Intra-individual variability in the effects of transcranial direct current stimulation on free choice saccade behaviour

Abstract

Transcranial direct current stimulation (tDCS) is used as a tool to causally influence neural activity in humans non-invasively. Although most studies recruit a large number of participants in order to uncover population-level effects, growing evidence suggests that tDCS may be expected to induce different effects in different individuals, leading to large inter-individual variability and confounds in population-level testing. Alternatively, this variability may arise from intra-individual sources that are difficult to assess in standard designs. Here, we performed between 8 and 10 sessions of tDCS within individuals to understand how intra-individual variability impacts the inference of tDCS effects. We recruited 5 participants who underwent functional MRI to localize the human frontal eye field (FEF) homologue. An HD-tDCS montage was used to stimulate the target location based on individual MRI localizations, alternating the polarity between anodal or cathodal current over 8-10 repeated sessions during a 5 week period. Participants performed a free choice task before and after stimulation while recording EEG activity. We then developed a difference-in-difference method based on permutation testing to assess the likelihood of a causal effect of tDCS at different levels of abstraction: group-level, inter-individual, and intra-individual. At the group-level, we found evidence for an influence of tDCS on choice reaction times, which followed a reaction-time dependent change in alpha-band activity, and on how choices depended on recent history. However, individuals showed heterogeneous, and often contradictory, effects. We then analyzed the distribution of session permutations at the intra-individual level, and found a discrepancy between the inter-individual effects that survived significance testing and the intra-individual effects that correlated on a session-session basis. We argue that, while the observed variability may have arisen from a combination of inter and intra-individual differences relevant to tDCS-dependent mechanisms of action, it may be equally well explained by spurious effects arising from history-dependence between repeated measures that are typically assumed to be independent. In light of this, we assess the counterfactuals that must be evaluated in order to make data-driven inferences about the causal effects of tDCS on free choice behaviour and its neural correlates.