Keywords

1 Introduction

The proliferation of smart devices in the workplace may serve to encourage multitasking among information workers. People’s focus duration on just the computer screen is found to be quite short. The average number of switches per minute between different contents is 0.95 and the average staying length on one content is 65 s [1]. Social media contributes significantly to multitasking in the workplace, e.g. a person switches from a Word document to Facebook. It was found that information workers check Facebook about 21 times a day in the course of working [2]. Among college students, who are potential future information workers, it was found that they check Facebook 52 times a day, and considering other types of social media, checking rises to 118 times a day [3]. Students who multitask due to checking social media most often report falling behind on schoolwork, and feel more distracted and lack control over online behavior [3].

Social media checking is associated with computer screen or device switching, and this type of multitasking has been found to be associated with stress [4], and negatively impact productivity [2] as well as people’s affect [5]. It has also been proved that multitasking is associated with the alteration of brain structures such that individuals performing high multitasking activities have smaller gray matter density in the anterior cingulate cortex [6]. The alteration of brain structures correlates with the decreased cognitive control performance and affect regulation among heavy multitaskers [6]. Thus, it is necessary to propose intervention methods to help information workers maintain productivity and focus duration and to promote workplace well-being. The current study aims to investigate whether mindfulness meditation can bring immediate benefits for people who work in a multitasking environment.

Neuroscience studies have shown that mindfulness meditation can increase control over the distribution of limited brain resources [7] and regulate affect [8]. Mindfulness is a state of consciousness in which attention is focused on present-moment phenomena [9]. Intensive meditation training can improve sustained attention against a decrement in vigilance [10]. In the workplace, mindfulness is associated with positive engagement in work and well-being [11]. Meditation training can also assist in treating for addiction [12]. The mediation practices are beneficial for controlling attention in an environment with distraction. Practicers are able to focus their attention on the primary task with the help of meditation and switch the attention back on the primary task from distracting interruptions.

The effects of mindfulness meditation on people’s attention and affect can come into play via continuous practices with a period of training. In the study by Nielsen and Kaszniak [8], the researchers recruited meditators following the Buddhist tradition, who had undergone long-term meditation practices. In another study, participants were given three months’ meditation training with five hours to meditate every day [10]. The results of the two studies both showed positive effect of meditation, either on emotional feelings or on task vigilance. In an experiment about multitasking and meditation [13], participants were given meditation training for eight weeks. The researchers found that after the training, participants stayed longer on the primary task and made fewer switches between tasks. These studies suggest that meditation training may be a feasible way to make people more focused in a distracting environment.

However, undergoing a long period of training in mindfulness meditation may not be practical for many people. While research is slowly suggesting that longer training can produce effects, many people may not be able to attend such training. At the same time, the human-computer interaction field is experiencing a surge of studies to investigate how short-term interventions can bring benefits [14]. However, it is unclear what effects a single experience of meditation can bring, in a person’s mental states or behavior. Understanding the effects of a single meditation session on mental states and performance is beneficial for proposing useful interventions that can counteract people’s stress and increase their focus.

In short, this study simulated an information environment where people multitask extensively and aimed to investigate the immediate effects of mindfulness meditation on people’s stress, focus, affect, workload, multitasking behavior and performance. The results will provide implications on designing intervention tools for information multitaskers to increase focus and reduce stress.

2 Method

2.1 Participants

Thirty-six students (17 males and 19 females)from a U.S. west-coast university were recruited in the experiment. A five-point Likert scale was used to measure the overall Facebook usage intensity (“to what extent do you agree or disagree you are a heavy Facebook user?”) in the recruitment questionnaire. Only “heavy Facebook users” (rating four or five points in this scale) were finally selected to participate in the experiment.

