Abstract: Literature indicates that there existed only low agreement among different informants in the assessment of children’s aggressive behavior (Karver, 2006; Ladd & Kochenderfer-Ladd, 2002), which reflects the fact that each informant offers his or her unique perspective on the target children’s behavior (Achenbach, 1995; Loeber et al., 2000). With the awareness of the disadvantages inherent in the assessment using single informant, MTMM (multitrait-multimethod) design is becoming increasingly popular in research on human development in recent years.
The present study investigated the convergent and discriminant validities of MTMM data on children’s aggression during middle childhood and the extent to which the validities of assessment varied between peers and teachers as informants.
The participants were 2695 children in their middle childhood (mean age = 10.06 ± 0.54 years) with approximately equal number of boys (51.87%) and girls (48.13%) and the children’s head teachers (n = 50). These students and teachers were from 50 classrooms in 14 schools in Jinan, capital city of Shandong Province of China. The MTMM data on children’s three types of aggression (i.e. physical, verbal and relational) was obtained from two types of informants (peer and teacher) and via three types of methods of data collection (i.e. peer nomination, peer rating and teacher rating). For peer nomination, the Children’s Social Behavior Scale (Crick, 1997) was used and the children were asked to nominate up to three children who best fitted the descriptions of items of physical and relational aggression from within their class. Information about children’s physical, verbal, and relational aggressions was also obtained via peer rating, in which children rated their peers on a 12-item Aggressive Behavior Questionnaire. In addition, teacher rating was conducted in which the head teacher of each of the participating classroom rated all the children in his/her class on the three types of aggression using a modified version of the aggressive behavior scale (Crick et al., 1997). All measures possessed acceptable reliabilities in this study.
Moderate to high correlations existed between the same trait measured using different methods (rs = 0.35 to 0.74). The CFA analysis indicated that all of the loadings of trait factors in the MTMM model were significant, and that the MTMM model better fitted the data than did the method model, reflecting the fact that the indicators measured via different methods concurrently converged on the common trait factors and increased the model fitting. High correlations were found between the different traits measured using the same method (rs = 0.74 to 0.92). Results of the CFA analysis demonstrated high loadings of each method factors of MTMM model (λs > 0.60), indicating a low discriminant validity of each of three assessment methods in assessing the three types of aggression.
In MTMM model, the loadings of the peer-reported indicators on the aggression factors were higher and companied with smaller errors as compared with those of the teacher-reported indicators. In the two-order model, the method factors of peer report explained a larger proportion of the variations of the second-order factor of aggression than did that of teacher report. Putting together, these results suggested that peer report (peer rating and peer nomination) served as more valid assessment of children’s aggression than teacher rating. Moreover, the method factor of peer rating explained a greater proportion of the variations of the second-order factor of aggression than did that of peer nomination, indicating that peer rating was more valid assessment than peer nomination.

Key words: middle childhood, physical aggression, verbal aggression, relational aggression, multitrait- multimethod analysis