The sample consisted of 84 children (mean age 11 years) who had been referred from shelters for battered women and child welfare agencies. Comparisons of those who met PTSD criteria (n=47) and those who did not meet PTSD criteria (n=37) found significant differences on variables of age, duration, and frequency of witnessing children. Child self-report instruments showed that the PTSD group differed significantly from the non-PTSD group on the basis of assault anxiety, depression, dissociation, anger, and negative attributions. Hierarchical multiple regression analysis showed that factors related to negative attributions and assault variables each contributed significantly to explaining 71 percent of the variance of PTSD symptoms, after the variables of age and gender were controlled. Discriminant function analysis correctly classified 84.3 percent of the respondents. The relevance of PTSD to child witnesses is considered, along with study limitations. 4 tables and 81 references