262 Some Consonant Alternations: Synchronic and Diachronic Views Wayne Leman Summer Institute of Linguistics

1. Introduction Several interesting consonant alternations occur in Cheyenne. Some find their sources of alternation in Proto- Algonquian. Others are apparently unique to Cheyenne. A preliminary inventory of Cheyenne is: p, t, (c), k, ?, s, s, (x), h, v, m, n, a, e, o. Disputed elements are parenthesized. This list is essentially that of Davis (1962)1 minus his voiceless m and n. are either phonemically high- or low-pitched. Pitch sandhi rules adjust phonemic pitches yielding five phonetic pitches (Leman, in press): raised high=*, high=', mid=~, lowered high=v, and low=v or unmarked. Vowels devoice (') predictably (Leman and Rhodes 1978). 2 . Status of c Frantz (1972) removed c (orthographic tjs) from the list of Cheyenne phonemes. It is an of (1) /t/ by dissibilation of /t/ to c (orthographic ts_) , then assimilation of this to a following s_ or k, or of (2) /k/ by palatalization and affrication of /k/ to c, apparently as a relic of Cheyenne male speech. Note these processes below (disregard, for now, other details of the "derivations", such as pitch sandhi processes): (1) /etehke?6/ > etsehke?o > etseske?o > etseske?o [fc-ki?6] 'It is little.' /ete?kehahe/ > etse?kehahe > etse?kehahe [fci?kfhah] 'He is young.' (2) /hohkeha?e/ > hohkeha?e [h6hkhya?i] ~ h6htseha?e [hohchya?i] 'hat' 3. of /h/ The /h/ can be shown to have the preconsonantal allophones [h], [s], [s], and [x]. Note the allophonic alternations, below, with the past-tense /-h-/ which occurs immediately following the third person prefix /e-/: (3) emane 'He drank. He is drinking.' ehmane 'He drank.' esto?e 'He got up (from bed).' essevano 'He skated.' eskoka?a?xe 'He jumped.' esseese 'He was prostrate.' exho?soo?e 'He danced.' exxaa?e 'He urinated.' 263

The basic process here is one in which /h/ assimilates (or dissimilates) to a following consonant. The advantages of treating these alternations as allophanic were recognized by oavis (1962), basically working within a structuralist framework, and by Frantz (1972), working within the frame­ work of generative phonology. Some versions of structuralist (classical) phonology would, of course, require calling the alternations in (3) "morphophonemic" since h, s, s, and x can be shown to contrast in some intervocalic-environments. 2 In this paper, however, I shall regard alternations that can be accounted for purely on phonetic grounds, such as those of (3), as being "allophanic", even if the segments involved are in clear phonemic contrast in some other environments. I shall reserve the term "morphophonemic" for alternations for which we require at least some reference to morphological (grammatical) information.

4. Status of x Davis and Frantz observed certain restrictions on the occurrence of s and x. Davis (1962:37) correctly stated that "s never occurs-preceding a other than e." Frantz (1972:13) formulated a rule stating that s is backed to ~preceding ~ and ~, equivalent to rule (7) below.

4.1 It can be seen from examples such as (4)-(6) that s and ~ regularly alternate.

(4) ne~e 'two (of)' ne~a ' two (times) ' eneseo?o 'They (animate) are two in number.' ene~6ho 'They () are two in number.' 'They (inanimate) are two in number.'

(Sl se?se 'duck' se?~o 'duck(s) (obv.)'

( 6) e6neseotse 'He is in pain. ' e6ne:§:e6hta6htse 'His leg hurts. ' ('leg' (medial)= / -6hta/) e6nexase6htse 'He has a bellyache.' ('belly' (med.) =/-ase/) e6nex6hta6h tse 'He has a stomach ache.' ('stomach' (med.) =/-ohta/)

S occurs only before e, while x occurs only before ~ or ~, that is, only before a [+back]-vowel. We, therefore, can formulate rule (7):

