This study aimed to investigate whether there are group differences in the performance of heavy, intermediate, and light media multitaskers on a measure of implicit learning. It was predicted that heavy media multitaskers would demonstrate a greater degree of implicit learning in the contextual cueing paradigm, as they are theorised to have a wider breadth of attention (Cain & Mitroff, 2011; Lui & Wong, 2012; Ophir et al., 2009; Uncapher et al., 2016). However, results found the opposite: The speed at which implicit learning occurred was slower for heavy media multitaskers compared with intermediate or light media multitaskers. The current study also aimed to explore whether differences in visual working memory performance affect performance in the contextual cueing task in heavy, intermediate, and light media multitaskers. Working memory performance was found to be unrelated to performance in the contextual cueing task, and more importantly, unrelated to media multitasking behaviour. Finally, the study also aimed to investigate the performance of intermediate media multitaskers, given that this group had received little research attention to date. One previous study found that on some laboratory tasks, intermediate multitaskers out-performed both heavy and light media multitaskers (Cardoso-Leite et al., 2016). There was no evidence for superior performance of intermediate multitaskers in the current study.
Implicit learning and media multitasking behaviour
Initially, we sought to explore whether there may be hidden benefits to being a heavy media multitasker. One possibility was that heavy media multitaskers might actually be picking up more perceptual information than other groups, even unintentionally (Lin, 2009; Lui & Wong, 2012), and this would manifest in a greater extent of implicit learning during the contextual cueing task. Surprisingly, results were in the opposite direction, and in fact heavy media multitaskers performed worse than other groups. Light and intermediate media multitaskers showed a steady increase in contextual cueing throughout the task, but for heavy media multitaskers, the degree of contextual cueing did not increase during the task. While unexpected, this result does conform to a growing body of literature that indicates that increased media multitasking behaviour is associated with poorer performance on a variety of cognitive tasks (Cain & Mitroff, 2011; Cardoso-Leite et al., 2016; Minear et al., 2013; Mosiala et al., 2016; Uncapher et al., 2016).
Previously, it has been argued that the main underlying difference between heavy and light media multitaskers is their scope of attention (Cain & Mitroff, 2011; Lui & Wong, 2012; Ophir et al., 2009; Uncapher et al., 2016). Heavy media multitaskers are theorised to have a broader scope of attention relative to light and intermediate media multitaskers. To date, only one study has specifically examined the impact of attentional scope on performance in the contextual cueing task (Bellaera, von Mühlenen, & Watson, 2014). Bellaera et al. (2014) examined whether a tendency to process visual information using a broad or narrow attentional scope affects performance in the contextual cueing task. Scope of attention was measured using a shape detection task, wherein participants searched for a target shape (such as a triangle), which might appear as a local shape (e.g. small triangles arranged in the shape of a square) or a global shape (e.g. a large triangle made up of small squares). Reaction times to detect global targets were subtracted from reaction times for local targets for each participant in order to determine a preference for either broadly distributed or narrowly focused attention. Participants subsequently completed the contextual cueing task. Results showed that those with a broader scope of attention displayed a significantly reduced contextual cueing effect; on average those with broadly distributed attention showed a contextual cueing effect of 114 ms, compared to a contextual cueing effect of 213 ms for those with a narrow attentional scope.
Taken together, the results of the current study and those of Belleara et al. (2014) indicate that a broader scope of attention is associated with reduced contextual cueing. In other words, the heavy media multitaskers in our study were uniquely impaired on the contextual cueing task because their broad attentional scope was actually more of a hindrance than a help during the task. The results of the current study are consistent with the local hypothesis: A narrower scope of attention meant that attending to the local context facilitated search performance for light media multitaskers, whereas attending to the global display impaired the performance of heavy media multitaskers.
A number of previous studies illustrate that it is not necessary to attend to the entire display to obtain a contextual cueing effect, and in fact, focusing on the local context can facilitate search. It has been demonstrated that repeating only half of the display, just one quadrant of the display (Olson & Chun, 2002), or just two distractors in the same quadrant of the target produces a significant contextual cueing effect (Brady & Chun, 2007). Interestingly, adding empty distance between the distractor and target does not eliminate this local effect, suggesting that if there is no information close to the target, selective attention may extend further in order to search for useful context (Olson & Chun, 2002). The use of the local context also explains why contextual cueing survives in experiments where entire displays are rescaled (Jiang & Wagner, 2004) but is extinguished when ‘new’ distractors are inserted in between the target and ‘old’ distractors (Olson & Chun, 2002). It has been suggested that during the contextual cueing task, the repeated presentation of the target and a subset of distractors forms a visual ‘chunk’ (Gobet et al., 2001; Manelis & Reder, 2012). If local context drives contextual cueing effect, this implies that heavy media multitaskers could be less proficient in the chunking of visual information, and less proficient in the use of these perceptual chunks to guide visual search.
