Morphophonemics of Ikema Miyakoan 1 Takubo, Yukinori NINJAL ytakubo@ninjal.ac.jp 1. Introduction This paper deals with the morphononemics of Ikema, a variety of Miyako Ryukyuan, spoken in Ikema Island, Sarahara in Irabu island and Nishihara in Miyako Island. I will first give a short sketch of the phonology of the language in order to provide the transcription system. I will then discuss the morphophonemic system of the topic and the accusative forms, which make use of similar rules. I will also make references to verbal morphophonemics when necessary. 2. 2.1. Ikema Phonology and the transcription system Phonology Vowels Four short vowels are phonemically distinguished : /a,i,u,ɨ,/. Long vowels are phonemically distinct from the corresponding short ones: /a:,i:,u:,ɨ:/. The distribution of the long counterparts of /o/ and /e/, i.e., /o:/ and /e:/, is very restricted, appearing only in such words used as answer particles, 'e: (Yes)' and 'o: (Yes (honorific))' or as a sentence final particle, 'do:'. 2 [ɨ] and [i] appear to show complementary distribution, [ɨ] only appearing after /s, z(=dz), c(=ts)/in Ikema. The change from proto Miyako /i/ to /ɨ/ probably occurred in all the consonants in Ikema, as were the case in other dialects in Miyako but was lost and returned to /i/ in consonants other than the sibilants. We have to treat them as the realization of two distinct phonemes because they contrast in the same (albeit limited) environment, constituting a minimal pair as in (1). The contrast was created by the historical change from /e/ to /i/ which occurred in all the varieties of Ryukyuan. (1) /siba/(worry) : /sɨba/(lip) ：/suba/(soba noodle) , /muus-ɨ /(burn-CONCL) :/muus-i/(burn-IMP) Consonants The following consonants are distinguished phonemically. 1 2 The data in this paper is based on the fieldwork that I have conducted at Nishihara since 2006. The main consultant has been Mr. Hiroyuki Nakama (born in 1947) . I also checked the data with other speakers in the village (Age 70-75). For the actual pronunciation and the discourse data, the readers are referred to the Digital Museum site that I constructed: http://kikigengo.jp/nishihara/doku.php?id=start. Parts of the paper include materials I presented in Takubo (2015, 2018), Celik and Takubo (2013, 2014). Short /e/ and /o/ can appear in borrowings. (2) /p, b, t, d, k, g, c (=ts), z (=dz) , s, f, r, m, n, ɴ̥, ɴ, j,w, h, j/ /hu/[ɸu:] and /fu/[fu:] contrast: /hu:/(do-progressive) vs./fuu/(come-CONCL) . In other varieties of Miyako, there is a contrast between /m/ and /ɴ/(realized as [n] or [ŋ]) in syllable final position. In Ikema this contrast has been lost to become /ɴ/ in generations younger than 70. /ɴ/ is usually realized as homo-organic nasals depending on the following consonants. There is no phonemic contrast between [dz] and [z]. They show free variations. /c/, /z/ become palatalized before vowels except before /ɨ/. {/c/, /z/}+{/u/, /a/}and{/c/, /z/}+{/ju/, /ja/}, therefore, neutralize and lose contrast. (3) /cɨ:/(breast milk) vs. /cju:/(dew) *[tsu], *[tsa]vs.[tɕu], [tɕa]: e.g./cju:/ [tɕu:](dew) , /acja/ [atɕa](tomorrow) For other consonants /u/ and /ju/, /a/ and /ja/ contrast. (4) a. /ku:/(suffer) /kju:/(today) b. Geminate /zz/ do not palatalize: e.g. /zzu/(fish) c. /cc/ palatalizes. E.g. [umattɕu] (=/umac+u/). There is no [ttsu]. /p/ only appears in loan words and in very limited native words. /p/ in other Miyako dialects corresponds mostly to /h/ in Ikema． (5) /pa:/ (grandpa), /piiki/(drill a hole) Syllable Structure Nasal consonants appear syllable initially in addition to syllable finally. If they appear as geminates and precede a vowel or obstruents /mm/ and /nn/ do contrast as in (6). When they appear as an independent syllable or appear before obstruents, /m/ and /n/ get neutralized and do not contrast as in (7). There is no contrast between /m/, /n/ before obstruents. [nta]～[mta] are free variations. (6) /nna/ (turban shell), /nnna/ (all) vs. /mma/ (mother), /mmma/ (the head of female priests) (7) /nta/[nta]~[mta] (earth), /nnta/[n:ta]~[m:ta] (frog), /nndi/[n:di]~[m:di](yes) vs. /nbu/[mbu](heavy), /nba/[mba][disagree] The combination of nasals with distinct place of articulation does not exist, that is, there is no [nm] or [mn]. The syllable structure of Ikema is as in (8)．