The study was preregistered on the Open Science Framework website (https://osf.io/rzt6j). Data and code for the present article are available at the Open Science Framework website https://osf.io/e9up2/. The University of Salzburg Ethics committee approved all procedures (Ethical Application Ref: EK-GZ 32/2023).

We defined a target sample size of N = 20, because we obtained significant results with 20 participants in a previous study using similar procedures (Seegelke et al. 2021). However, results were statistically inconclusive with the pre-registered sample size (see Supplementary Information); therefore, we collected data from another 20 participants. Thus, our final sample comprised 40 students (self-reported gender: 27 female, 12 male, 1 non-binary, mean age = 24.18 years, SD = 9.60) which participated in exchange for course credit. 38 participants were right-handed (mean handedness score = 91.87, SD = 11.79), two participants were left-handed (mean handedness score = −60, SD = 56.57; Oldfield). All participants were physically and neurologically healthy and had normal or corrected-to-normal vision according to self-report.

Participants sat in front of a 24-in computer monitor (1920 × 1200 pixels; 60-Hz refresh rate) that was positioned on a table at a viewing distance of about 75 cm. Two square PVC blocks (10 × 10 x 3 cm) with centrally embedded round pushbuttons (7 cm in diameter) served as start location for the left and right hand. They were positioned about 40 cm apart (center-to-center distance) at the table’s front edge equidistant with respect to the body midline within comfortable distance.

Participants performed two successive movements (i.e. a prime and a probe action) in response to visual white shapes (413 × 413 pixels) presented centrally on the monitor against a black background (Fig. 1). Each shape defined one of four responses: an inward left-hand movement, an outward left-hand movement, an inward right-hand movement, or an outward right-hand movement, with respect to the body midline. To avoid stimulus repetition between prime and probe actions, each of the four actions was associated with two visual specific shapes, one for the shape as prime and one for the shape as probe, with the specific shape-action assignment randomized across participants. The experiment was controlled with MATLAB (version R2015; The MathWorks, Natick, MA) using the Psychophysics Toolbox (Brainard 1997). An optical motion capture system (Visualeyez II VZ4000v; Phoenix Technologies Inc., Vancouver, BC, Canada) recorded hand kinematics at 250-Hz sampling rate. We placed infrared markers on the dorsal side of each hand (distal end of the third metacarpal).

At the start of each trial, participants depressed the start buttons with their hands for 500 ms. The prime stimulus, i.e., one of the four shapes for prime actions, then appeared on the screen for 2500 ms and participants performed the respective action and then moved the hand back to the start position. After a 500 ms delay, the probe stimulus (one of the other four shapes for probe actions) appeared and participants then executed the respective probe action. A 2000 ms inter-trial interval followed, after which participants could initiate the next trial by depressing the buttons. If participants did not complete the respective action within the 2500 ms stimulus display interval or released the wrong response button, they received the error message “Too slow” or “Wrong effector”, respectively, on the screen. We instructed participants to perform inward and outward movements (with respect to the body midline) of about 20 cm at a comfortable speed. On average, participants moved their hands 19.3 cm and 19.4 cm for prime and probe movements, respectively. Movement distances were also similar across hands and movement direction (see Supplemental Table S2). The 16 possible prime-probe combinations were repeated four times in each block in a randomized order, and each participants completed five blocks, yielding 320 trials in total.

Prior to the experimental blocks, participants completed one to two practice blocks until they reported to be familiar with the stimulus–response associations. During practice blocks, stimulus–response associations were visible to the participants. The procedure was identical to the main experiment except that participants performed only one action in response to a single stimulus. Each stimulus was repeated eight times within a practice block for a total of 64 trials. On average, participants performed the wrong action in 4.9% of practice trials. RT data over the course of the practice trials are shown in Supplemental Figure S3. It is evident that RTs similarly decreased for all responses and were about level (and equal for all responses) towards the end of practice trials The experiment took about 1.5 h to complete.

We processed kinematic data with customized scripts in MATLAB (version R2021b; The MathWorks, Natick, MA). We interpolated missing data points using the spring metaphor method in the inpaint_nan function (D'Errico 2021) and low-pass filtered the data using a second-order Butterworth filter with a cutoff frequency of 6 Hz. We determined movement onset as the time point at which the vectorial velocity of a marker exceeded 50 mm/s and reaction time (RT) as the time between stimulus onset and movement onset. We excluded trials in which no data were recorded and trials in which trajectories could not be reconstructed because too many data points were missing (286 trials, 2.2%). We removed trials in which participants performed the wrong prime or probe action or did not complete the respective action within the 2500 ms stimulus display interval (857 trials, 6.7%) and trials in which RT of probe actions was faster than 200 ms or slower than 1,500 ms (46 trials, 0.4%).

To analyze our data, we fitted Bayesian regression models created in Stan (http://mc-stan.org/) and accessed with the package brms version 2.16.3 (Bürkner 2017) in R (R Development Core Team 2020). In a first step, we fitted Bayesian regression models with RT as the dependent variable and the within-subject variables Hand (left, right) and Movement Direction (inward, outward), separately for prime and probe actions, to assess whether RT differed for the different movement types. For our main analysis, we fitted a model with probe RT as the dependent variable and the within-subject categorical variables Hand Repeat (repeat, switch) and Movement Direction Repeat (repeat, switch). We set orthogonal contrasts using the set_sum_contrast() command in afex version 1.0–1 (Singmann et al. 2020) and included random intercepts and slopes. We used a shifted log-normal distribution to estimate RTs and specified mildly informative priors for population-level (i.e., fixed) effects: For the intercept, the prior was a normal distribution with mean 6.4 and SD 0.5 (log-scale); for the effect of Hand Repeat, the prior was a normal distribution with mean 0.055 and SD 0.1 (log-scale). These settings reflect the prior knowledge of a main effect of hand repetition of about 70 ms, as we found in a previous study (Seegelke et al. 2021). To compare our results to findings of a previous study in which visual stimuli could be repeated between prime and probe (Seegelke et al. 2021), we additionally fitted a model with probe RT as the dependent variable, the within-subject categorical variables Hand Repeat (repeat, switch) and Movement Direction Repeat (repeat, switch), and the between-subject categorical variable Experiment (Present Experiment, Seegelke et al. 2021 Experiment 1; Seegelke et al. 2021 Experiment 2). The estimation of parameters’ posterior distributions was obtained by Hamiltonian Monte-Carlo sampling with 4 chains, 1,000 sample warmup, and 11,000 iterations and checked visually for convergence (high ESS and Rhat ≈ 1). We used the package bayestestR version 0.11.5 (Makowski et al. 2019a, b; Makowski, Ben-Shachar, & Lüdecke, 2019) to describe the parameters of our models. We report the median as a point estimate of centrality and the 95% credible interval (CI) computed based on the highest-density interval (HDI) to characterize the uncertainty related to the estimation. As an index of existence of an effect, we report the Probability of Direction (pd), representing the certainty associated with the most probable direction (positive or negative) of the effect (Makowski et al. 2019a, b). Following the recommendations of Makowski et al. (2019a, b), for interpretation we consider 95%. 97%, and 99% as reference points for the pd (pd <  = 95%: uncertain; pd > 97%: likely existing; > 99%: probably existing). In addition, as an index of the significance of a given effect, we tested whether the HDI excluded a region of practical equivalence (ROPE) of ± 0.1 effect sizes around 0. If the HDI is completely outside the ROPE, the null hypothesis for this effect is rejected. If the ROPE completely covers the HDI (i.e., all most credible values of a parameter are inside the ROPE), the null hypothesis is accepted.