(7) ~-Backing (regressive) 3

V s + X I (?) [+back] 264 4.2 Other forms (8)-(11) show that there is also a progressive s-Backing rule (12): (8) eoveSe 'He lay down, went to bed. ' eheamaxe •He lay on his side.' ehaoenaxe 'He prayed lying down.' (9) nemaese 'You barked at me.'1* nemeoxe 'You fought me.' (10) nee?ese 'You sliced me.' nea?taxe 'You accidentally cut me.' (11) nee?se 'You are afraid of me.' neho?xe 'You have me.' Following a [+back] vowel, s is backed to x: (12) s-Backing (progressive) v * / r (?) [+back' 4.3 Now, notice how similar rules (7) and (12) are. They are the same, except for the difference as to whether the conditioning vowel precedes or follows the segment to be conditioned. Because of this, it would be advantageous if we could collapse the regressive (7) and progressive (12) assimilation rules into one single assimilation rule (13): (13) s-Backing (?) x / V [+back] Rule (13) simply says that /s/ will be backed to x if it is preceded or foil owed (or both preceded and followed) by a [ -tback ] vowel . Neighborhood rules have a particularl y dubious status in phonological theory. There have been ma ny putative neigh- borhood rules which, upon closer examina tion, are found to be treated better as separate rules. Th ere is at least one complication of Cheye nne data that could cast doubt on the wisdom of collapsing (7) and (12) to the neighborhood rule (13). Progressive s Backing is blocked at a preverb boundary. Regressive s-Backing applies anywhere in a form. Note that the a of /ni.-/ 'first perso n prefix', the second a of /-saa-/ negative preverb', and the o of /mo-/ 'dubitative prefix' d o not cause backing of a following (14) naseese /s/, as seen in (14) 'I am prostrate.' nasaasdenohe 'I did not rob him. moseseotsehehe 'He must be free.' 265 The /s/'s in question occur immediately following a preverb boundary. It remains to be seen whether this exception to progressive assimilation is serious enough to keep us from capturing the facts of s-Backing in a single neighborhood assimilation rule. 4.4 There are a few lexical items which are potentially troublesome for our statement of progressive s-Backing. They can, however, be shown not to be exceptions to rule (12), after all. The for 'it is brown' is emoseskano. Superficially, it would appear that we have a case of a [+back] vowel, o, followed by s which should have backed to x according to- (12). But we must note that speakers alternate the pronunciation emoseskano with emoseskano. We know from other data that one phonemic ("phonemic" in the sense used in this paper) source of s is /s/. For instance, note (15) and (16) : (15) nemane?sese 'You (sg.) made me drink.' (16) nemane?seseme 'You (pi.) made me drink.' The first s of (16) is derived from /s/, as seen by comparison with (15), by assimilation to the following /s/. We now can conclude that the phonemic spelling of 'it is brown' is /emosehkano/. The pronunciation emoseskano is slightly closer to the phonemic spelling than is the alternate pronunciation emoseskano. The same two lines of argumentation allow us to conclude that the phonemic spelling for 'that's all, that's the end of the story' enehe?estosese ~ enehe?estosese is /enehe?ehtosese/. An AI final meaning 'to drink' is phonetically [asi?si]. It is spelled orthographically as -ase?se in Leman (1979: 157). There is, however, the possibility of spelling the final phonemically as /-ahse?se/ (Rhodes, personal communi­ cation) . (The /h/ preceding /s/ would prevent progressive s-Backing from occurring.) Whether by accident or design, that was the approach taken to spell one of the forms having this final, in the English-Cheyenne Student Dictionary (1976:135) . 'He's drunk' is spelled there as e-nono?tova-hse?se5. (However, the word for 'soda pop (cold drink)' in this dictionary (1976:104) was not written with an hs cluster in the AI final 'to drink', but, instead, with a single segment £6 . ) An hs cluster in this AI final would at first look unusual7, but its presence in phonemic spellings is not unprecedented. There is already a precedent for /h/ preceding /s/ in words containing a morpheme-final /h/ preceding root-initial /s/, e.g. (17) /e-h-seese/ esseese 'He (past tense) was prostrate.' (18) /mah-seseotet/ masseseotsestse 'when he is free' 266 4.5 When all the facts are considered, I am led to say that x is not a phoneme of Cheyenne. It is, instead, an allophone of /s/, or of /h/ by dissibilation in an /hh/ or /hx/ cluster, or of /h/ by assimilation in an /hs/ or /hs/ cluster. This reduces the phonemic inventory of Cheyenne to 13 segments: p, t, k, 7_, s_, s, h, v, m, n, a, e, o, plus a phoneme of pitch.T 5. Phonemic assignments Building upon the discussion above, and upon data and rules in Leman (1979), we can now observe some of the intricacies of Cheyenne phonemic assignment (19) . (19) Cheyenne phonemic assignments /W > [h].[si , s], [x] (See §3, above.) /s/ > [s],[s i (See §4.4, above. /£/ > [S],rx i A/ > [t], [te], [ts] (See §4, above.) A/ > [k]. [ts] (See §2, above.) (See §2, above.) It is important to notice that a Cheyenne phone can have more than one phonemic source: (20) Multiple phonemic sources [s] < /s/, /h/, /s/ [x] < /h/, /s/ [s] < /s/, /h/ [ts] < /t/, /k/ One must be careful, when phonemicizing, that phones are assigned to the proper phonemes. Study of the lexicon and of alternations within paradigms enables one to make these assignments without too much difficulty. 6. Diachronic support Finding x to be an allophone of s has strong diachronic support. Note the PA *s sources foF both s and x in the Cheyenne forms below. (PA data are from Aubin 1975 ) (21) *si:?si:pa /se?se/ Ie?se 'duck' ?*si:?s ? /se?xo/ se?xo 'duck(s) *seka:kwa /sa6n/ (obv.)' *sekiwa /esae/ xao?o 'skunk' ixaa?e 'He *wa:si /v6se/ urinated. ' *ni:wasi /navevose/' v°xe 'hole, den' *-eakesyi] /-tones/ naveyoxe 'my pack' -tonese,-tonexa_'bowel, natonese belly'my belly' ' he tonese 'his belly' hetonesev6tse 'their natonexan6tse bellies' 'our (excl.) 1 *-hsine /-se(na)/ -ie(na),-xe(na) bellies' 'fall, lie e.g. eovese (AI final)' 'He lay down.' 267