In addition to (8), syllables consisting only with nasals are possible as in (9). (8) (C1) (C2) (G) V1(V2) (C3) (9) /ɴɴ/(=/ɴ:/) (sweet potato) , /ɴ̥ɴ/[n̥n](step on, scoop (water)) Consonants with special distribution /ɴ/ appears before consonants except for /j/, /w/ and /h/ (10) /ɴta/(earth), /ɴsu/(miso), /ɴ:ku/(issue), /ɴccɨ:/(soup) /ɴ̥/ appears before /ɴ/, /n/, /m/. It does not appear word initially. /ɴ̥/ can, therefore, be thought of as a marker for devoicing nasals. (11) /ɴ̥na/(rope), /ɴ̥mu/(cloud) Geminates All the consonants except for /j,w,h, ɴ̥/ can be geminates. /r/ is realized as [ll] when it is geminated. (12) /t:, k:, c:, f:, s:, z:, r:, v:, m:, n:/ E.g. /t:a/(tongue), /k:unucɨ/(9 piecese), /c:jui/(break down), /f:a/(child), /mul:u/(Japanese horse mackerel) /s:a/(foot), /z:a/(father), /v:adi/(will.sell), /m:a/(mother), /n:a/(turban shell) There is a word minimality constraint in Ikema such that a free morpheme must have at least two morae. There are no morphemes consisting only of one mora. Words with syllable initial geminates and consonant clusters (C1) may contain only one vowel, which suggests that the first member of the geminates and consonant clusters form a syllabic consonant and is considered as constituting a mora. 2.2. Transcription system Below I will give examples in the phonemic transcription in italic without //. The following conventions will be adopted mostly for convenience.  Doubling of vowels and consonants will be used instead of /:/．  Voiceless nasals will be interpreted as a devoicing instruction and will be transcribed as /h+nasal consonant/. /h/ is interpreted as one mora long. (13) hnu [ɴ̥nu](horn, yesterday), hmu[ɴ̥mu] (cloud), hn[ɴ̥ɴ](scoop water, wear (a shoe)), step on)  Syllable initial /ɴ/, the neutralized form of /m/ and /n/, will be transcribed as /n/, because it can be distinguished from /n/ that contrasts with /m/ by the distribution, the former appears syllable initially, syllable finally and before obstruents, and the latter syllable initially. (14) a. nn[ɴɴ](sweet potato)，nta [nta](earth)，ssan [ssaɴ](louse) b. nna[n:a](turban shell), naa[na:](name): mma[m:a](mother), maa[ma:](trace) 3. 3.1. Morphophonemics of nouns: Focusing on topic and accusative forms Topic and accusative forms The topic morpheme in Ikema has the following allomorphs. (15) Topic forms a. After nouns ending in /a/: =a ffa (child) ffaa b. After nouns ending in /u/: =u zzu (fish) zzuu c. After nouns ending in /i/: =a saki (rice wine) d. After a long vowel or diphthong: =ja sakjaa suu (side dish) suuja, mai(rice) maija e. After nouns ending in Cɨ (Consonant+ɨ):(C)Ca dusɨ(friend) dus-sa (friend) f. After nouns ending in /ɴ/: na in (sea 3, dog) inna The distribution of allomorphs parallels that of the topic. (16) Accusative forms a. Nouns ending in /a/ or /u/: u b. Nouns ending in /i/: (j)uu c. ffa ffau, zzu zzuu saki sakjuu Nouns ending in long vowels or diphthongs: ju suu (side dish) suuju mai(rice) maiju d. After nouns ending in Cɨ (Consonant+ɨ):(C)Cu dusɨ (friend) dussu (friend) e. After nouns ending in /n/: (n)nu in (dog, sea) innu Previous studies 4 3 For older speakers (people in their late 80’s or older), /im/ is used for ‘sea’ as is the case with speakers of other parts of Miyako, in which case, the topic form is /imma/, the accusative form /immu/. 