The mental model theory postulates that reasoners build models of the situations described in premises. A conclusion is possible if it holds in at least one model of the premises; it is probable if it holds in most of the models; and it is necessary if it holds in all of the models. The theory also postulates that reasoners represent as little information as possible in explicit models and, in particular, that they represent only information about what is true. One (...) unexpected consequence of this assumption is that there should be a category of illusory inferences: they will have conclusions that seem obvious, but that are wholly erroneous. Experiment 1 established the existence of such illusory inferences about probabilities. Overall, 88% of the intelligent adult subjects chose as more probable an outcome that was impossible for at least one of the illusory problems. Experiment 2 corroborated the phenomenon and showed that illusory inferences include a wide variety of problems. Finally, the paper argues that current theories based on formal rules of inference are unlikely to be able to explain the illusions. (shrink)

It was hypothesized that the perceived irrelevance of the proposition “Some X are not Y” is a factor contributing to the difficulty of nearly all the determinate syllogisms classed as multiple model by Johnson-Laird and Byrne, according to mental models theory. Experiment 1 supported this hypothesis by showing that subjects frequently correctly evaluate valid “Some … not” conclusions but rarely produce them, even when they have evaluated them elsewhere. Explanations of these findings based on the use of superficial strategies were (...) ruled out. Experiment 2 further supported the hypothesis by showing that performance increased across the no-conclusion, multiple-choice, and evaluation task formats, and that this effect generalized to problems containing the quantifier “only”. However, the initial hypothesis was rejected in light of Experiment 3, which found no difference between multiple-choice and no-conclusion formats when the number of allowable conclusions was controlled for. Nevertheless, superior performance remained in the evaluation format, and it is suggested that offered conclusions may be used as a goal for the reasoning process. This interpretation is supported by the finding that subjects appear to search only for alternative conclusions that maintain the subject-predicate structure of the offered conclusion. (shrink)

A general two-stage theory of human inference is proposed. A distinction is drawn between heuristic processes which select items of task information as ‘relevant’, and analytic processes which operate on the selected items to generate inferences or judgements. These two stages are illustrated in a selective review of work on both deductive and statistical reasoning. Factors identified as contributing to heuristic selection include perceptual salience, linguistic suppositions and semantic associations. Analytic processes are considered to be context dependent: people reason from (...) experience, not from inference rules. The paper includes discussion of the theory in comparison with other contemporary theories of human inference, and in relation to the current debate about human rationality. (shrink)