¹ William L. Wonderly, Zoque Morphology, IJAL 17.1–9, 105–23, 137–62, 235–51 (1951); 18.35–48, 189–202 (1952); and with the separate papers bound together as a monograph (1953).

² Harry Hoijer, Navaho phonology, University of New Mexico publications in anthropology 1–59 (1945); Zellig S. Harris, Navaho phonology and Hoijer's analysis, IJAL 11.239–46 (1945).

³ Benjamin Lee Whorf, The Hopi language, Toreva dialect, VFPA 6.158–83 (1946). Whorf notes peculiarities of other dialects than Toreva but makes no reference here to his manuscript dictionary, which consists in part of a phonemic retranscription of Stephen's vocabulary, and otherwise gives forms obtained directly from a single informant. The entire retranscription appears in Elsie Clews Parsons, The Stephen's Hopi journal (Columbia University contributions to anthropology, Vol. 23, 1936), prefaced by Whorfs first published phonemic statement. Whorf's last publication to appear on Hopi sounds was his earliest version—it is inserted as §§1 and 2 in Clyde Kluckhohn and Kenneth McLeish, Moencopi variations from Whorf's Second Mesa Hopi, IJAL 21.150–6 (1955), a paper stimulated by the important manuscript dictionary already mentioned. (Another manuscript dictionary, more extensive than Whorf's, has been compiled by Harold S. Colton of the Museum of Northern Arizona).

⁴ Besides influencing Hopi specialists, including Alfred Whiting and Harold S. Colton, in their selection of orthography, Edward A. Kennard has produced two volumes of Hopi texts—Field Mouse goes to war (with Albert Yava and Fred Kabotie), and Litte Hopi Hopihoya (with Albert Yava and Charles Loloma), both published by The Education Division, United States Indian Service (n.d.).

⁵ Leonard Bloomfield, Language 85–92 (New York, 1933).

⁶ Before citing Trager and Smith, we distinguish between two not unrelated possibilities : (1) the applicability of any phonemicization of their kind for any English dialect; (2) the applicability of this or any single phonemicization for all English dialects. Both possibilities have been anticipated in previous papers—the first in George L. Trager and Bernard Bloch, The syllabic phonemes of English, Lg. 17.223–46 (1941) ; the second in Morris Swadesh, On the analysis of English syllablcs, Lg. 23.137–50 (1947), which sets up a composite system in a tentative way—i.e. with recognition that phonetic data of some dialects may not be accommodated by the composite system. See now George L. Trager and Henry Lee Smith Jr., An outline of English structure (Studies in linguistics, Occasional Papers 3; Norman, Okla., 1951). The authors have checked many English dialects, including (9): ‘Americans from all parts of the United States, Canadians from Eastern, Central, and Western Canada; Australians, New Zealanders; natives of Great Britain—speakers of Southern British, various British provincial speakers, Scotch speakers; speakers from Ireland and various British colonial regions; South Africans. Standard and various types of non-standard speakers were heard ... The analyses presented are based on these extensive observations, and are intended to be statements of the structure of the English language as a whole, with variations as stated.‘

⁷ Though the traditional point of reference system has not, orthography has been influenced by consideration of the contact language, as in Latin America where the affricate [ts] in Indian languages is often written ȼ instead of c to avoid confusion with the value of the latter in Spanish. Little more than this, apparently, even when scholars have opportunity to phonemicize de novo many dialects of one language: ‘The Conferencia de Linguístas y Filólogos adopted a set of principles regarding the choice of symbols, which took into account the needs of different dialects and, to some extent, the question of bilingual literacy. Such matters were applied in certain particulars in the Tarascan project. One odd instance that comes to mind was the use of turned v for high back unrounded vowel for the reason that it looks a little like a hand-written i, this being an advantage because there are some dialects which use i in place of the other vowel sound. Also y was convenient because it looks like n, and would hardly be noticed by the dialects which do not distinguish y from n. In practice these matters were connected with the choice of symbols rather than with phonemic analysis as such. I cannot think of any case where the analysis of a language was changed in order to conform with variant dialects.‘ (Morris Swadesh, letter dated 9 August 1954).

⁸ Hans Wolff, Comparative Siouan I, IJAL 16.61–66 (1950).

⁹ Zellig S. Harris and C. F. Voegelin, Hidatsa texts with grammatical analysis (Indiana Historical Society Prehistory Research Series, 1939); Florence M. Robinett, Hidatsa I: Morphophonemics, IJAL 21.1–7 (1955); see also Kenneth L. Pike for a very general discussion of this question in terms of a should-we-should-we-not choice (we should): Grammatical prerequisites to phonemic analysis, Word 3.157–72 (1947).

¹⁰ George L. Trager, An outline of Taos grammar, VFPA 6.195 (1946): ‘These are the only relations to be gotten from purely phonemic functioning [= distribution of unit consonants within the syllable]. When we consider the morphophonemic interchange (§2: 34) we see, however, that we may set up these further groupings: p, t, c, k alternate respectively with p’, t', c', k' ... b, d alternate with p, t; y with c and k; m, n with p, t.'

¹¹ Among regularities I would certainly include asymmmetries as well as symmetries in the sense in which these paired terms are used in my recent paper on models : Inductively arrived at models for cross-genetic comparisons of American Indian languages, University of California publications in linguistics 10.27–45 (1953). And of course I include common-denominator formulations, in the way that these are set up diachronically, by the comparative method; or synchronically, by Zellig S. Harris, Transfer grammar, IJAL 20.259–70 (1954).