4 Shimoji (2008, 2017, 2018) is an exception. He posits /a/ for the topic and /u/ for the accusative as the underlying forms of Irabu Miyakoan. Shimoji’s observations and the explanation thereof apply mutis mutandi to Ikema forms. The present paper is an attempt to explain the phenomena in terms of In previous works on Miyako dialects (Karimata (1992) a.o.), the basic (or underlying) form of the topic morpheme is usually hypothesized to be /ja/, that of the accusative morpheme /ju/, from which various allomorphs are derived 5. (17) /ja/ as the underlying form for the topic and /ju/ for the accusative a. Nouns ending in /a/ and /u/: delete /j/. ffa=ja>ffaa, zzu=ju>zzuu b. Nouns ending in /i/: delete /i/ and lengthen /a/ or /u/. saki=ja>sakja>sakjaa saki=ju>sakju>sakjuu c. Nouns ending in long vowels or diphthongs: no change. suu (side dish) suu=ja, suu=ju, mai=ja, mai=ju d. Nouns ending in Cɨ: delete /ɨ/ and undergo forward assimilation. dusɨ=ja>dussa dusɨ=ju>dussu e. Nouns ending in /n/: undergoe forward assimilation. in=ja>inna in=ju>innu In contrast to (17), I assume that /a/ is the underlying form for the topic and /u/ for the accustative. (18) Underlying form /a/ and /u/ as the underlying forms a. After nouns ending in /a/: /a/ is added without any change. ffa=a> ffaa b. After nouns ending in /u/: /ua/ is changed to /uu/ for the topic and /u/ is added for the accusative. zzu=a> zzu=a>zzuu zzu=u>zzuu c. After nouns ending in /i/: glidification of /i/ and lengthening of /a/ or /u/. saki=a>sakja>sakjaa saki=u>sakju>sakjuu d. After nouns ending in diphthongs and long vowels: insertion of /j/. mai=a>maija suu=a>suuja mai=u>maiju suu=u>suuju e. After nouns ending in Cɨ: deletion of /ɨ/, gemination of C and adding /a/. dusɨ=a>dussa dusɨ=u>dussu f. After nouns ending in /n/: gemination of /n/ and adding /a/. in=a>inna (partly) constraint based approach instead of derivational one with some changes in the description and the explanations. 5 Uemura(1992:802) seems to think that changes such as wa>a>ja occurred. But it is difficult to tell what exactly he was thinking when he says ‘incorporating the glide when it (Y.T. the topic form?) combines with front vowels (Y. T. He probably meant the change: ia>ja)’. 「前舌形の母音と結 びつくときのわたり音を取り込んで」 The derivation in (17) is not much more complex than (18): (17) needs the deletion of /j/ while (18) needs insertion of /j/. The diachronic perspectives clearly favor (15) over (17). If we posit the proto-Japonic form of the topic to be *pa, and that of the accusative to be *wo, the change *pa>ja or *wo>ju seem implausible. A much more natural change to be posited will be *pa>wa>a, *wo>o>u. It is, therefore more natural to posit /a/ for the topic, /u/ for the accusative as the underlying forms. In what follows, I will show that it is possible to give a simple account for the distribution of (15) and (16) by positing /a/ for the topic and /u/ for the accusative and assuming very general constraints in (19) 6 (19) Constraints (A）No three consecutive vowels：*VVV＞VVja 7 /j/ is inserted to avoid three consecutive vowels. The glide to be inserted can be either /j/ or/w/. /j/ is inserted because inserting /w/ violate constraint (B). (B) No rising diphthongs: Ikema does not allow the diphthongs /ia/ or /iu/. /i/, therefore becomes a glide. ia>ja, iu>ju (C) Constraint on the number of morae: The number of mora must be the same in the Input and the output. The glidification in (A) reduces the number of mora. So the vowel must be lengthened to keep the number of mora to be intact: Ci+a＞Cj+a>Cja>Cjaa (D) Syllable boundary must be kept: A geminate is formed by adding the same consonant as the preceding one to keep the syllable boundary: C+a>C+Ca Let us look at each case in detail. (A) is violated if /a/ or /u/ is added to nouns ending in a long vowel or a diphthong, so /j/ is inserted. （A）*VVV (20) Nouns ending in a long vowel suu (side dish)=a>suua>insertion of /j/>suuja. 