The aim of this study was to examine the relationship between social network variables and levels of and longitudinal changes in blood pressure in a middle-aged/older sample. The participants (50-75 years at baseline; n=1097) responded to questions concerning social relationships at baseline and their blood pressure (diastolic, systolic) was measured. Blood pressure levels were reassessed 5, 10, and 15 years later. Latent growth models with responses to questions concerning social relationships as predictors and basic demographic factors (age, sex) as covariates, unexpectedly indicated that a more limited social network (no close friend, few visits, little contact with friends in other ways, not living with someone, and a composite index based on all questions) was associated with significantly lower diastolic blood pressure levels. For systolic blood pressure a similar result was observed for one of the variables (lack of a close friend). In general, these effects diminished over time, as indexed by the positive relationship between several of the social variables and slope. The results were little affected by inclusion of additional covariates (e.g. measures of psychological distress, smoking/alcohol habits, and BMI) suggesting that the origins of this unexpected pattern of findings must probably be sought for in other subjectrelated factors, such as, for example, increased help seeking. Future studies should consider qualitative aspects (e.g. feelings of loneliness, quality of social relationships) in addition to structural aspects to provide a better understanding of these associations.