Residual Information of Previous Decision Affects Evidence Accumulation in Current Decision

TL;DR

This study shows that residual information from previous decisions influences current evidence accumulation, biasing perceptual choices, and improves modeling accuracy when incorporated into decision-making models.

Contribution

It introduces a novel hypothesis that residual decision-related neural activity from previous trials affects current decisions, supported by behavioral experiments and model modifications.

Findings

Residual decision information increases current decision accuracy.

Modified drift-diffusion model better fits behavioral data.

Previous decision activity influences evidence accumulation in current decision.

Abstract

Bias in perceptual decisions comes to pass when the advance knowledge colludes with the current sensory evidence in support of the final choice. The literature on decision making suggests two main hypotheses to account for this kind of bias: internal bias signals are derived from (a) the residual of motor response-related signals, and (b) the sensory information residues of the decisions that we made in the past. Beside these hypotheses, a credible hypothesis proposed by this study to explain the cause of decision biasing, suggests that the decision-related neuron can make use of the residual information of the previous decision for the current decision. We demonstrate the validity of this assumption, first by performing behavioral experiments based on the two-alternative forced-choice (TAFC) discrimination of motion direction paradigms and then, we modified the pure drift-diffusion model (DDM) based on accumulation to the bound mechanism to account for the sequential effect. In both cases, the trace of the previous trial influences the current decision. Results indicate that the probability of being correct in a current decision increases if it is in line with the previously made decision. Also, the model that keeps the previous decision information provides a better fit to the behavioral data. Our findings suggest that the state of a decision variable which is represented in the activity of decision-related neurons after crossing the bound (in the previous decision) can accumulate with the decision variable for the current decision in consecutive trials.