exchange for a $5 base payment. The procedure was identical to that employed in Study 2b, but while participants worked on the tasks, we used SMI RED-m eye-tracking equipment (iMotions, Boston, MA) to track their eye movements and pupil dilation. Participants were paid an additional $0.03 for each word that they correctly found in the puzzles.

To ensure that the use of different background colors did not impact the pupil dilation measures, we counterbalanced all background colors across conditions. Further, we verified that the level of luminance across conditions (which may affect pupil dilation) was nearly identical by calculating the mean luminance over the pixels² and scaling them from 0 to 255 (because the colors were counterbalanced, there were two task stimuli per condition—single task: 194.36, 194.44; multitask: 194.57, 194.63). Thus, this was unlikely to have caused a change in pupil dilation.³

Results

Consistent with eye-tracking research practices, we made several exclusions on the basis of data quality and criteria determined a priori. Participants were excluded if (a) their time on the task exceeded 2 standard deviations from the average time (5 participants; results held if included), (b) the eye-tracking device did not read their pupil dilation (2 participants), or (c) they had other technical difficulties (1 participant), leaving us with a total of 107 participants.

Performance

As predicted, participants submitted more words in the multitasking condition (M = 18.21, 95% CI = [15.22, 21.20]) than in the single-tasking condition (M = 10.65, 95% CI = [8.36, 12.95]), F(1, 105) = 13.84, p < .001, d = 0.73. Examining the number of correct words revealed a similar pattern (multitasking: M = 17.16, 95% CI = [14.30, 20.02]; single tasking: M = 9.78, 95% CI = [7.56, 12.00]), F(1, 105) = 14.36, p < .001, d = 0.74. Additionally, we found that most participants (97%) worked for the full 4 min. Thus, in this study, the framing manipulation had no significant effect on how long participants worked on the task (p > .25).

Pupil dilation

Participants’ average pupil dilation was larger in the multitasking condition (M = 3.90, 95% CI = [3.77, 4.03]) than in the single-tasking condition (M = 3.64, 95% CI = [3.50, 3.78]), F(1, 105) = 7.12, p = .009, d = 0.52. These results held when either median or maximum pupil dilation was analyzed (see Section SM 3 in the Supplemental Material). The effect of the multitasking framing on pupil dilation remained significant even after we controlled for the number of switches that participants made, F(1, 99) = 5.90, p = .017, d = 0.49.⁴ A mediation analysis using a bootstrap estimation approach with 5,000 samples (Model 4 from the PROCESS macro; Hayes, 2013) supported the assertion that the increase in pupil dilation mediated the impact of task framing on performance (b = 0.59, SE = 0.34, 95% CI = [0.08, 1.41]). Figure 2 depicts participants’ average pupil dilation over time across conditions.

Fig. 2.

Average pupil dilation across time in the multitasking and single-tasking conditions in Study 3.

Given the correlational nature of any mediation analysis, one should be cautious with causality interpretations between pupil dilation and performance. In particular, participants’ pupils might have been dilated because of happiness and excitement from finding more words (i.e., reverse causality). Although we cannot fully rule out this account, an additional analysis that controlled for the number of words found in each condition casts doubt on this interpretation. Specifically, when we analyzed average pupil dilation until participants found their first word in the puzzles, participants in the multitasking condition still had a greater average pupil dilation (M = 3.93, 95% CI = [3.80, 4.06]) than those in the single-tasking condition (M = 3.71, 95% CI = [3.55, 3.87]), F(1, 105) = 4.28, p = .041, d = 0.40. Similar patterns were found when examining pupil dilation until the second, third, fourth, and fifth words were found. Thus, regardless of the number of words found in the puzzles, participants in the multitasking condition exhibited greater physiological signs of engagement than participants in the single-tasking condition.⁵

Switching patterns

As would be expected, participants in the multitasking condition switched more (M = 11.83, 95% CI = [10.04, 13.50]) than those in the single-tasking condition (M = 6.62, 95% CI = [4.83, 8.82]), F(1, 100) = 14.44, p < .001, d = 0.76.⁶ However, although we did find a positive correlation between the number of words found and the number of switches made, r(100) = .55, p < .001, the directionality of the causal relationship between these two measures is unclear. Specifically, although one could argue that such correlation implies that switches impact the number of words found, one could also reasonably argue the reverse. That is, in the current paradigm, the likelihood of switching increases after a word is found. According to the latter account, because participants in the multitasking condition found more words, one should expect to see more switches in this condition and that the two would be highly correlated. Therefore, whether switches are the cause, as opposed to the outcome, of improvement in performance in this paradigm is not clear. Importantly, as discussed earlier, for this reason, we used other paradigms (see Studies SM 5 and SM 6) that show that the effect itself persists even after experimentally controlling for switching patterns, which is the strongest test of this account. Further, in the following meta-analysis, which includes 30 of the studies we ran, we provide an estimate for how much of the variance might be explained by shifts in switching patterns.