The average age of the participants was 21.4 (SD = 1.2) years old. All of them were native English speakers, with sufficient capability in English proofreading. The average Facebook age of all participants was 6.4 (SD = 1.6) years. According to their self-reported answers, they used Facebook for 3.7(SD = 2.3) hours daily and checked Facebook for 18.8 (SD = 19.9) times every day on average. Their Facebook intensity score as measured by with a five-point Likert scale was 3.9 (SD = 0.8). Participants were paid 20 dollars as the incentive.

2.2 Experiment Task

In this study, an information environment in the laboratory where multitasking occurs frequently were simulated, which was similar to previous studies [13, 15]. Participants were assigned to proofread Word documents on the U.S. California law codes about public utilities, which was the primary task. One 13-inch Dell Studio XPS laptop computer was provided to complete all the experiment tasks. The length of each document ranged from 1.5 pages to 2 pages, with Times New Roman as the font type and 12 points (pt) as the font size. The line spacing of the document was set 1.5 times. The page layout was A4 (21 cm × 29.7 cm). The top and bottom margins were 2.54 cm and the side margins were 3.18 cm. Each page contained nine spelling errors which were randomly distributed within the page. Participants were instructed to correct the errors directly in the document with the Word processor. Facebook browsing was the secondary task in the experiment. Participants were allowed to take a break whenever they needed it by switching to their own Facebook pages, but no other webpages should be browsed.

2.3 Design and Measurements

A between-subject experiment was designed to conduct this study. The independent variable was the intervention form with three levels: guided focus attention meditation, listening to a piece of arousing dance music remixes, and no intervention. Participants were randomly assigned to three groups corresponding to the three intervention forms. The dependent variables were stress, focus level, affect, workload, task switching behavior and task performance.

The participant’s stress level was measured by heart rate variability (HRV) with a digital heart rate monitor including a Polar RS800CX wristwatch receiver and a chest strap sensor. The sensor samples data with a frequency of 1 Hz and the HRV is calculated with the standard deviation of the R-R intervals between consecutive heart beats. The lower the standard deviation, the higher stress level is experienced [16].

The participant’s focus level was measured with EEG signals via a MuseTM brain sensing headband with seven electrodes [17]. The headband samples the EEG signal with a frequency of 220 Hz and the data are transmitted to a paired laptop computer. The developer kit of MuseTM supports exporting EEG data after the Fast Fourier Transform, status indicators of the four EEG data electrodes and the timestamp of each data frame. The status indicators of each electrode have three levels: 1 = good, 2 = ok, ≥3 bad. Only data frames with indicators of 1 or 2 were retained for further analysis. The EEG signals are categorized into different wave patterns with corresponding frequency intervals. In each data frame, the band powers of alpha (8–13 Hz), beta (13–30 Hz) and theta (4–8 Hz) waves were calculated for each electrode and the average band powers of the three types of brain waves were calculated across the four electrodes. The focus level is calculated according to Eq. (1), where alpha, beta and theta represent the average band powers of different waves [18].

$$ Focus = beta/\left( {alpha + theta} \right) $$
(1)

Affect was measured via a two-dimensional grid scale which was based on Russell’s Circumplex model which includes valence and arousal [19]. The scale has been proved to have adequate reliability, convergent validity and discriminant validity in measuring people’s affect [20]. Two orthogonal rating scales (610 px × 610 px) were represented on the laptop screen to measure the subjects’ valence and arousal respectively (Fig. 1). The horizontal scale measures valence from unpleasant (0 px) to pleasant (610 px). The vertical scale measures arousal from deactivated (0 px) to activated (610 px). In the experiment, participants were required to click the mouse in the two dimensional map presented in the browser and to click on the position that best represented their affect at that moment. The coordinates in pixels were then normalized to values ranging from 0 to 1 for the data analysis.

Fig. 1.
figure 1

Affect measurement map based on Russell’s Circumplex model.

Full size image

Workload was measured with a seven-point Likert scale adapted from the NASA-TLX [21], which contained six dimensions: physical demand, mental demand, time pressure, frustration, performance and effort. The scale about workload was produced and presented with Google Forms.

