, 2011).

The architecture of the SSP for the Simon task is identical to that of the Eriksen, except that the Gaussian spotlight centers on the relevant color feature of the stimulus. The color region is defined as 1 unit wide, and the remaining attention is allocated to the irrelevant spatial feature. Alternative versions of the SSP and DSTP are respectively characterized by a lack of attentional shrinking and a lack of late stimulus Selleck Galunisertib selection in compatible trials only. 80,000 Trials per experimental condition and fit cycle were simulated. Different starting points were used to ensure that the SIMPLEX gradient descent does not reach a local minimum in the parameter space. No parameter was allowed to vary between compatibility conditions. Boundary separations

were fixed across chroma levels due to the randomized design of the experiments. The non-decision time Ter and the drift rate for the response selection process in phase two urs2 in the DSTP were also fixed since variations VX-809 mouse of these parameters do not necessarily lead to Wagenmakers–Brown’s law (see Wagenmakers & Brown, 2007 and Section 2.2). To account for the experimental manipulation, parameters related to the perception/identification of the relevant stimulus attribute (prel 5 in the SSP, μrel and μss in the DSTP) were allowed to vary across chroma levels. A model variant of the SSP in which the spotlight shrinking rate rd was allowed to vary was also fitted to data. Because rd variations were very small and had a negligible impact on the fit quality (see Appendix F), rd was fixed. Best-fitting parameters

and fit statistics of the models are summarized in Table 4. Parameters are evolving as expected across chroma levels. The performance of the models can be graphically appreciated in Fig. 8. Original versions of the SSP and DSTP capture the main patterns of the data. However, the SSP overestimates the skew (i.e., tail quantile) of RT distributions SB-3CT for correct responses as chroma lessens. By contrast, the DSTP captures fairly well the variations of RT distribution shape for correct and error responses across conditions, although predicted errors are too fast for the lowest chroma level in the compatible condition (see Appendix E, for additional model analyses based on CAFs). Consequently, the DSTP provides a superior goodness-of-fit compared to the SSP, quantified by lower G2 values. The BIC also favors the DSTP, despite a higher flexibility (17 free parameters for the DSTP against 10 for the SSP). Focusing on mean RT for correct responses reveals an interesting phenomenon. Fig. 10 shows the predicted Wagenmakers–Brown’s laws from best-fitting models. As can be seen, the compatibility effect predicted by the SSP increases monotonically from 41 ms (80% chroma) to 54 ms (15% chroma), and the compatibility factor affects both the slope and the intercept of Wagenmakers–Brown’s law, consistent with our initial simulation of the model (see Section 2.1).