The Stochastic Early Reaction, Inhibition, and Late Action (SERIA) Model for Antisaccades

The antisaccade task is a classic paradigm used to study the voluntary control of eye movements. It requires participants to suppress a reactive eye movement to a visual target and to concurrently initiate a saccade in the opposite direction. Although several models have been proposed to explain error rates and reaction times in this task, no formal model comparison has yet been performed. Here, we describe a Bayesian modeling approach for the antisaccade task that allows us to formally compare different models on the basis of their model evidence. First, we provide a formal likelihood function of actions (prosaccades or antisaccades) and reactions times based on a recently published model. Second, we introduce the Stochastic Early Reaction, Inhibition, and late Action model (SERIA), a novel model that postulates two different types of mechanisms that interact in the antisaccade task: a race-to-threshold decision process and a binary, time-insensitive decision process. Third, we apply these models to a data set from an experiment with three mixed blocks of pro- and antisaccade trials. Bayesian model comparison demonstrates that the SERIA model explains the data better than competing models that are based only on race-to-threshold processes. Moreover, we show that the race-to-threshold decision processes postulated by the SERIA model are, to a large extent, insensitive to the cue presented on a single trial. Finally, we use the same inversion technique to infer upon model parameters and demonstrate that changes in reaction time and error rate due to the probability of a trial type (prosaccade or antisaccade) are explained mostly by faster or slower inhibition and the probability of generating late voluntary prosaccades. Author summary One widely replicated finding in schizophrenia research is that patients tend to make more errors in the antisaccade task, a psychometric paradigm in which participants are required to look in the opposite direction of a visual cue. This deficit has been suggested to be an endophenotype of schizophrenia, as first order relatives of patients tend to show similar but milder deficits. Currently, most statistical models applied to experimental findings in this task are limited to fit average reaction times and error rates. Here, we propose a novel statistical model that fits experimental data from the antisaccade task beyond summary statistics. For this, we suggest that antisaccades are the result of several competing decision processes that interact nonlinearly with one another. Applying this model to a relatively large experimental data set, we show that mean reaction times and error rates do not fully reflect the complexity of the processes that are likely to underlie experimental findings. In the future, our model could help to understand the nature of the deficits observed in schizophrenia by providing a statistical tool to study the biological processes from which they arise.