The addition of syllable-final stops in Ganan

Abstract

This paper discusses a secondary addition of syllable-final glottal stops in Ganan (Sino-Tibetan > Sal > Jingpho-Luish). In particular, it deals with the phenomenon where words ending with i or u in Luish languages Cak and Kadu have an additional glottal stop in Ganan. This study found that words ending with i or u can be reconstructed as either *i or *iy or *u or *uw respectively, and the secondary glottal stop is added in Ganan when the reconstructed form is *iy or *uw and does not have a high tone.

1. Introduction

1.1 Objectives

This paper reports a phenomenon found in Ganan (ISO 639-3 zkn), a Luish language of Tibeto-Burman (TB), which adds a syllable-final glottal stop unseen in other Luish languages, and discusses its historical development.

1.2 Ganan and the Luish languages

Ganan is a Luish language of Tibeto-Burman. The Luish languages include Cak (ISO 639-3 ckh) spoken in the Chittagong Hill Tracts in Bangladesh, Sak in tRakhine State in Burma, Chakpa in the Imphal basin in India, Kadu (ISO 639-3 zkd) in Sagaing Region of Burma, Ganan, etc.

Of these languages, Cak and Sak are almost identical as they are mutually intelligible insofar as native words are concerned. However, this is not always the case due to the large number of loanwords in Sak that originate from Burmese and its Arakanese dialect, as well as those from Bangla in Cak. Linguistic studies of Cak have been published by Huziwara (2008) (a grammatical description), and Huziwara (2016b) (a dictionary). Additionally, Löffler (1964) dealt with Cak's relationship to the Chakma language, Bernot (1966) reported several hundred basic words, and Maggard et al. (2007) compared four dialects. As for Sak, Thun Shwe Khaing (1988) provides an ethnography with a simple grammatical sketch and a list of basic vocabulary, and Moe Sandar (2010) presents a descriptive grammar of Sak. Luce (1985 vol. II, Chart K, L) has a few hundred basic words from two dialects, Bawtala and Dodem.

Chakpa is already a dead language in terms of its daily use, and is only occasionally used in rituals. For linguistic resources on Chakpa, in addition to McCulloch (1859), which contains 423 words from the Andro and Sengmai dialects recorded in the mid-nineteenth century,¹ the language is sometimes mentioned in ethnographies on the Chakpa people (Rorendrajit 2006; Basanta 2008).

Kadu is further divided into subclasses of Setto Kadu (STK), Moteik Kadu (MTK), Molang Kadu (MLK), Mokhwang Kadu (MWK), of which STK, MTK, and MLK are similar varieties and, although each has unique characteristics, are mutually intelligible. Nevertheless, MTK is not used daily anymore and is hardly passed down to the next generation. STK and MLK, similarly, have only a few villages passing down the varieties. MWK, as mentioned in Huziwara (2015), has distinct characteristics from other Kadu varieties, and is relatively close to Ganan concerning linguistic features. However, MWK is mutually intelligible with neither Ganan nor other Kadu varieties. For studies on Kadu, Sangdong (2012) is the most detailed research with texts and a vocabulary list, and Khin Moe Moe (2004) and Huziwara (2013) deal with the phonetics of Kadu. Other than these, Brown (1920) is one of the primary studies of Kadu, and Luce (1985, vol. II, Chart K, L) contains some 300 words of MLK.

Compared to Kadu, Ganan has less dialectal difference, and speakers of Ganan varieties do not have difficulty in understanding each other. The addition of a syllable-final glottal stop discussed in this paper is found in all Ganan varieties, based on the author's data. Previous works on Ganan are Ma Myo Myo (2006), which described the grammar of the Nanzar dialect, and Huziwara (2012a) on the phonology of the Shwegyaung dialect. Luce (1985, Vol. II, Chart K, L) lists some 300 Ganan words.² This paper treats the Shwegyaung dialect as representative of Ganan, taking into consideration the amount of data owned by the author.