6 The descriptions that follow are based on works with Kenan Celik (Celik and Takubo (2013). The original approach was optimality theoretic, but I changed it to the mixture of derivational and constraint based approaches to the ease of exposition. 7 This constraint does not apply when the word final vowel is /i/. (Michinori Shimoji: personal communication: July 30th, 2018): /faai/ (can be eaten)，/dooi/ (doo+i). If /j/ is inserted to avoid three consecutive vowels, the results are /faaji/ and /dooji/ respectively. I assume a constraint that dictates the avoidance of /ji/, a stronger constraint than (A). suu=u>suuu> insertion of /j/>suuju. (21) Nouns ending in a diphthong mai (rice) = a>maia>insertion of /j/>maija inau(tornade)=a>inaua>insertion of /j/>inauja mai=u>maiu>insertion of /j/>mai=ju inau=u>inauu>insertion of /j/>inau=ju (B) bans rising diphthongs. In order to avoid rising diphthongs /i/ becomes a glide. 8 Glidification of /i/ leads to a decrease of the number of morae. The vowel is lengthened by compensatory lengthening to keep the number of morae unchanged. (B) Glidification of /i/ in /iu/, /ia/ and (C) compensatory lengthening (22) banti(we) +a>bantia>glidification>bantja>compensatory lengthening>bantjaa banti+u>bantiu>glidification>bantju>compensatory lengthening>bantjuu As for (D), some assumptions will be necessary to apply this constraint. We assume that for nouns ending in Cɨ, /ɨ/ is inserted by some constraint dictating the ban on free morphemes ending in consonants. In Ikema, /ɨ/ is clearly heard in word final positions. In order to keep the constraint (D), a homo-organic consonant is inserted. (D) Syllable boundary must be kept (23) umac(ɨ) (fire) umac+a>*umaca>umac=ca umac+u>*umacu>umac=cu If we assume (D), words ending in /n/ can be explained in the same way 9. (24) in (dog) in+a>ina>inna in+u>inu>innu 8 *ua can also be part of this constraint. I make it a separate constraint because we need to explain why /ua/ does not become /wa/. 9 If /inna/ is derived as the result of a forward assimilation from /in/+/ja/, we have to posit an assimilation rule /n/+/j/>/nn/, which is implausible and not attested in Ikema because /nja/ is allowed and very common in Ikema: /anjaru/ (such), /njaan/ (not exist). There is, therefore, no reason to choose /inna/ over /inja/, which does not seem to violate any constraints. Hayashi (2013: section 2.1.3) posits /ɨ/ as underlying, disallowing words ending in consonants except for /N/. She posits a rule that /ɨ/ deletes because the diphthong *ɨV is not allowed in Ikema and word final C is geminated to keep the number of morae unaltered. (25) dusɨ=a＞dusa＞dus-sa One of the merits of this analysis is that it can account for the following alternation in verbal inflection. Ikema has the following alternation in negative forms. (26) Conclusive forms Negative forms sɨɨ (know) ss-an cɨɨ (angle, wear) cc-an cufɨɨ (make) cuff-an Hayashi assumes the basic verbal roots of each verb to be /sɨ-/, /cɨ-/, and /cufɨ/ and derives the conclusive forms as in (27) and negative forms as in (28). (27) Conclusive forms a. sɨ-i> forward assimilation>sɨɨ b. cɨ-i>forward assimilation> cɨɨ c. cufɨ-i>forward assimilation> cufɨɨ (28) Negative forms: a. sɨ-an>deletion of /ɨ/> consonant gemination>ssan b. cɨ-an>deletion of /ɨ/> consonant gemination > ccan c. cufɨ-an>deletion of /ɨ/> consonant gemination> cuffan The motivation for the two rules is the ban on the vowel combination *ɨV, but the means to avoid it differ between the conclusive and the negative forms, forward assimilation