Media Multitasking Scores and Media Use Hours: A Comparison Across the Standard Stroop Task and an Emotional Stroop Task

ABSTRACT

Heavier media multitasking and media use have been linked to poorer inhibitory control. This study examined the association between media multitasking and media use on performance in the standard and Emotional Stroop tasks. Participants completed the Media Multitasking Index and the two Stroop tasks. Depression, anxiety, and stress were measured to control for the effect of mood on Stroop performance. Within the standard Stroop task, after controlling for affect, higher media multitasking scores and media use hours showed a trend toward shorter overall RTs. For the emotional Stroop task, after controlling for affect, more media use hours were associated with shorter overall RTs and higher media multitasking scores were associated with more overall errors. While this may suggest an association between greater media engagement and more efficient task performance (or a processing speed advantage), the effect sizes were small. These results emphasize the importance of controlling for participant affect to minimize a potential confound when examining the association between media multitasking and inhibitory control. The results also indicate that measuring both media multitasking and media use hours will assist in further clarifying the relationship between media multitasking and cognition.

Introduction

The increased access to portable internet capable devices has enabled us to spend more time using media. A recent study shows that 93% of 18 year-olds in the USA now own a smartphone and that 13–18 year-olds spend 8.39 h daily using screen media, which often includes using multiple screens or media simultaneously (Rideout et al., 2022). This multiple media use is known as media multitasking, and includes the simultaneous use of multiple media or the constant switching between using different media types (Minear et al., 2013; Ophir et al., 2009). In many cases, media multitasking (e.g., posting on social media while doing homework) draws significantly on our Executive Functions (EFs). EFs refer to effortful top-down cognitive processes (i.e., cognitive flexibility, working memory, and inhibition) that allow us to complete mentally complex tasks (Diamond, 2013; Friedman & Miyake, 2017; Miyake et al., 2000). During media multitasking, our cognitive flexibility (shifting) moves our attention between tasks or devices, and working memory is used to mentally store, update, and manipulate information in each task (Baddeley & Hitch, 1994; Friedman & Miyake, 2017; Miyake et al., 2000). Inhibition or inhibitory control is used to attend to relevant information while ignoring irrelevant information (filtering) and enables us to prevent a response (response inhibition) when required (Diamond, 2013; Miyake et al., 2000).

As media multitasking requires the continual monitoring of multiple information sources, this has led researchers to examine if greater media multitasking is linked to a breadth-bias in cognitive control (Ophir et al., 2009). This broad attention to multiple streams of information, known as the scattered attention hypothesis (Ophir et al., 2009; van der Schuur et al., 2015), suggests that undertaking more media multitasking might be associated with greater difficulty filtering out irrelevant information as this processing style treats all information as potentially relevant. In contrast, the trained attention hypothesis (Ophir et al., 2009; van der Schuur et al., 2015), suggests that more frequent media multitasking may result in superior ability to cope with distracting information (i.e., better filtering ability) due to the constant exposure to multiple streams of information, only some of which are relevant. This study sought to further investigate the relationship between media use and implications for performance on a filtering task.

Several studies have investigated the relationship between media multitasking and filtering. Ophir et al. (2009) used a change detection task that required participants to detect the red rectangle that changed orientation during the display while ignoring the irrelevant distractors (i.e., participants attended to all red rectangles displayed to detect the orientation change whilst ignoring the blue distractor rectangles). The target detection accuracy for heavy media multitaskers (HMMs) declined as the number of distractors increased. Light media multitaskers’ (LMMs) performance was unaffected by the number of distractors. Using this same task, Wiradhany and Nieuwenstein (2017) reported poorer overall performance for HMMs than LMMs (Experiment 1). AX-Continuous Performance tasks (AX-CPT) have also been used to examine filtering for LMMs and HMMs. In an AX-CPT, participants respond yes when presented with an A followed by an X and respond no to any other letter-probe pair. Studies using AX-CPTs have reported slower RTs for HMMs than LMMs when the distractors were present (Ophir et al., 2009; Wiradhany & Nieuwenstein, 2017 Experiment 2) and in another study HMMs performed worse than average media multitaskers (AMMs) and LMMs regardless of the presence or number of distractors (Cardoso-Leite et al., 2016).