¹² David Olmsted, Ethnolinguistics so far (Studies in linguistics, Occasional Papers No. 2; Norman, Okla., 1950) ; Charles Osgood and Thomas A. Sebeok (edd.), Psycholinguistics (Indiana University publications in anthropology and linguistics 10,1954).

¹³ C. F. Voegelin, Culture, language, and the human organism, Southwestern journal of anthropology 7. 357–73 (1951).

¹⁴ Alfred S. Hayes, Field procedures while working with Diegueño, IJAL 20.185–94 (1954) ; Florence M. Robinett, First report on the archives of languages of the world, UAL 20.241–7 (1954), and Second report, IJAL 21.83–8 (1955). Magnetic recording was employed particularly for use in analysis of natural languages in 1952 and 1953, when five informants (speaking Arapaho, Blackfoot, and Shawnee; Crow and Hidatsa) were brought to the Linguistic Institute at Indiana University; this stimulated Hayes (see above) and others to make tape recordings of the same magnitude on subsequent field trips. In my own field work with Hopi in 1954 I obtained a 100-per cent sample of recording on tape; that is to say, I did not write anything on paper in the presence of an informant without also recording the informant's voice on tape. However, it was possible to hear Hopi spoken in the homes of informants and in public places where it would have been inappropriate to record either on paper or tape.

¹⁵ If we seek illumination in deductive systems of terminology like Hjelmslev's Prolegomena, we find a definition of ‘analysis’ ('description of an object by the uniform dependences of other objects on it and on each other') which appears to exclude free variants from ‘analysis’ unless the last of three classes of dependences ('A and B do not depend on one another') is applicable to free variants ; but an ‘analysis’ which includes free variants by such definition will fail to reach its goal (‘to construct ... texts in the same language‘).

The phrasing of this footnote is influenced by Einar Haugen's review of Prolegomena to a theory of language (IJAL 20.247–51).

¹⁶ The ‘preaspirated stops’ of Second Mesa (Mishongnovi) would accordingly be written -hS; ‘clusters with h’ would be written -h·S (S for stop, C for any consonant, V for vowel). One of the differences between two-syllable sequences of the shape CVhSV and CVh·SV is that h belongs with the second syllable in the former (CV + hSV) but with the first syllable in the latter sequence (CVh·+ SV) ; in general, consonants which have mora value are in syllabic final. In sound correspondences between Hopi dialects, however, such a sequence as CVh·SV is apt to be matched by CV·SV in another dialect, with the same sum in moras, namely three, but with moras differently distributed; stated in the terms used in §8, frame total may remain constant though frame shape changes.

¹⁷ Each of the three subclasses shares habits in respect to consonant clusters ; loose contact continuants are not involved, the other two subclasses of continuants are involved in the free-variant and/or phonemic-distinction problem treated in §5.2. Other subclasses of continuants could also—or alternatively—be set up, for example /v h/. In dialects in which /h/ occurs infrequently in clusters, /v/ appears as [v] more often than as [f] ; and, conversely, where /h/ occurs more frequently in clusters, /v/ appears as [f] more frequently also.

¹⁸ Even if it should turn out that the distinction between -n and -nh is nowhere phonemic —not even in Second Mesa, where two series of continuants are set up by Whorf—the alternative orthography proposed here (-n vs. -nh) is still not embarrassed by any additional phonemes, since both n and h are otherwise attested; the reductionism then obtainable would be morphophonemic rather than phonemic.

¹⁹ Because there is phonetic parallelism between kw, ηw, and ηy (not to mention [ky], an allophone of k), it is tempting to seek one answer for all—clusters or unit phonemes. But there are better reasons for considering each on its own merits.

Since the w in kw is voiced or voiceless as a positional variant in dialects in which the unit phoneme /w/ is not, /k^w/ is written as a unit phoneme rather than as a cluster of two consonants. In respect to voicing, /k^w/ is then analogous to /v/, whose allophones—[f] ~ [v]—though differently distributed, are complementary in all Hopi dialects.

Since the w in ηw behaves in respect to voicing as does /w/ otherwise, yw is written as a cluster of two phonemes. In this respect, ηy is analogous to ηw, hence ηy is also written as a cluster of two phonemes.

Of the three considered, k^w appears commonly in frame initial (§8), ηw rarely, ηy never. The following can be said for the alternative solution which takes each of the three to be unit phonemes: (a) that all three may occur in syllabic initial, and (b) that, accordingly, syllabic boundaries will be clearly indicated if they are written as single letters with superscript. But other consonant clusters than ηw and ηy may also occur in syllabic initial (for examples see §8). Syllabic boundaries are for the most part unambiguous, according to either solution.

The following can be said for the preferred solution : (a) that it takes into account the statistics as well as the algebra of distribution (only /k^w/ appears frequently in frame initial) ; and (b) the phonetic data given on voicelessness favor only /k^w/ as a unit phoneme,

²⁰ Whorf's transcription for Second Mesa vowel length distinguishes long vowels and short vowels, or rather two degrees of short vowels, short and overshort (‘clipped‘). This is comparable to my own earlier phonetic transcription of Tübatulabal, in which I distinguish short vowels and long vowels, or rather two degrees of long vowels, long and overlong. Since he did not recognize length phonemes as co-occurring with consonants as well as with vowels, Whorf's transcription of Hopi length is not strictly comparable to the present phonemicization.