Arbitrary conditional discriminations are difficult to obtain in non-human subjects. This study evaluates the efficacy of sample stimulus shaping procedure. Two experiments were carried out. In Experiment I two capuchin monkeys, M12 and M15, acquired quickly the discriminations, and a third subject, M09, required longer training. In Experiment II, the procedure was modified and a new training with M09 was carried out, and M09 performance improved. However, lack of stimulus control coherence was found, precluding the conclusion of the shaping process. Key procedural variables in stimulus control research and intervention with non-humans and people with developmental disabilities are discussed.