Psychiatric epidemiology has not until now reached the stage at which the individual's immediate social environment can be examined as a principal independent variable. That handicap may now be overcome by a recent development in social psychiatry: the systematic study of social bonds. Work in this area is providing a method for identifying those elements in social relationships which, when deficient, may be causally related to neurosis. There are considerable technical difficulties in teasing apart cause and effect. These can be tackled by designing intervention trials and by fitting structural equation models to longitudinal data. The evidence so far is at least consistent with, but cannot yet prove, the hypothesis that a deficiency in social bonds is a cause of neurosis.