Event-related potentials

Figure 10 shows the grand average cue-locked ERP. For the CNV, the ANOVA revealed no main effect in Task, Group, or interaction of both factors, Fs (1, 28) ≤ 1.21, ps ≥ .28, ηp2 ≤ .04.

Figure 10. Grand average cue-locked ERPs for the color-naming and word-reading tasks at FCz for the less active and more active participants. The dashed vertical line at time point 0 ms reflects the onset of the cue stimulus. The shaded area indicates the time window that was used for the CNV analysis

Figure 11. Grand average target-locked ERPs for the color-naming and word-reading tasks at F3, Fz, F4, C3, Cz, C4, P3. Pz, and P4 in congruent and incongruent trials for the less active and more active participants

3.3.2 Target-locked

Figure 11 illustrates the grand average ERP waveform for each physical activity group as a function of task and condition. For the P2, the ANOVA yielded a significant

main effect for Group, F(1, 28) = 9.18, p < .01, ηp2 = .25 with more active group yielding larger P2 amplitude than less active group (6.93 ± 2.63 vs. 4.38 ± 2.86 µV). No other main effects or interactions were observed for P2 amplitude, Fs (1, 28) ≤ 2.94, ps ≥ .09, ηp2 ≤ .09. In terms of P2 latency, no main effects or interactions were observed, Fs (1, 28) ≤ 3.19, ps ≥ .09, ηp2 ≤ .1.

For the N2 amplitude, the ANOVA revealed a marginally significant Task × Group interaction, F(1, 28) = 3.75, p = .06, ηp2 = .12. Decomposition of this interaction revealed that N2 amplitude was larger in the more active group relative to the less active group in the word-reading task (p < .05), whereas no difference was observed in the color-naming task (p = .88) (Figure 12a). No other main effects or interactions were observed for N2 amplitude, Fs (1, 28) ≤ 1.76, ps ≥ .20, ηp2 ≤ .06.

Figure 12. Interaction effect between group and task for the N2 and P3, and group and condition for the N450. Mean amplitudes and standard errors of the N2 (A) at Cz, P3 (B) at centroparietal area averaged across Stroop trials in the color-naming and word-reading tasks, and N450 (C) at frontocentral area averaged across Stroop tasks in the congruent and incongruent trials for the less active and more active participants

For the N2 latency, the ANOVA revealed a significant main effect of Task, F(1, 28) = 6.44, p < .05, ηp2 = .19, with color-naming task yielding shorter N2 latency than word-reading task (242.90 ±. 22.52 ms vs. 253.57 ± 27.46 ms). No other main effects or interactions were observed for N2 latency, Fs (1, 28) ≤ 2.14, ps ≥ .15, ηp2 ≤ .07.

For the P3 amplitude, the ANOVA revealed a significant main effect for Task, F(1, 28) = 6.09, p < .05, ηp2 = .18, with color-naming task yielding larger P3 amplitude than word-reading task (7.94 ± 3.62 vs. 6.89 ± 2.89 µV), and Condition, F(1, 28) = 21.60, p < .01, ηp2 = .34, with larger P3 amplitude in the congruent (7.81 ± 3.37 µV) relative to the incongruent (7.03 ± 3.22 µV). There was also a marginally significant interaction between Task × Group, F(1, 28) = 4.16, p = .05, ηp2 = .13. Post-hoc tests revealed that the P3 component had a significantly larger amplitude in the more active group compared with the less active group (9.24 ± 3.82 µV vs. 6.65 ± 2.93 µV), in the color-naming task, (p < .05), whereas this pattern was not repeated in the word-reading task (p = .41) (Figure 12b). No other main effects or interactions were observed for P3 amplitude, Fs (1, 28) ≤ 3.64, ps ≥ .67, ηp2 ≤ .12.

For the P3 latency, the ANOVA revealed a significant main effect for Task, F(1, 28) = 7.64, p < .05, ηp2 = .21, indicating shorter P3 latency in the color-naming task (362.43 ± 37.09 ms) relative to word-reading task (374.06 ± 39.02 ms). No other main effects or interactions were observed for P3 latency, Fs (1, 28) ≤ 4.08, ps ≥ .05, ηp2

≤ .13.