Replications and Meta-Analysis

Overall, we conducted 32 studies (5 of which are fully reported in the main text and an additional 6 are fully reported in the Supplemental Material). The studies we fully report in the main text and Supplemental Material were selected because they are representative of the methodology across all of the studies we conducted, employ a variety of paradigms, and address important rival accounts. Specifically, the studies fully reported in the Supplemental Material explored different facets and potential drivers of this effect such as productivity mind-set (see Study SM 5), sense of agency (see Study SM 7), individuals’ tendency to habituate and adapt to different stimuli (see Study SM 8), and locus of control and self-efficacy (see Study SM 9). Of the 32 studies, 1 study (Study 2a) was correlational by design and 1 study (Study SM 9) did not include performance measures. Therefore, both studies were excluded from the internal meta-analysis, leaving us with a total of 30 studies.⁷

Method

Of the 30 studies, 24 employed the word-puzzles task (described in Study 2b), 1 used a transcription task (described in Study 1a), and 1 used a note-taking task (described in Study 1b). Two additional studies used a count-locate task, which involved (a) counting the number of times a specific letter appeared in a passage and (b) locating a specific word in a passage on the basis of the number row and where it appeared in that row (the full description of this paradigm is reported in Study SM 5). Finally, 2 additional studies used an online-game task, which involved answering pattern-recognition questions that would sporadically appear in messenger windows (the full description of this paradigm is reported in Study SM 6). Table 1 lists and summarizes all studies we conducted, the paradigm employed in each, and descriptive statistics for the main effect of multitasking perception on performance. Table SM 4 provides further methodological details for all the studies we conducted that are not fully reported.

Table 1.

Comparison of Performance in the Multitasking and Single-Tasking Conditions of Each of the Studies

Study	Paradigm	Multitasking condition			Single-tasking condition			F	p value correct^a	p value attempts^b
Study	Paradigm	n	M	SD	n	M	SD	F	p value correct^a	p value attempts^b
Studies reported in the main text
1a	Transcription	82	223.77	108.89	80	177.2	110.97	F(1, 160) = 7.27	.008	.033
1b	Note taking	128	3.76	1.63	145	3.32	1.77	F(1, 271) = 4.55	.034	—
2a	Word puzzles	—	—	—	—	—	—	—	—	—
2b	Word puzzles	120	13.65	7.5	117	7.5	4.76	F(1, 235) = 44.6	< .001	< .001
3	Word puzzles (with pupil-dilation measures)	61	17.16	11.39	46	9.78	7.69	F(1, 105) = 14.36	< .001	< .001

Studies reported in the Supplemental Material
SM 4	Word puzzles (with pupil-dilation measures)	48	16.17	11.29	52	8.48	8.26	F(1, 98) = 15.26	< .001	< .001
SM 5	Count-locate	71	12.32	6.24	82	9.79	5.75	F(1, 151) = 6.81	.010	.069
SM 6	Game	51	11.14	6.04	48	9.5	6.67	F(1, 97) = 1.64	.200	.002
SM 7	Word puzzles	80	12.78	6.21	74	7.31	4.12	F(1, 152) = 42.17	< .001	< .001
SM 8	Word puzzles	301	11.72	7.75	305	7.61	5.99	F(1, 604) = 53.47	< .001	< .001
SM 9	Word puzzles	—	—	—	—	—	—	—	—	—

