In this study, Don't Know/Not Applicable (DK/NA) responses on a measure of academic behaviors associated with college readiness for high school students were treated with: (a) casewise deletion, (b) scale inclusion at the lowest level, and (c) imputation using E/M algorithm. Significant differences in mean responses according to treatment method resulted. Comparisons of survey completers and noncompleters revealed first generation students have a significantly higher number of DK/NA responses. Results indicate a potential external validity threat in treating DK/NA responses with casewise deletion or imputation because noncompleters would be removed or reduced, and the unique needs of first generation students could be overlooked. Findings support scale inclusion at the lowest level as the most appropriate treatment of nonresponse items. (Contains 3 tables and 4 figures.)