The data of the computer activity included the window switching frequency (window switching times divided by the time length of computer use) and the average duration of time on the proofreading windows. KidLogger was used to record the participant’s computer activity. The log records of KidLogger contain the start time and duration of the active window, the name of the window and a URL if applicable. The timestamp was recorded to the second. Computer activity data was calculated based on the log records. In addition, the accuracy rate of the proofreading task was adopted as the performance indicator.

2.4 Procedure

The experiment was conducted in a university laboratory and lasted about 1.5 h. Participants were randomly assigned to three groups which corresponded to the three intervention forms and were numbered group M, group A and group C respectively. In group M, participants were given a 12-minute session of guided focus attention meditation using the breath as the focus point. The meditation session was instructed by a pre-recorded gentle female voice accompanied with a sound of a sea wave. “Sit in a comfortable position and close your eyes. Allow your back to be straight and relax your shoulders. Take a few moments to relax by becoming aware that you are breathing. No need to change your breathing. Your body knows how to breathe.” This paragraph of instructions was adapted from those used in the MuseTM headband’s accompanied meditation application [17]. Participants were required to follow the instructions from the audio file to meditate until the instructions ended. In group A, participants listened to a piece of arousing club dance music remixes for 12 min, without engaging in any other activities until the music ended. Group C was a control group without any interventions.

The experiment procedure was illustrated in Fig. 2. Upon arrival, participants signed a consent form, read the instructions and filled out a general survey including their demographics, computer and Facebook experience, Facebook usage intensity. The Facebook usage intensity was measured with a five-point Likert scale. Afterwards, they put on the headband and heart rate monitor with the assistance of an experimenter.

Fig. 2.
figure 2

The experiment procedure.

Full size image

Then, participants were informed of the experiment instructions. In order to ensure the ecological validity and avoid bringing too many distortions on the behavior, the instructions omitted the real purpose of the experiment. Participants were informed that the purpose of the experiment was to test the impact of environment settings on university students’ working performance and workload. A paragraph was provided in the instructions as follows:

“We are interested in observing people’s work patterns. People often work on one task and take breaks when they need it. During the task, you can freely take a break at any time by switching to your own Facebook page. NO other new windows or webpages are allowed to open during the experiment. Please work with your natural pattern.”

What followed next was a training session with five sample documents to proofread. During the training session, the experimenter left the laboratory and the participants were not allowed to use Facebook. Upon the completion of the training task, the participants needed to inform the experimenter with a microphone. Then the experimenter would come and guide the participants to start the first formal session, in which participants needed to proofread the assigned documents and browse their own Facebook pages when they needed to relax. Participants were asked to pause the proofreading task after 30 min from the beginning of the formal session (session 1). Then affect and workload were measured in a questionnaire for the first time (measure 1). Participants in group M and group A were required to either did the meditation or listened to the club music. Afterwards, they were required to continue the proofreading task. Participants in group C were required to return to the proofreading task after the measure of affect and workload. After 30 min from the beginning of the second half session (session 2), participants finished the second half session and affect and workload were measured for the second time (measure 2). At the end of the experiment, participants rated their satisfaction level with the environment. In order to avoid the situation that participants did not browse Facebook during the experiment, they were asked to report one interesting Facebook post they had read at the end of the experiment. The recordings of the bio-signals and the computer activity was conducted throughout the formal session.

3 Results

The average age of participants in the three groups was 21.4 (SD = 1.1), 21.6 (SD = 1.0) and 21.3 (SD = 1.4) respectively, showing on significant inter-group difference (F (2, 31) = 0.32, p = 0.729, η2 = 0.02). There was also no significant inter-group difference on the gender ratio (χ2(2) = 0.22, p = 0.895), with 6, 5 and 6 male participants in the three groups respectively.

