A review was conducted of 585 articles in the medical literature, and the articles were categorized by study design and by statistical techniques employed. In the majority, statistics were utilized primarily for data description and hypothesis-testing. Descriptive studies were most frequent, and only a minority were epidemiological investigations. A questionnaire designed to assess knowledge of biostatistics and epidemiology relevant to this literature was administered to internists and medical house staff at a teaching hospital. Of 229 questionnaire recipients, 141 responded and achieved a mean score of 7.4 with 10.0 possible. Practicing physicians had a significantly lower mean score than house staff and full-time faculty. Scores were higher for the 44 percent of respondents who reported prior training in epidemiology or in biostatistics. The results of the journal review suggest that knowledge of basic biostatistics and of study design is important for reading medical literature. Physicians with prior training in epidemiology and in biostatistics appear to be better prepared for this task.