Inhibiting responses to difficult choices.

Matzke, Dóra; Curley, Samuel; Gong, Charlene Q.; Heathcote, Andrew · 2018 · Journal of Experimental Psychology General

DOI: 10.1037/xge0000525

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses a critical limitation in the stop-signal paradigm, the standard method for studying response inhibition. The traditional race model assumes a single "go" process, which fails to account for errors in choice-based tasks where participants must select between alternatives. Consequently, existing models often treat go errors as contaminants or ignore them, leading to biased estimates of stop-signal reaction time (SSRT), particularly in difficult choices or clinical populations with higher error rates. The authors propose a unified parametric framework that extends the race model to include multiple runners for different choices, thereby explicitly modeling go errors, go failures (omissions), and trigger failures (failures to initiate the stop process). The study employs a Bayesian parametric approach called BEESTS, which estimates the full distribution of SSRTs using an ex-Gaussian distribution. The authors first conducted simulation studies to demonstrate that applying standard two-runner models to difficult choices severely biases conclusions about inhibition. Specifically, they showed that even infrequent errors (approximately 2.5%) can lead to underestimations of stopping latencies if ignored. They then developed and tested the new unified model, which integrates the treatment of go errors with mechanisms to account for go and trigger failures. This framework was validated through parameter-recovery studies and applied to novel empirical data featuring a manipulation of task difficulty. The empirical application involved relatively small sample sizes with high error rates, testing the model's ability to provide precise estimates with limited data (e.g., 168 stop-signal trials per participant). The results indicate that the proposed framework provides accurate characterizations of behavior and precise SSRT estimates, even in conditions with high error rates and limited trial counts. Simulations revealed that ignoring go errors can create fictitious inhibitory differences between groups, potentially misleading researchers into attributing performance differences to inhibition deficits when they actually stem from choice accuracy or attentional failures. The unified model successfully disentangled these contaminants, showing that go failures and trigger failures have opposing effects on SSRT estimates—go failures masquerade as increased inhibitory ability, while trigger failures inflate apparent deficits. By accounting for these factors simultaneously, the model prevented the spurious inflation or deflation of SSRT estimates that occurs when these contaminants are ignored or handled inconsistently. The significance of this work lies in expanding the scope of the stop-signal paradigm to include difficult choices and populations prone to errors, such as children or those with clinical conditions like schizophrenia. The authors argue that inhibition is not a unitary construct that generalizes automatically from easy to hard tasks; therefore, accurate measurement requires models that reflect the complexity of real-world choice tasks. By providing a flexible, unified modeling framework that accounts for choice errors and various failure modes, this research offers a more complete characterization of performance in the stop-signal paradigm. This advance allows experimental psychologists to investigate response inhibition across a broader range of tasks and populations with greater precision, reducing the risk of erroneous conclusions driven by unmodeled contaminants.

Key finding

Extending the stop-signal race model to account for choice errors and trigger failures prevents severe biases in estimating stopping latencies that occur when using standard models on difficult choice tasks.

Methodology

modeling

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• gap acceptance