Mental Models, Deductive Reasoning, and the Brain

This chapter considers the two main approaches to deductive thinking: theories based on formal rules of inference postulate that deduction is a syntactic process akin to a logical proof; the mental model theory postulates that it is a semantic process akin to the search for counterexamples. Experimental evidence bears out the predictions of the model theory: the more models needed for a deduction, the harder it is; erroneous conclusions are consistent with the premises; and ^general knowledge affects the process of search. Recent neurological evidence bears out, as the model theory predicts, a significant involvement of the right hemisphere in reasoning. If deduction is a purely verbal process then it will not be affected by damage to the right hemisphere. It is affected by such damage. It is not a purely verbal process. This argument is an example of a valid deduction: Its conclusion must be true if its premises are true. (They may not be, of course.) Deductive reasoning is under intensive investigation by cognitive scientists, and more is known about it than about any other variety of thinking. The aim of this chapter is to explain its nature and to relate it to the brain. "The cerebral organization of thinking has no history whatsoever," Luria remarked (1973, 323); and Fodor (1983,' 119) suggested that nothing can be known about the topic, because thinking does not depend on separate "informationally encapsulated" modules (but cf. Shallice, 1988, 271). Many regions of the brain are likely to underlie it, but as we shall see, a start has been made on the neuropsychology of reasoning. Many cognitive scientists have argued that deductive reasoning depends on formal rules of inference like those-'of a logical calculus, and that these unconscious PHILIP N. JOHNSON-LAIRD Department of Psychology, Princeton University, Princeton, N.J. rules are used to derive conclusions from the representations of premises. These "prepositional" representations are syntactically structured strings of symbols in a mental language, and the chain of deductive steps is supposedly analogous to a logical proof (see, e.g., the theories of Braine, Reiser, and Rumain, 1984; Osherson, 1974-1976; Rips, 1983). An alternative account postulates a central role for mental models. This account does not reject prepositional representations, but it treats them as the input to a process that constructs a mental model corresponding to the situation described by the verbal discourse. The process of deduction—as well as induction and creation (Johnson-Laird, 1993) —is carried out on such models rather than on prepositional representations. Models are the natural way in which the human mind constructs reality, conceives alternatives to it, and searches out the consequences of assumptions. They are, as Craik (1943) proposed, the medium of thought. But what is a mental model? The underlying idea is that the understanding of discourse leads to a model of the relevant situation akin to one created by perceiving or imagining events instead of merely being told about them (Johnson-Laird, 1970). Experimental studies have indeed found evidence for both initial propositional representations and mental models (see e.g., Johnson-Laird, 1983; van Dijk and Kintsch, 1983; Garnham, 1987). The same idea has led to the model theory of deductive reasoning. The theory was not cut from whole cloth, but was gradually extended from one domain to another. From a logical standpoint, there are at least four main domains of deduction: 1. Relational inferences based on the logical properties of such relations as greater than, on the right of, and after. 2. Propositional inferences based on negation and on such connectives as if, or, and and. JOHNSON-LAIRD: MENTAL MODELS 999