Functional cue learning under different probabilistic feedback conditions