in the former, and deletion followed by consonant gemination in the latter. It is not quite obvious what the conditions are in the two ways of derivation. Another problem in her explanation is that she treats these verbs as ending in a vowel. The negative morpheme in Ikema has two allomorphs, -an for verb stems ending in a consonant and -n for those ending in vowels as shown in Table 1. Clearly as shown in (26) the negative forms for sɨɨ, cɨɨ and cufɨɨ behave as consonant stem verbs. stem type vowel stem consonant stem mii- (look), negative forms nii-(cook) mii-n, nii-n yum-(read), kak-(write) yum-an, kak-an Table 1 Derivational suffixes such as causative, passive/potential and honorifics have different allomorphs depending on whether the stems they attach to are consonant or vowel ending. as in Table 2. vowel stem verbs consonant stem verbs 10 causative -ssas-: mii-ssas-, nii-rai- -as: kak-as-, yum-as-, tur-as-, passive/potential -rai-:mii-rai-, nii-rai- -ai: kak-ai-, yum-ai, tur-ai- honorifics -samai-:mii-samai-, nii-samai- -amai-: kak-amai, yum-amai, tur-amai Table 2 As can be seen Table 3, the suffixes used to derive causative, passive/potential and honorifics for the verbs in (26) are clearly those of consonant stems. consonant stem verbs causative -as: ss-as-, cc-as-, cuff-as- passive/potential -ai: ss-ai-, cc-ai-, cuff-ai- honorifics -amai-: ss-amai-, cc-amai, cuff-amai Table 3 If we assume the deletion of /ɨ/ before V, the rule (29) cannot apply, because one cannot choose which suffix is added to the stem. In order to add consonant stem suffixes, (29) must be applied to delete the vowel to make the stem ending in a consonant. But if we do that, (29) loses the environment for the application. (29), therefore, is not applicable as it is. (29) ɨ→φ/_V In order to get out of this paradoxical situation, the stem of the verbs in (26) must be consonant ending. 10 Most consonant stem verbs in Ikema also have vowel stem alternants. So cuffi- is a possible stem taking suffixes for vowel stem verbs: cuffi-n, cuffi-ssas-, cuffi-rai-, cuffi-samai. In Takubo (2013, 2015) I have posited the following underlying forms for these verbs. They can account for all the forms except for the conclusive forms, for which constraint like (31) and a rule (31) to avoid it must be assumed. (30) ss- ss-an ss-ai ss-amai cc- cc-an cc-ai ss-amai cuff- cuff-an cuff-ai (31) a. b. cuff-amai *CCɨ CCɨ>Cɨɨ Alternatively, we can posit (32) 11 as the underlying forms. Gemination of consonants is necessary to avoid the constraint D 12. (32) s- (know) ，c- (wear, angle) ，cuf- (make) s-an>s-san c-an>c-can cuf-an>cuf-fan (32) cannot explain why the conclusive form of cuf- has to be cufɨɨ instead of cufɨ because cufɨ does not violate the minimality constraint. If we assume that the underlying form is cuff, the form cufɨɨ can be accounted for by the constraint (31) and the rule (31). 3.2 The alternation ‘ua>uu’ and the verbal morphophonetics In this subsection, I will discuss the alternation /ua/>/uu/. The change is seen as a forward assimilation. The alternation can be taken care of by the constraint (B) dictating the ban on rising diphthongs but an account must be provided why the glidification /ua/>/wa/ does not occur in this case. We will see below that the same applies to negative forms of verbs in Ikema. There are verbs in Ikema which have stems ending in /u/. Fau (eat) and kau (buy) behave like consonant stem verbs. (33) a. fau (eat) fa-an fa-ai- b. kau (buy) ka-an 11 ka-ai- faa-skaa-s- (32) is the approach that I adopted in Takubo (2018). I discard it in favor of Celik and Takubo (2014) and Takubo (2015) because of the reason given. 