Models Rule, OK? A Reply to Fetzer

How do logically-untrained individuals make deductions? A prevalent view in thepsychology of reasoning is that they rely on tacit rules of inference akin to those ofa formal logic. In Deduction (Johnson-Laird and Byrne, 1991), we argued insteadthat the untrained mind is not equipped with formal rules of inference, but relieson the general semantic principle of validity: A deduction is valid if the conclusionmust be true given that the premises are true. Reasoners put this principle into prac-tice in the following way. They construct mental models of the situations describedby the premises, formulate a conclusion that holds in these models – if none isprovided by a helpful interlocutor – and check its validity by ensuring that it holdsin all possible models of the premises. This account has theoretical advantages. Itdovetails with other parts of mental life – perception delivers models of the world(Marr, 1982), and comprehension of discourse delivers models of what is described(Garnham and Oakhill, 1996). And it provides a unitary explanation of inferencesyielding necessary, probable, and possible conclusions. A necessary conclusionholds in all the models of the premises, a probable conclusion holds in most ofthem, and a possible conclusion holds in at least one of them. The account alsohas empirical advantages. It predicts robust phenomena. Reasoners are faster andmake fewer errors with deductions that require them to construct only one modelthan with deductions that require them to construct multiple models. And theycharacteristically err by drawing conclusions that are supported by one model ofthe premises.James H. Fetzer has been kind enough to review Deduction twice (Fetzer, 1993,1998), which is going well beyond the bounds of duty, especially as he does notseem to have a good opinion of it. We are grateful to him for the chance of replyingto his latest salvo. It is always tricky for individuals in one discipline to reviewwork in another discipline (or indeed to reply to such reviews). We hope that thisreply will help cross-disciplinary understanding.Fetzer argues that the tenability of our theory depends on the meaning of thephrase, ‘mental model’. We agree. And that is why much of our book is devoted tocharacterizing mental models and why much of our research is devoted to devel-oping computer implementations of mental models. We stress that our polemicalstance concerns only the merits of current psychological theories, not the merits of