Supplemental Materials 8
Supplemental Methods
Procedure. Participants were screened for color-blindness (PseudoIsochromatic Plate Ishihara Compatible Color Vision Test). All participants reported to have full vision and hearing. Participants performed all tasks and questionnaires in a testing room with minimal environmental distractions: they were instructed to power down (not simply silence) cellphones and store all potential distractions out of reach and eyesight. Whether questionnaires or cognitive tasks were performed first was counterbalanced across participants. Cognitive tasks and questionnaires were performed on a 15” Macbook Pro laptop. Tasks were presented and responses collected using MATLAB (R2011b) and PsychToolbox (www.psychtoolbox.org) software. The SurveyMonkey online platform was used to present and collect responses for questionnaires.
Working memory task: rectangles. On half the trials, one of the target rectangles rotated 45°, either clockwise or counterclockwise (with equal probability), between encoding and test. Distractor rectangles never changed between encoding and test. Participants performed the task in two runs, each consisting of 100 trials with equal numbers of each distractor load condition (0, 2, 4, or 6 distractors). Thus there were a total of 50 trials per distractor load.
Working memory task: objects. Stimuli were comprised of line drawings derived from common objects contained in the International Picture Naming Project (Szekely et al., 2004), supplemented by a set used by Uncapher and Rugg (2009). Objects were nameable with non-overlapping identities. Images were transformed into line drawings and were colored red or blue using Adobe Photoshop CS4. Stimuli were counterbalanced across target and distractor conditions, such that each stimulus was equally likely to be viewed as a target (in red) or a distractor (in blue) across participants. To do this, we created 10 lists of 50 objects each (randomly selected from the object pool described above), and rotated each list across all conditions (including the memory test conditions, see below), across participants. Thus, each object appeared with equiprobability in each of the 4 target conditions, 3 distractor conditions, or 2 new item conditions.
Timing in this task was determined by piloting 20 independent participants for ability to identify the target objects during the encoding period (as indexed by object naming). Additionally, test array duration was reduced such that incidental encoding of the distractor objects was minimized during the test array. When a target object rotated, it did so only by 10°, in order to encourage participants to pay greater attention to the objects. Again, distractors did not change between encoding and test arrays. Participants performed one run of this task containing 200 trials, with 50 trials of each distractor load condition.
Recognition memory task: target objects. To mitigate fatigue, only half (100) the target objects were presented in the memory test, interspersed with 50 new (unstudied) objects. Of the 100 studied objects, equal proportions were from each WM distractor load condition (0, 2, 4, or 6 distractors; 25 items from each condition). All items (old and new) were presented in red, in the center of the screen, for 2000 ms each, with a 2000-ms fixation cross between trials.
Recognition memory task: distractor objects. A similar protocol was used to determine recognition memory performance for objects appearing as distractors in the WM objects task. Parameters that differed were as follows: (1) all items were displayed in blue, and (2) because no distractor objects appeared on ‘0 distractor’ WM trials, only 75 old objects were included in the recognition memory test (25 from each distractor load condition).
Questionnaires. All questionnaires were presented on the same laptop used to perform the cognitive tasks, and were self-paced. Questionnaires were completed in the following order: MMI, ADHD, BIS-11, videogame usage.
MMI status was calculated for each participant according to Ophir et al. (2009). Thresholds for HMM and LMM classification were defined as the top (HMMs) and bottom (LMMs) quartiles of all 139 participants. ADHD tendency was calculated according to Part A of the ADHD-ASRS v1.1 (Kessler et al., 2005). Impulsivity was calculated according to the BIS-11 (Patton et al., 1995). Finally, participants indicated their video game usage during the past year; the threshold for “action video game player” was >5 hrs/wk of action video game play at some point within the last 12 months (Green & Bavelier, 2007). The action video game questionnaire was included based on findings showing that action video game players display a better ability to resolve visually presented targets from distractors (Green & Bavelier, 2007). This finding contrasts with the initial observation that HMMs demonstrate diminished performance in the face of distraction (Ophir et al., 2009). Those studies motivate the question of whether there is the relationship between heavy media multitasking and video game playing.
Supplemental Data Analysis Information
Statistical modeling. R 3.1.0 (R Foundation for Statistical Computing) and RStudio (RStudio, Inc) were used for behavioral analysis and graphing. Mixed-effects multiple regressions were run using the lme4 (Bates, Maechler, Bolker, & Walker, 2014: http://CRAN.R-project.org/package=lme4) and lmerTest (Kuznetsova, Brockhoff, & Christensen, 2014: http://CRAN.R-project.org/package=lmerTest) packages. Graphs were rendered using the ggplot2 package (Wickham, 2009: http://had.co.nz/ggplot2/book).