According to the self-reported data, there was no significant difference among participants in the three experimental groups on Facebook age (F (2, 31) = 1.02, p = 0.373, η2 = 0.07), daily usage frequency (F (2, 33) = 2.15, p = 0.132, η2 = 0.12), daily usage time span (F (2, 33) = 0.70, p = 0.505, η2 = 0.04), as well as their Facebook usage intensity (F (2, 33) = 0.82, p = 0.450, η2 = 0.05).

3.1 Stress and Focus

For participants in the two experimental groups (group M and group A), their EEG focus level and HRV were compared between session 1 and the intervention break in order to check the effect of mindfulness meditation and listening to the arousing club music. The HRV data in group A passed the normality test while the other data failed to pass the normality test.

For participants in group M, the EEG focus level and the HRV were 0.65 (SD = 0.42) and 73.74 ms (SD = 36.58 ms) on average in session 1, and were 0.40 (SD = 0.17) and 86.07 ms (SD = 36.15 ms) during the meditation. Wilcoxon signed rank test showed that the EEG focus level decreased and the HRV increased significantly during the meditation (W = 12.0, p = 0.034, r = 0.43; W = 8.0, p = 0.015, r = 0.50).

For participants in group A, the EEG focus level and the HRV were 0.53 (SD = 0.25) and 71.62 ms (SD = 17.18 ms) on average in session 1, and were 0.51 (SD = 0.22) and 85.26 ms (SD = 16.61 ms) while listening to arousing club music. Wilcoxon signed rank test and pairwise T-test showed that the HRV data increased significantly and the focus level remained the same while participants were listening to music (W = 29.0, p = 0.433, r = 0.16; t(11) = 3.42, p = 0.006, d = 0.34).

The above comparison means that both mindfulness meditation and listening to arousing club music took effect on the subject’s bio-signals when they were undergoing the intervention. What follows next is the comparison between session 2 and session 1 in the EEG focus level and the HRV level.

In the aspect of the EEG focus level, the data failed to pass the normality test. In session 1, the EEG focus levels of participants in group M, group A have been reported as above and that in group C was 0.49 (SD = 0.23). There was no significant inter-group difference (H(2) = 0.32, p = 0.854, η2 = 0.01) for the EEG focus level. In session 2, the EEG focus levels of participants in group M, group A and group C were 0.73 (SD = 0.57), 0.54 (SD = 0.29) and 0.48 (SD = 0.22). Wilcoxon signed rank test showed there was no significant differences between session 1 and 2 in terms of the EEG focus level in three groups (group M: W = 30.0, p = 0.480, r = 0.14; group A: W = 36.0, p = 0.814, r = 0.05; group C: W = 36.0, p = 0.814, r = 0.05) (Fig. 3a).

Fig. 3.
figure 3

EEG focus and HRV level (ms) in the three groups.

Full size image

In the aspect of the HRV level, the data failed to pass the normality test. In session 1, the HRV data in group M and group A have been reported as above and that in group C was 72.83 ms (SD = 36.49 ms). There was no significant inter-group difference (H(2) = 0.25, p = 0.881, η2 = 0.01). In session 2, the HRV of participants in group M, group A and group C were 79.55 ms (SD = 31.98 ms), 89.23 ms (SD = 39.74 ms) and 73.76 ms (SD = 40.73 ms). Wilcoxon signed rank test indicated that the HRV level in session 2 increased marginally significantly in group M (W = 16.0, p = 0.071, r = 0.37), increased significantly in group A (W = 4.0, p = 0.006, r = 0.56) and remained the same in group C (W = 37.0, p = 0.875, r = 0.03) (Fig. 3b).

3.2 Affect and Workload

The data of self-reported valence and workload passed the normality test and the test of variance homogeneity while the data of self-reported arousal failed to pass the normality test.