Additional studies summarized in Table SM 4 in the Supplemental Material
SM 10	Game	50	12.67	8.3	48	10.26	8.2	F(1, 96) = 2.08	.152	.047
SM 11	Count-locate	78	10.84	6.55	76	12.27	5.62	F(1, 152) = 2.11	.149	.087
SM 12a	Word puzzles	101	12.53	9.48	99	5.28	7.76	F(1, 198) = 34.95	< .001	< .001
SM 12b	Word puzzles	153	11.22	6.55	146	7.09	6.04	F(1, 297) = 32.01	< .001	.008
SM 12c	Word puzzles	144	13.04	8.37	157	8.39	5.90	F(1, 299) = 31.40	< .001	< .001
SM 12d	Word puzzles	98	12.23	8.95	102	7.95	4.55	F(1, 198) = 18.40	< .001	< .001
SM 12e	Word puzzles	101	13.13	6.26	99	9.46	6.76	F(1, 198) = 15.84	< .001	.001
SM 12f	Word puzzles	45	14.49	10.71	46	9.63	8.28	F(1, 89) = 5.88	.017	.021
SM 12g	Word puzzles	101	11.33	8.94	100	7.82	4.93	F(1, 199) = 11.83	.001	.001
SM 12h	Word puzzles	153	12.01	7.86	88	8.56	5.73	F(1, 239) = 12.97	< .001	< .001
SM 12i	Word puzzles	94	12.83	9.13	105	8.38	9.36	F(1, 197) = 11.46	.001	.001
SM 12j	Word puzzles	284	11.58	6.59	285	8.75	5.61	F(1, 567) = 30.29	< .001	< .001
SM 12k	Word puzzles	86	14.00	8.35	84	10.44	7.95	F(1, 168) = 8.10	.005	.022
SM 12l	Word puzzles	151	11.97	7.55	149	8.77	7.23	F(1, 298) = 14.07	< .001	.001
SM 12m	Word puzzles	101	11.26	6.61	99	8.74	5.96	F(1, 198) = 8.02	< .000	< .001
SM 12n	Word puzzles	142	11.32	5.80	161	8.95	6.13	F(1, 301) = 11.90	.001	.001
SM 12o	Word puzzles	722	11.62	7.42	697	9.28	5.68	F(1, 1417) = 44.08	< .001	< .001
SM 12p	Word puzzles	191	10.55	6.60	203	8.48	5.58	F(1, 392) = 11.42	< .001	< .001
SM 12q	Word puzzles	96	17.49	9.39	95	15.08	9.37	F(1, 189) = 3.14	.078	.074
SM 12r	Word puzzles	38	12.42	7.58	46	10.35	6.64	F(1, 82) = 1.79	.185	.072
SM 12s	Word puzzles	139	10.34	6.74	136	8.98	4.53	F(1, 273) = 3.84	.051	.076

Note: Means for studies that used the transcription paradigm are for transcription accuracy, means for studies that used the note-taking paradigm are for codings of note-taking quality, means for studies that used the word-puzzles paradigm are for correct words found, and means for studies that used the count-locate paradigm and the game paradigm are for the number of questions answered correctly. In Study 2a, perception was measured and not manipulated, and in Study SM 9, no performance measures were taken; therefore, both of these studies were excluded from the meta-analysis. ^aThis column shows the p value for the comparison of the number of correct responses across conditions. ^bThis column shows the p value for the comparison of the number of submitted responses across conditions (F values not shown).

Of the 30 studies, 6 studies were conducted in a behavioral lab at a northeastern university (n = 884; 38% male; average age = 22.36 years, 95% CI = [22.34, 22.37]), and 24 studies were conducted online (n = 7,096; 43% male; average age = 36.33 years, 95% CI = [36.33, 36.33]). These samples had significantly different gender distributions, χ²(1, N = 7,980) = 10.82, p = .001, and average age distributions, t(7978) = 35.32, p < .001.

Dependent variable

Across all of the studies, performance was calculated on the basis of the specific paradigm that was employed. Specifically, in the transcription task, we used correct words transcribed (as described in Study 1a). In the note-taking task, we used coders’ scores (as described in Study 1b). In the word-puzzles task, we measured the number of correct words identified (as described in Study 2b). In the count-locate task and the online-game task, we measured the number of correct multiple-choice responses (as described in Studies SM 5 and SM 6, respectively).

