This study compares the performance of the Cornell Scale for Depression in Dementia (CSDD) and the Hamilton Depression Scale (HDS) in detecting Research Diagnostic Criteria (RDC) major depression in subjects with mild-to-moderate Alzheimer's disease (AD). Thirty-four subjects with this diagnosis and their caregivers were interviewed. The senior author conducted a diagnostic interview to determine RDC diagnosis. An investigator, blind to diagnosis, obtained demographic information and administered the Mini Mental State Examination, Global Deterioration Scale, CSDD and HDS. For each depression scale, the correlation with the RDC diagnosis of major depression was calculated, as were the sensitivity and specificity at various cutoff scores. Nonparametric receiver operating characteristic analysis was used to compare the performance of the two scales. The area under the receiver operating characteristic curve was .91 for the CSDD and .87 for the HDS. This differed from chance to a highly significant degree for both the CSDD and the HDS but the difference between the two scales was not statistically significant. Although the precision of the present study is limited by the small sample size, a cutpoint of 7 provided reasonable performance for both the CSDD and the HDS, yielding a sensitivity of .90 for both scales and a specificity of .75 for the CSDD and 0.63 for the HDS. Although the CSDD and the HDS are rating scales rather than diagnostic instruments, receiver operating characteristic analysis indicates that both demonstrate statistically significant discriminating ability for RDC major depression in mild to moderate, probable AD.(ABSTRACT TRUNCATED AT 250 WORDS)