This use of local context in the contextual cueing task also explains why Cain et al. (2016) did not observe any correlation between media multitasking behaviour and performance on a probabilistic classification task (i.e. the weather prediction task), as the implicit learning involved in the task is less dependent on local contextual learning. The current study indicates that frequent media multitasking behaviour is related to a specific deficit in implicit learning for spatial context. This is a novel finding, and extends on previous laboratory studies examining media multitasking, which have primarily examined conscious cognitive processes.
While it has been argued that the differences in implicit learning between media multitasking groups occurred due to differences in attentional scope, other explanations may also be considered. Overall, it is unlikely that other cognitive factors can account for reduced contextual cueing in heavy media multitaskers. For example, the observed differences cannot be explained by variability in psychometric intelligence, given that implicit learning is theorised to be independent of intelligence (Gebauer & Mackintosh, 2007; Kaufman et al., 2010). Moreover, differences in working memory were controlled for in the experiment, and did not interact with implicit learning. Previous research has found a small correlation between cognitive processing speed and implicit learning (Kaufman et al., 2010). However, this association has only been demonstrated using a measure of implicit sequence learning, the serial reaction time task, and has never been replicated within the contextual cueing paradigm (Bennett, Romano, Howard, & Howard, 2008). Furthermore, previous research has failed to find a relationship between multitasking behaviour and processing speed (Cain et al., 2016), so it is unlikely that our three groups of media multitaskers would vary systematically on a measure of processing speed.
Another possibility is that features of the contextual cueing task itself influenced the result. It could be contended that shortening the contextual cueing task impacted on the expression of implicit learning within this sample. However, this is unlikely because the contextual cueing effect emerged fairly rapidly for all groups during the experiment. In addition, previous research has found a greater magnitude of implicit learning when using a shortened version of the task (Bennett et al., 2009). Another potential concern is that the stagnant contextual cueing effect observed in heavy media multitaskers could reflect a ceiling effect for this group. However, this is improbable, as there was no indication of particularly rapid response times for heavy media multitaskers. Heavy media multitaskers showed somewhat slower mean reaction times overall (952.1 ms) compared to light (909.6 ms) and intermediate (912 ms) media multitaskers, though these group differences in reaction time were nonsignificant (p > .05). Therefore, there was no indication that heavy media multitaskers failed to improve during the task due to having quickly reached the upper limits of what is attainable in terms of response speed.
Alternatively, there is also some evidence that the use of different search strategies during the task can influence the development of contextual cueing. Lleras and von Mühlenen, (2004) found that participants instructed to be as receptive as possible during the search, and to just allow the target to ‘pop’ into their mind, demonstrated a robust contextual cueing effect. In contrast, those instructed to deliberately and actively direct their attention to search for the target displayed no contextual cueing, and even negative cueing effects (increased search times for repeated displays; Lleras & von Mühlenen, 2004). One possibility is that the heavy media multitaskers in our sample were more prone to using an active search strategy, which impaired their performance on the task. However, this is doubtful because substantial literature indicates that heavy multitaskers demonstrate reduced attentional control compared to other groups (Cain & Mitroff, 2011; Lui & Wong, 2012; Ophir et al., 2009; Ralph et al., 2015; Uncapher et al., 2016), which suggests they would actually be less likely to adopt a top-down search strategy. In addition, the pattern of our results differed from that of Lleras and von Mühlenen (2004) in that heavy media multitaskers did show a contextual cueing effect, but it was to a lesser extent than other groups. While differences in search strategy probably do not explain this result, future studies could question participants about any strategies used during the task in order to further investigate this possibility.
