Postscript: Still in search of a good theory of reasoning--Rejoinder to Barrouillet, Gauffroy, and Lecas (2008).

Abstract

In Barrouillet, Gauffroy, and Lecas's postscript (see record 2008-09896-014) to the current authors original comment (see record 2008-09896-009) on Barrouillet, Gauffroy, and Lecas's original article (see record 2008-09896-008), they made four clearly argued points. First, they argued that they had provided a clear rationale for truth value gaps. This misses the point of what a computational-level explanation means. Such an explanation would answer why the hypothetical processes are highly rational, why these initial models are the most plausible and relevant, and what constitutes an error according to this theory. Second, they argued that a putative definitional statement defeats our argument concerning negated conditionals. Indeed, the idea that logic can be viewed as a limiting case of probability when probabilities are either zero or one is a commonplace (Dale, 1976; Hajek, 2001). Consequently, we do not see such cases as placing much of a restriction on the probabilistic approach. Third, they argued that a conditional does not fall into a truth value gap when its antecedent is false. To avoid this happening, they endorsed the Ramsey test as a process to evaluate conditionals (Barrouillet et al., 2008, p. 772). We acknowledged this in our comment's footnote 3 (Oberauer & Oaksford, 2008, p. 775) and also pointed out that the Ramsey test is difficult to reconcile with the mental model framework. Finally, they argued that a good theory should explain the data, especially their new developmental data. While wholeheartedly agreeing with this sentiment, we suggest that good theories must meet more rigorous standards. The explanations they offer must not be ad hoc, and a good theory of reasoning, in particular, must provide an answer as to why the processes invoked are rational.