AbstractAn experiment was conducted to investigate the extent of caloric compensation to differential preloading among normally restrained and unrestrained female subjects. Subjects' eating was either explicitly observed or unobserved by the experimenter. Unrestrained eaters compensated appropriately for the preload, whether or not they were observed. Restrained eaters failed to compensate when unobserved, but compensated strongly when observed. A subsequent unobserved eating task revealed that restrained eaters failed to compensate for prior consumption when the observer was withdrawn. Implications for research and therapy are discussed.