The impersonalizing role passive voice plays in scientific discourse is well known. Analysis of the Methods sections of nine medical research articles shows that metonymy is another frequent strategy used to create anonymous authors/agents. Discourse agents were categorized into four semantic domains: familial lay, nonfamilial lay, authorial professional, and nonauthorial professional. Agents were investigated in relation to impersonalization and social identity. Results show that although possessive/causative metonyms produce generic participants and reduce most rival researchers to "previous studies," significant health professionals are often referred to in terms of representational/locative metonyms, highlighting their authoritative social identities. Additionally, authors are either highly visible or, if they choose to disguise themselves, they do so quite drastically using impersonalization devices or agentless passives. In contrast, for other researchers and health professionals, co-occurrence of metonymy and passive voice is generally avoided; nevertheless, these agents are usually more hidden than are the present authors. (Contains 4 figures, 5 tables, and 9 notes.)