Most studies regarding the role of teacher-child relationships for children's early school adjustment use only teacher ratings of relationship quality. The current study examines (a) the agreement between teacher and observer ratings, (b) whether similar patterns of relations with behavioral engagement are obtained across informants, and (c) which informant matters the most in the prediction of engagement. Teacher and observer ratings of teacher-child closeness, conflict, and dependency were gathered for a sample of 148 kindergartners and their teachers at three measurement occasions. Teacher and observer reports converged to a moderate degree but only when considering multiple occasions and ruling out occasion-specific variance. Although some relations with behavioral engagement were similar, only teacher ratings had unique, added value in this prediction. (Contains 1 figure and 3 tables.)