1.3 Data and abbreviations

Below is a list of the sources of data and the abbreviations used in this paper. Data is taken from STEDT unless otherwise noted.

*A

A is a reconstructed form

(A/B)

A and B are allomorphs

{A}

A is a spelled form

A

Andro, taken from McCulloch (1859).

ANDV

andative marker

C

Cak, taken from primary sources collected by the author.

CKP

Chakpa

cl

classifier

cmpl

completive marker

G

Ganan, taken from primary sources collected by the author.

h

prefix having a high tone in Proto-Luish

J

Jingpho

K

Kadu, in particular the Takotta dialect of STK unless specified; from primary sources collected by the author

Lui

Luish

MT

Moteik Kadu

num

numeral

OB

Old Burmese (taken from Nishi 1999)

PLu

Proto-Luish (taken from Huziwara 2012b, 2014, 2016b)

PKG

Proto-Kadu-Ganan

pred

predicate marker

PTB

Proto-Tibeto-Burman, taken mainly from STEDT

S

Sak

Se

Sengmai (taken from McCulloch 1859)

STEDT

Sino-Tibetan Etymological Dictionary and Thesaurus

TB

Tibeto-Burman

WB

Written Burmese, based on the transcription rules by Duroiselle (1916)

WT

Written Tibetan (Wylie transliteration system)

1.4 Notes on transcription

The primary source used by the author is written in simplified phonological transcription. For Kadu and Ganan, readers should pay heed to the points in (1):

1. (1)

   1. a. Tones: ´ stands for a high tone (H), ` for a low tone (L), ^ for a falling tone (F). No tonal symbols are attached to unmarked mid tone (M).

   2. b. The low tone in Kadu and Ganan developed as a result of a tonal change of a mid tone that had followed a high tone. Words starting with a low tone show that they used to have a prefix with a high tone.

   3. c. The falling tone in Kadu developed as a result of a tonal change of HM or ML, and regularly occurs as MF.

2. Subclasses of the Luish languages and Ganan

Before discussing the epenthesis of the syllable-final glottal stop in Ganan, I shall note how Luish languages are treated in the Tibeto-Burman language group, and how Ganan is treated in the Luish languages. Matisoff (2013) claims the Jingpho-Asakian languages as a subgroup of Tibeto-Burman languages, where Asakian languages have traditionally been called Luish languages. The fact that the Luish languages belong to the Tibeto-Burman language group is evident from the comparison of their basic vocabulary, as exemplified in (2).

1. (2)

   1. a. “I” WT nga, WB ngaa; PLu *ŋa; C ŋa, K ŋa, G ŋa; J ŋai³³;

   2. b. “pig” WT phag, WB wak; PLu *wak; C vaʔ, K waʔ, G waʔ; J waʔ³¹.

The Luish languages are classed under the Sal group (Burling 1983) as evidenced by the following special diagnostic lexicons (cited from Huziwara 2020: 46, with some modifications), which in turn is based on Benedict (1972: 7, 34 fn. 108) except for Luish data).

1. (3) “sun” PLu *ca-mík, C cəmíʔ, S səmíʔ, K/G səmíʔ, A/Se chameet (camit); Chairel sal; Taman pupek cf. J dźān; Namsang (Northern Naga) san, Moshang (Northern Naga) sár; Garo sal.

2. (4) “fire” PLu *wal, C/S vaiN, K/G wan, A/Se wal; Chairel phal; Taman vè cf. J ʔwàn; Namsang (Northern Naga) van, Moshang (Northern Naga) var; Garo waʔl.

3. (5) “foot” vs. “hand/arm”

   PLu *ta vs. *tak-, C ʔáta vs. taʔmiŋ “nail”, K/G ta vs. taʔmiŋ “nail”, A/Se ta- vs. takmeng “nail”; Chairel la vs. lak; Taman — vs. la < *lak; cf. Garo dzá vs. dzák, Dimasa ya vs. yau; Tableng ya vs. yak, Tamlu la vs. lak, Banpara tśia vs. tśak, Namsang da vs. dak, Moshang ya vs. yak.

