Interactive race model: As suggested by the reviewers, we now include a comparison to the interactive race model proposed by Boucher and colleagues (2007).
For the reactive task (see Figure 1 ), we compared three models, a dependent process model (C ), a traditional independent race model (D ), and interactive race model (E ).
The interactive race model is composed of a single absorbing bound (a), marking the distance from 0 that the execution process must travel in order to elicit a response.
We now implement a model with similar properties as the interactive race model in our diffusion process framework (see Experimental Procedures).
(6) d θ b = υ b d t σ d W (7) θ b (S S D ) = θ e (t r ) = 0 Interactive race model: We modified a model originally proposed by Boucher et al., 2007 in which a stochastically drifting process executes a response upon reaching a threshold.
Indeed, our DPM is conceptually similar to the interactive race model (Boucher et al., 2007), whereby a response is suppressed by the onset of a braking signal that directly inhibits the execution process, preventing it from reaching the action threshold.
The interactive race and DPMs make fundamentally different assumptions about the nature of the interaction between the execution and braking processes.
In contrast to, for instance, interactive race [33] or serial (exhaustive) processing architectures (e.g., Feature Integration Theory [34]), the DWA assumes that selective (focal) attention operates on a map of cross-dimensionally integrated (summed) feature contrast signals, with the most active location on this overall-saliency map determining the allocation of focal attention.
It is also very similar in structure to the interactive race model that we now include in the comparisons.
Overall the dependent process model (DPM; green) explains speed and accuracy in the behavioral data (gray histograms in A, black line in B ) better than the independent race model (Ind-RM; red) and interactive race model (Int-RM; blue).
