Rhesus monkeys were trained in a two-lever drug-discrimination procedure with several discrete trials per session. During training sessions, either a sham injection or a subcutaneous injection of the training drug was administered ten minutes prior to each trial. During each trial, completion of 100 consecutive responses on the lever appropriate for the animal's pharmacological condition (e.g., left for sham, right for drug) resulted in the delivery of three grams of food. Training sessions consisted of from zero to four sham trials that preceded two consecutive drug trials. The number of sham trials varied unsystematically to preclude discrimination of the drug trials on the basis of the number of preceding trials. Discriminations were established with each of the training drugs employed (codeine, methohexital, and ketamine). During dose-effect evaluations of the training drug or other drugs (test sessions), a progressively large dose of drug was injected prior to each trial and 100 consecutive responses on either the sham- or drug-appropriate lever resulted in the delivery of food. Test sessions continued until either drug-lever responding or a marked suppression in the rate of responding occurred. Thus, a cumulative dose-effect curve for each drug was generated within a single session. Preliminary findings suggest that the pattern of cross-drug generalization generated by this cumulative-dosing procedure is similar to that obtained with procedures that evaluate only a single dose of drug per session.