¹ I am indebted to Edward Sapir and Morris Swadesh for instruction in phonemic theory and its application to the Shawnee language. This language belongs to the Algonquian linguistic stock and is at present spoken in Oklahoma between Norman and Shawnee (Absentee group), in the neighborhood of Miami (Eastern Shawnee), and in the hills south of Vinita (Loyal or Cherokee Shawnee). The speech of informants studied includes especially that of James Clark (55 years old), Mary Williams (26), Thomas Wildcat Alford (74) of the Absentee Shawnee; Carrie Bluejacket (75) of the Eastern Shawnee, with which group Nancy Sky (67) now lives, although she formerly belonged to the Black Bob band of Shawnee, now extinct; and Frank Daugherty (70) of the Cherokee Shawnee. The three dialects show some lexical differences, very few phonetic, and no phonemic differences.

² The defective phonemes of this class are inclosed in parentheses in the table of phonemes.

³ This phoneme cannot be associated with any other phonemic class because its positional variants are so diverse phonetically.

⁴ Neither the semantic content of a phrase nor the considerable variation observed in phrase length are really pertinent to the definition of the phrase as a syllabic matrix. Observations on these points are difficult without the aid of phonograph records of natural speech. In artificially slow dictation some informants will make every word a phrase—i.e., will unvoice the final syllable of every word. More commonly, when an informant regards dictation as unnatural, he will not speak in phrases at all—i.e., he will voice the final syllable of every word. Phrasing is the first prosodic feature to be disturbed in artificial conditions of speech.

⁵ The word-stress, if one can speak of such a stress in words which have every doubled vowel and every syllable closed with a glottalic phoneme stressed, falls on the final vowel when the word is in phrase-medial and on the vowel preceding the unvoiced vowel(s) when the word is in phrase-final.

⁶ Ought one perhaps regard interjectives as not necessarily bound by phonemic conditions which apply to the general body of a language? A viewpoint such as this would be highly convenient in the phonemic treatment of certain languages. For example, in Tübatulabal, a Shoshonean language, a vowel is consistently nasalized only in the exclamation ma? ‘hello, greetings’. In English itself, a variety of clicks are used only in exclamations. Edgerton finds that in the Rigvedic dialect of Sanskrit the sacrificial exclamation svaha ‘was not subject to the phonemic variation in question’ (Language 10. 263) for reasons which include by implication the one suggested here (‘The word was a mere interjection.‘).

⁷ A view which regards these variants as constituting two independent phonemes suffers in theoretical and practical respects in having the glottal stop defective (not found initially), in having h as well as the glottal stop anomalous among consonants (one syllabic position would be left unfilled, syl. final and syl. initial respectively), in having a peculiar awkwardness when Shawnee is used comparatively. It would be necessary to postulate a pre-Shawnee ^? to relate the facts to Primitive Central Algonquian:

⁸ In view of the fact that in Primitive Central Algonquian h is thought not to occur initially as a phoneme, ‘though an on-glide of h is common before vowel initials’ (Bloomfield, On the Sound System of Central Algonquian, Language 1.146), it is worth pointing out that Shawnee ? in syl. initial, corresponding to the on-glide of h before vowel initials in other dialects, is a phoneme and is preserved as such in certain compounds. When preceding certain vowels, or when pre-consonantic in phrase initial, ? in syl. initial is replaced by t after personal pronoun prefixes.

⁹ This must not be taken to mean that Shawnee shows a great divergence from the vocalic systems of other Central Algonquian languages which are recorded with long and short vowels. The vocalic correspondences of these languages with Shawnee are in fact quite regular when the long vowels of the published texts are equated to the doubled vowels in Shawnee. Where Shawnee differs from PCA quantitatively, the divergence can usually be attributed to a loss of a vowel in Shawnee before ? in syl. final. Compare PCA (Bloomfield 135) *wē?θenιwa: Shawnee wi?θeni ‘he eats’.