For the N450 amplitude, the ANOVA revealed a significant main effect for Group, F(1, 28) = 7.27, p < .05, ηp2 = .21, with more active group yielding smaller N450 than less active group (4.06 ± 2.13 vs. 2.24 ± 1.13). There is also a significant Condition

× Group interaction, F(1, 28) = 4.67, p < .05, ηp2 = .14. Decomposition of this interaction revealed that the N450 amplitude had significantly larger amplitude in the

incongruent condition compared with the congruent condition in the more active group (p < .05), whereas no difference in condition was found with the less active group (p

= .50) (Figure 12c). No other main effects or interactions were observed for the N450 amplitude, Fs (1, 28) ≤ 2.43, ps ≥ .14, ηp2 ≤ .08.

Figure 13. Topographic distributions of the P2, N2, P3, and N450 components in color-naming and word-reading tasks for each condition and task for the less active group and more active group

4. Discussion

The present study assessed the relation of physical activity to interference processing in healthy young adults, using Stroop and reverse Stroop tasks. The ERP approach was used to improve characterization of the association between physically

active and executive function in this population. Participants were separated into more and less-active groups according to their regular physical activity. Both groups were matched with respect to gender, age, educational level, and health status, but they clearly differed in the level of physical activity. The key findings were that young adults with more physical activity generally have shorter RT and less IIV of RT on the Stroop tasks compared with young adults with less physical activity. Additionally, the ERP analysis revealed that more active young adults had enhanced P2, N2, P3, and reduced N450 amplitudes compared with less active young adults. Lastly, physical activity did not have a meaningful influence on CNV.

The RTs were significantly longer, and the error rates were higher for the incongruent trials compared with congruent trials in both Stroop (i.e., color-naming) and reverse Stroop (i.e., word-reading) tasks, which reflects the typical “Stroop interference effect” and suggests a greater amount of inhibitory control needed to resolve prepotent incorrect responses (Chang et al., 2017a; Chang et al., 2014b; Ludyga et al., 2018). This finding also demonstrated the appropriateness of our task manipulation. The RTs were significantly longer in the word-reading task than the color-naming task. Differences between the two tasks were also observed at the neuroelectric level. ERPs measures supported the behavioral findings due to the word-reading task inducing delayed and larger N2, and delayed and smaller P3 than the color-naming task. In light of the classical findings of Stroop effect, Stroop and other researchers (LaBerge & Samuels, 1974; Posner & Snyder, 1975; Shiffrin & Schneider, 1977; Stroop, 1935) described this interference with the automaticity hypothesis (e.g., word reading is more automatic than color naming). Based on this account, the demanding process of naming a color "vocally" is hampered by the more automatic process of word meaning; however, the opposite does not hold true (MacLeod, 1991;

Stroop, 1935). Contrary to predictions from the word automaticity account, Stroop interference can be strongly observed from color distractors in reverse versions of the Stroop task where manual responses were required (Blais & Besner, 2006; Durgin, 2000, 2003; Luo, 1999). This asymmetric nature of the Stroop effect suggests that computing the linguistic labels from perceived colors is not automatic and is highly task-specific (Trueswell & Papafragou, 2010). The modulatory effects of response types on Stroop performance have been described in various studies, suggesting that different response types lead to different Stroop interference effects (Henik et al., 1999;

Logan & Zbrodoff, 1998; White, 1969). A more general explanation of the Stroop effect simply focuses on the inability to ignore distractor information, which may vary in salience depending upon response types. Thus, Stroop interference effect occurs whenever participants fail to inhibit distractor information that is incompatible with the target task and response modality.

The first aim of the current study was to explore whether physical activity is generally or specifically associated with cognitive processes as assessed by Stroop and reverse Stroop tasks. We found that, regardless of the tasks and conditions, young adults with more levels of physical activity showed better cognitive performance, indexed by shorter RTs and less IIV of RTs. Further, the lack of significant differences in accuracy between the more active group and the less active group demonstrates that the facilitation in cognitive performance was not the result of a speed-accuracy trade-off.

