This study used poems written by fifth graders to determine any differences in the quality of children's writing which might have been induced by the "broad exposure" approach to teaching poetry, the "analytical" approach, or the "non-exposure" approach. A story-writing pretest was given to 93 fifth grade children who were then randomly assigned on the basis of their T-scores to one of the three instructional groups. After undergoing one of the three treatments for 14 days, the children wrote poems to be rated by pairs of judges on the basis of their artistic significance, originality, word choice, imagery, and unity. The judges' ratings were converted to T-scores, and coefficients of correlation were computed for each pair. When the Spearman-Brown formula was applied to the judges' ratings to estimate the reliability of the paired comparisons, only the ratings on artistic significance and originality exceeded a value of .70. On the basis of these and other findings from the study, it was concluded that (1) the evaluation of poetry is a personal and emotional process subject to the preferences of the reader and (2) no one of the three instructional methods being measured seems to be substantially better than the others when viewed in relation to the children's final written product. (Author/JB)