For this study of children's selection of key words, children in the first, third, sixth, ninth and twelfth grades were asked to identify which words, in their judgment, were key words in written materials. Relationship of these choices to skilled readers, to words derived from passage analysis, and to random selections of hypothetical subjects were explored. For matches to skilled readers and to the analytical model, grade level and passage difficulty and length were significant effects. Although the youngest group showed variability of word choice comparable to a purely random selection, further analysis revealed considerable preference for selected grammatical classes in contrast to random choice. Recall protocols after one week and six months showed vocabulary usage frequency parallel to initial key word identification frequency. (Author)