Studies employing Flanker Tasks to examine filtering skills in media multitaskers have reported inconsistent results. In a typical Flanker task, participants see an array of five centrally located arrows and must indicate the pointing direction of the center arrow. The two outer pairs of distractor arrows either point the same direction as the central arrow (congruent condition) or the opposite direction to the central arrow (incongruent condition). Baumgartner et al. (2014) reported a marginally significant trend for HMMs to perform better than LMMs particularly when distractors were incongruent and a general tendency for HMMs to perform better overall in the task than LMMs. HMMs also showed shorter overall RTs than LMMs, indicating superior performance in a flanker task (Rogobete et al., 2020). Other studies have not found any link between media multitasking and flanker task performance (e.g., Minear et al., 2013; Murphy et al., 2017; Rogobete et al., 2020; Seddon et al., 2018). Thus, various paradigms have been used to assess the link between filtering and media multitasking, producing heterogeneity in the outcomes (for recent meta-analyzes or reviews see Parry & le Roux, 2021; Uncapher & Wagner, 2018; Wiradhany & Nieuwenstein, 2017).

The heterogeneity of the outcomes may be due to the change detection, AX-CPT, and Flanker tasks presenting target and distractors as separate stimuli, which may facilitate ignoring irrelevant information and therefore overestimate filtering task performance. Using a task that presents relevant and irrelevant information integrated within a single stimulus would present a greater challenge for participant’s filtering skills. Thus, some studies have used the Stroop task to examine the relationship between media multitasking and filtering.

In a Stroop task (Stroop, 1935) participants identify the ink color of a printed word while ignoring the word meaning. When the word meaning, and the ink color are congruent (“RED” in red), or the meaning is neutral (“XXX” in red) responses are expedited compared to when the ink color and word are incongruent (“GREEN” in red). This Stroop effect is proposed to result from the fast automatic operation of word reading competing with the slower process of color identification, resulting in interference in the incongruent condition but not in the neutral or congruent conditions (Diamond, 2013; C. M. MacLeod, 1991). Studies examining the link between media multitasking and Stroop task performance have produced mixed outcomes. For example, Swing (2012) reported that higher levels of media multitasking were associated with better Stroop performance (i.e., smaller Stroop effect). Other studies have reported no performance differences between HMMs and LMMs (Luo et al., 2022; Müller et al., 2021), or less frontal region activation (i.e., lower levels of neural activation in task relevant areas) during the Stroop task for those undertaking more media multitasking (Luo et al., 2021). Further, a study using the Spatial Stroop task (in which participants indicated the pointing direction of an arrow while ignoring its location) reported that higher levels of media multitasking were associated with a trend for lower accuracy in the incongruent condition (Murphy & Creux, 2021). Although Stroop tasks may be better suited than other tasks to examine filtering abilities, the mixed results obtained in these Stroop task studies suggest that additional research is required to determine the relationship between media multitasking and filtering abilities.

Studies have shown that an emotional context affects filtering capabilities (Dresler et al., 2009) in the spatial cueing and dot probe tasks (Bar-Haim et al., 2007; C. MacLeod et al., 1986). In the spatial cueing paradigm, a cue is presented at one of the two locations followed by a target appearing at one of these locations. For most trials, the cue correctly indicates the target location (valid cue condition) and for the other trials the cue incorrectly indicates the target location (invalid condition) (Bar-Haim et al., 2007). RTs are shorter to targets presented in the valid than invalid cue condition. When the cue is emotional in nature, the magnitude of the cue validity effect (i.e., RT difference to targets shown on invalid cue and valid cue trials) is larger for negative (e.g., threat-related words) than neutral or positive cue conditions (Bar-Haim et al., 2007; Fox et al., 2001). In a typical dot-probe task, a threat and a neutral stimulus are presented, followed by a dot at the location of one of the stimuli, and participants indicate the location of the dot. The results usually show shorter RTs for the dot replacing the threat compared to neutral stimulus (Bar-Haim et al., 2007; C. MacLeod et al., 1986). This indicates the participant’s attention has been preferentially directed (i.e., an attentional bias) to the location of the previously presented emotional stimulus.

This bias in processing emotional information has been explained in terms of greater attentional capture (Hodsoll et al., 2011) or higher levels of activation (McKenna & Sharma, 1995) for emotional compared to neutral stimuli. Shukla (2016) employed a dot-probe task to investigate attentional bias for emotional stimuli in relation to media multitasking. It was found that HMMs were more attuned toward positive stimuli and more avoidant of negative stimuli compared to LMMs. More recently, it has been shown that compared to LMMs, HMMs took a shorter amount of time to categorize positive words compared to negative words (Shukla, 2022). Given only a few studies have examined the link between media multitasking and filtering for emotional stimuli, this study further investigated this issue using a variation of the Stroop task – the emotional Stroop task (Watts et al., 1986; Williams et al., 1996). In the emotional Stroop task, participants generally take longer to respond to the ink color of emotional words (e.g., happy in red font) than neutral words (e.g., paper in red font) and this is known as the Emotional Stroop effect (Watts et al., 1986). Thus, this study sought to examine if the magnitude of the Emotional Stroop effect was associated with media multitasking.

