Forty-seven subjects (19 elderly dements, 18 elderly depressives, 10 elderly physically ill) were given a continuous false recognition technique (FRT) in a test battery including a measure of short-term memory and the Kendrick Synonym Learning Test. Results were analysed according to signal detection theory and interpreted in terms of an encoding breakdown. Depressives were found to be characterized by conservative error-free performance whereas the demented made more errors, partly because of their more liberal criterion and partly because of their greater loss of encoding characteristics. From these results, an optimal cut-off score was chosen for differential diagnosis using the FRT. This was shown to hold more promise than the SLT in discriminating between dementia and depression.