Searching for Affective and Cognitive Restoration: Examining the Restorative Effects of Casual Video Game Play

Introduction

Cognitive fatigue (i.e., declines in working memory and decision making) due to sustained cognitive effort is a concern in many domains, occupational and academic (e.g., Demerouti, Bakker, & Leiter, 2014; Palmer et al., 2013). Fatigue is especially problematic due to its prevalence: 60% of Americans regularly experience stress, including feelings of frustration, anxiety, and fatigue (Anderson et al., 2015). Individuals with safety-critical occupations (e.g., air traffic controllers, pilots, transportation security officers, medical professionals) may experience greater cognitive fatigue due to their increased workload and may commit more errors (Mueller-Leonhardt, Stroebaek, & Vogt, 2015; Terte & Stephens, 2014).

Due to the high costs of fatigue, finding empirically validated methods of reducing its effects are vital. Solving this problem has the potential to provide benefits both on the job and in daily life. Previous research has investigated several methods to mitigate these symptoms and their impact on performance. Some studies have examined the effect of rest breaks following demanding tasks (e.g., Arrabito, Ho, Aghaei, Burns, & Hou, 2015; Helton & Russell, 2015, 2017). Others have examined the effects of guided relaxation techniques on cortisol levels associated with increased stress (Cruess et al., 2015). Another line of research has examined the phenomenon of cognitive restoration in order to understand which tasks can “recharge” cognitive functioning and why. Restoration has not been clearly defined, but studies have frequently examined both affective (e.g., mood or feelings of stress) and cognitive (e.g., working memory or executive attention) processes (Berman, Jonides, & Kaplan, 2008; Berto, 2005; Emfield & Neider, 2014). Previous research has also shown a dissociation between affective and cognitive restoration. In other words, feeling better is not the same as performing better (Finkbeiner, Russell, & Helton, 2016), making it critical to study both sides of restoration. We will refer to this composite approach as affective and cognitive restoration.

Methods

This research complied with the American Psychological Association Code of Ethics and was approved by the Institutional Review Board at the University of Central Florida where the study took place. Informed consent was obtained from each participant.

Participants

Sixty-six undergraduate students, 20 men and 46 women, with normal or corrected-to-normal acuity and color vision participated in the current study. Participants did not self-report any additional sensory or motor deficiencies. Participants self-reported their caffeine consumption within the 12 hours prior to their arrival at the lab as no more than two eight-ounce beverages. Participants received course credit for participation. Their ages ranged from 18 to 33 years of age, with a mean age of 20.28 (SD = 2.72). Overall, participants averaged 3.57 (SD = 4.82) hours of weekly casual video game play but did not report having any previous experience with Sushi Cat 2, the casual game used in the current study. Due to technical issues, we omitted data from 1 person, leaving 65 participants for further analysis. Table 1 shows demographic breakdown by condition.

OPEN IN VIEWER

Table 1:

Means and Standard Deviations of Demographic Information of Study Participants

	Break (n = 22)	Relaxation (n = 21)	Game (n = 22)
Age (years)	19.36 (1.29)	21.14 (3.92)	20.36 (2.17)
Casual video game hours per week played	4.55 (5.54)	3.19 (4.80)	2.95 (4.08)
Gender
Men (N)	7	6	7
Women (N)	15	15	15

Materials and Apparatus

The computer-based tasks employed in this study were performed on a Dell computer running Windows 7. Participants wore noise-canceling headphones and sat 66 cm from a Dell E176FP 43.2 cm flat panel LCD display (1,280 × 1,024 pixel resolution; 60 Hz refresh rate). The display thus subtended a diagonal visual angle of approximately 14.7°. The testing room was illuminated with common fluorescent lighting, and the workstation was situated so that glare was minimized. Experimental tasks were presented using E-prime v.2.10 (Psychology Software Tools, Pittsburgh, PA; Schneider, Eschman, & Zuccolotto, 2002).

We used the Temple et al. (2000) abbreviated vigilance task to induce fatigue. Participants’ head movements were unrestrained. The task consisted of the rapid presentation of either a low probability target (letter O) or two distractors (forward or backward letter D) presented as 8 × 6 mm light grey capital letters in 24-point Avant Garde font centered on the display screen and displayed with a visual mask so that the letter stimuli appeared to be underneath due to interposition. The mask consisted of 1 mm diameter unfilled dark grey circles uniformly spaced on a white background. Each circle was outlined in a .25 mm–thick black line. For all conditions, both practice and experimental, the proportion of target signal trials was held constant at .20, and the proportion of non-target stimulus trials was held constant at .80. Stimulus presentation was 40 milliseconds on each trial, with an inter-stimulus interval of 1000 milliseconds (Figure 1). During the task, participants were instructed to monitor the center of the display for the target stimulus and press the computer space bar when a target appeared and to not respond to non-target stimuli.