12 D in this case may be interpreted as 'keep the morpheme boundary' instead fo 'keep the syllable boundary.' We can posit a deleted stem final consonant such as /w/ in the underlying forms, e.g. faw-，kaw-, as is done in Standard Japanese for verbs like kaw- (buy) or aw- (meet). To do that in Ikema, we would have to posit an underlying /w/ which never realizes in any surface forms unlike in Standard Japanese, where /w/ appears as one of the allomorphs, as in the negative form, /kaw-anai/ 13． There are verbs ending in u such as fuu (come) and umuu which behave like vowel stem verbs. (34) a. fuu (come): kuu-n 14 kuu-rai- kuu-ssas- b. umuu (think): umuu-n umuu-rai umuu-ssas- Umuu has another paradigm (35). (35) umuu: umuun umuui umuus (35) can be thought of as a consonant ending paradigm. The passive/potential form in (35) can be derived by positing a stem final consonant /w/ which is deleted intervocalically. The following derivation can be posited 15． (36) umu(w)-ai->deletion of w>umu-ai->forward assimilation of ua>umu-uiThe same derivation can be observed in causatives. (37) umuw-as->deletion of w> umu-φ-as->forward assimilation of ua>umuus umu- can, therefore, behave like a consonant stem verb, which leads us to posit a change from /ua/ to /uu/ for verbal derivation, as is done in Topic formation. 4. Conclusion We have seen the morphophonemic system for topic and accusative forms in Ikema. We have argued that the underlying topic form must be /a/ and accusative form /u/ instead of /ja/ and /ju/ and that the actual realization forms can be derived by positing a small number of very general constraints. 13 For this reason, Shimoji (2008 ) avoids the term consonant and vowel stems and uses instead class 1 and class 2 verbs. In Ikema (and in other varieties of Miyako) /ku/ changes to /fu/. So /fu-u/ is the conclusive form of /kuu/(come). The negative form is /kuu-n/, which may correspond to /ko-n/ in Japanese．The change from /ko-/ to /fuu-/ cannot be explained because /ko-/ does not generally correspond to /fu/. 15 I owe this observation to Kenan Celik (personal communication). 14 We have also seen the constraints can account for verbal morphophonemics of Ikema. References Celik, Kenan, Yukinori Takubo (2014) Ikema topic and accusative marker, an OT analysis， FAJLposter sessions, NINJAL. Hayashi, Yuka (2009) Ryukyu Miyako Ikema hoogen no danwasiryoo (Discourse data on Ikema variety of Miyako Ryukyuan). In: Onishi, Masayuki and Kazuya Inagaki (eds.) Tikyuuken gengo kizyutu ronsyuu１, 153-199. Gengo kizyutu kenkyuukai. Hayashi,Yuka (2010) Ikema. In: Michinori Shimoji and Thomas Pellard (eds.) An Introduction to Ryukyuan Languages, Research Institute for Language and Cultures of Asia and Africa, Tokyo University of Foreign Studies, pp.167‐188. Hayashi Yuka (2013) Minami Ryukyuu Miyako Ikema hoogen no bunpoo (Grammar of the Ikema variety of Miyako, A Southern Ryukyuan Language) Ph. D. Thesis, Kyoto University. Karimata, Shigehisa (1992) Ryukyu rettoo no gengo (Miyako hoogen), Miayko Languages of the Ryukyu Archepelago. Gengogaku daiziten, vol. 4. 848-863. Tokyo: Sanseido Shimoji, Michinori (2008) A grammar of Irabu, A Southern Ryukyuan Language. Ph. D. thesis.Australian National University. Shimoji, Michinori (2017) A grammar of Irabu, a Southern Ryukyuan language. Fukuoka: Kyushu University Press. Shimoji, Michinori (2018) Minami Ryukyu Irabuzima hoogen (A grammar of Irabu, a Southern Ryukyuan language), ILCAAI. Takubo, Yukinori (2015) Issues in the verbal morphophonemics of Ikema Ryukyuan. Paper presented at Oninron Forum 2015. Takubo, Yukinori (2018) Ikema hoogen no keitai oninron (Morphophonemics of Ikema Miyakoan) Paper presented at 32nd Phonetics Society of Japan. Okinawa International University. September 16, 2018. Uemura, Yukio (1992) Ryukyu Rettoo no genngo (General overview) Gengogaku daiziten, vol. 4. 848-863. Tokyo: Sanseido. 771-814．