It has been suggested that the current calculation for the most widely used media multitasking inventory (Ophir et al., 2009) might not be the most suitable measure as it is based on the proportion of media use time that a person spends using multiple media (Wiradhany & Nieuwenstein, 2017). Therefore, a person spending 50 h per week using media can have the same media multitasking index as someone who spends 5 h a week using media. Thus, the calculation fails to take into account the amount of time spent in media use (Uncapher & Wagner, 2018; Wiradhany & Nieuwenstein, 2017). This is an issue as time spent engaged with media is related to neurocognitive functioning and therefore measuring media use time may further our understanding of the relationship between media multitasking and cognitive processing (Uncapher & Wagner, 2018). Given what is known about transfer and practice effects of EFs (Diamond, 2013; Klingberg, 2010) it is possible that more media use time (and more media multitasking) may lead to more practised filtering skills (Elbe et al., 2019; Lui & Wong, 2012). Therefore, in this study both media use hours and media multitasking were measured to further elucidate the link with filtering abilities (i.e., standard and emotional Stroop Task performance).

Previous research has shown that negative state mood can directly cause poorer performance on inhibition tasks (e.g., De Houwer & Tibboel, 2010). Interestingly, more media use or media multitasking has also been associated with higher levels of negative moods (Becker et al., 2013; Lee et al., 2019; S. Li & Fan, 2022; Mark et al., 2014). As affect has not always been measured or controlled in studies examining the association between media multitasking and filtering it is possible that this may have contributed to the heterogeneity of the results reported in this area (i.e., negative mood may have been the main driver of the association or group differences between HMMs and LMMs, rather than media use or media multitasking itself in some of the previous studies). Therefore, it is vital to control for participant affect when examining filtering task performance to reduce measurement noise in media multitasking research (Uncapher & Wagner, 2018). Thus, this study measured and controlled for participants’ state depression, anxiety, and stress scores to isolate and examine the relationship between media multitasking, media use hours, and performance on the standard and emotional Stroop tasks. This allowed for examination of the link between media multitasking and media use and filtering for neutral and emotional stimuli, independent of participant’s current emotional state (i.e., after controlling for affect as a confound). Thus, if frequent media multitasking or media use is associated with poorer filtering capacity, it would be expected that higher media multitasking scores or weekly media use hours would be associated with larger Stroop interference effects. In contrast, if greater media use and more frequent media multitasking enhances filtering skills, then this should be reflected in smaller Stroop interference effects.

Method

Participants

One hundred and fifty one second-year university students (114 females, 37 males, aged 18 to 57 years, M = 23.56, SD = 7.84) from various degrees (e.g., psychology, exercise science, criminology, law, business) voluntarily completed the study after providing written informed consent. (The study was approved by Griffith University Human Research Ethics Committee reference number PSY/03/14/HREC). An a priori power analysis determined this sample size sufficient to detect a small effect (f2 of 0.12, significance level 5% and 80% power). Participants completed the study in groups of 10 to 25 during a computer lab session within their course. All participants had normal or corrected-to-normal vision and normal self-reported color vision.

Measures

Depression, anxiety, and stress scale-21

Participant’s state mood was measured using the Depression, Anxiety, and Stress Scale-21 (DASS-21) (Lovibond & Lovibond, 1995). The DASS-21 asks participants to rate their feelings (e.g., I found it hard to wind down) over the past week on a 4-point scale [0 = Did not apply to me at all to 3 = Applied to me very much, or most of the time]. Higher subscale scores indicate higher levels of depression, anxiety, and stress. The DASS-21 has good test-retest reliability (α = .71–.81 depending on subscale) (Brown et al., 1997), good convergent validity (Antony et al., 1998) and high divergent validity (Gloster et al., 2008). Internal consistency for the subscales is excellent (Depression α = .94–.96, Anxiety α = .87–.89, Stress (α = .91–.93) (Antony et al., 1998; Brown et al., 1997).

Media-multitasking inventory

The Media-Multitasking Inventory (MMI) (Ophir et al., 2009) was used to measure participants’ media multitasking per hour of media use. Participants indicated how many hours per week they spent using each of 11 primary media and their number of texts per day. They also rated the frequency of other media use at the same time [Never = 0, A little of the time = 0.33, Some of the time = 0.67, Most of the time = 1.0]. The MMI media included print media, watching television, computer-based video, music, non-musical audio, video/computer games, telephone/mobile phone voice calls, instant messaging, text messaging, e-mail, web surfing, and other computer applications (e.g., word processing). The Ophir et al. (2009) formula was used to compute the MMI score for each participant with text messaging only included as a secondary media use option as it was not measured in hours. Higher MMI scores indicate more media multitasking per hour of media usage. Weekly media use hours were calculated by summing the hours each media was used during the week.