OPEN IN VIEWER

Figure 1.

Display of the abbreviated vigilance task. Example of target stimuli for three sample consecutive trials. ISI = 1000 milliseconds.

Participants first completed 5 minutes of practice trials with knowledge of results feedback, the purpose of which was to familiarize participants with the vigilance task. Thus, during practice, participants received auditory feedback as the words hit, miss, and false were spoken aloud for hit, miss, and false alarm responses, respectively; no feedback was given for correct rejections. We did not set a minimum hit or false alarm rate that participants needed to achieve prior to continued participation in the study. Following practice, participants completed six continuous 2.5-minute periods of watch without knowledge of results feedback. Each continuous period of watch consisted of 120 trials (24 critical; 96 neutral). The proportion of target signal trials to non-target trials was consistent within periods.

The Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988) was given to assess both positive and negative affect. A shortened, 20-question version of the Dundee Stress State Questionnaire (DSSQ-S; Matthews, Emo, & Funke, 2005; Matthews, Joyner, Gilliland, Huggins, & Falconer, 1999) was used to assess three aspects of stress: task engagement, distress, and worry. The PANAS and DSSQ-S instruments were administered to participants at pre-vigil, post-vigil, and post-intervention. These were administered via pencil and paper and scored according to the original authors’ methodology.

A computerized adaptive version of a backward digit-span (BDS) task was administered to each participant at pre-vigil, post-vigil, and post-intervention as a measure of working memory span. The test used a staircase-type procedure to present a series of numbers via the headphones that varied from three to nine digits in length, after which the participant would use the computer keyboard to type in the digits in reverse order. The BDS started with a three-digit numerical series, increasing in length by one digit each time participants consecutively answered two trials correctly and decreasing by one digit following each incorrect trial. After 14 trials were administered, participant’s digit-span was recorded as the length of the final series performed correctly for two consecutive trials. This method was modeled after previous research (Emfield & Neider, 2014).

The independent variable of interest was the type of intervention applied after the vigilance task. Participants played Sushi Cat 2 (Armor Games, Irvine, CA), engaged in a guided relaxation exercise, or sat quietly. Sushi Cat 2 (Figure 2) is a pachinko-type game and was used as it was a prototypical casual game that while being novel to participants, was simple with easy to understand instructions. It did not contain specific memory demands or violence, and minor increases in difficulty with increasing levels helped keep participants engaged without being overwhelmed. At the beginning of a level, Sushi Cat appears near the top of the pegboard. The horizontal position of the character is determined by the on-screen cursor position, as controlled by the computer mouse. Players use the mouse to move Sushi Cat to a desired position at the top of the pegboard and click the left mouse button to drop the character onto the pegs. Once released, Sushi Cat bounces around, ultimately landing in one of five bins with assorted point values. As Sushi Cat falls, it collects sushi items that are placed throughout the pegboard. The goal is to complete the level by collecting all of the sushi. In each level, players have a total of five drops of Sushi Cat in order to collect as much sushi as possible before continuing to the next level. As the Sushi Cat eats more sushi, it increases in size and weight, increasing gameplay complexity as the level progresses. Prior to play, participants in this condition were shown a PowerPoint tutorial regarding how to play the game. During play, a black mask was placed over the computer screen so that only the game window was visible within a 20 cm × 30 cm rectangle centered on the screen.

OPEN IN VIEWER

Figure 2.

Screenshot from Sushi Cat 2 (Armor Games, Irvine, CA).

The guided relaxation intervention was a 5.5-minute procedure developed by Souders et al. (2010) and was shown to be effective in reducing self-reported and physiological indicators of stress. The activity consisted of an audio recording of a woman’s voice that instructed participants through a deep breathing meditation and body awareness exercise by directing attention to individual body parts in a sequential fashion. Finally, the break condition consisted of a quiet period where participants were told to wait a moment before the next part of the study. They were not able to leave the lab or use their phone or the lab computer during that time. All interventions lasted for an equal amount of time (5.5 minutes).

