Ego Depletion May Explain Gender Differences in Multitasking

Introduction

An age-old claim is that females outperform males when multitasking (Mantyla, 2013), but only very recently scientific research supporting this claim has emerged (Stoet, O’Connor, Conner, & Laws, 2013). Multitasking is a relatively broad concept (Salvucci & Taatgen, 2010), but there are at least two main distinct types (Stoet et al., 2013): (1) tasks that need to be carried out sequentially; and (2) tasks that need to be carried out simultaneously. Here, we focus on the first type of multitasking, in particular, task switching (Monsell, 2003). This refers to gauging the difficulty of rapidly switching attention between two or more tasks. Responses tend to be slower and more error-prone immediately after a task switch, but this ‘switch cost’ can be reduced by an opportunity for preparation (Monsell, 2003). For this reason, here, we hypothesize that performing well in task switching depends on individual cognitive reflection. Automatic responses are error-prone, and slow thinking is necessary for one to reach the correct response.

Two mental systems compete for control in tasks that involve judgement and choice (Evans, 2003). Intuitive decisions that require little reflection use ‘System 1’. Decisions that use mental models engage ‘System 2’. Actually, System 1 is a set of autonomous subsystems that comprise input modules related to specific-domain knowledge. System 2 is related to abstract reasoning and the way we think using hypotheses. A simple test can measure how individuals differ in cognitive ability in terms of the relative powers of their System 1 and System 2: the cognitive reflection test (CRT) (Frederick, 2005). Individuals scoring higher on the CRT display enhanced ability for using their System 2 to override System 1 proclivities.

Cognitive reflection is affected by ‘ego depletion’ (Kahneman, 2011; Muraven, Tice, & Baumeister, 1998). Ego depletion refers to the fact that an effort of will or self-control is tiring, and that if you have had to force yourself to do something, you are less willing or less able to exert self-control when the next challenge comes around (Kahneman, 2011). Ego depletion impairs performance on tasks that require controlled processing, such as active problem-solving, but it does not impair performance on other tasks that involve basic forms of information processing without active self-control (Schmeichel, Vohs, & Baumeister, 2003). In other words, ego depletion impairs performance on tasks that use System 2, but not on tasks that are automatically performed by System 1.

Here, we apply the CRT in two experiments. In a pilot pretest (n = 460), male versus female performance is recorded. In the second experiment (n = 60), the CRT is applied after a computer-based switching task. We then record male versus female CRT performance, arguably under ego depletion. Both experiments allow us to compare male versus female performance with and without ego depletion and to collect information regarding the role ego depletion plays in male versus female switching task performance.

As for the first type of multitasking we study, there are two conflicting results in the literature. Stoet et al. (2013) argue that females are better than males at multitasking, while Buser and Peter (2012) disagree. Our aim is to take a side in this debate.

The rest of the article is organized as follows: The following section details the materials and methods, the Results section presents the results, the Discussion section discusses them and the Conclusion section shows the conclusion.

Materials and Methods

In a pilot experiment, the experimenter (L.B.) recruited 541 volunteers, mainly students from the Federal University of Santa Catarina, southern Brazil. He obtained 460 valid responses from 257 females and 203 males. Only the CRT was applied in this pretest (Experiment 1), between March 2018 and May 2018. The dataset is available at Figshare (http://doi.org/10.6084/m9.figshare.7668218.v1).

An Experiment 2 took place the next academic semester, from June to August, at the same premises. This time, the experimenter (R.L.) recruited, on a one-to-one basis, 30 females and 30 males who did not participate in the pretest. The smaller sample reflects the difficulties in applying the computer-based switching task, explained later. This task had to be applied online on an internet site and in the presence of the experimenter, and only those volunteers who had a good understanding of English could participate. Also, this task was time-consuming. The dataset is also available at Figshare.

The CRT (Frederick, 2005) applied in both Experiments 1 and 2 encompasses three questions that are designed to elicit automatic responses that are compelling, but wrong.

The experimenters instructed participants to respond to the three questions in less than 30 seconds. This assured them they made an automatic choice. They also asked whether each participant already knew any of the questions. Those who reported to know at least one were left out from the experiments.

As observed, task switching experiments are designed to measure the difficulty of rapidly switching attention between two or more tasks. Performing a task consists of a simple response to a simple stimulus according to simple rules. In Experiment 2, the experimenter (R.L.) employed the software PsyToolkit and administered to the participants its Task Switching (available at https://www.psytoolkit.org/experiment-library/taskswitching.html). Here, the participants had to perform three steps of the task, all of which displayed in a large square with four quadrants inside, where in each of them a combination of a letter and a number would appear.

In Step 1 of the task, a participant had to focus only on the letters that would appear along with the numbers only in the top quadrants. Whenever a consonant appeared, the participant should press the letter B of the computer keyboard. Whenever a vowel appeared, the participant should press the letter N (Figure 1(a)).

In Step 2, letters and numbers would continue to appear, but now in the bottom quadrants. In this case, a participant had to focus only on the numbers. Whenever an odd number appeared, the participant should press the letter B on the keyboard. Whenever an even number appeared, the participant had to press the letter N (Figure 1(b)).