Stroop Tasks

After completing the above measures, participants undertook the standard Stroop and emotional Stroop tasks, which were run on an Optiplex 755 PC (20 inch LCD monitor) using DMDX software (Forster & Forster, 2003). On each trial, a fixation cross appeared (500 ms) followed by a colored-letter string for 500 ms. Participants were instructed to respond to the ink color of the letter string as soon as possible, using the appropriate color-coded keyboard keys (red, yellow, blue, and green). There were 12 practice trials followed by 72 experimental trials. Half of the experimental trials were from the standard Stroop conditions (congruent – the meaning of the color word and color match; incongruent – the meaning of the color word and its color mismatch; neutral – row of XXXs in color; 12 trials per condition). The other half of the trials presented emotional or neutral words in color (negative valence – kill, war, fatal; positive valence – love, laugh, happy; neutral valence – clock, zone, notch; 12 trials per condition). Words for the emotional Stroop task were taken from a database noting the ratings of word valence and arousal (Bradley & Lang, 1999). These words were also rated in a pilot study, verifying that they belonged to either the negative, positive, and neutral word emotionality categories.

In both Stroop tasks, the dependent variables were response time (RT; ms) for correct responses and errors (%). To control for outliers, any participant’s RT more than 2.5 SDs away from their mean was replaced with a 2.5 SD RT value (Andrews & Murphy, 2006; Archibald et al., 2015; Murphy et al., 2017).

Results

Table 1 presents relevant descriptive statistics and correlations for the measures used within this study. Linear mixed-effect models were used to examine RTs (ms) and errors (%) for the standard and emotional Stroop tasks. Step one of the analysis ran a control model which included stress, anxiety, and depression as predictors, and participants as a random effect. The control model was compared to a null model. In step two, the following factors were included; the three conditions for each task (i.e., standard Stroop – congruent, incongruent, and neutral conditions; emotional Stroop positive, negative, and neutral conditions), MMI score or Weekly Media Use Hours, and the Stroop condition × MMI score (or Weekly Media Use Hours) interaction. This model was compared to the control model to determine if after isolating and controlling for affect, the main effects and interaction provided additional explanatory power in Stroop task performance. Models were run using lmerTest package (Kuznetsova et al., 2017) in R Studio (R Core Team, 2022), and were compared using the anova function with χ² tests. Pairwise comparisons for significant main effects used the emmeans and pairs functions from the emmeans package (Lenth, 2022). Conditional R² values for the whole model were calculated using the r2glmm package (Jaeger, 2017) using the method proposed by Nakagawa and Schielzeth (2013). Due to the way that variance is partitioned in mixed effects models, and there being no standard reporting of effect sizes for individual model terms and interactions (see Rights & Sterba, 2018 for a discussion of these issues) both unstandardized Beta values (B) (to indicate direction and overall of effects in the original measurement scales) and standardized β (with interpretation based on Cohen’s proposed cutoffs of 0.10–0.29 for small, 0.30–0.49 for medium, and 0.50 or greater as large effects; Cohen, 1988) are presented. To quantify the effects of Stroop task condition on the dependent variables, marginal means with 95% confidence intervals are reported.

Table 1. Descriptive statistics and Pearson correlations for the IVs (DASS, MMI, weekly media use hours) and the DVs (Stroop task measures).

	Stress	Anxiety	Depression	MMI Score	Media Use Hours	Mean	SD
Stress	1					13.66	9.58
Anxiety	.714**	1				6.56	8.31
Depression	.659**	.594**	1			7.71	8.79
MMI score	.042	.141*	.024	1		3.84	1.53
Media Use Hours	−.008	.169*	.049	.355**	1	73.31	41.78
Standard Stroop task
Congruent RT (ms)	.101	−.011	−.108	−.152*	−.214**	593	118
Incongruent RT (ms)	.109	.022	−.095	−.088	−.126^#	653	145
Neutral RT (ms)	.121	.039	−.080	−.128*	−.158*	604	107
Congruent Errors (%)	.008	.093	.095	.088	.146*	5.04	7.06
Incongruent Errors (%)	.007	.001	.065	.046	.125^#	8.35	12.17
Neutral Errors (%)	.065	.168*	.244**	.004	.090	6.46	9.22
Emotional Stroop task
Positive RT (ms)	.037	−.070	−.145*	−.108	−.215**	596	110
Negative RT (ms)	.134*	.049	−.064	−.093	−.194*	607	111
Neutral word RT (ms)	.146*	.046	−.081	−.122*	−.195*	609	111
Positive Errors (%)	.048	.093	.190**	.164*	.123	5.34	7.69
Negative Errors (%)	.086	.156*	.151*	.009	−.083	6.04	7.94
Neutral word Errors (%)	.022	.090	.141*	.203*	.062	6.37	8.52

#p < .06; *p < .05; **p < .01: Sub-scale scores for Depression, Anxiety and Stress for DASS-21; MMI score – Multimedia Multitasking Index score; Media Hours – hours per week media use.

