This study is intended to illuminate the highly recessive nature of the ‘New England Short o.‘ The varying stability of the feature from word to word and from region to region and the extent of divided and unsettled usage from community to community and speaker to speaker provide an unusual glimpse of a linguistic phenomenon long recognized but never adequately documented.

As has been pointed out on several occasions, varieties of American English spoken in the Pacific Northwest are largely derived from eastern sources. A good deal of information is already available on the geographical distribution of vocabulary throughout the United States, so that dialectologists now have a fairly clear idea of lexical continuity in the vast areas of western settlement. For this reason it has become standard procedure to identify dialectal types within these areas in accordance with the criteria presented by Hans Kurath in his Word geography of the eastern United States (Ann Arbor, 1949). A monumental work by Kurath and McDavid on the pronunciation of English in the Atlantic States has now been published, and new comparisons of dialect distribution east and west are therefore in order.

In the introduction to their recent book on eastern American dialects, Kurath and McDavid have a short section headed ‘Vowels in Unstressed and Weakly Stressed Syllables’. The first sentence is: ‘All dialects of American English have a free vowel /ə/ that occurs only in unstressed syllables.‘ The question here is whether or not it is proper or advantageous or both to include in a phonemic inventory of English such a vowel or possibly more than one. The alternative to something like /ə/ restricted to occurrence in unaccented syllables is of course identification of every vowel appearing in an unaccented syllable with one or another of the vowels appearing in accented syllables.

The identification of the morphemes and morpheme classes of a language frequently presents difficulties. Some of these involve problems in morphemic theory itself. This is true of two problems in Ocaina, one concerning morphemic segmentation and fusion, and the other concerning morphemic compounding as it distorts morpheme distribution classes. The data are presented and interpreted in §§1–2. The theoretical implications are summarized in §3, where it is pointed out that the interlocking of the lexical and grammatical hierarchies must be distinguished from effects of the interlocking of the lexical and phonological hierarchies.