LAUDABLE GOALS, INTERESTING EXPERIMENTS, UNINTELLIGIBLE THEORIZING: A CRITICAL REVIEW OF RELATIONAL FRAME THEORY

Behavior and Philosophy, 2003 by Burgos, Jose E

. . . Dymond and Barnes predicted that if picking stimulus B1 after making one response was reinforced, a subject, without further training, would then choose the following: . . . .All four subjects performed as predicted. (p. 60, emphasis mine)

. . .generic predictions have been made with regard to the types of histories that are required for relational framing to emerge (p. 148, emphasis mine).

According to their own characterization of hypothetico-deductive theories, then, there is something quite hypothetico-deductive about RFT (and behavior analysis in general) after all. Their attempt to distance RFT from hypothetico-deductive theorizing is thus inconsistent with their characterization of theories as inductive generalizations used for prediction, unless "prediction" takes different meanings throughout their discourse.

A way to resolve the above contradiction is to clarify that an inductive generalization can be as predictive as a hypothetico-deductive theory. The difference is the way in which predictions are justified. Under an inductive methodology (now I use "methodology" in its standard philosophical sense; see Note 5), predictions are justified by appealing to past experience.12 In contrast, under a hypothetico-deductive methodology, predictions are justified as logical consequences (or "implications") of certain premises adopted as working hypotheses. A famous example is Einstein's (correct) prediction of the discrepancies between the apparent and real positions of certain stars during the solar eclipse of 1919.

If RFT were inductive in the standard philosophical sense of the term, its predictions would be justified inductively. However, the justifications of the authors' predictions seem to be anything but inductive. For instance, consider the continuation of the penultimate quotation above:

Dymond and Barnes predicted that if picking stimulus B1 after making one response was reinforced, a subject, without further training, would then choose the following:

1. C1 following "one response." This would happen because C1 and A1, and B1 and A1 were in frames of coordination and thus C1 would acquire the same function as B1 by virtue of a transfer of function through the frame of coordination.

2. B2 following "no response." This would happen because B2 was less than A1 and A1 and B1 were equivalent. Thus, B2 would acquire a response function that is less than the B1 function. (p. 60, emphases mine)

Clearly, such justifications do not have the inductive form "because we have observed many times in the past that subjects under such and such conditions have responded in such and such ways." Hence, they are not inductive generalizations, at least in the standard technical sense of the term. On the contrary, they are deductive, in that they are logical consequences. This interpretation is consistent with the way the authors introduce Part II of the book: "In Part II we attempt to explore some of the implications of Relational Frame Theory in specific domains" (p. 155, emphasis mine). The predictions in question seem to be as much "implications" of RFT as those explored in Part II. RFT does not qualify as an inductive generalization insofar as "implication" means (as it usually does) "logical consequence."