Standard Stroop Task Results for MMI Score and Hours Media Use

RT data

Table 2 presents the results of the analysis including MMI score and Media Use Hours for the RT data for the standard Stroop task. The control variables (stress, anxiety, and depression) significantly predicted RTs in the standard Stroop task compared to a null model (χ² (3) = 8.50, p = .037). Shorter RTs were associated with lower stress scores (medium effect size) and higher depression scores (small effect size). The models including MMI Score (χ² (5) = 107.66, p <. 001) and Media Use Hours (χ² (5) = 109.20, p < .001) provided a significantly improved model fit over the control model, but there was no significant difference between these models in explanatory power (p ~1).

Table 2. Standard Stroop task RT and error model outputs from the mixed-effect models.

		RT						Errors
Model	R^2	Beta	β	F	DF	p	R^2	Beta	β	F	DF	p
Control Model	0.79						0.31
Stress		3.99	0.30	6.33	1, 146	.013*		−0.23	−0.24	6.76	1, 146	.010*
Anxiety		−1.07	−0.07	0.39	1, 146	.536		0.13	0.12	1.83	1, 146	.179
Depression		−3.45	−0.25	5.71	1, 146	.018*		0.25	0.24	9.19	1, 146	.003**
MMI Score × Stroop Condition	0.85						0.33
Stress		3.63	0.28	5.40	1, 145	.022*		−0.23	−0.24	6.59	1, 145	.011*
Anxiety		−0.39	−0.03	0.05	1, 145	.825		0.13	0.11	1.68	1, 145	.197
Depression		−3.55	−0.25	6.12	1, 145	.015*		0.25	0.24	9.13	1, 145	.003**
MMI Score		−13.28	−0.16	3.37	1, 145	.068		0.33	0.06	0.01	1, 145	.923
Stroop Condition		-	-	60.89	2, 296	.001***		-	-	5.20	2, 296	.006**
MMI Score × Stroop Condition		2.06	0.03	0.59	2, 296	.557		−0.51	−0.11	0.44	2, 296	.646
Media Use Hours × Stroop Condition	0.85						0.33
Stress		3.43	0.26	4.73	1, 145	.031*		−0.20	−0.21	5.13	1, 145	.025*
Anxiety		−0.30	−0.02	0.03	1, 145	.862		0.09	0.08	0.81	1, 145	.368
Depression		−3.41	−0.24	5.66	1, 145	.019*		0.24	0.24	9.15	1, 145	.003**
Media Use Hours		−0.54	−0.18	3.46	1, 145	.065		0.02	0.09	3.25	1, 145	.074
Stroop condition		-	-	61.18	2, 296	.001***		-	-	5.20	2, 296	.006**
Media Use Hours × Stroop Condition		0.18	0.06	1.30	2, 296	.273		0.02	0.08	0.60	2, 296	.550

*p < .05, **p < .01, ***p < .001. Beta and β included for continuous variables only and represent Congruent vs Incongruent comparisons in the interaction term.

Stroop condition was a significant predictor of RT, with significantly shorter RTs for congruent trials ($\bar{X}$ = 591 ms, 95%CI = 571–610 ms) than incongruent trials ($\bar{X}$ = 650 ms, 95%CI = 631–669 ms; t(298) = 10.52, p < .001), and slightly shorter RTs than for the neutral trials ($\bar{X}$ = 604 ms, 95%CI = 585–623 ms; t(298) = 2.13, p = .055). RTs for neutral trials were shorter than for the incongruent trials (t(298) = 8.19, p < .001).

There were also marginally significant trends for higher MMI scores (p = .068) and more Media Use Hours (p = .065) to be associated with shorter overall RTs; however, effect sizes were small (βs = −0.16 and −0.18 respectively). In both models, the interaction terms were not significant.

Error data

The results of the analysis for the standard Stroop task error data are shown in Table 2. The control model significantly predicted errors in the standard Stroop task compared to a null model (χ² (3) = 13.26, p = .004). Lower stress scores were associated with more errors (small effect), while higher depression scores were associated with more errors (small effect). The MMI Score (χ² (5) = 13.26, p = .047) and Media Use Hours (χ² (5) = 13.26, p = .011) models were significantly better than the control model and were not significantly different from each other (p ~1).