An important meta-analysis reported that cognitive functions requiring effortful processing, such as executive functions, may gain a greater benefit from chronic physical activity when compared with other, simpler functions (e.g., processing speed), showing the selective benefits of physical activity (Kramer et al., 1999). Interestingly, the group differences in response time were generally found in both Stroop conditions,

rather than being selective for the incongruent condition. Gajewski et al. (2020) found also an unspacifically enhanced performance (in congruent and incongruent trials) and larger ERP in high vs. low performing old adults in a Stroop task corroborating the present results and indicating that superior executive functioning is due to larger availability of cognitive resources. Another possible explanation of the present finding is that the cognitive load in the congruent trials was relatively high, due to the participant having to make a choice from four buttons corresponding to the given sensory or verbal information. Further, the participant may apply the same task rules by paying attention to either the sensory or verbal information and inhabiting the irrelevant information even if there was no interference in the congruent condition.

What supports this assumption is that the mean RTs in the congruent condition was around 600 ms, which is much longer compared to a simple or yes/no reaction time task (for more information about the RT on these tasks, see Cirillo et al. (2017). Thus, a general effect of physical activity on congruent and incongruent trials was observed in the present study.

Apart from speed, measures of response variability provide a reliable index of cognitive function beyond that of mean RT, with IIV being widely used as a behavioral marker of neurological health (MacDonald et al., 2006). IIV, as indexed by standard deviation of RT, characterizes the within-person fluctuations in behavioral performance.

This fluctuation affords an additional measure by which to understand behavioral development (Moore et al., 2013), and is separable from more enduring changes in learning and development (MacDonald et al., 2006). For instance, increased IIV of RTs has been found reliably in older than younger individuals (Berchicci et al., 2012), under more demanding task conditions than less demanding (Gajewski et al., 2011, 2012).

More importantly, IIV of RT was found to be lower in physically fit than unfit children

(Moore et al., 2013; Wu et al., 2011) and physically high active than low active older adults (Gajewski & Falkenstein, 2015), in an interference task. The present study shows that the more active young adults, exhibited decreased IIV of RT relative to the less active counterparts. This finding extends the results in the same vein with the young adult population during Stroop and reverse Stroop tasks.

A novel aspect of the current study was the monitoring of the time course ERPs, including both early (i.e., P2, N2) and late components (i.e., P3, N450) of the target-related potential as well as the cue-target-related potential (i.e., CNV). According to the literature, successful interference resolution should be expressed in smaller P3 in the incongruent condition relative to the congruent condition (Ila & Polich, 1999), which has been also found in the present study. More importantly, our finding revealed that P3 was larger in the more active group compared to the less active group in the color-naming task. This finding replicates previous studies in which higher physical activity levels is associated with a larger P3 amplitude (Chang et al., 2013; Dai et al., 2013;

Fong et al., 2014; Hawkes et al., 2014; Hillman et al., 2006; Huang et al., 2014; Tsai et al., 2016; Chun-Hao Wang & Tsai, 2016). Thus, the findings of this study imply that enhanced cognitive performance in young adults with higher physical activity levels can be attributed to an increase in attentional resource allocation and provide further evidence for the beneficial relation of physical activity to Stroop performance in young adults.

N450 showed a similar pattern to that of P3. The more active group had a smaller N450 amplitude compare with the less active counterparts. N450, a specific component evoked by the Stroop paradigm, is considered to originate in the anterior cingulate cortex (ACC) and to indicate a conflict detection processes (Minzenberg et al., 2014). Our finding of a larger N450 amplitude following incongruent trials than

following congruent trials supports evidence of the role of N450 in conflict monitoring activity (West, 2004). The association between N450 and the ACC provides an alternative explanation for the beneficial association between physical activity and cognitive interference control. Colcombe et al. (2004) indicated that older adults with higher physical fitness exhibits better executive function, with lower activation in the ACC than older adults with lower physical fitness. Accordingly, from a neuroelectrical perspective, our result of a decrease in N450 amplitude is consistent with the superior performance of the more active group on the Stroop task because of the reduction of ACC activation.