Results

We quantified participants’ performance during the vigil as a manipulation check to ensure it produced the desired cognitive fatigue (Figure 3). Evidence for fatigue was defined in terms of a vigilance decrement, as measured by a decline in correct target detections over time. The 15-minute vigil was divided into six continuous 2.5-minute intervals or periods of watch. We then conducted a 3 (intervention) × 6 (periods of watch) mixed ANOVA on the target detection rate, correcting for violations of the assumption of sphericity using the Box correction. Data from Period 1 and Period 2 did not meet assumptions for normality. Because of this, we compared the results of the data analysis conducted on both arcsine transformed data and the untransformed values. The data transformation did not affect the interpretation of the results. Thus, we report only the analysis conducted on the raw scores. We found a significant effect of time on watch, F(4.3, 264.5) = 33.21, p < .001, η_P² = .35. Bonferroni corrected t tests were conducted to compare performance in the first block with each subsequent block to quantify the change in performance over time. Our post hoc comparisons revealed that performance significantly declined with period on watch. Correct detections for Periods 2 (p = .002), 3 (p < .001), 4 (p < .001), 5 (p < .001), and 6 (p < .001) were all significantly lower than Period 1. The average correct detection rate was 94.16%, 95% CI [0.92, 0.96] in Period 1 and declined to 76.28%, 95% CI [0.71, 0.81] by the end of the vigil. This decline is apparent in Figure 3, which also depicts data from the practice session (not analyzed) for comparison. We did not find any effects of intervention condition (p = .14) or a period by intervention interaction (p = .82), indicating that participants in all three conditions experienced similar levels of fatigue. Performance during the abbreviated vigil was similar to Temple et al. (2000), replicating the vigilance decrement.

OPEN IN VIEWER

Figure 3.

Percentages of correct detections during the vigilance task for participants assigned to the break, guided relaxation, and casual video game conditions plotted by periods of watch. Error bars indicate 95% confidence intervals. Results are listed for the practice period (Prac) as well as the six 2.5-minute watch periods.

For all subsequent analyses, participants’ scores on each component of the ACA were converted into standardized change z scores (Post-Pre/σ_Pre). The pre-vigil scores were used as the pre-condition for the post-vigil scores, and the post-vigil scores were used as the pre-condition for the post-intervention scores. To examine if each ACA component significantly changed from pre- to post-vigil (Figure 4) or from post-vigil to post-intervention (Figure 5) we examined 95% confidence intervals of the standardized change scores against a hypothesis of no change. When the error bars do not overlap zero, we consider this evidence of a significant non-zero change. One-way ANOVAs were used to investigate differences between each condition.

OPEN IN VIEWER

Figure 4.

Standardized change scores indicating change from baseline as measured after the vigilance task (pre-intervention) for each of the affective and cognitive assessment (ACA) components as measured by the shortened version of the Dundee Stress State Questionnaire, Positive and Negative Affect Schedule, and backward digit-span. Error bars represent 95% confidence intervals. An average, collapsed across all conditions, is included to aid interpretation of the results.

OPEN IN VIEWER

Figure 5.

Standardized change scores indicating change following the three intervention tasks (post-intervention) for each of the affective and cognitive assessment (ACA) components as measured by the shortened version of the Dundee Stress State Questionnaire, Positive and Negative Affect Schedule, and backward digit-span. Error bars represent 95% confidence intervals.

Because all participants were assigned at random to each condition and received the same vigilance task, changes pre- to post-vigil were examined by using the average of each condition instead of the score for each condition separately (included in Figure 4). Based on the average confidence intervals, there was a significant decline in positive affect (M = −1.32, 95%, CI [−1.57, 1.07]) and an increase in negative affect (M = 0.70, 95% CI [0.48, 0.92]) on the PANAS. For the DSSQ-S scale, there was a decrease in worry (M = −1.23, 95% CI [−1.54, −0.92]), an increase in distress (M = 2.20, 95% CI [1.73, 2.67]), and a small decline in engagement (M = −0.40, 95% CI [−0.71, −0.09]). These results are thus consistent with prior vigilance research with regard to declines in affect. However, scores on the backward digit-span did not decrease (M = −0.37, 95% CI [−0.64, 0.10]) post-vigil.