Distinctive characteristics that separate the Luish languages from other Tibeto-Burman languages including Jingpho are: (a) the order of affixes “cl-one”, while “num-cl” from two onward;³ (b) a set of directional auxiliary verbs; and (c) a negative prefix a-. Examples of these features are shown in (6).

1. (6)

   1. a. Order of a classifier and “one” cl-num

      WT No classifier, WB num-cl

      J cl-num; however, the classifier is rarely used. Lui cl- “one”; for other numbers num-cl.

      “cl: man-one” C hú-wa, MT/G hɔ̀-wa

      “two-cl: man” C níŋ-hú, MT kleiŋ-hú, G kɛ-hɔ́;

   2. b. Having an andative auxiliary verb (andv) *-a and a completive auxiliary verb (cmpl) *-aŋ

      WT, WB, J. No such auxiliary verbs

      “drink-andv=pred” C Ɂu-wa=heʔ, MT/G ʔu-wa=ma

      “drink-cmpl=pred” C Ɂu-waŋ=heʔ, MT/G ʔu-waŋ=ma;

   3. c. Negative prefix form; WT ma, WB ma, J ń- “negative prefix-” PLu *á-;⁴ C ʔá/ʔa-, K ʔə-, G ʔə-, A/Se a-.

The Luish languages are divided into Cak and Chakpa-Kadu depending on the innovations they have undergone shown in (7)–(10), namely (A) PLu *ti > kyi;⁵ (B) deletion of PLu *r; (C) PLu *khy > ʃ; (D) marked linker for borrowed verbs.

1. (7)

   1. a. “sweet” PLu *ti; C kyi,⁶ K ti, G ti; A/Se tī; J tui³¹;

   2. b. “egg” PLu *ti; C ʔákyi, K təti, G titti; J ti³¹;

   3. c. “penis” PLu *tí; C ʔakyí, K tí, G tí.

2. (8)

   1. a. “thick” PLu *r^H-thay; C rət^he, K t^hɛ̀ ~ ʔət^hɛ̀, G t^hɛ̀; A/Se the; J that³¹;

   2. b. “cord” PLu *ri; C rɨ, K ʔi, G ʔi; J ʒi³¹;

   3. c. “buffalo” PLu *k-réy;⁷ C krí̵, K cɛ́, G cé; A/Se ké; J wă³³loi³³;

   4. d. “crab” PLu *a/n-har < *a/n-khar; C nəhaiŋ, K ʔəha, G ʔəha; A aha, Se niha; J tʃă⁵⁵khan⁵¹.

3. (9) “red” PLu *khyá; C ʃá, K há, G há; A/Se ha; J khje³³.

4. (10) “loan verb marker”⁸ PLu *-(t/l)ó; C —, K -tɔ́, G -lɔ́; A/Se -to.

The linguistic innovations dividing Chakpa, Kadu, and Ganan are: (A) PLu *-l > -n; (B) development of an infix -l-;⁹ and (C) change from the two-tone system to the three-tone system.¹⁰ Examples are shown in (11) and (12):

1. (11)

   1. a. “fire” PLu *wan < *wal; C vaiŋ, K wan, G wan; A/Se wal; J wan³¹;

   2. b. “tree” PLu *phón < *phól; C (ʔap^háŋ, púŋp^háŋ), K p^houŋklon, p^hón “firewood”, G p^hɔ́ntòn; A phol; J phun⁵⁵.

2. (12)

   1. a. “fat” PLu *sáw; C ʔasá, K s^həlɔ́, G s^həlɔ́; A/Se sa; J sau⁵⁵.

   2. b. “leaf” PLu *tap; C ʔátaʔ, K təlap ~ tətap, G təlap; A/Se tatup (tatap); J lap³¹.

3. Point at issue

The issue discussed in this paper is the phenomenon in which Ganan occasionally has an additional glottal stop consonant in a position where other Luish languages do not. This point is prominent evidence for dividing Ganan and Kadu.

