When discriminated IRT procedures have been used to determine preference relations among temporally extended operants, deviations from predictions of the matching law have been found (Hawkes and Shimp, 1974). Using a yoked-control procedure, the present study shows that keypecking in the discriminated IRT procedure has two sources of strength, that arising from the stimulus-reinforcer contingency and that arising from the response-reinforcer contingency Three out of four yoked birds autoshaped to the keylight, and all lead birds showed evidence of control by the keylight under some conditions. As any control of keypecking by the keylight, either discriminated or autoshaped, contributes to deviations from matching, the discriminated IRT procedure does not permit one to draw strong conclusions about preference relations among IRTs.