OPEN IN VIEWER

Figure 1.

Source: PsyToolkit

Step 3 of the task switching is full-blown multitasking. Participants had to consider letters and numbers together. Whenever a combination of a letter and a number appeared in the top quadrants, a participant had to focus only on the letter. Whenever a combination appeared in the bottom quadrants, the participant had to focus only on the number (Figure 1(c)).

Each of the three steps of the task switching experiment was played 40 times by every participant using his or her own computer. This circumstance is likely to have generated a cognitive load for him or her. The CRT was administered after the task switching through Google Docs and also using a participant’s own computer. This was purposeful. The very fact that the CRT came second ensured the participants responded to it after cognitive load and, therefore, under ego depletion. Thus, we expected impaired CRT performance in Experiment 2 relative to that in Experiment 1. Nevertheless, what mattered most for us was to investigate the relative male versus female performance in Experiment 2.

Results

Table 1 shows the results of the CRT in the pretest, by gender. More female participants failed to score, and more males correctly responded to the three questions. Thus, our result replicates the established fact in the literature that males score significantly higher on the CRT than females do (Frederick, 2005; Toplak, West, & Stanovich, 2011). We will be back to this issue in the Discussion section.

In Table 1, Pearson’s chi-square statistics with three degrees of freedom pointed to a significant difference between the CRT distributions, by gender (χ²₍₃₎ = 20.1, p-value = 0.00016). Considering that the CRT distributions according to gender presented different variances, a t test indicated a significant difference between their mean CRTs (t = −4.3405, d.f. = 427.01, p-value = 0.000012) and Cohen’s d with 95 per cent confidence was −0.409 ± 0.187. By coding ‘female = 0’ and ‘male = 1’, Pearson’s correlation computed from Table 1 was approximately equal to 0.20 (p-value = 0.000017). The female CRT distribution showed higher concentration of zeros (χ²₍₃₎ = 44.8, p-value < 0.00001), therefore, being nonuniform. Yet, the male CRT distribution was statistically uniform (χ²₍₃₎ = 1.0, p-value = 0.80), thus suggesting male performance was more randomly distributed.

OPEN IN VIEWER

Table 1.

CRT Performance (Percentage Scoring 0, 1, 2 or 3) Without Ego Depletion, by Gender

Gender	n	0	1	2	3
Female	257	42	23	20	15
Male	203	24	26	22	28

Table 2 shows the results of the CRT applied after the task switching experiment. The CRT performance was affected by ego depletion for both males and females. None of the 60 participants now scored three correct responses.

OPEN IN VIEWER

Table 2.

CRT Performance (Percentage Scoring 0, 1, 2 or 3) with Ego Depletion, by Gender

Gender	n	0	1	2	3
Female	30	57	30	13	0
Male	30	60	33	7	0

Table 2 shows concentrations of zeros for both males and females. Now, Pearson’s chi-square statistic with two degrees of freedom did not indicate a significant difference between the CRT distributions, by gender (χ²₍₂₎ = 0.75, p-value = 0.688). By coding ‘female = 0’ and ‘male = 1’, no significant Pearson’s correlation was computed in Table 2 (null hypothetical correlation with p-value = 0.57). Considering that the CRT distributions according to gender presented equal variances, a t test did not show a significant difference between their mean CRTs (t = 0.57, d.f. = 58, p-value = 0.571). Similarly, Cohen’s d was also statistically null, since its 95 per cent confidence interval 0.147 ± 0.518 encompasses the value zero.

Figure 2(a) and (b) compare female and male performance, respectively, considering the results in the pretest. As can be seen, multitasking exponentially impaired cognitive reflection for both males and females. Yet, it seems that males suffered more from the ego depletion induced by multitasking because they came from relative success in Experiment 1 as compared to females to similar poor performance in Experiment 2.

OPEN IN VIEWER Figure 2.

(a) Female CRT Performance (Percentage Scoring 0, 1, 2 or 3) With (Yellow Bars) and Without (Red Bars) Ego Depletion; (b) Male CRT Performance (Percentage Scoring 0, 1, 2 or 3) With (Yellow Bars) and Without (Red Bars) Ego Depletion

As for the task switching experiment, we computed the errors made by the participants after 40 rounds of each of the three steps. Regardless of gender, the mean error was 1.6833 in Step 1; 2.1167 in Step 2 and 4.9667 in Step 3. As expected, the mean error in Step 3 was larger. More importantly, the mean error for females in Step 3 was 4.0000 (SD = 0.0263), while that for males was 5.9333 (SD = 0.0613). Thus, females outperformed males in this type of multitasking. This favours the finding of Stoet et al. (2013) over that of Buser and Peter (2012).

