The notion of the "psychosomatic" family continues to enjoy uncritical acceptance in the absence of promised data and despite its dependence on an outmoded view of how psychosocial factors are involved in illness. We review the decline of psychosomatic models of illness that assume that arousal is the only or primary means by which psychosocial factors influence illness. Focusing on brittle diabetes, we note the potential for family theorists to develop more adequate models of poor self-care and medical crises as interactional tactics, as dynamic efforts to solve problems, define relationships, and influence others, even if they are costly and self-defeating. In an appendix, we note the inadequacy of Rosman and Baker's (1988) reanalyses of the Minuchin, Rosman and Baker (1978) data.