The effect of glottal stops on tone in Burmese∗

Seunghun J. Lee Rutgers University

1 Introduction The aim of this paper is to show that Burmese has a glottal stop coda and that the glottal coda affects the preceding vowel’s tone. After demonstrating the existence of a syllable-final glottal stop using both phonological and phonetic evidence, the effect of glottal stops on the tones of the preceding vowel is accounted for.

Burmese has two tones: high and low. High tone and low tone occur in open syllables and in syllables with a nasal coda when these syllables have modal phonation. These two tones, however, neutralize to high tone when syllables have creaky phonation or a glottal stop coda.1 This neutralization is analyzed as due to the interaction of two types of constraint. The markedness constraint *ʔ/L bans low tone on a glottal stop. The *ʔ/L constraint dominates the faithfulness constraint IDENT-T, which preserves an underlying tone in the input. In syllables with creaky phonation, the markedness constraint *V/L bans low tone on creaky syllables dominates IDENT-T. This analysis crucially assumes that consonants are tone bearing units (TBU’s) and that faithfulness constraints such as IDENT-T only preserve tones on moraic or syllabic TBU’s. As a result, no faithfulness constraints evaluate underlying tones on consonants.

The effect of coda consonants on tones of the preceding vowel has been reported in other tonal languages. For example, Lahu, a Lolo-Burmese language, has five tones in open syllables, but only two in syllables with a glottal stop coda (Matisoff 2003). San Martín Itunyeso Trique, a Mixtecan language, has only level tones in syllables with a glottal stop coda, while other syllable types allow contour tones (DiCanio 2006). Phonetic studies on Eastern Cham (Brunelle 2005), and a comparative study on Kuki-Chin languages (Hyman 2007) also discuss various issues about the relationship between coda consonants and tone.

This paper is organized as follows. Section 2 presents phonological and phonetic evidence that shows that Burmese has a glottal stop coda. In section 3, neutralization of tones in syllables with a glottal stop coda and syllables with creaky phonation will be analyzed. Section 4 presents general remarks on the relationship between syllable types and their interaction with tones.

∗I would like to thank Paul de Lacy and Akinbiyi Akinlabi for their valuable insights and comments. I would also like to thank to participants of CLS 43 for their comments. I also thank James Bruno, José Elías-Ulloa, Patrick Houghton, Paula Houghton, and Brian McHugh for comments. Any error or shortcomings are my own. Comments are welcome: juliolee@rutgers.edu. 1 Low tone has no marking, high tone is marked with an acute accent, and creaky phonation is marked with a tilde under the vowel.

Draft

2 A glottal stop coda in Burmese This section argues that Burmese allows glottal stops in coda position. This claim is controversial in the sense that other descriptions have treated the glottal stop as a type of allotone (Okell 1968). However, there is a variety of phonological and phonetic evidence to support the claim that codas can contain a glottal stop segment. Glottal stops assimilate in place of articulation with the following onset (section 2.1); they block obstruent voicing (section 2.2); co-occurrence restriction of nasals and glottal stops in coda position also provides support (section 2.3). An asymmetry of vowel qualities in open and closed syllables shows that syllables with a glottal stop pattern with closed syllables (section 2.4). Finally, phonetic studies reveal that the duration of syllables with a glottal stop is shorter than other syllables (section 2.5). An analysis set within Optimality Theory will be presented in section 2.1 and 2.3 (Prince and Smolensky 1993/2004). 2.1 Glottal stop assimilation Glottal stop codas assimilate to the following onset. A glottal stop becomes a voiceless obstruent that agrees in place and continuancy with a following obstruent.2 In (1a-b), the final glottal stop /ʔ/ assimilates to the following onset and becomes /k/ or /t/. Glottal stop codas also assimilate to a voiced fricative /z/ and become a voiceless homorganic fricative [s] as in (1c).

