Figure 3. Median RTs of the sequences are best predicted by joint entropy following sigmoid function.

(a) Calculation of joint entropy (JE) and conditional entropy (CE). The equations describe how the entropy values can be calculated directly from the joint and conditional probability (described in Fig. 2a). The JE is the entropy of the set of transitions representing previous and present locations, while the CE is the entropy of previous locations subtracted from the JE (see diagram in the left of the panel). (b) RTs as a function of CE and JE for Experiment 1. Each data point indicates the sequence entropy and the median RT of each subject of Experiment 1. (c) Explained median RTs variance by JE and CE, under sigmoid and linear fit, for Experiment 1. The variance of median RTs was best or equally well described by the sigmoid fit on the JE values, for each number of previous locations used to constitute the transitions. (d) Same as Fig. 3b, for Experiment 2. Data points indicate the sequence entropy and the median RTs of each session. (e) Same as Fig. 3c, for Experiment 2. Again, JE values under a sigmoid fit explains maximally the variance of median RTs.