Two groups of male rats were placed on a feeding regimen in which a fat/protein diet and a carbohydrate/protein diet were available ad lib. Naltrexone was infused via osmotic minipumps either at the time the diets were introduced or after one week of adaptation to the diets. In rats adapted to the diets, naltrexone caused a decrease in the intakes of fat/protein and carbohydrate/protein diets. Relative preferences for the two diets were generally unchanged. In contrast, when naltrexone was infused at the time of introduction of the diets, a polarization phenomenon was observed: rats tended to consume nearly all of their daily calories from either one diet or the other. Six rats (out of 10) showed a stronger preference for the carbohydrate/protein diet than did any of the saline-treated rats, while 3 showed a stronger preference for the fat/protein diet than did any of the saline-treated rats. Thus, the effect was not diet- or macronutrient-specific. These preferences became significantly less extreme after termination of naltrexone infusions. Conditioned aversions and naltrexone-induced reductions in exploratory behavior are discussed as potential explanations for this polarization effect. These results indicate that naltrexone has differential effects on the development versus the maintenance of diet preferences. Further, they emphasize the importance of examining individual differences as well as baseline preferences in studies on the control of intake and diet selection.