ehaoenaxe 'He prayed lying down.' eheamaxe 'He lay on his side.' *ko?si /e?seha/ e?seha 'Fear him!' *mi:ka:si /meoseha/ meoxeha 'Fight him.; ' *ne?si /na?seha/ na?xeha 'Kill him:' Space does not allow a full treatment of PA:Ch correspondences here. Segmental correspondences should be fairly obvious. Vowel correspondences are PA *i, *o: Ch e; PA *a: Ch o; PA *e: Ch a. The Cheyenne reflex of PA vowel length is phonemic high pitch. 7. Morphophonemic alternations Now let us consider examples of alternations between full phonemes, namely, instances of morphophonemic alternation. The alternations here will be ones which cannot be predicted solely on phonetic grounds. 7.1 /£/ - /a/ We saw in (4) that the /§/ of the Cheyenne root 'two' has the allophone x preceding a [+back] vowel. We can say that s and x allophonically alternate in the root 'two'. But some Cheyenne forms with the meaning ' two' have neither s_ nor x, but, instead, s_. The s appears preceding a L+back] vowel. Compare the forms of (4) with those of (22). (22) enes66hne 'She two-birthed (gave birth to twins, or, has borne children twice).' neso?e '20' tsenesoxe?ohe '2 o'clock' enesovavenestse 'He is bilingual (two-speaks).' The morpheme 'two' in (4) is phonemically /nes(e)/ while it is /nes-/ in (22) . Can we predict which spelling will occur in a form? My first guess is that /nes(e)/ occurs as a preverb or root, while /nes-/ occurs as an initial (accommodation would need to be made for noun constructions). Perhaps the tighter "bonding" that an initial has with the following medial or final requires the morpheme-final /s/ rather than /s/. Or, there may be a diachronic explanation for the morphological alternation of /s/ ~ /s/ in the morpheme 'two'. Another /s/ ~ /s/ alternation occurs in transitive finals meaning 'to cut', as seen in (23) and (24). In paradigms of of cutting /s/ occurs whenever the Cheyenne TA theme vowel /-6/ follows the final (indicating a direct person relationship). The phoneme /s/ (with allophone [x]) occurs elsewhere. (23a) nae?eso 'I sliced him.' ,' ' (23b) nee?ese 'You sliced me. (23c) e?eseha 'Slice him: 268 (23d) nae?esestse 'I sliced myself.' (23e) nae?exam6ho 'I sliced him (obv.).' (23f) ee?exa 'He sliced it.' (23g) e?ex6tse 'Slice it:' Note that it is only the TA theme vowel /-6/ that requires the 'to cut' final /s/ instead of /s/. x (< /s/) appears in (23g) preceding the /6/ of the TI imperative suffix /-6t/. (24) na?tas_o 'I accidentally cut him.' nea?taxe 'You accidentally cut me.' naa?taxestse 'I accidentally cut myself.' naa?taxamoho 'I accidentally cut him (obv.).' ea?taxa 'He accidentally cut it.' An /s/ ~ /s/ morphophonemic alternation occurs with the morpheme meaning 'hole, den' (2 5). (25) voxe /vose/ 'hole, den' evosphone /evosohone/ 'He dug.' evosana /evosani/ 'He dug it by hand.' ev6s_6ha /evosoha/ 'He dug it by tool. ' Interestingly, PA forms with a meaning of 'hole, den' have an alternation, involving PA *9_ and its palatalized counter­ part *s. Following the pattern in this paper, I would call the PA alternation an allophonic one. (26) wa: si 'hole, den' wa:6ali 'holes, dens' wa:9ehke:wa 'He digs a pit 7.2 /h/ - /s/ Many Cheyenne forms illustrate morphophonemic alternations between /h/ and /s/. Note Cheyenne forms, below, with some PA sources.12 The Cheyenne alternations reflect allophonic alternations of PA. While a phonetic connection may be seen in the Cheyenne alternations, Cheyenne forms often require some reference to to account for the alternations. (27) nae?ho *neko?£a:wa 'I am afraid of him.' nae?ha *neko?9ekwa 'He is afraid of me.' nee?se *keko?si 'You are afraid of me. ' e?seha *ko?5i 'Fear him'.' (28) etahoo?e /etahoe/ *te:h9apiwa 'He is seated upon, riding. ' etaxeoo?e /eta^eoe/ *te:hjika:piwa 'He stands upon.' (29a) namane?seho 'I made him drink.' (29b) nemane?sese 'You made me drink.' (29c) emane?sehe 'He was made to drink.' 269 (29b) and (29c) show /s/ and /h/ in clear phonemic contrast in Cheyenne .

