Original key findings replicated: stronger implicit interference in CH-PATs

The analyses below are reported in the format of the mean (standard error of the mean [SEM]) unless otherwise specified. The implicit interference was deemed the target reaction time slowing caused by an incongruent implicit distractor, as compared to a congruent implicit prime. Because only thirty-six out of forty participants from the previous study were considered here, we repeated the same analyses to ensure that the results were similar. Indeed, the mean accuracy of the awareness test was 44.58% (2.32%) and slightly below chance (compared to 50%, t(29) = −2.33, p = 0.03), confirming the implicit nature of the distractor. No difference in distractor awareness was found between CH-NATs (44.26 (3.8) %) and CH-PATs (45.00 (2.21) %) (two-sample t-test, t(28) = 0.1543, p = 0.88). Similarly, we reproduced the key finding that during the high-load, color-naming condition, CH-PATs had more interference from an implicit response-incongruent distractor than CH-NATs (CH-PATs: 4.21 (1.51) % change in reaction time, t(18) = 2.79, p = 0.01; CH-NATs: −0.31 (1.1) % change in reaction time, t(16) = −0.28, p = 0.78). A direct comparison between the two yielded t(34) = 2.37, p = 0.02, Cohen’s d = 0.81.

Practice has distinct effects on CH-PATs and CH-NATs

Both groups exhibited a practice effect in the second session. Overall, between the two sessions, the average reaction time decreased by 6.19% (3.04%), while the average accuracy increased by 6.09% (1.40%). The time interval between the two sessions was comparable across the two groups (CH-PATs: 15.84 (2.87) days; CH-NATs: 15.29 (1.68) days, t(34) = 0.16, p = 0.87).

We investigated whether practice over time differentially affects performance of the low-risk and high-risk participants. To this end, we first performed two separate mixed-effect analyses of variance (ANOVAs) on the mean accuracy and on the reaction time with one between-subject factor (participant status: CH-NATs/CH-PATs) and one within-subject factor (session number: one/two). The following data are collapsed across all three factors in the original design (task switching vs. non-switching, word-naming vs. color-naming, and incongruent vs. congruent distractors). Analysis of the accuracy yielded a main effect of session number, F(1, 34) = 21.66, p < 0.0001 , ηp2 = 0.39, and an interaction between session and participants’ status, F(1, 34) = 9.03, p = 0.005 , ηp2 = 0.21. The analysis of reaction time yielded a main effect of session, F(1, 34) = 4.20, p = 0.048 , ηp2 = 0.11. Post hoc analyses revealed that CH-PATs exhibited a stronger practice effect in terms of accuracy (CH-NATs: 2.08% (1.36%) vs. CH-PATs: 9.67% (2.05 %), t(34) = −3.01, p = 0.005) but not reaction time (CH-NATs: 4.06% (4.44%) vs. CH-PATs: 8.10% (4.24%), t(34) = -0.66, p = 0.52).

To address the impact of practice on implicit interference, we ran a correlation analysis (Pearson’s correlation) between the implicit distractor interference effect and the practice effect in CH-NATs and CH-PATs, respectively. For the purpose of this analysis, the practice effect was defined as the percentage decrease of reaction time in the second session compared to the first because the interference effect was only observed in the domain of reaction time. Please note that the implicit distractor interference was calculated entirely based on the results of the second session. By converting reaction times into percentages for both interference and practice effects, baseline performance differences and the long tails in typical reaction time distributions were eliminated. Analysis of the CH-NATs yielded a marginal, negative correlation between practice and implicit interference (r = −0.46, p = 0.06). However, the same analysis on CH-PATs yielded a positive correlation between the two (r = 0.50, p = 0.03). We calculated z-scores for the r values and directly compared the two groups’ correlations with a two-tailed test, which resulted in a p-value of 0.004. These results showed that the interaction between practice and the implicit interference was distinct between CH-NATs and CH-PATs: better performance in the second session was correlated with less implicit distraction in CH-NATs, while for CH-PATs, better performance in the second session led to stronger distraction (Fig. 2, left).

Fig. 2

Left Distinct correlations between practice effect (percentage decrease of reaction time in the second session; positive values indicate a stronger practice effect, i.e., more decrease in reaction time, x-axis) and implicit interference (percentage decrease of reaction time between incongruent distractor and congruent prime; positive values indicate stronger interference, i.e., more increase in reaction time, y-axis) in CH-NATs (low-risk) and CH-PATs (high-risk). Right Positive correlation between lower alpha (8-11 Hz) ERD (event-related desynchronization) difference (x-axis) and the implicit interference (y-axis) was observed only in CH-NATs. Each dot represents a participant. The dotted lines were linearly fitted to the two correlations

Greater low-range alpha ERD difference between task-switching and non-switching trials on EEG in session one is correlated with interference in CH-NATs

In the first session, EEG data was analyzed for the majority of participants (14 of 17 CH-NATS and 16 of 19 CH-PATs), which allowed us to determine the correlation between prior EEG data and later behavioral performance in these two groups. To minimize multiple comparisons and to focus on an attention component, we examined the most distinctive alpha event-related desynchronization (ERD) difference between the CH-NATs and CH-PATs. In the original study, we calculated the difference in the central area (C3, Cz, C4) alpha ERD between high-load task-switching trials and low-load non-switching trials (alpha ERD (switching)–alpha ERD (non-switching)). Note that ERD is a reduction of power, and a stronger (more negative) ERD is therefore a stronger reduction. For CH-PATs, the alpha ERD difference was negative, indicating that these participants had stronger low alpha ERDs during switching trials than non-switching trials. For CH-NATs, the opposite was observed: their alpha ERD difference was positive, showing stronger low alpha ERDs during non-switching trials. Together, these results suggest that CH-PATs have stronger attentional engagement during task switching than on non-switching tasks, whereas CH-NATs have more engagement when not changing tasks. These results were also found with the reduced cohort in our study (low alpha ERD mean difference (SD), CH-PATs: −0.26 (0.63) vs. CH-NATs: 0.37 (0.67), t(28) = −2.68, p = 0.01, Cohen’s d = 1.00) (Fig. 3).

Fig. 3

Alpha ERD difference between task-switching and non-switching trials in session one in CH-NATs and CH-PATs. On average, CH-PATs exhibited a stronger ERD (reduction in power) during task-switching than non-switching trials, while CH-NATs showed the opposite

We ran a correlation analysis between the aforementioned low alpha ERD difference in the first session and the implicit interference in the second session in both groups. CH-PATs had no correlation between their low alpha ERD difference and implicit interference (r = −0.16, p = 0.54). In contrast, there was a positive correlation between low alpha ERD difference and implicit interference in CH-NATs (r = 0.56, p = 0.04) (Fig. 2, right). A further z-score comparison between the two correlations revealed a p-value of 0.05. These results showed that the CH-NATs who exhibited a higher low-range alpha ERD difference in the first session also experienced stronger distraction in the second session. This finding suggests that in low-risk participants, having lower attentional engagement in switching trials was associated with stronger implicit interference. A framework for how CH-PATs and CH-NATs utilized attentional resources differently is provided in "Discussion" section.