Table 3 shows the number of errors in Step 3 of the task switching experiment, by gender. No statistically significant difference was detected between the averages of the distributions of number of errors (t = −1.5878, d.f. = 39.30, p-value = 0.12). Cohen’s d statistic was also not different from zero, presenting a 95 per cent confidence interval −0.410 ± 0.522. However, the standard deviations (2.63 for females and 6.13 for males) were statistically distinct, according to Levene’s test for equality of variances (p-value = 0.002). This suggests males presented a more heterogeneous pattern in relation to females, having a greater frequency of extremes (higher incidence of both few errors and many errors). Table 3 shows that almost all the errors made by females were concentrated in the interval 1 to 10, and Fisher’s exact test allowed us to infer that the distributions of errors between males and females were indeed different (p-value = 0.0013).

OPEN IN VIEWER

Table 3.

Number of Errors in Step 3 of the Task Switching Experiment, by Gender (Percentage)

Gender	No Error	1 to 10 Errors	>10 Errors
Female	0.0	96.7	3.3
Male	16.7	60.0	23.3

Discussion

The result that males present superior cognitive reflection does not refer to IQ. The CRT measures the ability to resist reporting the response that first comes to mind (cognitive reflection). It thus tracks ‘broad rationality’, not instrumental rationality or algorithmic intelligence, which is what conventional measures of intelligence gauges (Stanovich, 2004). For this reason, the CRT is arguably better than the conventional measures for predicting decision-making (Frederick, 2005; Toplak, West, & Stanovich, 2011).

System 1 is similar to a Swiss army knife, in that each of its modules is evolutionarily adapted for solving a different problem, and System 1 is evolutionarily older than System 2. Many decisions based on System 1 that seem irrational from one individual’s perspective ultimately have an evolutionary logic (Da Silva et al., 2018).

Thus, what we found was simply that males are more likely to reflect on their responses and less inclined to go with their intuitive responses. Females are more automatic; they have a stronger System 1, which in a sense means they are relatively more evolutionarily adapted than males. This brings females advantages in tasks that require (System 1 based) domain-specific knowledge, although this very fact also exposes them to more predictable biases, especially in context-free environments. The CRT is context-free and thus of limited ecological validity.

Our result that females outperform males in multitasking was obtained using a computer-based switching task, which is arguably of limited ecological validity too. However, using our context-free switching task is likely to even underestimate female advantage on this issue. This is so because females are evolutionarily adapted for gathering, and males for hunting. And females’ gathering needed to be combined with looking after offspring. This possibly deeply explains that females are better than males at multitasking (Ren, Zhou, & Fu, 2009).

Neuroscience evidence suggests that retrospective memory is key to multitasking, where both the left anterior and posterior cingulate play a main role (Burgess, Veitch, Costello, & Shallice, 2000). Secondly, prospective memory and planning also matter for multitasking, and, here, the left brain areas 8, 9 and 10, and the right dorsolateral prefrontal cortex are important. Thus, searching for gender differences in the functioning of these brain areas could ultimately reveal the neural substrates for female advantage in multitasking.

We focussed on the first type of multitasking, that is, tasks that need to be carried out sequentially. As for the second type (simultaneous tasks), Mantyla (2013) found that, contrariwise, males had an advantage over females, mainly due to male advantage in spatial skills. However, this second type of multitasking is of less relevance to daily life contexts, where people often carry out tasks sequentially (Stoet et al., 2013).

Strayer and Watson (2012) argued that individual differences in executive attention are most likely to underlie the ability to multitask, and this holds true regardless of gender. However, this stance misses the point that executive attention is an ingredient of cognitive reflection, which is not gender neutral. Never-theless, Strayer and Watson are correct at indirectly pointing to the mediation of cognitive reflection on multitasking, which ends up related to cognitive load and ego depletion, a case we elaborated in this study.

Finally, we turn to our finding that multitasking exponentially impairs cognitive reflection for both males and females, though males are more affected by multitasking-induced ego depletion. Wenzel, Zahn, and Kubiak (2018) argue that practice counteracts the effects of diminished willpower in the long run. This may help explain our finding that males are more affected by ego depletion. Indeed, not all System 1 modules are wired, some modules result from overpractice (Kahneman, 2011). And as observed, ego depletion impairs performance on tasks that uses System 2, but not on tasks that are automatically performed by System 1. So, as females have a stronger System 1, they are supposedly more prone to incorporate learned practices, and this renders them more resilient to the negative effects of ego depletion.

Conclusion

Are females better than males at multitasking? Tentative scientific answers have been produced only very recently. For tasks that need to be carried out sequentially—arguably the most relevant type of multitasking—the answer is ‘yes’ for some (Stoet et al., 2013), but ‘no’ for others (Buser & Peter, 2012). Here, we set two experiments only to find support for the ‘yes’. Yet, this is a qualified ‘yes’ because males presented a higher incidence of both many errors and few errors in the task switching experiment.

Thus, we roughly replicate the finding that females outperform males in multitasking. Considering the task switching type of multitasking, our experiments show that multitasking impairs cognitive reflection through ego depletion, regardless of gender. However, the cognitive reflection of males seems to be more negatively affected after multitasking. This suggests ego depletion may be an interesting candidate mechanism to explain gender differences in multitasking performance.