(1) Assimilation to onset: obstruents (Okell 1969: 7) a. jaʔ + kweʔ → jak.kweʔ area, quarter

b. seiʔ + thin → seit.thin opinion c. louʔ + zaʔ → lous.zaʔ fictitious story Glottal stops also become a homorganic nasal [m] when followed by a nasal

onset /m/ as illustrated in (2a-b). (2) Assimilation to onset: nasals (Okell 1969: 7)

a. ʔeiʔ + me → ʔeim.me (I) will sleep b. ʔeiʔ + meʔ → ʔeim.meʔ dream

Glottal stops are deleted when followed by a voiceless nasal onset /n/ or a

palatal /ɲ/ as in (3a-b). (3) Deletion of a glottal stop: voiceless nasals3 (Okell 1969: 7)

a. mjeʔ + na → mje.na face b. kauʔ + ɲín → kau.ɲín sticky rice

2 A historical study by Maran (1971) shows that final glottal stops in Modern Burmese result from obstruent codas in Old Burmese. 3 Voiceless nasals are probably specified with [+spread glottis] rather than [-voice] because they pattern on par with aspirated consonants in active/passive verbal pairs (Green 2005: 5).

Draft

The presence of final glottal stops is explained by neutralization of obstruent codas in final position. Any consonant that surfaces in coda position in the output neutralizes to a glottal stop unless it is followed by another consonant (see de Lacy 2006 for a recent OT account of coda debuccalization).

The constraint ranking that accounts for the glottal assimilation is shown in (4). An input /VʔkV/ with a glottal stop followed by a velar stop becomes [Vk.kV] because of AGREE(PLACE) (=AGREE(PL)), which requires the same place of articulation between adjacent segments across syllable boundaries. The faithful candidate [Vʔ.kV] in (4d) cannot be the optimum because AGREE(PL) dominates IDENT(PLACE) (=ID(PL)), which ensures identity between the coda and the following onset. IDENT(PL) penalizes place features in the output that are not identical to the correspondent segment in the input, and it is violated when /ʔ/ becomes [k] as in (4a).

(4) Glottal Assimilation4 /VʔkV/ MAX ID-ONS AGREE(PL) *{P, T, K} ID(PL) a. [Vk.kV] ** * b. [V.kV] W * L * L c. [Vʔ.ʔV] W * L * d. [Vʔ.kV] W * L * L

The assimilation cannot occur regressively as [Vʔ.ʔV] due to the ranking of

IDENT(ONSET) (=ID-ONS) over *{P, T, K}. This ranking blocks the change of the onset segment to a glottal stop. Moreover, AGREE(PL) cannot be satisfied by deleting the glottal stop due to the higher ranked MAX constraint (4b).

The ID-ONS constraint is violated when an onset in the output is not identical to the correspondent in the input (Beckman 1997). The *{P, T, K} constraint is violated when consonants with either labial, coronal, or dorsal places are present (de Lacy 2006). The MAX constraint assigns violation marks when an input segment does not have a correspondent in the output.

A hypothetical input /Vk/, assuming Richness of the Base, neutralizes to a glottal stop in word-final position. The neutralization ranking is shown in (5), in which *{P, T, K} dominates IDENT(PLACE).

(5) Neutralization to a glottal stop /Vk/ *{P, T, K} ID(PL)a. [Vʔ] * b. [Vk] W * L

4 All tableaux in this paper are comparative tableaux, following Prince (2000). While ‘L’ means that the optimum loses against a given candidate under a constraint, ‘W’ means that the optimum wins over a given candidate under a constraint. If the optimum has the same violation profile with a candidate under a constraint, the cell is left blank. In a given row in the comparative tableaux, all L’s must be dominated by at least one W.