There was a significant interaction effect of session and group on the valence of participants in the three groups (F(2, 33) = 5.94, p = 0.006, η2 = 0.27), as well as a marginally significant main effect of session (F (1, 33) = 3.42, p = 0.073, η2 = 0.09). In session 1, the valence levels of participants in group M, group A and group C were 0.56 (SD = 0.20), 0.61 (SD = 0.19) and 0.68 (SD = 0.10), showing no significant inter-group difference (F(2, 33) = 1.46, p = 0.246, η2 = 0.08). In session 2, the valence levels of participants in group M, group A and group C were 0.62 (SD = 0.21), 0.57 (SD = 0.19) and 0.55 (SD = 0.09). Simple effect analysis showed that mindfulness meditation and listening to club music had no significant influence on valence (F (1, 33) = 1.95, p = 0.172, η2 = 0.06; F (1, 33) = 1.27, p = 0.269, η2 = 0.04). However, the valence level of participants in group C decreased significantly in session 2 (F (1, 33) = 12.07, p = 0.001, η2 = 0.27). The comparison results of valence are depicted in Fig. 4a.

Fig. 4.
figure 4

Self-reported affect in the three groups.

Full size image

The arousal levels of participants in group M, group A and group C were 0.57 (SD = 0.19), 0.63 (SD = 0.23) and 0.63 (SD = 0.20) in session 1, showing no significant inter-group difference (H (2) = 0.65, p = 0.723, η2 = 0.02). In session 2, the arousal level of participants in group M increased to 0.67 (SD = 0.24), which was a marginally significant effect on increasing the arousal level (W = 15.0, p = 0.060, r = 0.38). The arousal levels of participants in group A and group C were 0.59 (SD = 0.24) and 0.57 (SD = 0.21) in session 2, showing no significant difference compared with the arousal levels in session 1 (W = 29.5, p = 0.456, r = 0.15; W = 25.5, p = 0.289, r = 0.22). The comparison results of arousal are depicted in Fig. 4b.

The workload levels of participants in group M, group A and group C were 3.6 (SD = 0.8), 3.4 (SD = 0.9) and 3.3 (SD = 1.0) in session 1, showing no significant inter-group difference (F(2, 33) = 0.35, p = 0.705, η2 = 0.02), and were 3.4 (SD = 1.1), 3.7 (SD = 1.1) and 3.3 (SD = 1.1) in session 2. There was no significant inter-session difference in workload among the three groups (F(2, 33) = 1.48, p = 0.242, η2 = 0.08). With a further analysis on the six dimensions of NASA-TLX, it was found that interventions had different effect on frustration among the three groups (F(2, 33) = 4.85, p = 0.014, η2 = 0.23), as shown in Fig. 5. Mindfulness meditation significantly decreased frustration (F(1, 33) = 4.68, p = 0.038, η2 = 0.12), from 4.1 (SD = 1.6) to 3.1 (SD = 1.9), while listening to club music significantly increased frustration (F(1, 33) = 4.68, p = 0.038, η2 = 0.12), from 3.2 (SD = 1.3) to 4.2 (SD = 1.3). For participants in group C, the frustration level was 2.8 (SD = 1.6) in session 1 and 3.2 (SD = 1.5) in session 2, showing no significant difference (F(1, 33) = 0.52, p = 0.476, η2 = 0.02). Inter-group comparison showed that there was no significant difference in frustration in session 1 among the three groups (F(2, 33) = 2.24, p = 0.123, η2 = 0.12).

Fig. 5.
figure 5

Self-reported frustration in the three groups.

Full size image

3.3 Task Switching Behavior and Performance

There was no significant inter-group difference in session 1 in terms of the window switching frequency (F (2, 33) = 0.08, p = 0.919, η2 = 0.01), the average staying time on the proofreading windows (F (2, 33) = 0.24, p = 0.789, η2 = 0.02) and the proofreading accuracy (F (2, 33) = 1.19, p = 0.318, η2 = 0.07). In session 2, task switching behavior of participants did not change significantly among the three groups in terms of the window switching frequency (F (2, 33) = 0.27, p = 0.763, η2 = 0.02) and the average staying time on the proofreading windows (F (2, 33) = 0.11 p = 0.893, η2 = 0.01). Different interventions had no significantly different effect on the proofreading accuracy among the three groups (F (2, 33) = 0.99, p = 0.384, η2 = 0.06).

