An Empirical Exploration of Fast and Slow Errors at Group and Subject Levels in Speeded Decision-making

Abstract

Qualitative benchmarks in empirical data constrain the theories and models researchers generate about any given phenomena. In the field of decision-making, two key qualitative benchmarks are fast and slow errors. To account for these benchmarks researchers have added two complicated between-trial variability mechanisms to prominent decision making models, but the reliability of these benchmarks is yet to be determined. Our study aimed to provide an assessment of the reliability of fast and slow errors at both the group and individual subject level. We used archival and newly collected data and analysed these data under Bayesian frameworks. We found inconsistencies in the benchmarks between experiments and between subjects within the same experiment. Our work not only questions the reliability of fast and slow errors, but also the minimum requirements for a pattern of data to be classed as a qualitative benchmark that forms the basis for psychological theory.

Hidden Markov Models of Evidence Accumulation in Speeded Decision Tasks

“…The results demonstrate that the model shows a good fit to the data and is able to capture most of the patterns in the data. However, the model also showed a slight systematic misfit because the predicted error responses under the controlled state were slower than that of the data (a typical example of a phenomenon known as fast errors; Tillman and Evans, 2020). The results suggested quite strong consistency between participants in terms of the speed-accuracy trade-off -suggesting that the inaccessibility region (i.e., a region of speed of accumulation and response caution which "cannot be accessed", resulting in switching between two discrete states) predicted by the phase transition model could be qualitatively similar across participants (see Fig.…”