According to the social withdrawal hypothesis, a personal network becomes smaller when a person starts dating, cohabitates and marries. This phenomenon is widely established in the literature. However, these studies were usually done with cross-sectional data. As a consequence, it is still unclear whether or how personal networks actually change after the formation of a romantic relationship (i.e. dating), after starting cohabitation and after getting married. It is also unclear how long and to what extent social withdrawal continues. To overcome these shortcomings, we examine how the size and composition of personal networks change after relationship formation. We use two waves of the PAIRFAM dataset (2008 and 2011), which include information about 6640 Germans who were between 16 and 39 years of age at the time of the second interview in 2008. Results from fixed effects regression models underscore that the association between romantic relationships and changes in personal networks is more dynamic than previous studies suggested. For example, after the formation of a romantic relationship people show a decrease in non-kin contacts, while an increase in non-kin contacts is observed after two years of dating, as well as after two years of cohabitation. These network changes suggest that people adapt their social networks to the demands and constraints of each phase of a romantic relationship. Because the decline in network size after dating is not stable, there is no need to be afraid that those who have a romantic partner remain isolated from other relationships.