OBJECTIVE--To test the hypothesis that symposia on environmental tobacco smoke (ETS) are more likely to present unbalanced data and be authored by tobacco industry-affiliated individuals than journal articles on ETS. To compare the publication records and affiliations of authors of symposia with the authors of scientific consensus documents on ETS. DESIGN--Content analysis of articles; computerized literature searches of English-language publications (except for one symposium) supplemented with additional sources. PARTICIPANTS (ARTICLES)--All 297 symposium articles on ETS and a random sample of 100 journal articles on ETS published between January 1, 1965, and March 31, 1993; the 1986 Surgeon General's report on ETS; and the 1986 National Research Council's report on ETS. MAIN OUTCOME MEASURES--For each article, regardless of whether it had a methods section, agreement with the tobacco industry position that ETS is not harmful; topic; funding source(s); affiliation(s) of author; and publication records of authors. RESULTS--Of the symposium articles 41% were reviews, compared with 10% of journal articles. A total of 83% of original symposium articles and 100% of journal articles contained methods sections (P = .0001). Symposium articles were more likely to agree with the tobacco industry position (46% vs 20%), less likely to assess the health effects of ETS (22% vs 49%), less likely to disclose their source of funding (22% vs 60%), and more likely to be written by tobacco industry-affiliated authors (35% vs 6%) than journal articles (P = .0001). Symposium authors published a lower proportion of peer-reviewed articles (71% vs 81%) (P = .0001) and were more likely to be affiliated with the tobacco industry (50% vs 0%) than consensus document authors (P = .0004). CONCLUSIONS--Symposium articles on ETS differ from journal articles and consensus documents in ways that suggest that symposia are not balanced.