Modeling Statistical Learning and Response Inhibition with the Change Signal Task

The change signal task is a two alternative forced choice task with the addition of a change signal presented on 1/3 of the trials at some delay relative to the initial stimulus. The change signal indicates to participants that they should inhibit their initial choice and respond with the other alternative. It provides an opportunity to examine the cognitive mechanisms involved in statistical learning and response inhibition. Within the task, change signal delays are associated with stimulus color, and are adjusted independently with a step function to produce high and low error conditions. Observed data show a significant difference in reaction times between these two conditions. In this paper we propose a model for the change signal task that leverages existing declarative memory mechanisms in ACT-R as a surrogate for the implicit contextual learning observed in human trials. We compare the mechanisms in this model briefly to an existing neural account, and use the model to predict the consequences of cue-conditional reversal.