Impact of tDCS on working memory training is enhanced by strategy instructions in individuals with low working memory capacity

Interventions to improve working memory, e.g. by combining task rehearsal and non-invasive brain stimulation, are gaining popularity. Many factors, however, affect the outcome of these interventions. We hypothesize that working memory capacity at baseline predicts how an individual performs on a working memory task, by setting limits on the benefit derived from tDCS when combined with strategy instructions; specifically, we hypothesize that individuals with low capacity will benefit the most. Eighty-four participants underwent two sessions of an adaptive working memory task (n-back) on two consecutive days. Participants were split into four independent groups (SHAM vs ACTIVE stimulation and STRATEGY vs no STRATEGY instructions). For the purpose of analysis, individuals were divided based on their baseline working memory capacity. Results support our prediction that the combination of tDCS and strategy instructions is particularly beneficial in low capacity individuals. Our findings contribute to a better understanding of factors affecting the outcome of tDCS when used in conjunction with cognitive training to improve working memory. Moreover, our results have implications for training regimens, e.g., by designing interventions predicated on baseline cognitive abilities, or focusing on strategy development for specific attentional skills.

Demographic characteristics and descriptive statistics of the overall sample, divided by groups. For each group, we report the count N and the average score, together with its standard error, the F statistics, corresponding p-value and effect size ( ! ! " ) from a 1-way ANOVA between groups.    in the adaptive spatial nBack task (aNback) as the average difference between the mean 'n' within a session (excluding the first block) and the mean 'n' at baseline (∆n # = n # − n # !"#$%&'$ ).
We conducted a 3-way mixed ANOVA with two between-subject factors (STIMULATION: ACTIVE, CONTROL x STRATEGY: STRATEGY, NoSTRATEGY) and one within-subject factor (TIME: change at DAY 1, DAY 2). We found a main effect of TIME (

POSSIBLE ADVERSE EFFECTS OF BRAIN STIMULATION
Possible adverse effects were collected from participants after each stimulation session, together with the likelihood of such effects being caused by stimulation.

MOOD AND ATTITUDE TOWARD THE STIMULATION
Mood and attitude towards the intervention were monitored before each testing session, see Table S6. A 2-way mixed ANOVA (between subjects: STIMULATION: ACTIVE, CONTROL x STRATEGY: STRATEGY, NO STRATEGY; within-subject: TIME: T0,T1 on DAY 1, T2,T3 on DAY 2) revealed no effect on ALERTNESS, MOTIVATION or SADNESS (all ps > 0.1). We did find a main effect of TIME on EXPECTATION ON WM PERFORMANCE (F(3,231) = 3.24, pGG = 0.031, n2 = 0.01).
We also found a significant effect of TIME (F(3,234) = 9.79, pGG < 0.001, n2 = 0.025), and a significant TIME x STIMULATION x STRATEGY interaction (F(3,234) = 4.711, pGG = 0.009, n2 = 0.01) on EXPECTATION ON tDCS with the ACTIVE -NO STRATEGY group having higher expectation towards the end of the intervention. We found no effect of the intervention on NEGATIVE ATTITUDE scores (as self-reported on the PANAS). We did find a significant effect of TIME on POSITIVE ATTITUDE scores (F(1,80) = 8.40, p = 0.005, n2 = 0.01), with scores higher before than after the intervention (Table S7).

RETROSPECTIVE STRATEGY QUESTIONNAIRE
In order to analyze self-reported feedback on strategy use, the following questions were asked at the end of the second day of testing. The STRATEGY group was asked if they used strategy explained to them at the beginning of the experiment, while the No STRATEGY group was asked if they used a strategy and how they would classify it according to the following. In between brackets we report how each strategy was classified by us in the following analysis. We asked ourselves if motivation could be a factor for an individual in developing a more efficient strategy on their own. We followed this up with two additional analysis: 1) Correlation between motivation and performance. Following the assumption that using a more effective strategy would lead to better performance, we analysed the correlation between performance at baseline (as the average n at baseline on the aNback) and motivation (as measured by either Positive Attitude (PANAS), expectation towards Cognitive training and General motivation (both measured on a Likert scale). All correlations were not significant (ps > 0.1). Therefore, more motivated individuals were not more likely to develop an effective strategy than less motivated individuals.
2) Correlation between strategy effectiveness and motivation. At baseline, before being instructed with a strategy, it is possible that individuals would still develop a strategy on their own. We collected this information during the experiment (see STRATEGY questionnaire in the Supplementary Material), however, the following analysis could only be qualitative as this kind of feedback is not very reliable (individuals often have difficulties verbalising the strategy used). First, we classified the strategy reported by the participant as EFFECTIVE (those similar to the one provided by us, e.g., ASSOCIATION + GROUPING) or INEFFECTIVE (different from the one provided by us), then we visualised the distribution of motivation scores across the two strategy categories (see here below) and found that there was no visual relationship between strategy effectiveness and motivation. Figure S1 Qualitative comparison of motivation as a function of strategy efficacy.