Draft

Alternatively, the glottal assimilation in Burmese can be analyzed with CODACONDITION (Itô 1989) under the assumption that glottal stop is a placeless consonant (Green 2005). In Green’s analysis, CODACONDITION is violated if a consonant with a place feature occurs in syllable-final position. Thus, by definition, placeless nasals and glottal stops are the only segments allowed in the coda. However, CODACONDITION is not violated when codas occur in a position in which their place feature is licensed by the following onset as above. For similar analyses regarding placeless glottal stops, see also Broselow (2001) on Sulawesi and Elías-Ulloa (2006) on Capanahua.

Either neutralization analysis works because the glottal stop presented here is assumed to be a full segment. If it were not a full segment, but instead a vowel feature of some sort, there is no reason why a coda consonant should appear when followed by another obstruent.

2.2 Obstruent voicing The analysis above relies on the assumption that the first segment in an assimilated cluster is a full segment that inhabits a coda. Evidence for this view is found in post-sonorant and intervocalic obstruent voicing. In (6a), a velar stop /k/ becomes voiced to [ɡ] when it occurs after a nasal coda [ŋ]. The aspirated velar stop /kh/ becomes voiced to [ɡ] between vowels. All obstruents undergo voicing regardless of their place of articulation as in (6a-d). As shown in (6e), fricatives such as /s, sh/ also become voiced to [z].

(6) Obstruent voicing in Burmese5 (Okell 1969: 12-13)

a. Velars: /k, kh/ → [ɡ] (ə)ka dance kəjiŋɡa Karen dance (ə)kha charge khəjíɡa fare b. Alveolars: /t, th/ → [d] té hut boudé rest house thá native əɲadá native of Upper Burma c. Labials: /p, ph/ → [b] poún can shiboún oil can (ə)phoún cover səlaúmboún saucepan lid d. Palatals: /ʧ, ʧh/ → [ʤ] ʧaʔ rupee ŋáʤaʔ five rupees ʧhaʔ flat thing ŋáʤaʔ five flat things e. Fricatives: /s, sh/ → [z] si each təkhṵzi one each shi oil oúnzi coconut oil

5 The symbol [ə] stands for a toneless syllable.

Draft

The examples in (6) also reveal that neither tones nor syllable types are irrelevant to this process. The preceding syllables can have low tone (e.g. boudé ‘rest house’ in 6b) or high tone (e.g. ŋáʤaʔ ‘five rupees’ in 6d). The syllable can also have creaky phonation as in (7). This demonstrates that neither tones nor phonation affects the obstruent voicing process.

(7) Creaky tone and obstruent voicing (Okell 1994: 164)

/-pa-tɛ/ ‘honorifics + 3rd person singular’ a. ka to dance ka badɛ (he) dances b. ma to lift ma badɛ (he) lifts

The obstruent voicing demonstrated in (6) and (7) does not occur when an assimilated glottal stop coda is present. The assimilated stop blocks obstruent voicing as the examples in (8) show. Additionally, they assimilate to the following onset as we have already discussed in section 2.1.

(8) Glottal stop codas blocks obstruent voicing (Okell 1969: 13) a. təjouʔka *təjouʔɡa [təjoukka] Chinese dance b. leʔkha *leʔɡa [lekkha] wage c. ʧhauʔʧaʔ *ʧhauʔʤaʔ [ʧhauʧʧaʔ] six rupees

The assimilated glottal stop blocks obstruent voicing because it destroys the voicing environment – it makes the onset obstruent neither intervocalic nor post-sonorant.

2.3 Co-occurrence restriction of glottal stop Glottal stops do not co-occur with nasal rhymes. Syllables with nasal codas have low tone, high tone, or creaky phonation. As shown below in (9), however, no attested syllable has a nasal rhyme in addition to a glottal stop.

(9) Nasal rhymes a. Low tone khan to undergo (cf. kha shake) b. High tone khan to dry up (cf. kha be bitter) c. Creaky phonation khan to appoint (cf. kha fee) d. Glottal stop coda *khaʔn (cf. khaʔ to draw off)

The reason can be found in the phonotactics of Burmese syllable structure. A

canonical syllable is formed as C(G)V(V)C (Green 2005). A syllable must have an onset, but a syllable cannot have consonant clusters in the coda. Thus, the restriction observed in (9d) is explained if the phonotactics does not allow coda clusters.

