Inhibiting Responses to Difficult Choices

The stop-signal paradigm is a widely used procedure to study response inhibition. It consists of a 2-choice response-time task (a "go" task) that is occasionally interrupted by a stop signal instructing participants to withhold their responses. The paradigm owes its popularity to the underlying race model that enables estimation of the otherwise unobservable latency of stopping. As the race model assumes a single go runner that produces the response unless it is beaten by an inhibitory stop runner, it cannot account for errors on the go task. We propose a parametric framework that extends the standard 2-runner race model to account for go errors, and hence expand the scope of the stop-signal paradigm to the study of response inhibition in the context of difficult choices. We combine our treatment of go errors with the ability to address 2 common contaminants in stop-signal data: failure to trigger the go or the stop runner. We show with simulations that applying 2-runner parametric race models to difficult choices can severely bias conclusions about response inhibition. Notably, we also show that even infrequent errors, which have been common in previous stop-signal studies, can result in underestimation of stopping latencies. We demonstrate that our framework enables researchers to study difficult-choice inhibition even in relatively small samples by applying it to novel stop-signal data with high error rates and a manipulation of task difficulty, showing that it provides an accurate characterization of behavior and precise stop estimates.