Fig 2. Subjects made decisions more quickly as they progressed through the experiments and during trials when they used the object-based rather than the feature-based model.

(A-D) Plotted is the average z-scored RT (across all subjects) as a function of the trial number within a session of an experiment separately for trials in which the object-based or feature-based model was used. Panels (A-D) correspond to Experiments 1 to 4, respectively, and shaded areas indicate s.e.m. The gray dotted line shows the time point during a session (in terms of trial number) at which the experiment was paused for a short rest. In Experiments 3 and 4, a different set of objects was introduced at this point. (E-H) The same as in (A-D) but showing the average actual RT across all subjects. Overall, subjects made decisions 300-400ms faster at the end relative to the beginning of the experiments. Nevertheless, they were consistently faster on trials when they adopted an object-based rather than feature-based strategy.