Stroop Condition was significantly associated with errors. Fewer errors were made in the Congruent ($\bar{X}$ = 5.07%, 95%CI = 3.67–6.47) than the Incongruent condition ($\bar{X}$ = 7.83%, 95%CI = 6.43–9.24; t(296) = 3.22, p = .004). There was no difference in errors between Congruent and Neutral conditions ($\bar{X}$ = 6.50%, 95%CI = 5.10–7.90; t(296) = 1.67, p = .221) or between Neutral and Incongruent conditions (t(296) = 1.56, p = .264). Higher media use hours were marginally associated with more errors (p = .074, β = 0.09, small effect size). MMI Score was not a significant predictor and neither interaction term was significant.

Emotional Stroop Task Results for MMI Score and Hours Media Use

RT data

Table 3 shows the results of the RT data analysis for the emotional Stroop task. The control model was a significantly better fit of RTs in the emotional Stroop task conditions than the null model (χ² (3) = 9.13, p = .028). Lower stress scores (medium effect size) and higher depression scores (small effect size) predicted shorter RTs.

Table 3. Emotional Stroop task RT and error model outputs from the mixed-effect models.

				RT					Errors
Model	R^2	Beta	β	F	DF	p	R^2	Beta	β	F	DF	p
Control Model	0.88						0.25
Stress		3.83	0.33	7.08	1, 146	.009**		−0.14	−0.17	3.46	1, 146	.065
Anxiety		−1.24	−0.09	0.62	1, 146	.431		0.12	0.12	1.89	1, 146	.171
Depression		−3.21	−0.26	5.85	1, 146	.017*		0.19	0.20	6.85	1, 146	.010*
MMI Score × Emotional Stroop Condition	0.89						0.27
Stress		3.60	0.31	6.24	1, 145	.014*		−0.13	−0.15	2.74	1, 145	.100
Anxiety		−0.74	−0.05	0.21	1, 145	.645		0.08	0.08	0.87	1, 145	.352
Depression		−3.28	−0.26	6.16	1, 145	.014*		0.19	0.21	7.38	1, 145	.007**
MMI Score		−8.36	−0.12	2.17	1, 145	.143		0.74	0.14	3.97	1, 145	.048*
Emotional Stroop Condition		-	-	5.44	2, 296	.005**		-	-	0.93	2, 296	.395
MMI Score × Emotional Stroop Condition		−1.00	−0.02	0.27	2, 296	.761		−0.78	−0.19	2.42	2, 296	.091
Media Use Hours × Emotional Stroop Condition	0.89						0.26
Stress		3.25	0.28	5.09	1, 145	.026*		−0.14	−0.17	3.29	1, 145	.072
Anxiety		−0.38	−0.03	0.06	1, 145	.813		0.11	0.12	1.73	1, 145	.190
Depression		−3.16	−0.25	5.84	1, 145	.017*		0.19	0.20	6.80	1, 145	.010*
Media Use Hours		−0.51	−0.19	5.37	1, 145	.022*		0.02	0.09	0.01	1, 145	.939
Emotional Stroop condition		-	-	5.43	2, 296	.005**		-	-	0.93	2, 296	.396
Media Use Hours × Emotional Stroop Condition		0.05	0.02	0.13	2, 296	.880		−0.04	−0.20	2.11	2, 296	.123

*p < .05, **p < .01, ***p < .001. Beta and β included for continuous variables only, and represent Positive vs Neutral comparisons in the interaction term as the largest effect.

The full models for MMI Score (χ² (5) = 13.59, p = .018) and Media Use Hours (χ² (5) = 16.52, p = .006) were both significantly better than the control model, but not different from each other (p ~1). Emotional Stroop condition was significantly associated with RT (p = .005). RTs were significantly longer in the Negative ($\bar{X}$ = 605 ms, 95%CI = 588–623 ms) than the Positive condition ($\bar{X}$ = 595 ms, 95%CI = 578–612 ms; t(296) = 2.52, p = .033) and RTs in the Neutral condition ($\bar{X}$ = 608 ms, 95%CI = 591–612 ms) were longer than in the Positive condition (t(296) = 3.10, p = .006). There was no difference in RTs between the Neutral and Negative conditions (p = .830). Higher Media Use Hours was associated with shorter RTs, showing a small effect size (β = −0.19). MMI score was not significant and neither interaction term was significant.

Error data

The analysis for errors in the emotional Stroop task is shown in Table 3. The control model provided a significantly better fit of errors for the emotional Stroop task conditions compared to the null model (χ² (3) = 11.60, p = .009). Higher depression scores (small effect) were associated with more errors. Lower stress scores were not significantly associated but approached significance with a small effect size (p = .065, β =-0.17).

The model containing MMI Score approached significance (χ² (5) = 10.762, p = .056). The model containing Media Use Hours was not significantly better than the control (χ² (5) = 6.11, p = .296). Emotional Stroop condition did not predict errors in this task (i.e., % errors were similar for all conditions). Higher MMI scores were associated with more errors in the task (small effect size). Media Use Hours were not associated with task errors. The interaction term for Condition and Media Use Hours or MMI score was not significant.