1. (13)

   1. a. “bamboo shoot” G kəmiʔ; K kəmi, C kəmɯkaiŋ;

   2. b. “bone” G maŋkuʔ; K maʔku, C (ʔáməra);

   3. c. “cat” G hánsìʔ; K hanɕî, C háiŋ; Se huljeek (haljik);

   4. d. “comb (v)” G s^hiʔ; K ɕi, C sɨ;

   5. e. “correct” G hiʔ; K c^hi, C heʔ, huʔ “copula”;

   6. f. “elbow” G táʔs^hùʔ; K taʔkəs^hû, C (táiŋdoŋ);

   7. g. “faeces” G hiʔ; K c^hi, C ʃi;

   8. h. “horse” G s^həpùʔ; K s^həpù, S sapú (Hodgson 1853: 5); A/Se shoorook (shuruk);

   9. i. “medicine” G s^hiʔ; K ɕi, C sɨ;

   10. j. “porcupine” G kətùʔ; K kətù, C (phaiŋ); A/Se kootook (kutuk);

   11. k. “rat” G cùʔ; K kəyù, C kəyvu; A/Se kooyook (kuyuk);

   12. l. “smoke” G wann̥uʔ; K (wans^huŋ), C vaiŋhvu; A walkhoo (walkhu), Se walhoo (walhu);

   13. m. “steal” G kuʔ; K ku, C kvu; A/Se kook (kuk);

   14. n. “vine” G yəluʔ; K yəlu, C (ʔárəkuʔ); A loohook (luhuk);

   15. o. “wall of a house” G címkùʔ; K cémkù, C kíŋhvu.

From the examples in (13), we find that Ganan has a glottal stop after i (13a, c, d, e, g, i) and u (13b, f, h, j, k, l, m, n, o), whereas Kadu and Cak do not. Based on this point, Hypothesis 1 (14) is established.

1. (14) Hypothesis 1. When a syllable has i or u at the syllable-final position in the Luish languages (e.g. Kadu and Cak), its corresponding form in Ganan has an additional glottal stop to it.

Nevertheless, there are many cases with no glottal stop after i or u, as shown in (15) and (16).¹¹

1. (15)

   1. a. “barking deer” G ŋəhí; K ʔəc^hí, C ʔiʃí;

   2. b. “buy” G mí; K mí, C mərí̵;

   3. c. “come” G li; K li, C (vaiŋ)

   4. d. “cord” G ʔi; K ʔi, C rɨ;

   5. e. “die” G s^hí; K ɕí, C sí̵;

   6. f. “dog” G ci; K ci, C kvu;

   7. g. “egg” G titti; K təti, C ʔákyi, S wa-tí (Hodgson 1853: 8);

   8. h. “elephant” G ʔəcí; K ʔəcí, C ʔukvú ~ wvukvú;

   9. i. “four” G pí; K pí, C prí̵;

   10. j. “fruit” G s^hiʔs^hi; K ɕəɕi, C ʔásɨ;

   11. k. “give” G ʔi; K ʔi, C ʔi;

   12. l. “let out (fart)” G p^hí; K p^hí, C p^hí;

   13. m. “penis” G tí; K tí, C ʔakyí;

   14. n. “smooth” G pit; K pi, C prɨ;

   15. o. “sour” G hí; K c^hí, C hrí̵;

   16. p. “sweet” G ti; K ti, C; kyi;

   17. q. “wash (clothes)” G hi; K c^hi, C hrɨ;

   18. r. “woman” G ʔínáʔs^ha “girl”; K ʔiɕî, C ʔísa “old lady”;

   19. s. “younger brother” G nəs^hì; K nəɕì, C ʔanésɨ.

