Abstract

The diagnostic evidence model gives a computational account of how people classify items in single categories and in category combinations (complex categories formed by combining two or more single categories). This model sets out to explain generativity in category combination (the fact that people can classify items in new category combinations even if they have never seen any examples of those combinations). The model also aims to explain context effects such as overextension in category combination. In an experiment people learned to identify imaginary diseases from artificiallyconstructed patient descriptions, and then classified new patient descriptions into combinations of those disease categories. The model accurately predicted people's classification scores for patient descriptions in these disease combinations, requiring no free parameters to fit the experimental data. The experiment showed that both generativity and overextension can occur in combinations of artificially-constructed disease categories, and confirmed the model's predictions about when overextension and generativity will occur.