Two experiments were carried out to examine the influence and limits of training on rats' preference. In the first experiment, rats' preference for barpress food versus freely available food was alternated between the two choices via interspersed training. The second experiment demonstrated that the preference could not be alternated indefinitely but eventually remained constant, with a strong free-food preference. Consumption rates during training sessions with barpressing and with free food showed that the number of pellets an animal consumed per minute increased with training. Both choice alternatives yielded similar learning curves. Free food demonstrated a slight initial advantage over and reached a higher asymptote than did the barpress. Preference during training could be predicted on the basis of which alternative was more efficient at that time.Since Jensen's (1963) initial experiment, a number of experimenters have documented and explored the "contrafreeloading" or "free food" phenomenon. Jensen found that when rats were given a choice between barpressing for food or eating it directly from a bowl of freely available food, the percentage of food obtained via the barpress increased monotonically with increases in the amount of barpress training prior to the test. Much of the research since 1963 has been directed toward examining the variables that affect whether animals prefer to barpress for food or eat "response-independent and continuously available food ," often referred to as free food (Osborne , 1977, p. 222).A range of training procedures has been utilized in preparation for the choice situation. Some evidence indicating that training procedures may contribute to the animal's preference has emerged. One such procedure was utilized by Carder and Berkowitz (1970), who gave rats 3 days of free-food training and then 6 days of barpress training. Each choice day began with 25 earned food pellets prior to the insertion of free food into the Skinner box. This procedure resulted in an overwhelming preference for food obtained through barpressing.Using the same training schedule, Tarte and Snyder (1973) replicated the results achieved by Carder and Berkowitz (1970). However, when Tarte and Snyder provided equal amounts of counterbalanced training, a clear preference for free food was observed. In his review article, Osborne (1977) suggested that animals seem to prefer barpress food only after they have received asymmetrical training. He defines asymmetrical training as the absence of prechoice training with freely Requests for reprints should be sent to: George Wolford, Psychology Department, Dartmouth College, Hanover, New Hampshire 03755. 57 available food and/or unequal pretraining experience with the choice alternatives.A number of other researchers have examined the effects of training on the choice situation. Jensen (1963) found that the percentage of food obtained via the barpress increased monotonically with increases in the amount of barpress training prior to the preference test. On the basis of...