Stimulus-fading (SF) involves the presentation of increasingly difficult sets of discriminations arrayed along some continuum, where progress from one to the next depends on mastering the current discrimination. Response accuracy is correlated with two reinforcement contingencies: (1) the trial-based presentation of a consequential stimulus (e.g., a short-duration video-clip for a correct response), and (2) the response-contingent change in discriminative stimuli (RPCIDS) produced by correct responding on the trials at a fading level. This experiment isolated the effects of both contingencies by comparing learning in SF and in yoked-control (YC) version of fading wherein video-clip reinforcers were presented independent of response accuracy, but response accuracy determined advancement across fading levels. Discriminations at all fading levels were acquired by 90% and 10% of participants in the SF and YC conditions, respectively. Although many of the YC participants did not learn all of the discriminations, they did learn some of them, which reflected the reinforcing effects of response-contingent-change in discriminative stimuli. Because this contingency occurs in all fading protocols, it could influence learning in all of them. At some fading levels in the YC condition, response accuracy was unexpectedly correlated with video-clip reinforcement: a finding that did not compromise the effect of response produced change in discriminative stimuli on discrimination learning in the YC condition. This methodological limitation might be cured by substituting a truly-random-control version of fading for the YC condition.