Behaviorist theorizing occurs in many forms. At the two extremes are the hypothetico-deductivism of Hull and the descriptivism of Skinner.
The status of Hullian theoretical concepts is somewhat ambiguous. Some are clearly hypothetical constructs. Others appear to be intervening variables but operate as hypothetical constructs.
Hull's hypothetico-deductive method is not like deductive systems in logic and mathematics. Instead, his postulates define concepts as hypothesized quantitative functions of empirically defined variables, and they state hypotheses as to how these concepts combine to determine behavioral variables.
The purpose of a theory, according to Hull, is to mediate the deduction of theorems about observable behavior. Theorems are experimentally tested and either add confirmation to a theory or force its revision. Theory develops through this process of deduction, test, and revision. Explanation is achieved when the description of a phenomenon is deduced as a theorem from a theory.
Skinner's theoretical concepts consist of intervening variables, such as drive, and private events, consisting of covert stimuli and responses. Although the status of the latter is not clear, they are best viewed as hypothetical constructs.
Skinner objects to theories on the grounds that they inhibit fruitful research. Instead, he advocates a descriptivism with an emphasis on discovering orderliness in behavioral data. Theory, for Skinner, consists of economical descriptions of functional relations which subsume a number of behavioral regularities.
Both Hullian and Skinnerian meta-theory are open to criticisms from a nonlinear model of the development of science. This model questions behaviorist assumptions about the continuity of theory development and the relationship of theory to data.
Despite behaviorist attempts to link concepts to the behavioral data language, the relationship of concepts to observation is underdetermined. This linkage is ultimately psychological. Within this context, behaviorist meta-theory can be viewed as a decision to specify concepts precisely enough for scientists to learn to achieve intersubjective agreement, prediction, and control while using them.