We describe methodological steps in the selection of questions on social networks and support for a cohort study of 4,030 employees from a public university in Rio de Janeiro. First, group discussions with volunteers were conducted to explore the adequacy of related concepts. Next, questions in the Medical Outcomes Study questionnaire were submitted to standard "forward-" and "back-translation" procedures. The questions were subsequently evaluated through five stages of pre-tests and a pilot study. No question had a proportion of non-response greater than 5%. Pearson correlation coefficients between questions were distant from both zero and unity; correlation between all items and their dimension score was higher than 0.80 in most cases. Finally, Cronbach Alpha coefficients were above 0.70 within each dimension. Results suggest that social networks and support will be adequately measured and will allow for the investigation of their associations with health outcomes in a Brazilian population.