When compared to standard logic, research in the psychology of deductive reasoning has found that people make large and systematic (i.e., non-random) errors (Manktelow 1999), which suggests that humans may be irrational (Stein 1996; Stich 1985). However, the probabilistic approach argues against this interpretation. Rather than view this behaviour as errorful, it is argued that performance may have been compared to the wrong normative standard. When compared to probability theory rather than logic, participants' reasoning may be seen in a more positive light.

The probabilistic approach contrasts with mental logic (e.g., Rips 1994) and mental model theories (e.g. Johnson-Laird and Byrne 1991) which both argue that systematic deviations from logic represent unavoidable performance errors. In both theories working memory limitations restrict people's reasoning abilities. These approaches are hard to reconcile with the high error rates seen in some tasks, for example, up to 96 percent in Wason's selection task, and the fact that everyday thought and action seems to be highly successful. How can this success be understood if peoples' reasoning system is prone to so much error? The probabilistic approach resolves this problem by adopting a different normative theory (Oaksford and Chater 1998b; Stanovich and West 2000) and by considering the role of the environment in reasoning (Anderson 1990, 1991; Chater and Oaksford 1999b; Oaksford and Chater 1998a).