Combining Speed and Accuracy in Cognitive Psychology: Is the Inverse Efficiency Score (IES) a Better Dependent Variable than the Mean Reaction Time (RT) and the Percentage Of Errors (PE)?

Bruyer, Raymond; Brysbaert, Marc · 2011 · OpenAlex-citations

DOI: 10.5334/pb-51-1-5

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 evaluates the utility of the Inverse Efficiency Score (IES) as a combined dependent variable in cognitive psychology experiments, which typically yield separate measures for reaction time (RT) and percentage of errors (PE). The authors address the interpretive difficulties arising from analyzing RT and PE separately, particularly when results indicate a speed-accuracy trade-off or divergent effects. IES, defined as RT divided by the proportion of correct responses, was proposed by Townsend and Ashby (1978, 1983) to integrate speed and accuracy into a single metric. The study aims to determine whether IES provides a more robust summary of experimental findings than analyzing the two variables independently. The authors analyzed existing datasets from various cognitive psychology studies, including face perception tasks and lexical decision experiments. A primary case study utilized data from the French Lexicon Project, involving lexical decisions for 38,335 French words assessed by 975 participants. The researchers compared statistical outcomes—specifically variance explained and significance of effects—between standard RT/PE analyses and IES analyses. They examined scenarios where IES might clarify data patterns, such as in emotional face morphing tasks, and scenarios where it might distort results, such as in studies with high error rates or speed-accuracy trade-offs. The findings reveal that while IES can occasionally provide a clearer picture of data trends, it frequently increases data variability to a degree that undermines its statistical utility. In the French Lexicon Project, although the word frequency effect appeared visually stronger in IES plots, the variance explained by frequency was significantly lower for IES ($R^2 = .12$) compared to RT ($R^2 = .33$). This discrepancy persisted even when restricting analysis to words with high accuracy rates. The authors identified that IES becomes problematic when the proportion of correct responses falls below 0.90, as low accuracy rates introduce instability into RT estimates and disproportionately inflate the IES value. Furthermore, the transformation can create spurious significant effects or mask existing ones, leading to potential Type I errors, particularly in interaction terms of multivariable experiments. The study concludes that relying solely on IES is inadvisable. The measure increases variability and reduces statistical power, especially in studies with small sample sizes per condition or high error rates. The authors recommend that IES should only be used when error rates are low and there is a high positive correlation between RT and PE. Even then, researchers should calculate and report RT and PE separately to ensure that IES results align with the underlying data. The paper advocates for caution against the "blind" use of IES, suggesting that further research is needed to identify potentially superior combinations of speed and accuracy metrics.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

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

• cognitive capacity variation