To determine the effect size on performance, we conducted an internal meta-analysis using a conventional approach (e.g., Borenstein, Hedges, & Rothstein, 2007; Cumming, 2014; Wilson, 2006). For each study, we calculated the Cohen’s d for the main effect of task framing (multitasking vs. single tasking) regardless of whether an additional factor was manipulated or whether we required all participants to work the same amount of time. Thus, we calculated a conservative estimate for the effect size. To calculate each study’s Cohen’s d, we subtracted the average performance in the multitasking condition by the average performance found in the single-tasking condition and divided the difference by the pooled standard deviation. We then weighted the Cohen’s d on the basis of the inverse variance of the study’s sample.

To determine whether a fixed-effects or random-effects model was appropriate, we conducted a test of homogeneity that revealed that the variability observed across the effect sizes exceeded what would be expected from sampling error (Q = 80.63, p < .001). Thus, we estimated an average Cohen’s d of 0.48, 95% CI = [0.40, 0.55], using a random-effects model. The random-effects model demonstrated that the effect of the perception of multitasking on performance was moderate in size and significantly greater than zero (z = 11.94, p < .001). Additionally, the I² statistic of 56.59%, 95% CI = [49.92%, 77.63%], suggests significant true heterogeneity of the effect size (for a forest plot, see Fig. 3).

Fig. 3.

Forest plot of the 30 observed effect sizes (Cohen’s ds) in each of the current studies, along with the overall size of fixed and random effects. Error bars show 95% confidence intervals.

Moderators

The 30 studies varied systematically on several dimensions. We explored some of these variations using a meta-analytic approach (Wilson, 2006). Because these analyses were exploratory and we did not make a priori predictions about these potential moderators, we adjusted the critical p value using a Bonferroni correction. Because we tested 6 potential moderators, the threshold for statistical significance was set at a p value of .009. In all of the analyses, we used a method-of-moments random-effects model (fixed-effects results are reported in the Supplemental Material). See Table 2 for a summary of results.

Table 2.

Size Estimates for Each Moderator Level

Moderator level	Cohen’s d	95% CI	z	p
Used term multitasking	0.49	[0.39, 0.59]	9.81	.00
Did not use term multitasking	0.44	[0.28, 0.60]	5.46	< .001
Lab	0.37	[0.18, 0.57]	3.80	< .001
Online	0.50	[0.41, 0.59]	11.29	< .001
Time controlled for	0.54	[0.20, 0.88]	3.12	.001
Time not controlled for	0.47	[0.39, 0.55]	11.27	< .001
Word puzzles	0.53	[0.44, 0.61]	12.72	< .001
Other paradigms	0.24	[0.06, 0.41]	2.58	.010

Note: CI = confidence interval.

Lab versus online sample

We first tested whether there was a significant difference in effect size as a result of running studies in the lab (6 studies; coded as 1) or online (24 studies; coded as 0). The analysis did not reveal a significant difference as a result of where the study was conducted (b = −0.12, 95% CI = [−0.34, 0.08]; z = −1.20, p = .230).

Including the term multitasking in the manipulation

In several of the studies (22; coded as 1), we used stronger manipulations in which we explicitly mentioned the word multitasking as part of the instruction in the multitasking condition. The rest of the studies (8; coded as 0) did not explicitly use this term. The analysis did not reveal a significant difference from using the term multitasking in the manipulation (b = 0.05, 95% CI = [−0.14, 0.23]; z = 0.48, p = .631).

Time on task

We next tested whether there was a significant difference as a result of exogenously controlling the time that participants spent on the task. Specifically, we compared studies (2; coded as 1) in which all participants were forced to work on the task for a specific amount of time with studies in which they were free to quit at any time (28; coded as 0). We did not observe a significant difference as a result of holding time constant (b = 0.07, 95% CI = [−0.28, 0.42]; z = 0.38, p = .702).

Switching and work sequence

As alluded to earlier, in the word-puzzles paradigm, participants were free to switch between the tasks whenever they liked. Thus, it is possible that part of the effect may be driven by a shift in work sequence due to different switch patterns across conditions. To approximate how much of the effect the switching account could explain, we coded whether each study employed the word-puzzles paradigm (24 studies; coded as 1) or used a different paradigm (6 studies; coded as 0). The analysis revealed a significant difference on this dimension, suggesting a stronger observed effect for studies that employed the word-puzzles paradigm (b = 0.29, 95% CI = [0.09, 0.49]; z = 2.90, p = .004). Note that this significant difference in effect size may be driven by factors inherent to the specific word-puzzles paradigm but may also suggest that participants’ difference in switching patterns across conditions is partly driving the effect. Because in our data, these two are fully confounded, we cannot disentangle these possibilities. Further, as alluded to earlier, one should be cautious in interpreting this analysis because the word-puzzles paradigm does not allow one to distinguish whether a task switch is an outcome of finding a word or the cause. Additional research is needed to more thoroughly and directly address this issue of causality. Having said that, we found that the effect persisted even when the analyses did not include all studies that employed the word-puzzles paradigm (d = 0.24, 95% CI = [0.06, 0.41]; z = 2.58, p = .010).

