The data used in the study were drawn from the national Incidence Study of Missing, Abducted, Runaway, and Thrownaway Children (NISMART). A smaller subset of the total sample (families in which any child had a MISMART event and a random sample of nonevent, or control families) responded to a long interview form that provided information on such issues as parent-child interaction, child stress, and parents' own childhood experiences. This study used 332 of these long- interview cases, comprised of 224 "control" children (those who experienced none of the NISMART focus events) and 108 runaways. All 332 children were aged 10 or older. Independent variables were in five categories: demographic characteristics of children, measures of parent-child conflict, measures of conflict or stress in the child's life, measures of parent-to-child violence, and a measure of parents' own childhood experiences with running away from home. Because the dependent variable in the model (running away from home) is dichotomous, logistic regression was used in the analysis. Findings show that parents' childhood experiences were significantly related to an increased likelihood that their children will runaway from home. This phenomenon can be explained by a pattern of family interaction that is being transmitted across generations; parents who grow up in disruptive families, and who may respond to this by running away from home, may recreate some of the same problems in their families of procreation. The same conflicts that existed in the family of orientation may well be played out in parent-child interactions in the family of procreation. 2 tables and 35 references