Aside from cognitive and task-related factors, it is also possible that affective features may have impacted on the result obtained. There is some evidence that depressed individuals are uniquely impaired on contextual cueing, to the extent that they do not show a contextual cueing effect (Lamy, Goshen-Kosover, Aviani, Harari, & Levkovitz, 2008). Interestingly, there is also evidence that frequent media multitasking behaviour is associated with psychological distress, including depression (Becker et al., 2013; Reinecke et al., 2016). Given the proposed link between media multitasking and depression, there may have been a higher incidence of subclinical symptoms of depression in the group of heavy media multitaskers, and this could have impaired their contextual cueing performance. However, it is unlikely that all of our heavy media multitaskers were depressed, because they did exhibit a contextual cueing effect overall, albeit a smaller effect than the other groups. Nonetheless, the impact of subclinical depression on contextual cueing remains untested. Future research could screen participants for psychiatric distress in order to better control for this possibility.
Working memory and media multitasking behaviour
Individual differences in working memory can influence the ability to control attention (Fukuda & Vogel, 2009; Vogel, McCollough, & Machizawa, 2005; Vogel, Woodman, & Luck, 2001). The n-back task was included in the current study to control for the possibility that these individual differences could be a confounding factor when measuring implicit learning. However, results found no relationship between working memory and implicit learning. It is worth noting that the n-back task used in the current study involved objects rather than spatial configurations, which may limit the conclusions that can be drawn regarding the relationship between visual working memory and implicit learning (of spatial context). However, performance on n-back tasks using objects and n-back tasks using spatial context are suggested to be highly correlated (Jaeggi et al., 2010). Overall, results suggest that the differences in implicit learning observed between heavy media multitaskers and other groups are unlikely to have occurred due to underlying differences in working memory.
Intriguingly, there was also no relationship between performance on the n-back task and media multitasking behaviour. This is at odds with previous studies using the n-back task. Originally, Ophir et al. (2009) concluded that while heavy and light media multitaskers do not fundamentally differ on the measure of working memory, heavy media multitaskers showed a disproportionate increase in false alarm rates during the more difficult three- and four-back levels of n-back task (i.e. they were misidentifying nontarget stimuli as being targets). Ophir et al. (2009) interpreted this as evidence for poor inhibitory control in heavy media multitaskers, as it represented difficulty managing the intrusion of irrelevant information in working memory. Using a standard visual working memory task, Uncapher et al. (2016) obtained a conceptual replication of this result, as they found that heavy media multitaskers had more difficulty screening out irrelevant information, and this extraneous information placed an additional burden on working memory during both encoding and retrieval.
Conversely, a study that directly measured interference in working memory found no differences in performance between heavy and light media multitaskers (Minear et al., 2013). There was no evidence that heavy media multitaskers were impaired on the recent probes task, a widely used measure of the ability to regulate information in working memory in accordance with task goals (Jonides & Nee, 2006). One possible explanation is that these conflicting results have occurred due to task differences in cognitive load. In Ophir et al. (2009), group differences only emerged at a higher cognitive load as demands on working memory increased. Minear et al. (2013) may have failed to reproduce Ophir’s result, as the recent probes task may not have made sufficient demands on cognitive load. A second possibility is that these different results reflect additional demands on the executive function component of working memory assessed during the n-back task. Both the recent probes task and the n-back task measure proactive interference, a reduction in accuracy and response time seen due to intrusion from previously relevant stimuli (Jonides & Nee, 2006). However, the n-back task used by Ophir et al. (2009) also assesses cognitive updating, the process of maintaining relevant information and deleting or replacing irrelevant information in memory (Carretti, Cornoldi, De Beni, & Romanò, 2005). It is possible that heavy media multitaskers differ from light media multitaskers in the cognitive updating of information, rather than the ability to inhibit distractors per se.
Further evidence that heavy media multitaskers may have difficulty with cognitive updating was seen in a recent study that used the n-back task (Cain et al., 2016). Increased media multitasking was linked to poorer performance on the n-back task. However, Cain and colleagues’ (2016) result was driven by a declining hit rate for heavy media multitaskers as the task difficulty increased (i.e. they were failing to identify targets). A tendency for heavy media multitaskers to miss targets suggests that increased media multitasking behaviour is linked to difficulties in the updating of information within working memory rather than difficulties managing interference. However, the present study failed to find evidence for either of these explanations, as there was no evidence that heavy media multitaskers performed poorly on the task overall, and there was also no evidence that heavy media multitaskers were specifically impaired as task difficulty increased in the three- and four-back conditions. Notably, one other recent study also failed to find group differences between heavy and light media multitaskers on the n-back task (Cadoso-Leite et al., 2016).
