Correct, you are just reinforcing approximations at that point. If the individual cannot perform the entire behavior, you have to reward the approximations to the end result.