Quantifying the effects of the differential outcomes procedure in humans: A systematic review and a meta‐analysis

Abstract

We present a systematic review and a meta‐analysis comparing the differential outcomes procedure to a nondifferential outcomes procedure among clinical and nonclinical populations. Sixty distinct experiments were included in the systematic review, 43 of which were included in the meta‐analysis. We calculated pooled effect sizes for accuracy (overall accuracy, test accuracy, transfer accuracy) and acquisition outcomes (latency, errors, and trials to mastery). The meta‐analysis revealed significant medium‐to‐large effect sizes for all three accuracy measures (pooled effect size range, 0.57 to 1.30). We found relatively greater effect sizes among clinical populations (effect size = 1.04). The single‐subject experimental literature included in the systematic review was consistent with the findings from the group studies, demonstrating improvements in accuracy and speed of learning for the majority of participants. Moderator and subgroup analyses suggest that discrimination difficulty may induce relatively larger differential outcomes effects. The results indicate that the differential outcomes procedure can be a valuable addition to reinforcement‐based interventions.

Vicarious value learning by differential outcomes training: A social transfer of control methodology

“…Standardly, the differential outcomes training procedure entails a single stage of training over a number of learning trials where the objective is for participants to learn to associate different stimuli with different responses following repeated alternating presentations of each of the stimuli and all of the response options. The use of differential rewarding outcomes (for each correct stimulus-response pair) has been robustly found to speed up learning as compared to a non-differential (or common) outcomes control [3] . This is known as the differential outcomes effect .…”

“…In memory training studies there can be a big difference in performance among target groups, e.g. young infants, elderly or adult students [3] .…”

A social differential outcomes learning task: Performance, EEG, and questionnaire data

“…In order to estimate the sample size needed, an a-priori power analysis was performed with G*Power [1] using the effect size specification option “as in SPSS”, the effect size found specifically for the transfer of control procedure obtained via the meta-analysis by McCormack et al. [2] , eta squared = 0.33, power = 0.8, and alpha = 0.05 resulted in N = 20. The participants were 9 male and 11 female students at the University of Gothenburg, with ages between 20 and 44 years ( M = 27.74).…”

A Clarification of Slope and Scale

“…Descriptions of various types of meta-analyses and how they combine and analyze effect-size measures across studies are outside the scope of this paper (e.g., Cowen et al, 2017;Van den Noortgate & Onghena, 2008;Moeyaert et al, 2014;McCormack, 2019;Parker et al, 2011;Pustejovsky, 2018;Valentine et al, 2016;Wolfe et al, 2019b). However, as Wolfe et al (2019a) noted, level, variability, immediacy, overlap, consistency, and trend should be accounted for if the standards for visual analysis outlined in WWC (2017) are to be incorporated.…”