Review Review Mental Models and Deduction

Philip N. Johnson-Laird Dept of Psychology, Princeton University, Green Hall, Princeton, NJ 08544, USA. e-mail: phil @ princeton.edu In a civil action, an expert witness conceded two points: If a pollutant (TCE) had come from the river, it would be in the river bed. TCE was not in the river bed. The opposing lawyer claimed: The pattern is consistent with the conclusion that no TCE came from the river. The witness agreed (Ref. 1, pp. 361–362). But the lawyer had made a mistake: the evidence is not merely consistent with the conclusion. It yields a valid deduction of the conclusion, that is, if the evidence is true then the conclusion must be true too. The example shows that human reasoning is not always successful. Cognitive scientists need to understand the causes of such failures. Reasoning lies at the core of human intelligence2. And it is central to science, society, and the solution of practical problems. It starts with premises, which can be statements, perceptions or beliefs. Ideally, it yields a valid conclusion that is not explicit in the premises. But the intervening processes are mysterious. Many theorists suppose that the mind constructs syntactic representations of the logical form of assertions and applies the rules of a formal logic to them3,4. There is another possibility: reasoners could rely instead on their grasp of meaning, their general knowledge, and principles akin to those for the semantics of a logic5. They construct mental models of the premises, which represent the situation, and draw conclusions from them6,7. This article outlines the theory and its recent developments.