The selectionist model of behavior is consistent with natural selection and its basis in physiology (Edelman, 1987; Palmer & Donahoe, 1992; Skinner, 1966). In natural selection, the characteristics of species are selected on the basis of their contribution to fitness (that is, survival). Among the characteristics of an organism that contribute greatly to fitness are special sensitivities to biologically significant things (e.g., food, water, sex) that enable survival. It is essential that these biologically significant elements, called primary values, have the effect of selecting (i.e., strengthening) those behaviors that produce them in particular environments. Thus are the two selection processes--evolution and learning--not only analogous but also linked. The primary values of an organism are selected by the evolutionary process due to their effect on survival. In turn, the behavior of an organism, for example, appropriate food-seeking behavior, which is selected based on its primary values, enables the organism to obtain food and survive.
As Donahoe et al. (1997) have argued cogently, the same value-based selectionist learning process applies whether there are discriminative stimuli which are predictive of the consequences of responses or not, and whether the conditioning is operant or classical. This generality in the learning process enables a relatively simple learning model to handle a wide range of situations.
Learning as Selection by Value
Many behavioral scientists such as ourselves characterize learning as a selection process analogous in a broad way to evolution, with the same three components of variation, selection, and retention. According to the selectionist value-based learning model, environment-behavior relations are what are selected and strengthened as a consequence of the value the behavior provides to the organism in that environment (Donahoe, Burgos, & Palmer, 1993; Donahoe & Palmer, 1994; Donahoe, Palmer, & Burgos, 1997). Through the selection process, the connections between all stimuli present and the responses that were recently done are strengthened. For example, if pressing a button is always followed by a positive value when the green light is on, the connection between pressing the button and seeing the green light will be strengthened until the button-pressing response is dominant in the presence of the green light. On the other hand, other irrelevant stimuli, such as noises, will adventitiously occur with button-pressing responses both when they are and are not followed by a positive value; those connections will sometimes be weakened and sometimes strengthened, usually producing small random connection strengths.