As revealed by one-way independent groups ANOVAs (Table 2), these effects did not vary as a function of the intervention conditions except in regard to the distress measure. In the case of that measure, participants in the guided relaxation condition experienced a larger increase in distress following the vigilance task than participants in the break and game conditions. Because all of the participants were assigned at random and experienced the same tasks at this point, we concluded that this was due to measurement error rather than a unique property of this condition. We note that the relaxation condition had a lower baseline score than the other conditions as well as a larger post-vigil score for distress, which when combined led to the significant change in distress found. Because we used standardized change scores, this finding does not affect the results of intra-condition differences, but by experiencing greater distress post-vigil, the relaxation condition may have been more likely to recover from their feelings of distress post-intervention and thus should be interpreted more conservatively.

OPEN IN VIEWER

Table 2:

Test Statistics for ANOVAs Conducted Both Following the Abbreviated Vigilance Task (Post-Vigil) and the Experimental Intervention of the Break, Relaxation, or Game Condition (Post-Intervention)

Source	Test	p	ηp2	Post Hoc t Tests
Change post-vigil
Positive affect	F(2, 62) = 1.40	.25	.04	NA
Negative affect	F(2, 62) = 2.51	.09	.08	NA
Engagement	F(2, 62) = 0.35	.71	.01	NA
Distress	F(2, 62) = 5.99	.004	.16	Relax > break (p = .009); relax > game (p = .015)
Worry	F(2, 62) = 0.067	.94	.002	NA
Backward digit-span	F(2, 62) = 0.93	.29	.04	NA
Change post-intervention
Positive affect	F(2, 62) = 23.12	<.001	.43	Game > break (p = .001); game > relax (p < .001)
Negative affect	F(2, 62) = 4.03	.02	.12	Game < break (p = .024)
Engagement	F(2, 62) = 28.74	<.001	.48	Game > break (p < .001); game > relax (p < .001)
Distress	F(2, 62) = 1.43	.25	.04	NA
Worry	F(2, 62) = 14.70	<.001	.32	Game < break (p < .001); relax < break (p = .005)
Backward digit-span	F(2, 62) = 1.30	.28	.04	NA

Note. Results show differences between each experimental condition. p values were tested against a Bonferroni adjusted alpha level. NA = not applicable.

Next, we analyzed the change scores as measured post-intervention to determine the degree to which each intervention restored each component of the ACA. Hours of casual video game experience was evaluated as a possible covariate but did not meet inclusion criteria due to not being correlated with the dependent variables (DVs) and was excluded from analysis. We conducted a one-way ANOVA on each ACA component using the standardized post-intervention change scores. We found significant effects of intervention for all DVs except distress and digit-span (Table 2). In these two instances, the differences between conditions were not significant.

Using the same approach we used to analyze change post-vigil using confidence intervals (in Figure 5), the video game condition was the only intervention to increase engagement, while the break condition was associated with a decrease in engagement. All three conditions led to a decline in distress. The game was also the only intervention to decrease worry, while the break condition was associated with an increase in worry. Further, the game was again the only intervention to increase positive affect. Both the game and the relaxation condition reduced negative affect. Finally, the relaxation condition was the only condition to show an increase in backward digit-span and indicate some level of cognitive restoration, while the break condition led to a decline in this measure.

Discussion

First, our results demonstrated that in addition to the well-known vigilance decrement, there are affective and cognitive consequences of a short vigilance task. We have also demonstrated that breaks from work can counteract these consequences, supporting prior work (Finkbeiner et al., 2016). Differing from this previous work, however, our results suggest that what one does to rest during this break matters. Tasks that continue to deplete cognitive resources will not be restorative (Helton & Russell, 2017). While each intervention led to some affective restoration, the CVG was the only condition associated with restoration on all five affective components. In summary, after playing the game, participants exhibited greater positive affect and task engagement as compared to other conditions. Negative affect and worry also decreased more following the game than after the break. On the other hand, in terms of feelings of distress and working memory span, restoration associated with the game was statistically equivalent to the other interventions. The increase in digit-span from the game was not significant, but it is worth noting that this finding is not consistent with prior work reporting declines in cognitive performance following game play (Kuschpel et al., 2015; Liu et al., 2016).

