The Discrimination Generalisation Task (DGT) was developed as 1 of a set of tests for use in the comparative study of intelligent behaviour. In this task, S learns a simple discrimination of an X from 4 Os, and then receives 10 test problems that call for generalisation and learning set. The performance of rats, rabbits and cats was compared in the expectation that cats would be superiour to rabbits and rabbits superior to rats. This expectation was realised. Rabbits and cats were highly consistent in the ordering of problems by difficulty, both within and between species. Rats were much less consistent probably due to the fact that their spread of scores was much less.