The use of a teaching machine in the assessment of mental function has been described by Gedye and his colleagues. This report concerns its use in testing 56 elderly patients suffering from functional psychiatric disorders or dementia. After a diagnostic interview patients were tested on the teaching machine as well as on the Mill Hill, Progressive Matrices, Paired Associate Learning Test and Digit Copying Test. After a fixed period all tests were repeated and clinical progress rated. Results showed a significant correlation (P>0.01) between teaching-machine scores and those on the standard psychological tests. The correlation between the machine scores on the two occasions was high (r=0.58) but this was complicated by practice effects and by the low ‘ceiling’ of the test. It is concluded that the test provided a relatively reliable and valid measure of cognitive function but that its value would be increased by providing a parallel form and by extending its level of difficulty.