For more information on experiments see Shevinsky and Reinagel 2019.
For more information on modeling see Nguyen and Reinagel 2022.
For more on both, watch this recent seminar

We previously reported that rats take more time to make a sensory decision when the visual ambiguity is greater (1,2). When cued to prioritize accuracy, rats take more time to decide and are more accurate for any given stimulus difficulty. In these respects rat decision-making resembles that reported for monkeys and humans. But among the interleaved trials within a block, for rats the probability of their response being correct is higher in the trials with later decisions. Primates, however, are widely reported to make more errors in later responses. The Diffusion to Bound Model in its most basic form predicts neither of these. The "collapsing bound" variant of the DTB model elegantly explains the primate anomaly but does not generalize well to explaining the rat data.

Task differences between our experiments and the primate literature had to be ruled out. We have now confirmed that humans behave in our task as previously described by others: with increasing errors as trial time elapses. Thus task differences cannot explain the discrepancy between rats and primates. For example, for rats the mean reaction time (RT) is greater in the correct trials than the error trials (A), while it is the errors that have longer RT for our human subjects (B). We show that a stochastic Diffusion to Bound Model in which parameters vary from trial to trial, can qualitatively replicate both rat (C) and human (D) patterns in a single model with few free parameters.

A.Example Rat Data	B. Example Human Data
C. Rat-Like Model	D. Human-Like Model

There are many distinct parameter dimensions that can change the model from rat-like to human-like decision patterns, however. To distinguish among these alternatives we are currently developing a hybrid Diffusion-to-Bound model that attributes fluctuations in parameters to the predictable effects of recent reward history, recent behavior, static biases, and motivational arousal, all in the context of neural noise.

Numerically fitting this model to our behavioral data sets from both rats and humans provides a parsimonious account of how decision making differs algorithmically across species and individuals. The model parameters are directly relatable to neural candidates where trial-by-trial variation in task-related activity is predicted to explain trial-by-trial fluctuations in decision behavior.

(1) Speed and Accuracy of Visual Motion Discrimination by Rats (2013)
(2) Speed and Accuracy of Visual Image Discrimination by Rats (2013).