A 113-item survey instrument was designed to measure the class climate perceptions of 20,157 students in grades 7-12 as part of the "A Study of Schooling" research project. The 29 constructs of the questionnaire focused on students' interpersonal needs, classroom procedures, and instructional practices. Survey items were written in the first person singular, to measure affect or interpersonal relationships, as well as in the third person, to measure classroom instructional practices. Both factor and cluster analysis were used at the item level to explore the viability of the constructs. More than one correlation matrix was computed for appropriate units of analysis to identify conceptual differences between groups. Since a lack of consensus among students was viewed as leading to possible avenues for further study rather than as an error, the generally low percentage of between class variance was not disappointing. An analysis of the items revealed that differences in correlation ratios (eta squareds) can loosely be attributed to either the content or to the grammatical structure of the items. Constructs with lower correlation ratios asked questions that tended to divide classes into groups. Constructs with high correlation ratios tapped pervasive, salient features of the classroom, such as teacher characteristics that did not divide students into groups. The survey demonstrates the difficulties in writing meaningful, high inference items about classroom procedures on which students can agree. (FG)