To examine the quality of mental health research in nursing homes, 130 articles published in six geriatric specialty and health care delivery journals were reviewed. Thirty-nine (30%) articles used a mental health measure. Measures of cognitive function were most common, being used in 32 (25%) of the articles reviewed. Twenty-three (18%) studies measured abnormal mental experiences and 17 (13%) articles measured behavioral disorder. Many articles used measures or determinations with no established reliability. Twenty-six of the articles which used a mental health measure also used a measure of activities of daily living or physical function. Retrospective and prospective studies were similar in number. A minority of articles used control groups, random samples, or prepost measures while a majority (64%) identified an outcome measure. We conclude that nursing home research can be improved by the increased use of reliable measures of cognition and abnormal mental experiences and by the development of reliable measures of behavioral disorder. Study design can be improved by identifying a priori hypotheses and by the increased use of random sampling and control/comparison groups.