A study examined 3 grammar teaching methods to understand why some methods may carry over into writing better than others. E. Bialystok and E. B. Ryan's (1985) metacognitive model of language skills was adapted to plot traditional grammar, sentence combining, and the functional/inductive approach according to the amount of analyzed knowledge and cognitive control each method requires to raise metalinguistic awareness. In so doing, the cognitive demands asked of various kinds of writers by each method can be ascertained. Research results from several studies were analyzed. Results indicated that traditional grammar is not being blended into students' writing because it requires a great deal of cognitive control and analyzed knowledge to deploy. On the other hand, neither sentence combining nor the functional/inductive approach hamstring students with multiple terms and abstract concepts to memorize. Rather, writers focus their attention on some aspect of sentence error and correction by using implicit knowledge, thus keeping the value on the analyzed knowledge axis low. Findings suggest that the highly analyzed system of traditional grammar has had limited success in carrying over into writing because: (1) its goal is to shape mental representations, necessitating an extended period of time to acquire the system; (2) control is possible only after the writer knows the system; and (3) overemphasis on that skill can undermine the production side of things. (Contains 3 figures and 37 references; notes are appended.) (Author/CR)