2. (16)

   1. a. “bathe” G kú; K kú, C krvú;

   2. b. “burn” G hu; K hu, C hru;

   3. c. “burn/roast” G su; K su, C cu;

   4. d. “cut(vi)” G tu; K tu, C tvu;

   5. e. “emerge” G pu; K pu, C pru;

   6. f. “dig” G t^hu; K t^hu, C t^hu;

   7. g. “drink” G ʔu; K ʔu, C ʔu;

   8. h. “fishy” G s^hú; K s^hú, C svú;

   9. i. “fowl” G ʔu; K ʔu, C ʔu;

   10. j. “get” G lu; K lu, C lu;

   11. k. “grind/pound” G t^hu; K t^hu, C t^hvu ~ t^hu;

   12. l. “open (umbrella)” G p^hú; K p^hú, C ʔahvú;

   13. m. “mushroom” G kúʔmú; K kəmú, C kəmúkaiŋ;

   14. n. “rot” G mú; K kəpú, C ɓú;

   15. o. “seed” G tuttu; K tətu, C ʔátvu;

   16. p. “snake” G kəp^hú; K kəp^hú, C kəhvú;

   17. q. “watch” G yu; K yu, C yu.

However, there are cases where both Ganan and Kadu have a glottal stop after i and u, as exemplified in (17).

1. (17)

   1. a. “mosquito” G pəsíʔ; K pəsíʔsáuʔ, C pəcíʔ;

   2. b. “sun” G səmíʔ; K səmíʔ, C cəmíʔ;

   3. c. “belly” G púʔ; K púʔ, C ʔapí̵ʔ.

Therefore, it is far-fetched to conclude that a lexical form ending with i or u in the Luish languages unconditionally corresponds to a form with an additional glottal stop in Ganan. Huziwara (2012b, 2014) then assumed that a glottal stop in Ganan enumerated in (13) is a derived form originating from the Proto-Luish language, and proposed a reconstructed form *-k as in (18). Below are instances with a reconstructed PLu form.

1. (18)

   1. a. “bamboo shoot” PLu *k-muy-k;

   2. b. “comb (v)” PLu *si-k;

   3. c. “faeces” PLu *khyi-k;

   4. d. “horse” PLu *s^H-pu-k;

   5. e. “medicine” PLu *si-k;

   6. f. “rat” PLu *k^H-yuw-k;

   7. g. “smoke” PLu *wán-huw-k < *wál-khuw-k;

   8. h. “steal” PLu *kuw-k;

   9. i. “wall of a house” PLu *kím-(k/kh)uw-k.

However, the reconstructed form *-k is unnecessary if the syllable-final glottal stop in Ganan in (13) is predictable. This paper now goes on to discuss whether or not the secondary addition of the syllable-final stop in Ganan can be predicted.

4. Rhyme reconsidered

Supposing that the addition of syllable-final stops in Ganan is predictable, what conditions are at work?

To discuss this, it is helpful to look into similar phenomena found in neighbouring languages.¹² Burling (1966) reported a phenomenon where a stop consonant is secondarily inserted in Maru,¹³ one of the Burmish languages.¹⁴ According to Burling (1966), for words ending with i or u in Atsi,¹⁵ which belongs to the same language group as Maru, their corresponding form in Maru is it or uk when they correspond to {e} or {ui} in WB.¹⁶ A list of examples is given in (19):¹⁷

1. (19)

   1. a. “die” WB {se}, Atzi šî, Maru šit;

   2. b. “parrot” WB {kye:}, Atzi jì, Maru jìt;

   3. c. “in front” WB {hre.}, Atzi hı̌, Maru ɣ^ʔít;

   4. d. “horn” WB {khyui}, Atzi khyúi, Maru khyùk;

   5. e. “steal” WB {khui:}, Atzi kháu, Maru khúk;

   6. f. “breast” WB {nui.}, Atzi nàu, Maru núk.

In Maru, there are words ending with i or u as well, and they by-and-large correspond to WB {ii} or {uu}. Below (20) shows examples from Sawada (2017):

1. (20)

   1. a. “fire” WB {mii:}, Zaiwa mi²¹, Lhaovo mji^L;

   2. b. “to be white” WB {phruu}, Zaiwa phyu⁴¹, Lhaovo p^hju^F.

Below, (21) is a list of Ganan word forms shown in (13), together with corresponding Burmese forms.

