Prior (left) and Posterior (right) distributions of each model’s parameters.

<p>Bayes factors comparisons can be problematic if the posterior is radically outside the central coverage of the prior, though <i>c</i>.<i>f</i>. [<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0131984#pone.0131984.ref036" target="_blank">36</a>]. In this case, this does not appear to be an issue. 1b. The cup preference model does not strongly suggest than any cup has above-chance preference (blue line). Error bars are 95% CI. 2b. The Cue-Mixing Model fits fits a mixture that is about 50% of each cued location, which is needed to correctly predict that mean error will be at about the chance expectation. 3b. The Cue-Combination model fits a higher spread than the Cue-Mixing Model.