One of the methodological challenges facing any behavioral researcher is the possibility that the experimental subjects may show biases or preferences that cannot be accounted for by the study's programmed independent variables. Bias is a well-known phenomenon in behavioral studies of choice; in Baum's formulation of the generalized matching law (Baum, 1974) it is represented by a free parameter. However, it has not been a topic for systematic study in choice experiments, and outside of the

more »

... rature on matching, consistent bias for one operant response over another has hardly been mentioned by experimental behavior analysts. (This type of bias will be referred to in the present paper as "operant bias" rather than "response bias" in order to avoid confusion with the common cognitive usage of the latter term, which carries a different meaning.) Operant bias by its very nature is easy to observe but almost impossible to analyze quantitatively, let alone categorize or predict. Operant bias is not under the experimenter's control, and its existence is