Molar versus local reinforcement probability as determinants of stimulus value.