Objective: Controversy abounds regarding the process by which child sexual abuse victims disclose their experiences, particularly the extent to which and the reasons why some children, once having disclosed abuse, later recant their allegations. This study examined the prevalence and predictors of recantation among 2- to 17-year-old child sexual abuse victims. Method: Case files (n = 257) were randomly selected from all substantiated cases resulting in a dependency court filing in a large urban county between 1999 and 2000. Recantation (i.e., denial of abuse postdisclosure) was scored across formal and informal interviews. Cases were also coded for characteristics of the child, family, and abuse. Results: A 23.1% recantation rate was observed. Multivariate analyses supported a filial dependency model of recantation, whereby abuse victims who were more vulnerable to familial adult influences (i.e., younger children, those abused by a parent figure and who lacked support from the nonoffending caregiver) were more likely to recant. An alternative hypothesis, that recantations resulted from potential inclusion of cases involving false allegations, was not supported. Conclusion: Results provide new insight into the process by which children reveal interpersonal trauma and have implications for debates concerning the credibility of child sexual abuse allegations and treatment in dependency samples. (Contains 1 table.)