This restriction on multiple codas can be accounted for by the ranking of *COMPLEXCODA (=*CC) in the grammar of Burmese. Assuming Richness of the Draft

Base, a hypothetical input /khaʔn/ does not surface faithfully in the output due to *COMPLEXCODA. Depending on the ranking of other constraints, the glottal stop and the preceding vowel may coalesce into a vowel with creaky phonation as in (10a), the glottal stop may be deleted altogether as in (10b), or the nasal coda may be deleted as in (10c).

(10) /khaʔn/ →

a. [kha n] fusion of vowel and glottal stop (coalescence) b. [khan] deletion of glottal stop (deletion) c. [khaʔ] deletion of nasal (deletion)

A possible way to choose from among the possible options in (10) is

suggested in Green (2005: 3) in which he examines loanwords. In Burmese, a loanword with an obstruent coda in the source language is realized with a glottal stop coda. Moreover, loanwords with nasal plus obstruent cluster in the coda such as pint surface [pain] with creaky phonation and a nasal coda. See also Chang (2003) for a careful and detailed study on loanword adaptation in Burmese.

These studies suggest that the hypothetical input /khaʔn/ in (10) would become [kha n] in the output. If indeed the optimum is [kha n], which shows the coalescence of the glottal stop and the preceding vowel, the ranking is as in (11). The competing candidates [khan] (deletion of a glottal stop) and [khaʔ] (deletion of a nasal) violate MAX. While the faithful candidate [khaʔn] violates *COMPLEXCODA, the optimum [kha n] violates lower ranked UNIFORMITY constraint that bans coalescence in the output.

(11) Ranking: *COMPLEXCODA, MAX >> UNIFORMITY /khaʔn/ *CC MAX UNIFORMITYa. [kha n] * b. [khan] W * L c. [khaʔ] W * L d. [khaʔn] W * L

2.4 Vowel Quality This section is based on Watkins (2000). Open syllables with low tone, high tone, or creaky phonation have tense vowels as in (12). In contrast, vowels in closed syllables are realized with lax vowels or diphthongs as in (13).

The vowel /ɛ/ in syllables with a glottal stop coda and the vowel /a/ in closed syllables also appear in open syllables. A detailed phonetic study by Watkins, however, shows that the quality of these two vowels in closed syllables is different from when they occur in open syllables. When the first formant (F1) and the second formant (F2) are plotted, the vowel /a/ in closed syllables is more closed and front than the vowel /a/ in open syllables. The vowel /ɛ/ before a Draft

glottal stop is more open and back than the vowel /ɛ/ in open syllables (Watkins 2000: 147).

(12) Open syllables (Watkins 2000:145)

a. low b. high c. creaky /i/ miː (name) míː fire mí mother /e/ meː May méː ask mé forget /ɛ/ mɛː mother mɛː vote mɛ without /a/ maː hard ma ː towering ma female /ɔ/ mɔː look up mɔː tired mɔ tilt up /o/ moː heaped mo ː sky mo because /u/ muː nature mu ː drunk mu respect

(13) Closed syllables (Watkins 2000:145) Nasal Non-nasal a. low b. high c. creaky d. glottal stop /ɪ/ mɪ ː fond of mɪː king mɪ (particle) mjɪʔ river /eɪ/ meɪ ː season meɪː girl meɪ say meɪʔ friend /ɛ/ mɛʔ dream /aɪ/ maɪ ː mile maɪː mine paɪ pint maɪʔ stupid /a/ ma ː plaster ma ː recite ma jiggle ma ʔ March /aʊ/ maʊː younger

brother maʊː drive maʊ wait maʊʔ haughty

/oʊ/ moʊː storm moʊː hate moʊ flour moʊʔ pearl /ʊ/ mʊː Mon mʊː decorate tʊ wrinkle mʊʔ smooth

