Purpose: During recent decades, research institutions have increased collaboration with other institutions since it is recognized as a good practice that improves their performance. However, they do not usually consider external collaborations as a strategic issue despite their benefits. The purpose of this paper consists of identifying different patterns of collaboration and internationalization of universities, with the aim of helping managers and policy makers to take decisions related to their national research policies. Design/methodology/approach: Co-authorship analysis has been used in conjunction with social network analysis to model inter-institutional collaborations as networks, extracting these collaborations from the Web of Science database. Using several structural properties of the extracted networks and applying a statistical treatment, the main profiles of collaborations and internationalization have been obtained. Findings: Obtained results distinguish three patterns of collaborations according to the intensity and scope of collaborations. The statistical treatment also provides a segmentation of universities according to their collaboration profiles. Finally, universities are represented in bi-dimensional maps using external collaborations as a measure of similarity. Research limitations/implications: Although this study is restricted to English universities, it could probably be extended at least to other countries in the European Union or even other developed countries. Practical implications: Research and institutions productivity are usually linked to the amount of received funding. The use of indicators related to internationalization of institutions can help to avoid a bias in favour of research quantity rather than quality, and towards a short-term performance rather than a long-term research capacity. Originality/value: As a difference to previous works, this paper analyses networks of collaboration from the viewpoint of institutions. More specifically, the combination of social network analysis and factor analysis is used to identify patterns of collaboration among institutions. A longitudinal study is also included to demonstrate that the obtained categorization of universities is maintained over time. (Contains 7 tables and 4 figures.)