The study of testosterone is likely to be prominent in future epidemiological work on endocrine function and the clinical treatment of age-related diseases. Thus, understanding the hormonal changes involved in the normal male aging process will be critical. Using techniques of meta-analysis, the authors examined 88 published studies of the age-testosterone relation in men. These studies reported conflicting results: age-testosterone correlations ranged from -0.68 to +0.68. In cross-study comparisons, certain research design characteristics (e.g. time of day of blood sampling) and various sample characteristics (e.g. volunteers vs patients as subjects) were related to both mean testosterone level and the slope of the age-testosterone relation. For example, for subgroups of subjects that did not exclude ill men, the mean testosterone levels were low, and did not decline with age. Subgroups that included only healthy subjects, in contrast, had higher overall testosterone levels and showed a decline of testosterone with age. Implications of these results for design, analysis and reporting of future epidemiologic studies will be discussed. These results also illustrate the utility of meta-analysis for research with the aged.