CONTEXT: A number of general medical journals and the International Committee of Medical Journal Editors (ICMJE) request authors to disclose their contributions. Little is known about the effect of journal policies on authors' disclosure of their contributions. OBJECTIVE: To determine the number of named authors who do not meet ICMJE criteria for authorship, according to their published contributions, in 3 medical journals with different contribution disclosure practices. DESIGN: Observational study of authors' contributions in research articles published in 2002 in Annals of Internal Medicine (n = 72), BMJ (n = 107), and JAMA (n = 81). BMJ asks authors to describe research contributions in their own words; Annals asks authors to choose from a list of coded contributions; and JAMA uses a structured checklist with instructions on contributions that qualify for ICMJE authorship criteria. Honorary authorship was defined as the lack of contribution from the first ICMJE criterion (study conception and design, or acquisition of data, or analysis and interpretation of data) and/or second (drafting the article or critical revision for important intellectual content) ICMJE criterion. RESULTS: According to authors' published contributions, the number of honorary authors was highest in Annals (121/562 authors, 21.5%), followed by BMJ (46/482, 9.5%), and JAMA (3/641, 0.5%) (chi2(2) = 146.67, P<.001). The number of articles with honorary authors was 60% in Annals, 21% in BMJ, and 4% in JAMA. Honorary authors had fewer published contributions than authors who met ICMJE criteria and were positioned more toward the end of the byline. Honorary authors either lacked contributions for both ICMJE criteria (10% in Annals and 22% in BMJ) or contributions to the second ICMJE criterion (75% in Annals, 67% in BMJ, and 2 out of 3 in JAMA). CONCLUSIONS: General medical journals differed in prevalence of honorary authors according to published research contributions of named authors. Different authorship/contributorship policies and procedures should be explored as a possible explanation for the differences in contributions disclosed by authors among these journals.