This study, part of a larger ethnographic study of elementary students categorized as at-risk, attempts to determine whether the labelling of students affects teachers' expectations of, and behavior toward, at-risk students. The beliefs of a teacher sample of five women, three Hispanic and two Anglo, at two urban schools in the Southwest with large populations of minority and low-income students were assessed in two different ways. Teachers were asked first about the notions of "at-risk" in general, and then asked to identify and describe their at-risk students. The first set of questions was designed to elicit their "declared," or public beliefs; the second set was designed to elicit their more private beliefs by asking them to think of specific examples. Later in the year the teachers were asked if their concept of at-risk had changed and whether the students categorized as such still deserved the label. The major findings from analysis of the teachers' perceptions are the following: (1) their concept of at-risk was fluid, with the definition changing from a sociocultural perspective initially to one more focused on a student's ability to learn at the end of the term; (2) teachers' decisions to refer students to special education were based on expectations for academic performance and norms for appropriate classroom conduct rather than on the students' behavior; (3) teachers understood that at-risk status could be situation-specific, with the same student considered at-risk in one classroom but not in another; (4) they tended to blame students' parents and homelife for academic failure rather than the students' personal shortcomings or their own inability to teach effectively; and (5) they failed to see their own role in the creation of an environment and set of expectations that affected both the labelling of at-risk students and the students' behavior and that they themselves may limit the students' potential for academic achievement. The teachers' limited awareness of the problems of the social constructivist nature of their labelling can result in programs that deal with students' individual problems but fail to implement structural changes in classrooms and schools. Two tables and 30 references are appended. (WS)