Differentiating Computational Roles of the Superior Colliculus during Evidence Accumulation

Vaishnavi P. Kadambi, Michael D. Nunez, Michele A. Basso

We seek to understand the role of the superior colliculus in visual evidence accumulation during decision making.The Drift Diffusion model accounts for steadily accumulated evidence while the Urgency-Gating model, UGM, accounts for some decay or "leak" as well as an urgency signal. We wish to understand which decision-making model, whether it be a steady or "leaky" model, matches true decision-making response time, choice data and neural data from the superior colliculus. However, we must first understand which models can be recovered from recorded data. To do this, a simulation study of decision making behavior is necessary. Using code provided by Evans and Hawkins (2017), we built six statistical models of the drift-diffusion process that describe experimental choice, response times, and neural population dynamics. This program provides evidence for a specific theory of decision making by simulating and fitting neural data using Bayesian sampling techniques. Using these models, we were able to simulate data by using parameter intervals and plot our simulated data. Our next step is to use the python package HDDM, which is used for the hierarchical Bayesian parameter estimation of DDM. Using this new package and the simulated response times we are evaluating the recovery of the true parameter from the posterior distributions of parameters from sequential sampling models. Our goal for the quarter is first to fit both the drift-diffusion and the Ratcliff model, the drift-diffusion model with intrinsic trial-to-trial variability. After this, we are going to try to fit the urgency-gating model. Our ultimate goal is to fit the response times we collect from behavioral experiments and see which mathematical model the data fits to, especially if it fits a "leaky" or not leaky model.