Incentive strength

We next tested whether there was a significant difference in the effect size as a function of incentives strength. Because our studies differed in the level of monetary incentives, for each study, we calculated the expected amount of bonus per minute on the basis of the average correct performance and used a regression to examine its effect. No significant difference was observed as a function of the expected bonus incentives employed (b = −0.10, 95% CI = [−0.65, 0.45]; z = −0.36, p = .718). We did a similar analysis examining actual bonuses received, which differed from expected bonuses when real-time coding of correct responses could not be done (e.g., the number of correct words transcribed). Again, we did not observe a significant impact of the strength of incentives employed on the effect (b = −0.22, 95% CI = [−0.96, 0.51]; z = −0.59, p = .554).

We next examined the effect of the bonus incentives as a proportion of the base payment amount (i.e., what participants typically receive per minute of work in the lab or on MTurk without bonus payments). Because there are different base payment amounts typically used in the lab and in the online panel, the proportion yields a standardized measure. For each study, we calculated the expected bonus amount on the basis of the average correct submissions and divided it by the base bonus amount ($0.17 per minute for lab studies and $0.10 per minute for online studies). Again, we did not observe a significant difference as a result of the different bonus incentives employed (b = −0.02, 95% CI = [−0.11, 0.08]; z = −0.36, p = .720). Similarly, using the calculation for an actual bonus as a proportion of the different base amounts did not yield a difference in the size of the effect (b = −0.04, 95% CI = [−0.17, 0.09]; z = −0.55, p = .584).

Analyzing performance as a function of perceptions

In 16 of the studies, in addition to manipulating the perception of multitasking, we measured participants’ perceptions of multitasking versus single tasking at the end of the study. Given the time that elapsed between when the manipulation was administered (prior to the task) and when the manipulation check (“To what extent did you feel like you are multitasking on these studies?” 1 = not at all, 4 = somewhat, 7 = totally) was measured (only at the end of the study), this measure was suspected to be relatively noisy. Nevertheless, when pooling the results from these studies, we did find a significant positive relationship between perceptions of multitasking and answers submitted (using z scores calculated for each study: b = 0.11, 95% CI = [0.08, 0.14]), t(4797) = 7.59, p < .001. A mediation analysis using a bootstrap estimation approach with 10,000 samples (Model 4 from the PROCESS macro; Hayes, 2013) estimated the mediating role of the manipulation check (b = 0.013, SE = 0.006, 95% CI = [0.0005, 0.0246]), thus providing further evidence that the mere perception of multitasking improves performance.

General Discussion

Multitasking is often a matter of perception. Although the term refers to the concurrent execution of multiple tasks, most tasks that require attention cannot be done simultaneously. In this article, we demonstrated the malleability of people’s perceptions of multitasking by showing that the exact same activity may or may not be perceived as multitasking. Further, by both manipulating and measuring multitasking perceptions, we found that, holding the activity constant, the mere perception of multitasking improves performance and that heightened engagement is an important driver of this effect.

These findings do not suggest that multitasking is superior to single tasking. Voluminous research demonstrates that working on more than one task is detrimental to performance. However, we argue and demonstrate that, holding the task (or tasks) constant, the mere perception of multitasking is beneficial to performance. Stated differently, one implication of this research is that separating an activity into its components and merely creating the perception of multitasking could improve people’s performance. For example, if we were to mention that reading this article entails two distinct tasks (e.g., switching back and forth between the text and figures), to the extent that this framing would trigger a perception of multitasking, one should observe an improvement in performance (i.e., better comprehension). Furthermore, our findings suggest that if people are already engaged in multiple tasks, making them aware that they are multitasking should increase engagement and help them perform better. So, if you are doing other activities while reading this article, such as answering urgent e-mails, realizing that you are multitasking should improve your engagement and performance in each of these activities.