Individuals in the relaxation condition showed decreased negative affect and distress and improved digit-span performance when compared to this condition’s post-vigil ACA scores, but these improvements were not significantly better than any other conditions. This finding differed from prior work (Zeidan et al., 2010) that showed a mindfulness intervention both reduced anxiety and fatigue and also improved working memory and executive functioning. One reason may be that we used too short a duration, 5.5 minutes instead of the 20 minutes used by Zeidan and colleagues (2010). Moreover, our participants may not have been experienced enough. The long-term practice of relaxation techniques has been shown to improve health, enhance mood, and help individuals cope and self-regulate stress (e.g., Lykins & Baer, 2009), and so it may be the case that people would garner greater benefits from the relaxation exercise with more mindfulness experience.

In the passive break condition, we observed limited evidence of restoration. While participants in this condition reported reduced feelings of distress, they also reported decreased engagement and increased worry, findings that are counterproductive to restoration. This condition did not improve positive affect, and participants’ digit-span decreased from the post-vigil measurement, indicating that participants were still using cognitive resources preventing recovery, a finding consistent with ART (Kaplan, 1995). Although we do not know the content of our participants’ thoughts while sitting quietly, this finding may also be consistent with recent work showing that people find such tasks unpleasant (Wilson et al., 2014). It may also be the case that affective restoration requires some degree of engagement in a task (see also Finkbeiner et al., 2016). Our work highlights that a casual video game is at least one task that has this affective benefit, but there are surely many activities that may improve mood.

Although the game condition did not restore cognition, the improvements in affect may explain their popularity and still provide a workplace benefit (CGA, 2007; Reinecke, 2009a, 2009b). Disengaging from work has been cited as key to stress recovery (Sonnentag & Fritz, 2015). ART suggests that cognition can only recover when not in active use (Kaplan, 1995)—such as when one is not at work. This type of attention was originally proposed by James (1890) and elaborated by Kaplan (1995) to describe involuntary attention as a mode of functioning that is effortless, allowing for cognitive recovery. In contrast, Ulrich (1984) claimed it is not the separation from work itself but an effect that is mediated through engaging in relaxing activities. For example, staying active during a break (e.g., taking a walk, reading a book) has been found to increase post-break job performance and affect (Fritz, Ellis, Demsky, Lin, & Guros, 2013).

The degree of directed attention was not controlled across the three interventions, therefore it is not possible to assess the implications of our findings for Kaplan’s ART. In terms of Ulrich’s theory, while we did find a positive affective response to the video game, it was not associated with a corresponding improvement in cognition, at least as measured by the backward digit-span. However, a more comprehensive set of cognitive tests may need to be used in order to determine whether such findings are consistent with Ulrich’s theory. It may also be true, however, that affect and cognitive performance are not directly related. One recent study found active breaks created positive emotional reactions but did not improve task performance over and above passive breaks (Finkbeiner et al., 2016). These authors concluded that performance and well-being may be separate and independent capacities requiring true rest to replenish. This finding was supported by Helton and Russell (2017), who demonstrated breaks must fully disengage cognitive resources to be restorative. Our findings support this conclusion also; although video games were engaging and enjoyable, they did not improve cognition. Also, while guided relaxation did not increase positive affect, this is the only condition to show an increased digit-span following the vigilance task. However, more research is needed to investigate this potential dissociation.

Conclusions

We have discussed the critical issue of affective and cognitive fatigue at work, especially for those who work in critical jobs (e.g., security, medical, and military personnel). Our findings provide evidence that playing casual video games could be an effective and even fun way to recover from such fatigue, at least in the affective domain. However, it is important to make the distinction that decreasing stress, increasing mood, and having fun are not completely interchangeable. Though gaming may be fun and may reduce subjective reports of stress, there are many other activities that some classify as fun (e.g., exercise) that are also stressful. This is especially true for some video games. Players state that they are having fun even when playing video games that are stressful. We also note that games that are too difficult for players may also be frustrating (Przybylski, Deci, Rigby, & Ryan, 2014). Thus, these findings may not be generalizable to all games. Future research should be devoted to trying to tease out the specific characteristics a CVG must have in order to support affective restoration.

We generalize our findings to the context of work; however, the laboratory environment may not completely generalize to real work situations. Studies using a CVG intervention during shiftwork are needed to further support our findings.

Finally, because our guided relaxation condition outperformed the break or game conditions in terms of cognitive restoration, future research should focus on replicating and extending this finding in the workplace. Although games may make a worker feel better, how to make them perform better is still unknown. We also encourage future work leveraging the engagement of games to create more effective interventions to restore cognitive ability. The perfect workplace activity that is short and engaging and restoring affect and also cognitive ability has yet to be found.

Key Points