Whether students of various ages reported different self-perceptions of their writing because of survey design effects was studied. As part of a writing performance assessment, 1,972 students in an urban/suburban school district in grades 3 (n=736), 6 (n=636), and 9 (n=600) completed a brief self-perception survey that contained 5 items measuring students' perceived competence in writing. Students were randomly assigned to answer questions or to respond to statements regarding their beliefs about themselves as writers. Both versions of the survey contains response options that were ordered from positive to negative for all items, and items were presented in the same order on both versions. A univariate analysis that controlled for gender showed a two-way interaction between grade level and survey format. Students rated themselves higher on the question than statement format at all grade levels, but the difference between groups increased with the age of the student. Older students generated more accurate estimates of their competence than did younger students, but overall the correlations between perceived competence and overall score on the writing assessment were low. These results indicate that, by acknowledging the potential effects of survey design decision on data interpretation, researchers will improve the validity of their inferences in developmental studies as well as improving the comparability of data across researchers. (Contains 1 figure, 2 tables, and 12 references.) (SLD)