Why is this happening? We found that the perception of multitasking increases engagement in the activity. Consistent with prior literature demonstrating that individuals’ motives for investing effort and cognitive control increase with the difficulty of the task, we conjecture that because multitasking activities are naturally perceived as challenging, merely framing the task as multitasking may increase individuals’ motivation. In addition, in light of our findings that multitasking is a desirable trait, an additional reason for the improvement in performance may relate to one’s motivation to appear an adept multitasker. Admittedly, other reasons may underlie this boost in engagement, and additional research is needed to disentangle the different psychological mechanisms that drive this robust effect.

Of importance, this effect persisted even when the work sequence was externally imposed (see Studies SM 5 and SM 6) and when switches between tasks occur rapidly (see Studies 1a and 1b). Thus, differential switching alone because of multitasking cannot fully account for the focal effect. Still, in Study 3, we did find that individuals switched between tasks more frequently in the multitasking condition. Although in this particular study it is not clear whether the increase in the number of switches was the cause for or the outcome of the improvement in performance, additional research is needed to examine whether tasks involving higher switching costs would pose a boundary condition for the observed effect.

In sum, although the prevalence of technology is bringing multitasking to almost every aspect of life, social scientists have focused on the detrimental effects of doing multiple tasks rather than doing a single task. By contrast, we made a different comparison: Given that many activities consist of different components, we tested whether the mere perception of engaging in multitasking or single tasking impacts performance. We show that in this context, multitasking is a malleable perception that, on its own, benefits rather than harms performance.

Supplemental Material

SrnaOpenPracticesDisclosure – Supplemental material for The Illusion of Multitasking and Its Positive Effect on Performance

Supplemental material, SrnaOpenPracticesDisclosure for The Illusion of Multitasking and Its Positive Effect on Performance by Shalena Srna, Rom Y. Schrift and Gal Zauberman in Psychological Science

Supplemental Material

SrnaSupplementalMaterial – Supplemental material for The Illusion of Multitasking and Its Positive Effect on Performance

Supplemental material, SrnaSupplementalMaterial for The Illusion of Multitasking and Its Positive Effect on Performance by Shalena Srna, Rom Y. Schrift and Gal Zauberman in Psychological Science

Footnotes

Action Editor

Leaf Van Boven served as action editor for this article.

Author Contributions

S. Srna collected and analyzed the data. All the authors jointly developed the study concepts and designs, conducted the studies, wrote the manuscript, revised the manuscript, and approved the final manuscript for submission.

Declaration of Conflicting Interests

The author(s) declared that there were no conflicts of interest with respect to the authorship or the publication of this article.

Funding

This work was supported by the Wharton Risk Center, an Ackoff Doctoral Student Fellowship, the Jay H. Baker Retailing Center, and the Wharton Behavioral Lab.

Supplemental Material

Additional supporting information can be found at

Open Practices

All data and materials have been made publicly available via the Open Science Framework and can be accessed at osf.io/9ungb/. The design and analysis plans for the studies were not preregistered. The complete Open Practices Disclosure for this article can be found at https://journals-sagepub-com.web.bisu.edu.cn/doi/suppl/10.1177/0956797618801013. This article has received the badges for Open Data and Open Materials. More information about the Open Practices badges can be found at .

Notes

References

Beatty

Lucero-Wagoner

(2000). The pupillary system. In Cacioppo

J. T.

Tassinary

L. G.

Berntson

(Eds.), Handbook of psychophysiology (Vol. 2, pp. 142–162). New York, NY: Cambridge University Press.

Bianco

A. T.

Higgins

E. T.

Klem

(2003). How “fun/importance” fit affects performance: Relating implicit theories to instructions. Personality and Social Psychology Bulletin, 29, 1091–1103.

Borenstein

Hedges

Rothstein

(2007). Meta-analysis: Fixed effect vs. random effects. Retrieved from https://www.meta-analysis.com/downloads/M-a_f_e_v_r_e_sv.pdf

Borger

(1963). The refractory period and serial choice-reactions. Quarterly Journal of Experimental Psychology, 15(1), 1–12.

Botvinick

M. M.

Braver

T. S.

Carter

C. S.

Barch

D. M.

Cohen

J. D.

(2001). Conflict monitoring and cognitive control. Psychological Review, 108, 624–652.

