This study examined a conservation model (Allison & Mack, 1982) that predicts a linear relation between the weighted sum of two responses, autoshaped leverpressing and polydipsia, and the amount of food delivered on a variable-time schedule. Fifteen rats were assigned randomly to one of three groups. The rats in Group 1 were maintained at 80% of their free-feeding body weights. Those in Group 2 began at 100% but were allowed to lose weight during the experiment. The rats in Group 3 also began at 100%of their free-feeding weights and were maintained at this level. Each group was exposed to five conditions that delivered less food than that consumed during baseline and to one condition that delivered more food, The results did not support the conservation model. Contrary to the model, the decreasing linear relation between the individual responses, or the weighted sum of the responses, and the amount of food delivered was not found for all rats, and some rats responded more when an excessive amount of food was presented than during baseline, Conservation theory predicts how an animal will respond during schedules of reinforcement, by assuming that the responses involved can be scaled on some common dimension such as energy expenditure. It also assumes that this dimension is conserved across two sessions of the sameduration: an experimental session, in which the opportunity to performa particular response is on a schedule,anda baseline session, in which theresponses involved are freely available. For example, the total dimension apportioned to leverpressing and eating will be the same during baseline sessions, in which access to food and a lever are freely available, and experimental sessions, in which food is delivered on a specified schedule of reinforcement. This conserved dimension can be interpreted as a variable common to these responses; it is represented as a scaling parameter in the conservation models.Conservation models have successfully predicted responding during a numberof different schedules that use reciprocal contingencies. For example, there is considerable evidence supporting the model for the simple fixedratio schedule of this type (Allison, 1976;Allison, Miller, & Wozny, 1979). Such a schedule specifies that the subject must perform a fixed number of responses of one type (e.g., leverpressing) forthe opportunity to perform another type (e.g., drinking), and mustthenperform a fixed number of responses of the second type (drinking) for another chance to engage in the first, and so on.Recently, the conservation model has beenextended to account for autoshaping and schedule-induced polydip-The authors which to thank John M. Hinson for his invaluable assistancein preparing this manuscript. Reprints may be obtained from JenniferJ. Higa, Departmentof Psychology, WashingtonState University, Pullman, sia (Allison & Mack, 1982). Autoshaping wasdiscovered by Brown and Jenkins (1968). In one condition, Brown andJenkins presented, to food-deprived pigeons, a lighted key followed by access to grain on a vari...