A behavior needs to be put on a random rewardschedule before that behavior is linked to a second split behavior, When you stop and think about that statement it makes sense.
However, the monkey's response to being cut off from the reward is very different if you start out on a purely random schedule, instead of a consistent one.