Abstract
We examine theories of simple choice as a race among evidence accumulation processes. We focus on the class of deterministic race models, which assume that the effects of fluctuations in the parameters of the accumulation processes between-choice trials (between-choice noise) dominate the effects of fluctuations occurring while making a choice (within-choice noise) in behavioral data (i.e., response times and choices). The latter deterministic approximation, when combined with the assumption that accumulation is linear, leads to a class of models that can be readily applied to simple-choice behavior because they are computationally tractable. We develop a new and mathematically simple exemplar within the class of linear deterministic models, the Lognormal race (LNR). We then examine how the LNR, and another widely applied linear deterministic model, Brown and Heathcote’s (2008) LBA, account for a range of benchmark simple-choice effects in lexical-decision task data reported by Wagenmakers et al. (2008). Our results indicate that the LNR provides an accurate description of this data. Although the LBA model provides a slightly better account, both models support similar psychological conclusions.
Highlights
Humans and other organisms often have to respond to stimuli under time pressure that requires them to make choices in a few seconds or less
In the first variation rates for all accumulators were guaranteed to be positive on every trial by sampling them from univariate normal distributions truncated below at zero, and in the second at least one sampled rate was guaranteed to be positive on every trial by sampling the rates from a truncated multivariate normal distribution1. All of these linear ballistic accumulator (LBA) models produced similar fits and parameter estimates, so we report fits from the second variant for two reasons: (1) it is most directly comparable to the Lognormal race (LNR) model, which has positive rates for all accumulators on every trial and (2) it is of interest in itself as an alternative to the LNR for solving the issue of potential non-responding in the original LBA model
Recall that the LNR model we fit is simplified in two senses; it assumes no variability in residual time, and it assumes inputs and distances are uncorrelated across accumulators
Summary
Reviewed by: Patrick Simen, Oberlin College, USA KongFatt Wong-Lin, University of Ulster, Northern Ireland Rani Moran, Tel Aviv University, Israel. We focus on the class of deterministic race models, which assume that the effects of fluctuations in the parameters of the accumulation processes between-choice trials (between-choice noise) dominate the effects of fluctuations occurring while making a choice (within-choice noise) in behavioral data (i.e., response times and choices). The latter deterministic approximation, when combined with the assumption that accumulation is linear, leads to a class of models that can be readily applied to simple-choice behavior because they are computationally tractable.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