Mixed-effects models for working and recognition memory (d’) were estimated to examine fixed effects of group and distractor presence and random intercept effects of subject. The group factor was contrast-coded to test whether d’ for HMMs and LMMs separately differed from the grand mean, and the factor of distractor presence was also contrast-coded to test whether d’ for distractor-present target trials and for distractor-absent target trials differed from the grand mean. All p-values were estimated using a two-tailed threshold. All significant effects are reported.
Signal detection analyses. Outliers in each task were detected using Grubbs’ test for outliers (Grubbs, 1950), and a standard correction was implemented across all participants in both WM and LTM tasks to account for participants with 100% hit rates or 0% false alarm rates (Snodgrass & Corwin, 1988). The Grubbs’ test identified four participants (two LMMs, two HMMs) as outliers based on their performance (d’) in the WM rectangles task, and one (LMM) on their performance (d’) in the LTM targets task. No outliers were identified in the WM lines or LTM distractors tasks.
Supplemental Results
MMI. MMI scores followed a normal distribution (Fig S1), fitting the 68-95-99.7 rule, with 71.9% of participants falling within 1 SD, 95.0% within 2 SDs, and 99.3% falling within 3 SDs. The HMM/LMM thresholds in the present study were 5.60 and 2.96, respectively, which are similar to the 5.90 and 2.86 thresholds in Ophir et al. (2009).
Impulsivity index. Across-participant BIS-11 score subscale means were: attention, 16.64 (±3.94); motor, 21.36 (±4.38); and nonplanning, 23.37 (±4.90). Group-wise subscale means were as follows: HMMAttention=17.59 (±4.46), LMMAttention=15.71 (±3.63); HMMMotor=21.59 (±4.41), LMMMotor=20.91 (±4.73); HMMNonplanning=23.62 (±4.89), LMMNonplanning=23.23 (±5.01).
ADHD index. A total of 38 participants scored 4 or higher on subscale A of the ASRS v1.1—typically considered consistent with a clinical ADHD diagnosis (Kessler et al., 2005). Of these, 15 were HMMs and 7 were LMMs. Removing these participants from the analyses did not qualitatively change the pattern of results; thus, reported analyses include all participants.
Relationship between indices. Given the correlated nature of the MMI, BIS, and ADHD indices, it is important to identify whether there exists unique variance that is explained by MMI above and beyond that explained by impulsivity and ADHD indices. To do this, we implemented a multiple regression analysis and found that MMI score predicts a significant amount of unique variance in working memory performance for both tasks (K ~ MMI * BIS * ADHD): rectangles, effect of MMI, t = -3.23, p = .0016; objects, effect of MMI, t = -2.00, p = .048. This finding indicates that our media multitasking index captures information about a multidimensional construct that is not encapsulated by impulsivity and ADHD indices, and is thus useful to characterize beyond the other two measures.
Action video game playing index. Only nine participants in total qualified as action videogame players, with two also classified as HMM and none as LMM. These data suggest that media multitasking and action video game playing are independent behavioral patterns. As such, the present findings reflect differences related to media multitasking behavior, rather than to action video game playing.
Long-term memory: target objects. We examined the relationship between WM and LTM measures of discriminability and bias by assessing whether d’ and C in WM correlated with the same measures in LTM. LTM discrimination correlated with discrimination on both WM tasks (d’WM with d’LTM, rectangles: r132=.30, p=5.0*10-4; objects: r136=.29, p=5.4*10-4) but not with WM bias (CWM with d’LTM, rectangles: r132=.11, p=.20; objects: r136=.066, p=.44). Moreover, WM bias also did not predict LTM bias (CWM with CLTM, rectangles: r132=.091, p=.29; objects: r136=.10, p=.23).
Given the relationship between overall WM and LTM performance, and that the object WM task served as the study phase for the recognition memory task, we hypothesized that LTM performance for a target object would depend on whether the participant performed the WM trial correctly (i.e., correctly detected a change at study or correctly determined no change occurred). An ANOVA of d’LTM conditionalized on Study Accuracy, Study Category (change/no change), and study Distractor Load revealed that—for both HMMs and LMMs—Study Accuracy predicted LTM performance, in that d’LTM was significantly lower when participants failed on the WM task (main effect of Study Accuracy: F1,1589=37.9, p=9.5*10-10). Furthermore, HMMs showed worse later memory than LMMs regardless of Study Accuracy (main effect of Group: F1,1589=23.4, p=1.5*10-6). Whether the trial had been a change trial or a no change trial had no effect on later LTM performance (main effect of Study Category: F1,1589<1), nor did the number of distractors present during the study task (F1,1589<1).
