¹ Yale University dissertation (1958, unpublished). Part of an earlier draft of this paper was read to the departmental seminar of the Columbia University linguistics department in November 1958, and thanks are due for suggestions offered at that meeting. My Seneca field work during the summers of 1956-8 was sponsored by the New York State Museum and Science Service, William N. Fenton director. Any work that I do in the Iroquois field is built upon the solid foundation provided by Floyd G. Lounsbury.

² H. M. Hoenigswald, Internal reconstruction, SIL 2.78-87 (1944); id., Sound change and linguistic structure, Lg. 22.138-43 (1946); J. W. Marchand, Internal reconstruction of phonemic split, Lg. 32.245-53 (1956).

³ See references in Marchand, fn. 3.

⁴ Written phonemically.

⁵ The prime indicates palatalization. In Marchand's original article, palatalization is denoted by a comma.

⁶ Cf. G. Bonfante, On reconstruction and linguistic method, Word 1.144 (1945). Bonfante was referring only to the comparison of separate languages or dialects when he excluded comparison from the ‘chronology of the texts’ and described it as optional for certain others among his ten historical methods.

⁷ Although not discussed in precisely these terms, an extensive illustration of stylistic variants is provided by C. T. Hodge, Some aspects of Persian style, Lg. 33.355-69 (1957). Others prefer to regard such alternative forms as containing morphs in free variation; see for example P. L. Garvin, On the relative tractability of morphological data, Word 13.17 f. (1957).

⁸ The principal step in comparative grammar, Lg. 26.357-64 (1950).

⁹ This way of indicating the pertinent environments of a correspondence, often with the aid of abbreviations for phoneme categories (as V for vowel) will be followed throughout the paper. Thus, to write that the correspondence a/b occurs in the environments V-V/-x, y- means that a occurs intervocalically while b occurs before x and after y.

¹⁰ The term ‘analogy’ could of course be used in a wider sense to include both processes.

¹¹ Although of course the picture is often complicated by gradual rather than sudden and complete divergence.

¹² Hoenigswald (1946) uses the terms ‘compulsory’ and ‘noncompulsory’.

¹³ Cf. Hoenigswald 1946, §5.

¹⁴ s/t is in contrastive distribution with ø/s, so that the reconstruction of an *s is precluded. See fn. 15.

¹⁵ Zero is an acceptable reconstruction only if epenthesis is plausible as a historical explanation. Epenthesis is plausible whenever the environments in which it might have taken place are definable in strictly phonological terms. That is not the case here, so that we are led to assign ø/s to *s.

¹⁶ Cf. F. G. Lounsbury, Oneida verb morphology 14 f. (New Haven, 1953). Note that with chronologically and environmentally isolated changes it is sometimes impossible to distinguish phonemic from analogic change.

¹⁷ Published descriptions of Seneca have been CF. Voegelin and W. D. Preston, Seneca I, IJAL 15.23-44 (1949); and N. Holmer, Seneca II, IJAL 18.217-22 (1952), Seneca III, IJAL 19.281-9 (1953), and The Seneca language, a study in Iroquoian (Upsala and Copenhagen, 1954). These works were based on very little contact with the language and are at best preliminary sketches.

¹⁸ The indicated symbols for phoneme categories will be used below to abbreviate the descriptions of environments.

¹⁹ Additional prosodic phonemes, including other accents as well as pitches and junctures, are not involved in the discussion in this paper.

²⁰ I.e. it has a relatively small proportion of agglutinative morpheme junctures and a relatively large number of morphemes per word. See J. H. Greenberg, A quantitative approach to the morphological typology of language. Method and perspective in anthropology 192-220 (R. F. Spencer, ed.; Minneapolis, 1954).

²¹ C. F. Hockett, Problems of morphemic analysis, Lg. 23.333 (1947).

²² Cf. fn. 15.

²³ At which there is always a potentiality of juncture, although the juncture is not always present. We may infer that the prejunctural forms have been generalized.

²⁴ The correspondence ø/: after the æ's is discussed in §52.

²⁵ k, t, or s. j does not occur in this position.

²⁶ The correspondence at the beginning of these forms is discussed in §47.

²⁷ The correspondence ø/t at the end of these forms was discussed in §32.

²⁸ ‘Elsewhere’ here and in §§39 and 40 does not include all other environments; specifically, it does not include the limited environments described in the other two of these three paragraphs.

²⁹ Because of the nature of the environments, a three-way correspondence n/w/y would be impossible.

³⁰ For the correspondence :/ø after the o's, see §50.

³¹ Note that *r is one specific phoneme, while *R in reconstructions indicates either *n, *w, *y, or (see below) *r.

³² See §61 for the shift of accent.

³³ The environments of these changes have been stated as if they were more recent than the loss of intervocalic *h, but can equally well be stated in terms appropriate to earlier stages; e.g. *V_∊h(r)- rather than V_∊–.

³⁴ See §61 for the shift of accent.

³⁵ On this point see C. H. Borgström, Internal reconstruction of Pre-Indo-European word-forms, Word 10.276 (1954).