OBJECTIVE: To evaluate whether physicians' beliefs concerning episiotomy are related to their use of procedures and to differential outcomes in childbirth. DESIGN: Post-hoc cohort analysis of physicians and patients involved in a randomized controlled trial of episiotomy. SETTING: Two tertiary care hospitals and one community hospital in Montreal. PARTICIPANTS: Of the 703 women at low risk of medical or obstetric problems enrolled in the trial we studied 447 women (226 primiparous and 221 multiparous) attended by 43 physicians. Subjects attended by residents or nurses were excluded. MAIN OUTCOME MEASURES: Patients: intact perineum v. perineal trauma, length of labour, procedures used (instrumental delivery, oxytocin augmentation of labour, cesarean section and episiotomy), position for birth, rate of and reasons for not assigning women to a study arm, postpartum perineal pain and satisfaction with the birth experience, physicians: beliefs concerning episiotomy. RESULTS: Women attended by physicians who viewed episiotomy very unfavorably were more likely than women attended by the other physicians to have an intact perineum (23% v. 11% to 13%, p < 0.05) and to experience less perineal trauma. The first stage of labour was 2.3 to 3.5 hours shorter for women attended by physicians who viewed episiotomy favourably than for women attended by physicians who viewed episiotomy very unfavorably (p < 0.05 to < 0.01), and the former physicians were more likely to use oxytocin augmentation of labour. Physicians who viewed episiotomy more favourably failed more often than those who viewed the procedure very unfavourably to assign patients to a study arm late in labour (odds ratio [OR] 1.88, p < 0.05), both overall and because they felt that "fetal distress" or cesarean section necessitated exclusion of the subject. They used the lithotomy position for birth more often (OR 3.94 to 4.55, p < 0.001), had difficulty limiting episiotomy in the restricted-use arm of the trial and diagnosed fetal distress and perineal inadequacy more often than the comparison groups. The patients of physicians who viewed episiotomy very favourably experienced more perineal pain (p < 0.01), and of those who viewed episiotomy favourably and very favourably experienced less satisfaction with the birth experience (p < 0.01) than the patients of physicians who viewed the procedure very unfavourably. CONCLUSIONS: Physicians with favourably views of episiotomy were more likely to use techniques to expedite labour, and their patients were more likely to have perineal trauma and to be less satisfied with the birth experience. This evidence that physician beliefs can influence patient outcomes has both clinical and research implications.