Discussion

This study examined the relationships between media multitasking, media use habits and inhibitory control (i.e., the filtering of irrelevant information) using the standard color-word Stroop and emotional Stroop tasks after controlling for participant’s self-reported stress, anxiety, and depression. In both Stroop tasks, lower stress and higher depression scores were associated with shorter RTs and higher errors were linked with lower stress scores (approached significance in emotional Stroop) and higher depression scores. Anxiety was not significant for RT or errors for either task. The stress and depression results appear to indicate a speed-accuracy trade-off in performance for both Stroop tasks (i.e., lower stress scores shorter RTs and higher errors, higher depression scores shorter RTs and more errors). This could be taken as evidence of overall poorer task performance depending on participant mood state which is consistent with previous research examining the link between cognitive inhibition task performance stress and depression (Dell’acqua et al., 2022; Li et al., 2021; Sänger et al., 2014; Shields et al., 2016; Vinski & Watter, 2013). These results illustrate the importance of measuring and statistically controlling for state affect to ensure performance on filtering tasks is not affected by mood state and thus any relationships or group differences can be linked to media multitasking or media use alone. As affect was not measured or controlled for in many previous media multitasking studies, this may explain the heterogeneous results in this area of research. For example, it is possible that where group differences between HMMs and LMMs (e.g., HMMs better or worse performance) were evident for filtering tasks, this was an artifact of group differences in stress or depression rather than superior or poorer cognitive skills. Thus, it is imperative that media use and media multitasking studies measure and control for participant affect to ensure these studies specifically examine the link between media usage and filtering task performance.

For the standard Stroop task, there were longer RTs and more errors in the incongruent than congruent condition, replicating the typical task results reported in the literature (Logan et al., 1984; C. M. MacLeod, 1991). For the emotional Stroop task, RTs were shorter in the positive compared to neutral or negative conditions and there was no effect of emotional valence for errors. These results are consistent with a prior study that also used button-press responses in an emotional Stroop task (McKenna & Sharma, 1995; Sharma & McKenna, 2001). This suggests that there is less interference from the processing of the word meaning to the processing of the ink color when the emotional content is positive compared to when it is of neutral or negative valence.

In terms of media multitasking behavior, higher MMI scores showed a trend toward shorter RTs overall for the standard Stroop task, however the effect size was small. The magnitude of the Stroop effect was not related to MMI score. MMI scores were not associated with errors in any of the conditions in this task. The analysis for Media Use Hours showed trends toward significance such that shorter overall RTs and marginally more errors were associated with higher media use hours, but these effects were both small. As the magnitude of the Stroop effect was not associated with MMI score or Media Use Hours, these results support prior studies that have shown no Stroop task performance differences between HMMs and LMMs (Luo et al., 2022; Müller et al., 2021). These results are also consistent with recent meta-analyzes which demonstrated an overall small effect size and non-significant associations between filtering task performance (interference or distractor management) and media multitasking (Parry & le Roux, 2021; Wiradhany & Nieuwenstein, 2017). However, the relationship between filtering performance and media multitasking may be task dependent. For example, Uncapher and Wagner (2018) note that media multitasking was not related to performance on the change detection paradigm, however in tasks such as the AX-CPT, HMMs performed more poorly than LMMs. Given this variability across paradigms and as only a few studies have examined filtering performance and media multitasking using the Stroop task, replication of these results is required.

The results for the standard Stroop task do not support the assumption that more frequent media multitasking, and more hours of media use are linked to breadth-bias in cognitive control and scattered attention toward multiple streams of information (i.e., scattered attention hypothesis) (Ophir et al., 2009; van der Schuur et al., 2015). Nor do the results support the trained attention hypothesis (Ophir et al., 2009; van der Schuur et al., 2015), which proposes that more frequent media multitasking may result in superior ability to cope with distracting information (i.e., better filtering ability). As higher MMI scores and more weekly hours of media use showed a trend toward shorter overall RTs in the standard Stroop task, this may suggest that greater media engagement is linked with more efficient processing (Murphy et al., 2020) or is indicative of a processing speed advantage. While this is consistent with some previous media multitasking studies (e.g., Müller et al., 2021; Murphy & Creux, 2021; Rogobete et al., 2020), given the small effect sizes, these results should be interpreted with caution and further research conducted to replicate these outcomes.

MMI scores were not related to RTs overall or in any valence condition in the emotional Stroop task. Higher MMI scores were associated with more errors overall, although the effect size was small. These results are not consistent with the dot-probe study (Shukla, 2016) which noted that HMMs were more attuned toward positive stimuli and more avoidant of negative stimuli compared LMMs. Further research comparing performance of media multitaskers on emotional Stroop and dot-probe tasks would determine if these emotionality effects were task specific (i.e., only evident in the dot-probe task).

