In cognitive psychology, understanding the role of inhibition is crucial for explaining how individuals resolve conflict and make decisions. The Leaky Competing Accumulators (LCA), a widely used model within the evidence accumulation framework, posits that inhibition plays a vital role between two leaky accumulators. However, previous attempts to simultaneously recover both leakage and mutual inhibition parameters have proven challenging. Given the ongoing debate regarding the role of leakage in cognitive processes, this study investigates the recoverability of inhibition after removing the leakage parameter from the LCA, resulting in the Competing Accumulators (CA) model.

We use Mixed Neural Likelihood Estimation (MNLE) to fit the CA model to reaction time (RT) and choice data, obtaining Maximum A Posteriori (MAP) estimates of the model parameters. Our results demonstrate that inhibition can be effectively recovered for CA, which provides accurate predictions for behavioral data in both brightness detection tasks (2AFT and 4AFT) and conflict tasks. Notably, in conflict tasks, the data show that evidence in the congruent condition is stronger than in the incongruent condition, with active inhibition exerted by controlled processes on automatic processes in the incongruent condition. This work advances our understanding of decision-making processes, particularly the quantification of cognitive inhibition, and has important implications for future research in cognitive psychometrics.