This paper analyzes evidence related to the effectiveness of vowel generalization from two viewpoints: the nature of letter-sound and sound-letter relationships as derived from Dewey, Hanna, and Hillerich, and the nature of these relationships as derived from results of teaching vowel generalizations to children. From the evidence and the research studies reviewed the author concludes that generalizations about vowels can be grouped into two categories: generalizations which hold true most of the time but which include too few words to be worth teaching, and those which apply to many words but which are so unreliable that they are not worth teaching. Finally, the author suggests an alternative to the direct teaching of generalizations about vowels, which emphasizes the exploration of language and learning to read through reading. (Author/WR)