Response times—all tasks. A group difference in response times was observed only in the LTM task for target objects. In this task, HMMs were not only less accurate (Fig 3A) but also slower to respond than LMMs [ANOVA on RT by Group, Distractor Load, Response Accuracy, and Study Accuracy: F1,945=5.56, p=.019]. The pattern of findings—that d’LTM was lower for HMMs relative to LMMs—held when trials were subsampled to equate for RT across groups [F=7.99, p = .0048; effect of subsampling: F < 1].
Differences from Ophir et al. (2009). The WM rectangles task was designed to be similar to the task reported by Ophir et al. (2009), although in that prior investigation we identified group differences in WM performance only at high distractor load (6 distractors). By contrast, here the number of distractors did not influence group differences in K (group by distractor interaction: F<1). This discrepancy may be due to the present design incorporating four times as many trials than our prior design, and thus here we had more power to detect effects at all load levels. The group by distractor interaction also was not significant in the WM objects task (which also contained high trial numbers/condition).
Supplemental Discussion
Impulsivity. The findings further revealed that attentional impulsivity positively related to the degree to which participants multitasked with media. This relationship between impulsivity and MMI has been reported previously, with one group showing a similar correlation across participants on the attention subscale that extended to the motor subscale (Sanbonmatsu et al., 2013). Another study reported group-wise effects (HMMs vs. LMMs) on the motor subscale only (Minear et al., 2013).
Given prior evidence that HMMs self-report as more impulsive, we originally predicted that signal detection decomposition of WM performance would reveal that their performance deficits stem not only from reduced discrimination but also a more liberal response bias (making decisions using less evidence). This, however, turned out not to be the case. Both groups showed relatively little bias in the WM rectangles task, and were both slightly conservative in the WM objects task and LTM task (while there was a difference in the LTM task across all trials, importantly, the groups were equally conservative when investigating their most confident responses). Moreover, instead of predicting bias, attentional impulsivity negatively predicted discriminability on the WM tasks. This finding is consistent with a recent investigation of impulsivity in university students not assessed for their media multitasking habits (Smillie, Dalgleish, & Jackson, 2007). In that study, impulsivity score also negatively predicted discriminability and not bias, using a different cognitive task (a category-learning task) and a different impulsivity inventory (the I7 Impulsiveness Questionnaire; Eysenck, Pearson, Easting, & Allsopp, 1985) than used here. Thus, attentional impulsivity may more broadly impact task performance by reducing the ability to attend to goal-relevant information, rather than by biasing one to rapidly reach a decision using less evidence. This appears to fit with the interpretation that HMMs are continually distracted by external and/or internal information: to the extent that high self-reported attentional impulsivity reflects the inability to filter out distracting thoughts and information, this reduced filtering would lead to increased representational competition in WM and LTM between task-relevant and –irrelevant information, potentially resulting in lower fidelity representations and thus reduced discrimination in WM and LTM.
Supplemental References
Eysenck, S. B. G., Pearson, P. R., Easting, G., & Allsopp, J. F. (1985). Age norms for impulsiveness, venturesomeness and empathy in adults. Personality and Individual Differences, 6(5), 613–619. doi:10.1016/0191-8869(85)90011-X
Green, C. S., & Bavelier, D. (2007). Action-video-game experience alters the spatial resolution of vision. Psychological Science : a Journal of the American Psychological Society / APS, 18(1), 88–94. doi:10.1111/j.1467-9280.2007.01853.x
Grubbs, F. E. (1950). Sample Criteria for Testing Outlying Observations. The Annals of Mathematical Statistics, 21(1), 27–58.
Kessler, R. C., Adler, L. A., Ames, M., Demler, O., Faraone, S., Hiripi, E., et al. (2005). The World Health Organization Adult ADHD Self-Report Scale (ASRS): a short screening scale for use in the general population. Psychological Medicine, 35(2), 245–256.
Smillie, L. D., Dalgleish, L. I., & Jackson, C. J. (2007). Distinguishing between learning and motivation in behavioral tests of the reinforcement sensitivity theory of personality. Personality and Social Psychology Bulletin, 33(4), 476–489. doi:10.1177/0146167206296951