Another novel finding of the present study was that regular physical activity elicited similar modulation in earlier ERP components (i.e., P2, N2) as in later components (i.e., P3, N450). Specifically, P2 was larger in the more active group relative to the less active group. To our knowledge, these data are the first to show modulation of the P2 amplitude as a function of physical activity level. P2 is related to selective attention and basic perceptual processing (Luck & Hillyard, 1994). Thus, the difference in P2 amplitude between more and less active groups is considered to reflect selective inhibition processes. Based on this finding, we suggest that regular physical activity levels might influence the selective inhibition processes in young adults. This finding is also supported by a previous study using the same Stroop paradigm, that showed shorter P2 latency in physically high active elderly than low active elderly (Gajewski & Falkenstein, 2015). With regard to N2, our results show a marginally significant interaction between group and task. Young adults in the more active group exhibited a significantly larger N2 amplitude than those in the less active group in the only high interference task (i.e., reverse Stroop task). N2 is believed to be related to pre-response conflict monitoring (Yeung et al., 2004). To date, only two studies have

analyzed the N2 in physically active compared to low active individuals. Using a cross-sectional design, Taddei et al. (2012) and Di Russo et al. (2006) reported shorter RTs as well as larger N2 in fencers compared to non-fencers. These N2 findings, along with the decrease in the N450 amplitude in the more active group, suggest that regular physical activity is positively associated with improved top-down executive control and reduced conflict processing under high workload.

Regarding CNV, we hypothesized that the superior task performance observed in the more active individuals could be extended to motor preparation processes. Unlike the previous studies that mentioned in the introduction (Hillman et al., 2002; Kamijo et al., 2010), no statistically significant difference in CNV component has been reported between the groups. Our result suggests that the superior task performance observed in the more active individuals is not a secondary effect of an additional effort invested during task preparation or a more efficient preparatory activity, consistent with Cirillo et al. (2017) study. The differences between the results of Hillman et al. (2002) and Kamijo et al. (2010), and Cirillo et al. (2017) among the present study could be due to the different tasks and instructions used. In particular, Hillman et al. used a S1–S2–S3 paradigm with different task difficulties indicated by a cue, and Kamijo et al. used a Sternberg task with different levels of working memory load. In contrast, the cues conveyed no task-relevant information in Cirillo et al. and the present study. It is, therefore, possible that they were ignored by our subjects and not used for an active task preparation. As we did not find group differences in the CNV, we believe that the superior task performance observed in the more active group has primarily resulted from cognitive processes elicited by the targets. Taken together, this study further supports the results from previous studies that exhibit the facilitative effect of regular physical activity on attentional control (Winneke et al., 2012), executive control

(Kamijo & Takeda, 2009), and cognitive flexibility (Themanson et al., 2008). In addition, it extends to the ability to inhibit cognitive interference in young adults.

Although this study furthers the current knowledge of the relation of physical activity to interference processing in young adults using behavioral and neuroelectrical approaches, several limitations should be acknowledged. First, the positive associations between physical activity and cognitive performance during Stroop interference tasks observed in the present study was based on cross-sectional evidence, which limits the interpretation of causal relationships. However, along with recent neuroimaging investigations that used longitudinal interventions to show that exercise leads to modifications of brain function and structure (Broadhouse et al., 2020; Erickson et al., 2011), additional exploration of the effects of physical activity on executive function will be warranted using randomized controlled trials and neuroelectrical approaches.

Second, the significant findings in the present study could be affected by a selection bias between the higher and lower physical activity groups because some uncontrolled factors (e.g., social context of the sports club, nutrition, genetics) have been shown to influence neurocognitive performance. However, the effect of selection bias is diminished because nonsignificant differences were noted between the more and less active groups across several demographic characteristics and mental status. Lastly, although the more active group showed better cognitive performance relative to the less active group, which would be attributed to their more active lifestyle, physical fitness or the intensity of their daily physical activities could be potential factors in the physical activity cognition association. Thus, future studies are encouraged to consider the characteristics of physical fitness and the intensity of physical activity.