Barycentric extension of generalized matching.

In most studies of choice under concurrent schedules of reinforcement, two physically identical operanda are provided. In the "real world," however, more than two choice alternatives are often available and biases are common. This paper describes a method for studying choices among an indefinite number of alternatives when large biases are present. Twenty rats were rewarded for choosing among five operanda with reinforcers scheduled probabilistically and concurrently. Large biases were generated by differences among the operanda: two were levers and three were pigeon keys. The results showed that when reinforcer frequencies were systematically varied, an extension of Baum's (1974) Generalized Matching Model, referred to as the Barycentric Matching Model, provided an excellent description of the data, including individual bias values for each of the operanda and a single exponent indicating sensitivity to reinforcer ratios.