1. (21)

   1. a. “faeces” G hiʔ; K c^hi; WB {khye:}, OB {khliy};

   2. b. “medicine” G s^hiʔ; K ɕi, C sɨ; WB {che:};

   3. c. “smoke” G wann̥uʔ; K (wans^huŋ), C vaiŋhvu; WB {khui:}.

From the examples in (21), all word forms correspond to either {e} or {ui} in WB, which can be reconstructed as *iy or *uiw respectively (Hill 2019). As such, I shall modify Hypothesis 1 in (14) as (22).

1. (22) Hypothesis 2. Forms corresponding to {e} or {ui} in Burmese (*iy or *uiw in the proto-language) correspond with a form with an additional syllable-final glottal stop in Ganan.

That said, as we will see in (23), there are some cases where cognate Burmese forms have {e} or {ui} even though their corresponding Ganan forms have no glottal stop. For this reason, Hypothesis 2 in (22) still seems incorrect.

1. (23)

   1. a. “barking deer” G ŋəhí; WB {khye}, {gyii} (Judson 1893);

   2. b. “die” G s^hí; WB {se}, OB {siy};

   3. c. “four” G pí; WB {le:} PTB b-ləy;

   4. d. “mushroom” G kúʔmú; WB {hmui};

   5. e. “dog” G ci; WB {khwe:}, OB {khuy};

   6. f. “snake” G kəp^hú; WB {pui:} “insect”;

   7. g. “bathe” G kú; WB {khyui:}, OB {khluiw}.

From the words listed in (23), we see that all the Ganan word forms but (23e) have a high tone, while the Ganan forms with a syllable-final glottal stop in (13) have a mid or low tone. Therefore, Hypothesis 2 in (22) is revised as (24):

1. (24) Hypothesis 3. For word forms with {e} or {ui} in Written Burmese, the corresponding word forms in Ganan have an additional syllable-final glottal stop when they do not have a high tone in Ganan.¹⁸

Huziwara (2012b, 2014) reconstructed PLu *i and *uw for PTB *əy and *əw, namely {e} and {ui} in Written Burmese, and for words with a syllable-final stop in Ganan they assumed PLu *i-k and *uw-k, respectively. Given (24), however, it is possible to predict the addition of the syllable-final stop in Ganan by supposing, for example, *iy and *uw to PLu.

Thus, the PLu or PKG forms of the word samples in (13) can be reconstructed as (25). PLu is reconstructed if a Cak word is cognate with either Kadu or Ganan, and PKG if only Kadu and Ganan are cognate.

1. (25)

   1. a. “bamboo shoot” G kəmiʔ; K kəmi, C kəmɯkaiŋ; PLu *k-muy;¹⁹

   2. b. “bone” G maŋkuʔ; K maʔku, C (ʔáməra); PKG *má(k/ŋ)-kuw;

   3. c. “cat” G hánsìʔ; K hanɕî, C háiŋ; PLu *hán-(c/s)iy < *hál-(c/s)iy;²⁰

   4. d. “comb (v)” G s^hiʔ; K ɕi, C sɨ; PLu *siy;

   5. e. “correct” G hiʔ; K c^hi, C (heʔ, huʔ) “copula”; PKG *khyiy;

   6. f. “elbow” G táʔshùʔ; K taʔkəshû, C (táiŋdoŋ); PKG *ták-suw;

   7. g. “faeces” G hiʔ; K c^hi, C ʃi; PLu *khyiy;

   8. h. “horse” G s^həpùʔ; K s^həpù, S sapú (Hodgson 1853: 5); PLu *s^H-puw;

   9. i. “medicine” G s^hiʔ; K ɕi, C sɨ; PLu *siy;

   10. j. “porcupine” G kətùʔ; K kətù, C pədvu; PLu *k^H-tuw;

   11. k. “rat” G cùʔ; K kəyù, C kəyvu; PLu *k^H-yuw;

   12. l. “smoke” G wann̥uʔ; K (wans^huŋ), C vaiŋhvu; PLu *wán-khuw;

   13. m. “steal” G kuʔ; K ku, C kvu; PLu *kuw;

   14. n. “vine” G yəluʔ; K yəlu, C (ʔárəkuʔ); PKG *yəluw²¹ < *yuw;

   15. o. “wall of a house” G címkùʔ; K cémkù, C kíŋhvu; PLu *kím-(k/kh)uw.

