Reaction time signatures of discriminative processes: Differential effects of stimulus similarity and incentive

Abstract

In three experiments with pigeons, the similarity of unreinforced test stimuli to a reinforced stimulus and the frequency of reinforcement associated with a stimulus were varied. The stimulus on each trial was a small spot that appeared in different hues or, in Experiment 3, different forms. Differential response frequency and reaction time (RT) patterns emerged: Changes in similarity affected the percentage of stimuli responded to but left the shape of RT distributions about the same, whereas changes in reinforcement shifted RT distributions but had little effect on the percentage of responses. When the similarity and reinforcement variables were applied to the same stimuli (Experiment 2), their effects were largely independent. A generalization procedure (Experiment 3) replicated the similarity effects of the initial discrimination procedure. The RT distributions were modeled by a diffusion process, and implications for a memory-instance model were suggested.

A signal detection analysis of contingency data

“…Collins & Shanks, 2002;, 2004. That is, the observer's trial prediction responses closely match the actual programmed contingency between the cue and the outcome.…”

“…As the value of the theory was established for understanding simple detection, its application expanded within psychophysics to more complex tasks such as discrimination and identification. SDT has been influential outside psychophysics as well, providing a conceptual framework for understanding many findings in diverse areas such as recognition memory, medical decision making, pigeon choice behavior, and the placebo effect (e.g., Allan & Siegel, 2002;Blough, 2004;Swets, 1996).…”

“…In some contingency judgment experiments, observers have been asked to predict the outcome on each trial (e.g., , 2004 in addition to rating the relationship between the cue and the outcome at the end of a block of trials. On each trial, the cue is either presented (C) or is not presented (~C), the observer predicts whether the outcome will be presented (Y) or will not be…”

Explicit melioration by a neural diffusion model

“…Since we propose to model performance in typical VR and VI conditioning tasks (tasks in which no signal to respond is given) by restarting the DDM from 0 after every response, the same first-passage time distributions of the DDM serve as IRT predictions without any modifications (cf. a similar approach in Blough, 2004). These IRT predictions must be addressed before an adaptive DDM can be considered a plausible behavioral model of operant conditioning data.…”

“…Response times and inter-response times are a valuable dependent variable in such tasks that can help to elucidate the mechanisms underlying choice. Indeed, although choice-RT and IRT data in concurrent tasks seem to have received less attention than response proportions in the animal behavior literature, RT and IRT data have occasionally been used to distinguish between alternative models of operant conditioning (e.g., Blough, 2004; Davison, 2004). Data from both of these articles included long-tailed RT/IRT distributions that appear log-normal or ex-Gaussian — approximately the shape predicted by the DDM.…”