To improve identification of contributors to manuscripts, editors of medical journals have developed authorship responsibility criteria. Some have specified an acceptable number of authors per manuscript. We wanted to examine changes in patterns of authorship in the context of the development of these specifications. Therefore, we used a retrospective cohort design to calculate the average number of authors per manuscript and the prevalence of group and corporate authorship between 1980 and 2000 for original, scientific, non-serial articles published in four prestigious medical journals: the Annals of Internal Medicine, Archives of Internal Medicine, Journal of the American Medical Association, and the New England Journal of Medicine. Group authorship identifies individual authors in the byline who are writing for a group; in corporate authorship, contributors are not individually listed in the byline. We found that the number of authors per article increased dramatically over time in each journal, from an average of 4.5 in 1980 to 6.9 in 2000 across journals. As a proportion of published manuscripts, group authorship (authors listed in the byline) increased from virtually zero to over 15%, while corporate authorship (authors not listed in the byline) remained rare and stagnant. Manuscripts published by single authors all but vanished. Group authorship was most prevalent in journals that limited the acceptable number of authors per manuscript. These findings suggest that the number of authors per manuscript continues to grow. The growth in the number of authors on bylines and the proportion of group-authored manuscripts is likely to reflect the increasing complexity of medical research.