(30) ema?haata *me?9_ekekwa 'He is big. ma?xe- /ma?se-/ *me?si- 'big'

The Cheyenne alternations in (27)-(30) reflect the PA alternation *9_ ~ *s resulting from palatalization of *8 before a or . At least in the documented cases, above, Cheyenne /h/ is a reflex of pre- glottalized or preaspirated PA *9_, while Cheyenne /s/ is a reflex of preglottalized or preaspirated PA *s. Because of the complexities of the historical development of Cheyenne, reflexes of PA *8_ ~ *s do not have a purely phonetic connection. Forms such as (29b) and (29c) require reference to non-phonetic information to account for the occurrence of /s/ or /h/. Numerous forms show that /s/ and /h/ occur in exactly the same phonetic environment. 7.3 /t/ - /a/ (~ /h/) Other Cheyenne forms showing reflexes of the non-cluster PA *6_ ~ *s alternation consistently have a /t/ ~ /s/ morphophonemic alternation in the TA final.

(31) namedto /namedt6/ *nemi:ka:9a:wa 'I fought him. ' nemeoxe /nemedjie/ *kemi :ka: Si 'You fought me. meoxeha /medSeha/ *mi :ka :|i. 'Fight him: ' emeohe /emeohe/ ?*mi:ka:9a:wa 'He was fought.'

(32) ehet6ho *e9_e: wa 'He says so to him (obv.! heseha *e|i ' Say so to him'. '

(33) namaeta *nemeki9_ekwa 'He barked at me.' emaet6ho *mekie_e :wa 'He barked at him (obv.) maeseha *mekisi 'Bark at himl' Cheyenne TA verb finals with the /t/ ~ /s/ alternation require /t/ when followed by the /-6/ or /-a/ theme vowels (i.e. [+back] vowels), but /s/ when followed by the front vowel, /e/. The /t/, of course, is a reflex of PA *Q_, while /s/ is a reflex of PA *s, the palatalized counterpart of *9_.