Syllables with glottal stop codas illustrate that the vowel quality is the same as in other closed syllables. This similarity supports the claim that glottal codas are present in Burmese. 2.5 Duration: phonetic support Watkins (2000) also conducts phonetic studies on syllables with creaky phonation and syllables with glottal stop. The table in (14) shows the duration, the change in fundamental frequency (F0), and the closed quotient6 change of each category. The results demonstrate that syllables with creaky phonation show great similarities with syllables with a glottal stop.

6 Closed quotient is the percentage of closed glottal phase within one glottal cycle (p.c. Jonathan Wright).

Draft

(14) Comparing Burmese syllables (Watkins 2000) glottal

stop creaky phonation

high tone

low tone

a. duration short short long longb. F0 falling falling rising low c. closed quotient high high low low

The results of the change in F0 and the closed quotient change are reproduced

in (15).7 Syllables with a glottal stop are marked as ‘killed’, and syllables with creaky phonation are marked with ‘creaky’ in the graphs. F0 drops significantly in syllables with creaky phonation and a glottal stop. Note that the onset of F0 in these syllables is higher than the onset of high tone.

The closed quotient becomes higher in syllables with creaky phonation and syllables with a glottal stop. Higher closed quotient means the sound is more creaky, and lower closed quotient means that a sound is more breathy.

(15) Creaky tone and stop tone (Watkins 2000)

a. Fundamental frequency b. closed quotient

I recorded a male Burmese speaker in his 60’s from Nyaung Shwe to confirm

the measurements. The consultant read each word three times in isolation from Burmese writing to an electret condenser microphone (Sony ECM-F01). The recording was done at the sampling rate of 44,100 Hz. Using Praat 4.5.18 (Boersma & Weenink 2007), the duration was measured using visual cues from spectrograms and waveforms.

This experiment confirms results on duration from previous studies by Watkins (2000) and Mehnert and Richter (1976) as in (16). In all three experiments, syllables with a glottal stop (marked as ‘Stop’) are the shortest.

(16) Comparison of duration

a. Nyang Shwe speaker (left): Low > High > Creaky > Stop b. Mehnert and Richter (1976: 146): Low, Creaky > High > Stop c. Watkins (2000:142, right): Low, High > Creaky > Stop

7 I would like to thank Justin Watkins for granting permission to reproduce these graphs.

Draft

613

446

354

165

384

231

384

158

0

100

200

300

400

500

600

700

low high creaky stop

Lee M&R Watkins

Figure 1: The mean duration of Burmese syllables The short duration in glottal stop syllables is in accordance with the claim that

a phonological mora contributes to the duration of a syllable (Hubbard 1994) if we consider glottal stops as non-moraic segments. So, the presence of a glottal stop makes syllables shorter. Syllables with a glottal stop coda have a structure as in (17c). Open syllables (17a) or syllables with a nasal coda (17b) are longer than syllables with a glottal coda (17c) because they are bimoraic.

(17) Burmese syllable structure (see also Green 2005: 8)

closed open nasal coda glottal coda

σ μ μ

σ μ μ

σ μ

a. kh à b. kh à N c. kh a ʔ The shortness of syllables with creaky phonation is due to the phonetic effect of creakiness. The creakiness obscures the end of the F0 and results in shorter duration than syllables with high tone or low tone. Based on his findings, Watkins (2000) also suggests that the shortness of syllables with creaky phonation may result from glottal closure. See also Thein Tun (1982) for relevant discussion.