One explanation for these mixed results is that the current study included a more heterogeneous sample. While Ophir et al. (2009) administered the n-back task to a relatively homogenous population of university students, and Cain et al. (2016) studied adolescents in middle school, the current study was not limited to students. This has important implications, because the media use questionnaire measures the frequency of media multitasking behaviour and does not distinguish between choosing to media multitask and a requirement to media multitask due to circumstance. For instance, in many office roles, workers may be expected to be responsive to incoming e-mails and instant messages in combination with other computer tasks, even if this is not their preferred style of work. These individuals may indeed media multitask very frequently, yet differ fundamentally from those who multitask frequently as a personal preference. As an example, various studies have suggested that heavy media multitasking is linked to impulsivity and sensation seeking (Minear et al., 2013; Sanbonmatsu et al., 2013); however, these relationships may not hold for heavy media multitaskers whose multitasking behaviour is more driven by necessity.
Intermediate media multitaskers
The majority of media multitasking studies that use the media use questionnaire employ an extreme groups design, comparing the performance of heavy and light multitaskers and excluding the middle of the distribution. As a result, people who media multitask in moderation have received little research attention to date. However, one recent study included analysis of intermediate media multitaskers, and found that this group actually performed better than other groups on measures of working memory and proactive cognitive control (Cardoso-Leite et al., 2016).
The results of the current study found no evidence for superior performance of intermediate media multitaskers. While Cardoso-Leite and colleagues (2016) found that intermediate media multitaskers performed better than other groups on the n-back task, in the current study, the performance of intermediate media multitaskers on the n-back task did not reliably differ from that of other groups. On the measures of implicit learning, the performance of intermediate media multitaskers tended to fall between the two extreme groups. Unlike heavy media multitaskers, intermediate media multitaskers demonstrated a steady increase in the magnitude of the contextual cueing effect throughout the task.
One reason why these results differed from previous research may be that Cardoso-Leite et al. (2016) also targeted people who frequently play action video games, resulting in an increased proportion of gamers in their sample. This is of interest because unlike media multitasking, frequent use of action video games has been linked to beneficial outcomes for visual processing, attention, and decision making (Green & Bavelier, 2012; Spence & Feng, 2010). Therefore, it could be that superior performance in intermediate media multitaskers is only seen in association with increased use of action video games. Inclusion of intermediate media multitaskers in future studies is needed to clarify these findings.
Theoretical and practical implications
While previous studies have proposed that heavy media multitasking is often associated with poorer performance in various conscious cognitive processes, our finding that increased media multitasking is also linked to a disruption in implicit learning is new. This finding is important because the ability to quickly and implicitly learn the association between visual stimuli, as it is defined within the contextual cueing paradigm, is fundamental to efficient visual processing, including identifying relevant information, scene learning, navigation, and prediction (Bennett et al., 2009; Brady & Chun, 2007; Brockmole et al., 2006; Chun & Jiang, 1998; Couperus, Hunt, Nelson, & Thomas, 2010; Goujon et al., 2015; Jiang & Chun, 2003).
The current result is also of theoretical importance, as it affirms an established trend seen in a number of studies for heavy media multitaskers to demonstrate breadth-biased attention (Cain & Mitroff, 2011; Lui & Wong, 2012; Mosiala et al., 2016; Ophir et al., 2009; Uncapher et al., 2016). As the research evidence to date is cross-sectional and correlational, it is impossible to infer causality from these results, or determine the direction of the relationship. If increased media multitasking behaviour is linked to the development of breadth-biased attention, then our findings suggest that this tendency to allocate attention more widely is also related to a reduced ability to incidentally learn and make use of perceptual regularities. On the other hand, if the differences observed in heavy media multitaskers reflect preexisting individual differences, this suggests that people who are already less sensitive to detecting visual regularities in the environment are choosing to multitask more often. Ironically, these heavy media multitaskers may be particularly ill equipped to manage multiple streams of information, given that they show a deficit in implicit learning, which plays a fundamental role in visual processing.
In conclusion, the results of the current study indicate that frequent media multitasking behaviour is associated with a reduced magnitude of implicit learning, as measured within the contextual cueing paradigm. In contrast to several previous studies, media multitasking behaviour was not found to be associated with differences in working memory performance. The present study extends on previous research in three important ways: it lends further support to the theory that media multitaskers differ in their scope of attention, it extends on previous media multitasking research by examining nonconscious processes, and it contributes to a small number of studies exploring individual differences in implicit learning.