The analysis for Media Use Hours, revealed a different pattern of results for the emotional Stroop task compared to those evident for the MMI scores. More media use hours were associated with shorter RTs overall (although with a small effect size) and media use hours were not associated with errors. The difference in results for the MMI scores and Media Use Hours may indicate a distinction in the processing of emotional stimuli between media multitasking and general media use. That is, while MMI score is not related to the speed of processing of emotional information, those who frequently engage in general media consumption responded to emotional and neutral stimuli more efficiently. Given that prior studies have shown a link between high levels of media use (e.g. internet addiction, social media use) and alexithymia (loss of sensitivity to emotions) (Dalbudak et al., 2013; Mersin et al., 2019), it is possible that these results reflect the relationships between high levels of media use and a lack of sensitivity to emotional information which enabled overall greater processing efficiency in the emotional Stroop task. However, it has been shown that persons with alexithymia had longer RTs to all emotional words relative to those without alexithymia in an emotional Stroop task (Pandey, 1995). Thus, the emotional Stroop task results may demonstrate a processing advantage for emotional stimuli in high media users and not a lack of the processing on emotional content.

The different pattern of outcomes for the emotional Stroop task for the MMI and media use hours analysis provide further support for the suggestion of Uncapher and Wagner (2018) that measuring media use time will further our understanding of the relationship between media engagement and differences in cognitive functioning. Thus, future research should consider assessing both media engagement metrics in relation to performance on filtering tasks.

Larger Stroop effects have been reported for verbal color-naming compared to button-press response tasks (Klein, 1964; Majeres, 1974). Therefore, replication of this study using verbal responses could determine if the link between media multitasking, media use hours, and Stroop performance depends on task response modality. Moreover, questions have been raised about the utility of an emotional Stroop task as a measure of the Stroop effect within an emotional context (Algom et al., 2004). Therefore, future research could consider using alternatives such as the Posner or Flanker tasks to assess filtering in relation to emotional stimuli (Gibb et al., 2016), to gain greater clarity on the association between media engagement and the processing of emotional content.

While the MMI score (Ophir et al., 2009) is a standard measure used within the literature, its reliance on self-report means it is open to individual interpretation (e.g., the definition of what constitutes some of the time might vary across participants), which may affect the reliability of the measure. Using additional measures such as tracking apps on smartphones would provide an objective assessment of media multitasking and hours of media use and should be considered as an option in future studies.

A few limitations are evident in this study. One is that the cross-sectional design of the research does not inform about the casual link between media multitasking, general media use and filtering out irrelevant information. Another issue is sourcing participants from the University student population. To progress this area of research future studies could examine longitudinal media use and media multitasking behavior (e.g., Luo et al., 2022) within a community sample of various age ranges.

As this study used the original MMI (Ophir et al., 2009) to measure media multitasking, it may be argued that it does not adequately capture current media use. However, most media from the MMI (Ophir et al., 2009) are still relevant today (e.g., watching TV/video, playing games, reading print, listening to music). One notable exception is the absence of social media, however the MMI did include instant messaging which does involve social media use. Therefore, it would appear that the media categories of the MMI (Ophir et al., 2009) are still sufficient to cover the major aspects of today’s media engagement. This is further supported by the fact that the mean MMI score for this study (3.84) is only marginally lower than the mean MMI scores in studies that have used modified versions of the MMI which specifically include more recent media (e.g., 4.2 Murphy & Creux, 2021; 4.26; Shin & Kemps, 2020). This suggests the use of the original MMI (Ophir et al., 2009) to measure media multitasking did not affect the measurement of this construct within the study.

This study showed that state self-reported stress and depression are associated with performance in Stroop tasks, indicating the importance of measuring and controlling for affect in studies examining the relationship between media engagement and executive functions. The results for the standard Stroop task, showed that after controlling for participant affect there was a marginally significant trend toward more frequent media multitasking and media use being associated with shorter RTs. In the emotional Stroop task, frequent media multitasking was associated with poorer performance (more errors), while higher media use hours were associated with more efficient performance (shorter RTs overall). While there was no significant relationship between media engagement and filtering skills, the Stroop task results suggest that more frequent media multitasking and greater general media use are linked with better overall processing efficiency. However, given the small effect sizes, these results should be interpreted cautiously and further research conducted to provide replication. Further, it is recommended that future research assess and control for participant affect to clearly understand the nuanced relationship between media usage and the executive function of filtering.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

The data from this study cannot be made publicly available as participants did not consent to this option at the time of completing the study. Therefore, the authors do not have the participants’ consent to publish the data.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.