The study sample consisted of 35 abused (25 with friends and 10 friendless) and 43 nonabused children (33 with friends and 10 friendless). The abused group ranged in age from 4.3-11.5 years old (n=22 boys). The nonabused group ranged in age from 5.5 to 11.6 (n=23 boys). The study used a convenience sample of children with diverse maltreatment histories, with all abused children experiencing more than one type of abuse. All children in the 15 classrooms that contained a mix of abused/nonabused students were administered sociometric measures; however, only those children in friendship pairs who matched the abused children demographically (age, gender, majority/minority status) participated in the main friendship study. Children rated each of their classmates on a 3-point scale from "like" to "dislike." to play with them. At the end of the rating, the children were also asked to nominate the three children with whom they liked to play least in their classroom, as well as three children with whom they liked to play most in their class. The children were paired into friendships according to reciprocal nominations of "the child I like to play with most" during the sociometric assessments. Friendship pairs were determined by matching the top three "best liked" playmates. The friendship sample consisted of 25 abused and 33 nonabused children who had close reciprocal friendships. Some children had more than one best-friend match. "Friendless" status for both abused and nonabused children was determined when a child had no reciprocal nominations in his/her class; no one in his/her class nominated him/her as one of their top three best friends; and the child stated that he/she had no other "best" friends outside of the classroom. Friendship quality was partially determined by observations, which were coded live by extensively trained raters who observed the children's interactions from behind a one-way mirror. The study found that abused children were not rated significantly lower sociometrically, nor did they differ significantly from control children on several measures of friendship quality, such as resolving conflicts and helping each other; however, abused children were found to be more negative and less proactive in their interactions. They also reported their friendships as having more conflict and as having more experiences of betrayal and less that reflected caring. Only observational friendship variables predicted loneliness. These results challenge the general assumption that abused children's peer relationships are uniformly more maladaptive than those of nonabused children. This suggests the potential benefits of structured intervention designed to "normalize" their friendship interactions. The pattern of difficulties exhibited by abused children in their peer relationships suggests targets for more specific interventions. 6 tables and 46 references