7.4 The one exception to the /t/ ~ /s/ alternation in Cheyenne verbs is in the verb 'give to'. Here the alternation is t - ts.

(34) nameto /nameto/ *nemi : l_a : wa 'I gave it to him.' nemetse /nemete/ *kemi:li 'You gave it to me. ' emetse /emete/ ?*mi:l_a:wa 'It was given to him. ' 270 emetdho /emetdhd/ *mi:le:wa 'He gave it to him (obv.).' metseha /meteha/ *mi:li 'Give it to him: ' The most common Cheyenne reflex of PA *1 is /t/, as it is here. The t in this paradigm allophonically alternates with ts, consistently retaining the phoneme /t/ reflex of the PA *1 that appears throughout the PA paradigm of this verb. In §7.3 there was a morphophonemic alternation in the Cheyenne data, reflecting the allophonic alternation of the PA source segments. In §7.4 there was no PA alternation with *1, and the Cheyenne reflexes have only an allophonic alternation. 7.5 Cheyenne has other morphophonemic alternations which reflect PA alternations. A few forms are sufficient to show this. (35) /s/ - /ht/ (/s/ < *c; /ht/ < *t) -vese- /-vese-/ *wi:di 'along with, with (pre­ verb) ' evestoemoho /evehtoemoho/ *wi:tapime:wa 'He sits with him (obv.)' (36) /?s/ ~ /?t/ (/?s/ < *?c; /?t/ < *?t) -a?_se- /-a?se-/ *pe?ci 'by accident' ea?taxa /ea?taxa/ *pe?tesamwa 'He accidentally cut it.' The underlined Cheyenne segments in (35) and (36) reflect PA palatalization of *t to *c preceding a front vowel or semi­ vowel . The Cheyenne reflexes of the PA allophonic alternation alternate morphophonemically. 8. Conclusion Cheyenne has allophonic alternations with reflexes of PA *1 (t ~ ;ts) and of PA *£ (s_ ~ x) . There are morphophonemic alternations with reflexes of PA *6_ - *s and of PA *t ~ c. In general, it appears that we must account for Cheyenne morphophonemic alternations on diachronic grounds, while the allophonic alternations are accounted for solely on synchronic grounds. Such correlation of synchronic analysis with diachronic information is not to be rejected out-of-hand. The role of diachronic information in synchronic analysis has, of course, often been debated in the history of phonological studies. While few would feel comfortable in allowing diachronic factors to determine completely the course of synchronic analysis, it does appear that diachronic in­ formation must be more than just a step-child in synchronic phonological analysis. Speaking to this issue, Greenberg recently pointed out (1979:275): 271 ...recent years have been marked by an increased perception of the relevance of diachronic studies for the basic problems of linguistic science. As a minimum, the strict separation of synchronic and diachronic studies—envisaged by Saussure, but never absolute in practice--is now widely rejected. We have seen that there is a definite relationship between what I have called "morphophonemic alternations" in Cheyenne and historical information from Proto-Algonquian. Writing from the viewpoint of "natural phonology", Donegan and Stampe (1979:144) explain how "processes" (approximately equal to what I have called "allophonic alternations") differ from "rules" (approximately the same as my morphophonemic alter­ nations) : First, and most importantly, processes have synchronic phonetic motivation and represent real limitations on speakers' productions. Rules lack current phonetic motivation; they are sometimes the historical result of 'fossilized' or conventionalized processes which have lost such motivation... German umlaut is an example. On the other hand, processes lack positive semantic or grammatical functions, which some rules (like umlaut) [and many Cheyenne morphophonemic alternations—W.L.] do have. Hopefully, this description of Cheyenne consonant alter­ nations, with views both on synchrony and diachrony, can not only present the data, but also be a constructive addition to the discussion of the role of diachronic information in synchronic analysis.

