In a shock-escape T-maze task, rats were required to learn drug discriminations involving various doses of pentobarbital, phencyclidine and ketamine. Repeated training sessions were administered. Typically a right turn in the maze was required during drug sessions and a left turn was required during alternate no drug sessions. Results were as follows: 1. Linear dose effect curves relating discriminability (sessions to criterion) to dosage were obtained with all three drugs. 2. Phencyclidine and ketamine produced much more 'response randomization' than did pentobarbital at doses matched for discriminability. 3. Phencyclidine and ketamine tended to mimic each other's discriminable effects, but neither mimicked or was mimicked by pentobarbital. 4. Drug vs drug training demonstrated that phencyclidine and ketamine differed discriminably from each other.