Strong inference methods

The method of strong inference (Platt Reference Platt1964) rejects the traditional null-hypothesis model and instead tests between multiple, viable scientific hypotheses about a phenomenon. One can then devise experiments that adjudicate between these rival hypotheses (by identifying the situations under which the hypotheses made different predictions and conducting research on these situations). This process can be recycled as many times as needed to definitively exclude one or more of the alternate hypotheses and to evaluate subsequent refinements of remaining hypotheses. Using strong inference methods generates multiple empirical norms and would go a very long way towards avoiding the problems associated with prescriptive normativism. Strong inference methods also appear to occupy the top, left corner of E&E's Figure 1 (see target article), a position which they claim is impossible.

The idea of strong inference is hardly new (Anderson & Shanteau Reference Anderson and Shanteau1977), so why is it not a go-to option for avoiding the pitfalls of prescriptive normativism? One likely reason is the persistence of null-hypothesis statistical testing (NHST) ideology (Krueger Reference Krueger2001; Loftus Reference Loftus1996; Nickerson Reference Nickerson2000; S. Sun et al. Reference Sun, Pan and Wang2010). This use of NHST is not only entrenched in textbooks and pedagogy, but it also appears to be a product of what comes more easily to the human mind. As F. Scott Fitzgerald (Reference Fitzgerald1936) noted, “the test of a first rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to function” (p.44). We need our science to be inspired by first-rate intelligences, not by second-rate null-hypothesis testing practices.

A theory of “thinking”

The topic of this target article is described as the study of human thinking. “Thinking” is used in the title, in the abstract, throughout the text, and in the name of E&E's Hypothetical Thinking Theory model. It should not be news to anyone that human cognitive processes, including language, memory, reasoning, decision making, vision, hearing, attention, and problem solving, are not all accomplished by a single thinking system, and therefore are neither expected nor required to follow a single normative standard. It seems odd, therefore, to attempt the construction of a normative system that encompasses such a broad (and vague) realm of mental activities – it is almost sure to be impossible. Just as “living” entails the coordinated activities of many functionally specialized systems (e.g., for breathing, eating, digestion, excretion, etc.), the phenomena of “thinking” entail functionally specialized and coordinated systems such as those listed previously. In fact, thinking as a unitary process is contradicted by the neurosciences (e.g., by localization of function), computer sciences (e.g., in dealing with the frame problem), philosophy (e.g., see issues of indeterminacy), and biology (e.g., based on multiple adaptations designed to address multiple, discrete selection pressures).

Domain specificity

It would be much more productive, we think, to have serious discussions about the domain-specificity of rationality (and associated empirical norms). Issues such as domain-specificity, modularity of the human mind, and the principles by which the domains should be discerned are weighty and contentious (Barrett & Kurzban Reference Barrett and Kurzban2006; Carruthers Reference Carruthers and Stainton2006; Hagen Reference Hagen and Buss2005; Samuels Reference Samuels, Carruthers and Chamberlain2000). For the purposes of this commentary, however, it is sufficient to note that any degree of functional modularity (e.g., vision versus language versus reasoning) is troublesome for the monolithic E&E account of “thinking.” Even basic views of the functional modularity of the mind (e.g., Fodor Reference Fodor1983) create problems for any completely general model, such as Hypothetical Thinking Theory.

More ambitious views of “massive modularity” (e.g., Pinker Reference Pinker1997; Reference Pinker2002; Tooby & Cosmides Reference Tooby, Cosmides, Barkow, Cosmides and Tooby1992) are often based on extensive considerations of evolutionary principles, which are ill-represented by E&E. For example, the work of Oaksford and Chater, which is interesting in many respects, is presented as evolutionarily informed and even adaptationist because they argue “behavior is rational to the degree that it tends to increase inclusive fitness” (Oaksford & Chater Reference Oaksford and Chater2007, p. 26, cited in E&E's target article, sect. 5.1, para. 3). Such use of a vague “inclusive fitness” rationale does not qualify Oaksford and Chater as “adaptationist leaning[]” (see sect. 5.3, para. 1), just as acknowledging that there are neurological underpinnings to the mind does not make them or anyone else neuroscientists.

In summary, rather than a generic model of thinking (dual process or otherwise), we need models of human cognition that incorporate the numerous functionally distinct domains of human cognition. There are many models of how the modularity of mind is structured, and this situation should be taken advantage of by utilizing strong inference methods.