NOTES 1 Davis' tY = i. Davis' w = v used here. 2 Note forms such as nehe 'that one (an.)', hese 'fly', mese 'tick',- vose 'hill', voxe 'hole, den', nahkohe 'bear'. We shall see, below, that only /h/, /s/, and /s/ contrast phonemically. The x is an allophone of /s/. 3 Synchronically, s is chosen as the more "basic" of the two segments. It is, for example, the segment occurring in grammatically unmarked se?se 'duck', as opposed to x in grammatically marked (obviative) se?xo 'duck(s) (obv.)'. This conclusion is supported diachronically. Numerous cases can be found where both s and x are reflexes of the single PA segment *s. PA *x, on the other hand, does not have s or x reflexes in Cheyenne. 272

" Throughout this paper, 'you' in a gloss (whether expressed as here, or implied, in a command) is singular, unless other­ wise indicated. ' I would now spell this as enondtovahse?se. 6 I would now spell 'soda pop' as todomahse?sestotse [todomasi?sistoc] /todomahse?sehtot/. 7 Ideally, we would want comparative evidence to support the postulated /h/ of the cluster. 8 After having come to these conclusions about s and x, and written the basic draft of this paper, I discovered a brief workpapers article by Rhodes (1972) which basically said the same thing. Rhodes, too, listed only 13 segmental phonemes (he happened to list x instead of s), and said in a footnote (1972:56): "Space does not permit me to argue that x and s are the same underlying phoneme, more than to say that they never contrast in stems." 9 The Cheyenne form is reanalyzed, based on the PA form with a third person possessor, *wi:wasi. Cheyenne frequently reanalyzes possessed nouns like this, basing them on the PA form having a third person possessor. 10 From Siebert (1975:319). I have shown elsewhere (Leman, to appear) that PA *k: Cheyenne n is a common segmental correspondence. 1' /ni/ deletes in singular subject affirmative indicative forms. Compare forms in the text with eovesenahe 'Did he lay down?', nasaa?ovesenahe 'I did not lie down', naovesename 'we (excl.) lay down'. Pitch assignment and resultant pitch sandhi are somewhat complex, but need not concern us here (see the relevant footnote in Leman, in press). ' Some of the PA forms from here on are extrapolated from related forms in Aubin's list.

REFERENCES AUBIN, George F. 1975 A Proto-Algonquian Dictionary. Ottawa: National Museums of Canada. DAVIS, Irvine 1962 Phonological function in Cheyenne. IJAL 28-36- 42. 273 DONEGAN, Patricia J. and David Stampe 1979 The study of natural phonology. In D. Dinnsen (ed.), Current Approaches to Phonological Theory. Bloomington, Indiana: University of Indiana Press. ENGLISH-CHEYENNE STUDENT DICTIONARY 1976 Lame Deer, : The Language Research Dept. of the Northern Cheyenne Title VII ESEA Bilingual Education Program. FRANTZ, Donald G. 1972 Cheyenne distinctive features and phonological rules. IJAL 38:6-13. GREENBERG, Joseph H. 1979 Rethinking linguistics diachronically. Language 55:275-290. LEMAN, Wayne 1979 Cheyenne Grammar Notes. Crow Agency, Montana: Crow Bilingual Education Materials Development Center. (Retitled, A Reference Grammar of the Cheyenne Language. 2nd edition. Greeley, Colorado: University of Northern Colorado, Museum of Anthropology.) In press. Cheyenne pitch rules. (In press, IJAL). To appear Evidence for a PA *k: Cheyenne n correspondence (To appear, IJAL). LEMAN, Wayne and Richard Rhodes 1978 Cheyenne vowel devoicing. In W. Cowan (ed.), Papers of the Ninth Algonquian Conference. Ottawa, Carleton University. RHODES, Richard 19 72 Cheyenne vowel devoicing and transderivational constraints. In C. Warner (ed.), Work Papers of the Summer Institute of Linguistics. University of North Dakota, Grand Forks. SIEBERT, Frank T. 1975 Resurrecting Virginia Algonquian from the dead: the reconstituted and historical phonology of Powhatan. In J. Crawford (ed.), Studies in Southeastern Indian Languages. Athens, Georgia: University of Georgia Press.