Data Analyses

As behavioral parameters, we considered RTs, MTs and the stop-signal reaction times (SSRTs). RTs were computed as the difference between the time of the go-signal presentation and the onset of movement. MTs were determined as the difference between the time of movement onset and the time at which the peripheral target was touched. Trials with RTs shorter/longer than the mean minus/plus three SDs were excluded from the analysis. Overall 1.8 and 0.4% of the data were discarded in patients and controls, respectively. The SSRT represents the estimate of stop latency (25), and it was estimated exploiting the integration method, which gives the best estimate when proactive slowing occurs (26). The SSRT provides a measure of reactive inhibition, i.e., the ability of a subject to react outright to the stop signal. In contrast, proactive inhibition, i.e., the ability of subjects to shape their response strategy in anticipation of known task demands as the awareness of the fact that sometimes a stop-signal could have been presented, was assessed by comparing the RTs and the MTs of no-stop trials vs. those of go-only trials. In fact, it has been shown that when the subject executes a no-stop trial, its RT is lengthened and its MT is shortened with respect to situations in which the same movement has to be performed in the context of the go-only trial. This “context effect” (27) represents an optimization of costs and benefits because longer RTs are compensated by shorter MTs and vice versa.

Different types of analysis of variance (ANOVA) were used depending on the experimental design in order to assess changes in RTs, MTs, and SSRTs. Bonferroni corrections were applied for all multiple comparisons. To contrast cumulative distributions of RTs and of MTs obtained in no-stop and go-only trials, two-sample Kolmogorov–Smirnov tests were used. χ²-tests were used to determine whether there were significant differences between the occurrences of the context effects.

We measured the “effect-size” by computing the partial eta-squared (ηp2) for each ANOVA (with values of 0.01, 0.058, and 0.139 indicating small, medium and large effects, respectively), and Cohen's d for t-tests (with values of 0.2, 0.5, and 0.8 indicating small, medium, and large effects) (28). Finally, to quantify the strength of null hypotheses, we calculated the Bayes factors (BF10) with an r-scale of 0.707 (29). BF10 values < 0.1 and < 0.33 provide strong and moderate support, respectively, for a null hypothesis compared to the alternative hypothesis. For the sake of clarity, and to improve readability, we report just significant results, unless otherwise indicated. Data will be freely available from the Open Science Framework platform (https://osf.io/pfeqv/).