The research focused on children ages 2.5 to 7 years of age who were admitted to an inpatient assessment in Chicago between January 1988 and February 1989. The sampling strategy was sequential rather than random and resulted in 104 suitable interviews. The questionnaire responses of the children's interviewers were compared with those of unobtrusive observers. The particular items examined for reliability included children's specific disclosure statements, doll demonstrations associated with specific disclosure statements, and the affective/expressive behaviors of children that may be crucial considerations in a clinical assessment. Results revealed that interrater reliability was highest for questions addressing children's statements and lowest for those addressing affective/expressive behaviors. Results suggest specific areas of observation and interpretations that tend to be typically ambiguous as well as those that may be more dependent on the experience and skill of the interviewer. Figures, table, and 21 references (Author abstract modified)