This article discusses the use of`dangerousness' as a legal criterion for civil commitment. The notion of dangerousness is defined within the perspective of the relationship between judicial and medical-psychiatric institutions. By reviewing empirical evidence concerning the possibility of a link between mental illness and dangerousness, we critically evaluate the main postulate supporting the inclusion of this notion in the civil laws. We then examine empirical studies of psychiatric expertise in dangerousness assessment and risk prediction. By using observational studies of civil commitment proceedings, we examine how the legal criterion of danger to self or others is actually operationalised into a series of heuristic criteria. These criteria are teleological statements: if being mentally ill and being dangerous are, in this context, interchangeable, so are the finalities of treating and controlling. We conclude that, at the psychiatry-justice interface exemplified by civil commitment, treatment and control have been equated conceptually and in practice, even if the written law clearly distinguishes the concepts. With respect to civil commitment, the institutions of mental health and justice are not, as usually depicted in sociological analysis, two different systems that meet and compete at this junction. Rather, they join together ± becoming, in effect, one actor ± within a treatment-control system which has as its function and aim to`take care of' residual cases viewed as problematic for society.