Figure 4. Description of RTs across training blocks by joint entropy following sigmoid curve.

A sigmoid function was fitted to the distribution joint entropy values and median RTs for each block from 5 to 22, across all volunteers of Experiment 1 (a) or all sessions of Experiment 2 (b), respectively. The purple horizontal line represents the mean of median RTs of block 4, the last block of random sequences before the varying sequences. Inset panels show the estimated values (±standard error) of these three parameters along the training blocks. The increase of “Xhalf” values across training blocks is associated to the progressive acquisition of the sequential structure. Similarly, the reduction of “Ymin” is associated to the progressive automation of very simple sequences. The subtle or absent reduction of “Ymax” values indicates that the stimulus-action mapping was (mostly) learned before the blocks expressed on the analysis.