Bradley

M. M.

Miccoli

Escrig

M. A.

Lang

P. J.

(2008). The pupil as a measure of emotional arousal and autonomic activation. Psychophysiology, 45, 602–607.

Creamer

L. R.

(1963). Event uncertainty, psychological refractory period, and human data processing. Journal of Experimental Psychology, 66, 187–194.

Cumming

(2014). The new statistics: Why and how. Psychological Science, 25, 7–29.

Egner

Hirsch

(2005). The neural correlates and functional integration of cognitive control in a Stroop task. NeuroImage, 24, 539–547.

10.

Greene

B. A.

Miller

R. B.

(1996). Influences on achievement: Goals, perceived ability, and cognitive engagement. Contemporary Educational Psychology, 21, 181–192.

11.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis. New York, NY: Guilford.

12.

Higgins

E. T.

(2006). Value from hedonic experience and engagement. Psychological Review, 113, 439–460.

13.

Hoeks

Levelt

W. J. M.

(1993). Pupillary dilation as a measure of attention: A quantitative system analysis. Behavior Research Methods, Instruments, & Computers, 25, 16–26.

14.

Hommel

Fischer

Colzato

L. S.

van den Wildenberg

W. P. M.

Cellini

(2012). The effect of fMRI (noise) on cognitive control. Journal of Experimental Psychology: Human Perception and Performance, 38, 290–301.

15.

Jumio. (2013). 2013 Mobile Consumer Habits Study. Retrieved from http://pages.jumio.com/rs/jumio/images/Jumio%20-%20Mobile%20Consumer%20Habits%20Study-2.pdf

16.

Kahneman

(1973). Attention and effort. Englewood Cliffs, NJ: Prentice Hall.

17.

Kahneman

Beatty

(1966). Pupil diameter and load on memory. Science, 23, 1583–1585.

18.

Kieras

D. E.

Meyer

D. E.

Ballas

J. A.

Lauber

E. J.

(2000). Modern computation perspectives on executive mental processes and cognitive control: Where to from here? In Monsell

Driver

(Eds.), Control of cognitive processes: Attention and performance XVIII (pp. 681–712). Cambridge, MA: MIT Press.

19.

Kofman

Meiran

Greenberg

Balas

Cohen

(2006). Enhanced performance on executive functions associated with examination stress: Evidence from task-switching and Stroop paradigms. Cognition & Emotion, 20, 577–595.

20.

Kreckler

Catchpole

Bottomley

Handa

McCulloch

(2008). Interruptions during drug rounds: An observational study. British Journal of Nursing, 17, 1326–1330.

21.

Kukla

. (1972). Foundations of an attributional theory of performance. Psychological Review, 79, 454–470.

22.

Leroy

. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. Organizational Behavior and Human Decision Processes, 109, 168–181.

23.

Levy

Pashler

(2001). Is dual-task slowing instruction dependent? Journal of Experimental Psychology: Human Perception and Performance, 27, 862–869.

24.

Ophir

Nass

Wagner

A. D.

(2009). Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences, USA, 106, 15583–15587.

25.

Pashler

(1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116, 220–244.

26.

Plessow

Fischer

Kirschbaum

Goschke

(2011). Inflexibly focused under stress: Acute psychosocial stress increases shielding of action goals at the expense of reduced cognitive flexibility with increasing time lag to the stressor. Journal of Cognitive Neuroscience, 23, 3218–3227.

27.

Poynton

. (2002). Gamma FAQ—Frequently asked questions about gamma. Retrieved from https://poynton.ca/notes/colour_and_gamma/GammaFAQ.html

28.

Sanders

. (1983). Towards a model of stress and human performance. Acta Psychologica, 53, 61–97.

29.

Schrift

R. Y.

Netzer

Kivetz

(2011). Complicating choice. Journal of Marketing Research, 48, 308–326.

30.

Wang

Tchernev

J. M.

(2012). The “myth” of media multitasking: Reciprocal dynamics of media multitasking, personal needs, and gratifications. Journal of Communication, 62, 493–513.

31.

Wilson

D. B

. (2006). Meta-analysis stuff [Macros for SAS, SPSS, and Stata]. Retrieved from http://mason.gmu.edu/~dwilsonb/ma.html

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.80 MB

0.41 MB