5. Similar phenomena in related languages

Taman,²² reportedly closely related to Luish languages, including Chakpa, has phonological phenomena similar to Ganan.

In Chakpa (Andro and Sengmai), a stop consonant is found in cognate words ending with a secondary glottal stop in Ganan, as shown in (26).

1. (26)

   1. a. “cat” G hánsìʔ; K hanɕî, C háiŋ; Se huljeek (haljik);

   2. b. “medicine” G s^hiʔ; K ɕi, C sɨ; A/Se seek (sik) “tobacco”;

   3. c. “porcupine” G kətùʔ; K kətù, C (phaiŋ); A/Se kootook (kutuk);

   4. d. “rat” G cùʔ; K kəyù, C kəyvu; A/Se kooyook (kuyuk);

   5. e. “steal” G kuʔ; K ku, C kvu; A/Se kook (kuk);

   6. f. “vine” G yəluʔ; K yəlu, C (ʔárəkuʔ); A loohook (luhuk).

Occasionally, however, Chakpa has words without an additional stop where its Ganan counterpart does have it, as in (27):

1. (27) “smoke” G wann̥uʔ; K (wans^huŋ), C vaiŋhvu; A walkhoo (walkhu), Se walhoo (walhu).

Example (28) might be a case of a secondary stop insertion; however, it remains uncertain because few cognate forms can be found.

1. (28) “man” A tik “he/she”; G tìʔs^ha; Sema [Sumi] ti mi (cited from STEDT Database, which is based on Marrison 1967).

As for Taman, there are no words cognate with Ganan where a stop is added, but some words, as for example in (29), seemingly have a secondary stop added to the end of the syllable compared to other Tibeto-Burman languages.

1. (29)

   1. a. “cat” Taman mətʃeksɔ; C háiŋ, K hanɕî, G hánsìʔ; A hunggen (hanggen), Se huljeek (haljik).

   2. b. “horse” Taman tʃipòùk; C (məráŋ), S sapù (Hodgson 1853: 5) s^həpù, G s^həpùʔ; A/Se shoorook (shuruk).

In (30), Ganan and Chakpa do not have any secondary syllable-final stop added, while Taman does in comparison with other Tibeto-Burman languages.

1. (30)

   1. a. “man (human being)” Taman mek; C (lú), Kadu təmìs^ha, Ganan tìʔs^ha;

      A teeksahora (tiksahora), Se teekhora (tikhora).

   2. b. “write” Taman rek; C (rwé < Marma), K ʔəc^hìn, k^hù, G (yé < Burmese); A/Se —; OB riy (Nishi 1999: 39).

As is observed, an addition of a secondary stop after a high vowel does not consistently occur in cognate forms. This implies that this innovation did not happen in the proto-language which Ganan, Chakpa and Taman commonly share, but happened independently in each language.

6. Conclusion

This paper has discussed the phenomenon whereby Ganan has an additional syllable-final glottal stop in words whose Luish counterparts do not. As a result, it demonstrated that, by assuming *iy and *uw in PLu, such Ganan forms are mostly predictable if they do not have a high tone.²³

There are remaining issues as in (31):

1. (31)

   1. a. Rhyme system of PLu: If we reconstruct *iy for words with a glottal stop at their coda position in Ganan, there is a contradiction with the form *iy reconstructed earlier in Huziwara (2012b, 2014, 2016b). Therefore, it would be necessary to change *iy to *ey, and *ey to *ay. The problem remains unsolved as to how PLu rhymes should be reconstructed.

   2. b. It has not been explained yet why a glottal stop is added not after *i and *u but after *iy and *uw, and also why it is not added to high-tone words.²⁴