A Comparison of Single and Multiple Diffusion Process Models of Continuous-Outcome Decision Making

Continuous-outcome decision tasks, in which responses are made on continuous scales, are increasingly used to characterize the processes involved in speeded decision making. These tasks yield entire distributions of response errors rather than just error proportions. Two diffusion process models of continuous-outcome decisions have recently been proposed, the circular diffusion model (CDM), and the spatially continuous diffusion model (SCDM), which generalize the two main classes of diffusion process models of two-choice decisions to continuous spaces. The CDM, which generalizes the two-boundary Wiener model, assumes evidence is accumulated by a two-dimensional Wiener process with a vector-valued drift rate on the interior of a disk, whose bounding circle represents the decision criterion. The SCDM, which generalizes the class of multiple accumulator models, assumes evidence is accumulated by a Gaussian random process with a function-valued drift rate to a criterion. We compared the ability of the models to account for the distributions of errors and response times on three perceptual tasks: color identification, Landolt rings, and triangle orientation, in which the difficulty of the task was manipulated by varying the exposure duration. Despite some differences in their predicted joint distributions of errors and response times, both models provided good accounts of performance in all three tasks. Our findings reinforce and extend findings from two-choice studies showing that diffusive evidence processes and not model architecture is the main determinant of successful model performance, and show the theoretical unity of diffusion models across different domains.