The role of passive voice as a device used in medical and scientific discourse to mystify the author is clearly articulated and well-known. Through analysis of the Methods section of nine medical research articles, this paper shows that metonymy is another frequently used impersonalisation strategy in medical discourse. Furthermore, this paper argues that impersonalisation is not restricted to the authors and that two types of impersonalisation need be distinguished: generalisation and socialisation. Discourse agents were categorised into the ‘present authors’ versus ‘other researchers and health professionals not part of the research team’. Agents were investigated in relation to impersonalisation and social identity. Results show that possessive/causative metonyms are used to produce genderless, generic ‘present authors’ as well as ‘other researchers’. In contrast, more significant ‘health professionals’ are often referred to in terms of representational/locative metonyms highlighting their authoritative social identity. The study also shows that for these non–authorial professionals co-occurrence of metonymy and passive voice is generally avoided. Although ‘present authors’ are mainly absent, this analysis reveals a higher than expected author presence resulting in a significantly higher degree of impersonalisation for non-authorial agents.