In summary, there was no significantly different intervention effect on the subject’s task-switching behavior as well as the performance in the task.

4 Discussion

The present study investigated whether mindfulness meditation could have immediate effects on a person’s mental states or behavior in an information multitasking environment. The results showed some immediate effects of mindfulness meditation and listening to arousing club music on people’s stress, affect and frustration. A brief meditation session could decrease a person’s stress level and increase the arousal level. Meditation could also significantly decrease a person’s frustration level and retain the valence level, but it has no significant effects on a person’s focus level, workload, task switching behavior and task performance. Listening to arousing club music could decrease a person’s stress level, but the side effect was that the frustration level was increased. Besides, listening to arousing club music had no significant influence on a person’s affect, focus level or task switching behavior.

In the experiment, the valence of multitaskers was maintained by meditation, but tended to decline if no break was taken, despite the fact that the valence was still kept positive on average in the control group. This is consistent with previous findings [13], indicating that meditation can resist the decrease of positive affect. It is worth noticing that other forms of breaks might have effects in keeping the positive affect as well, such as exposure to music. This suggests that appropriate break may be beneficial to increase workplace wellbeing. In terms of arousal, meditation increased the arousal level, but no obvious change was found in the other two groups. The results suggest that meditation holds promise in refreshing and energizing people even though they have not experienced long-term training. It was also found that although listening to club music could significantly decrease people’s stress level in an information multitasking environment, it had a side effect of strengthening the frustration. The music type adopted in this experiment was club music which was arousing with a fast beat. Future research can investigate the effect of music type on improving people’s mental states in a multitasking environment.

We also found that the multitasking computer activity, primary task performance, the EEG focus level and workload were not influenced by mindfulness meditation nor exposure to club music. Considering that previous findings on meditation’s role in increasing attention and decreasing task switches were all based on long-term training [10, 13], we infer that the instant effect of a brief intervention may be limited in changing a person’s focus level or too frequent task switching behavior. That highlights the necessity of professional training and discipline with meditation. Meditation should not be counted on to be able to correct inappropriate multitasking behavior or increase attention without continuous and disciplined engagement. Another possible reason for the limited effect of meditation is that the multitasking in this experiment involved mainly internal interruptions [22]. Previous study proved that meditation could change task switching frequency given that the interruptions were external [13]. Due to different mechanisms behind the two types of interruptions, the effect of meditation may be different and further exploration is warranted.

Our results have implications for product design aiming to regulate people’s attention and behavior. As assistive tools for meditation, products need to provide scientific guidance to guarantee that users meditate in the right way. In addition, assistive products for meditation should increase user stickiness and continuous engagement by introducing some mechanism such as gaming or social media sharing. Otherwise, the influence of such products on mental health and behavior is limited for short-term users. In fact, the headband MuseTM has a corresponding mobile application to allow users to share their meditation progress, forming a mutual competition and motivation among their social relations. These functions are beneficial for motivating users to engage in long-term meditation practices and regulating their mental states. However, it is worth noticing that the function of gaming or social media sharing should not cause users to be addicted, otherwise it may exacerbate task switching or distracted mental states for information workers.

Our study investigated whether a meditation intervention can have immediate effects on people’s mental states and behavior, using a simulated work environment. Mindfulness meditation appears to show promise when training is undergone for extended periods. Yet despite the widespread interest in developing short-term interventions for improving workplace wellbeing, our results suggest that single instances of meditation may have only limited effects. We hope that our results can spark others to investigate ways to alleviate workplace stress and increase focus in the modern information work environment.