When you read Ivar Lovaas' work, you're always running into statements you can't forget, even though you might wish to.
This is from Lovaas et al. (1967):
Reinforcement was given in the form of small bites of the child's food, such as pieces of Sugar Flakes, a breakfast cereal. As learning progressed, evidenced by the fact that the child might occasionally emit an approximation to the current response without prompts, food reinforcement was withheld for prompted behavior and only delivered contingent upon unprompted behavior. This step is a rather important one in the training, since continual reinforcement for prompted behavior probably would prevent a shift into imitative responding.
The use of food as reinforcement apparently "forces" the child to behave correctly. We have observed that some children would not acquire new responses unless it was impossible for them to survive without the new learning, i.e., unless their sustenance was contingent upon the new learning. Stated differently, social rewards, such as attention and approval, which effectively control the behavior of most individuals were inadequate during these early stages of learning. The child not only did not learn, but became very uncooperative and tantrumous.
Reference:
Lovaas, O.I., Freitas, L., Nelson, K., & Whalen, C. (1967). The establishment of imitation and its use for the development of complex behavior in schizophrenic children. Behaviour Research and Therapy, 5, 171-181.