The evidence presented in this section supports the claim that Burmese permits glottal stops in codas. Glottal stop codas assimilate to the following onset, glottal codas block obstruent voicing, glottal codas cannot co-occur with nasal codas, and vowel quality in syllables with a glottal stop coda follows the pattern found in closed syllables. Moreover, the phonetic experiments support the presence of glottal stops. Draft

The disagreement in the Burmese tonal literature is due to the contentious status of final glottal stop and to the definition of what constitutes a tone. As a result, Burmese has been described as having four tones (Okell 1969, Thein Tun 1982), or three tones (Bernot 1963, Yip 1995). Bradley (1982), on the other hand, proposes that Burmese is a register language. Previous research has focused on the phonetic properties of tone to determine the tonal inventory. While phonetic properties need to be carefully examined in understanding tonal qualities, the inventory of phonological tones requires phonological evidence, as given here (see also Chen 2000: 17-19, Maddieson 1978: 44).

3 Tonal neutralization In section 2, phonological as well as phonetic evidence showed that Burmese has a glottal stop coda. In this section, the influence of glottal stop codas on tone is analyzed. Glottal stop syllables show phonetic similarities to creaky phonation syllables in duration, fundamental frequency (F0), and closed quotient (Watkins 2000). In the analysis proposed here, however, the constraints targeting glottal stops are different from the constraints on creaky phonation.

When a glottal stop coda is present in Burmese, tones are neutralized to high tone. This neutralization will be accounted for by proposing the markedness constraint *ʔ/L, banning low tone on a glottal stop. Tones on syllables with creaky phonation will also be explained as a neutralization process to high tone by proposing the markedness constraint *V/L that bans low tone on creaky syllables.

The neutralization to high tone before glottal stops is discussed in section 3.1. In section 3.2, the neutralization to high tone in syllables with creaky phonation is presented with a support from an alternation that shows the neutralization of low and high tone syllables to creaky phonation.

3.1 Neutralization to high tone before a glottal stop This section presents an analysis that neutralizes underlying high and low tones to high tone before a glottal stop coda. A key assumption is that consonants are Tone Bearing Units (TBU’s), and no faithfulness constraints preserve underlying tones on consonants.

Moras as well as syllables are argued to be TBU’s in the tonal literature (Bao 1999, Duanmu 1990, Yip 2002). I propose that consonants can also be TBU’s. The difference between consonantal TBU’s and other TBU’s lies in how tones are associated with TBU’s. While moraic TBU’s are associated with a tone via prosodic nodes such as moras or syllables, consonantal TBU’s are directly associated with tones via a root node (marked with a circle) as in (18). (18a) is a language that has moraic TBU’s, and (18b) is a language in which syllables are TBU’s.

Draft

(18) H tone syllable a. TBU = mora b. TBU = syllable σ

μH | o o oH [kh a ʔ]

σH μ | o o oH [kh a ʔ]

The proposal will be illustrated using the examples in (18). In a language with

moraic TBU’s as in (18a), the consonantal TBU’s are assigned high tone by being associated with a high tone on the root node. In a language with syllabic TBU’s, consonantal TBU’s are assigned high tone as shown in (18b).

Crucially, there are no faithfulness constraints that preserve underlying consonantal tone. Underlying tones on moraic or syllabic TBU’s, on the other hand, can be preserved by faithfulness constraints that only target TBU’s associated with a mora as in (19).

(19) Faithfulness constraints

a. DEP-T The tone associated with a moraic (or syllabic) TBU in the output has a corresponding tone in the input.

b. MAX-T The tone associated with a moraic (or syllabic) TBU in the input has a corresponding tone in the output.

c. IDENT-T The tone associated with a moraic (or syllabic) TBU in the output is the same correspondent tone in the input.

These faithfulness constraints do not target consonantal TBU’s. Thus, if

consonantal TBU’s have tones in the input banned by markedness constraints, assuming Richness of the Base, the tones will not be preserved because the optimal candidate of the particular input must obey markedness constraints.

For example, the markedness constraint *ʔ/L bans a glottal stop from being associated with a low tone. When *ʔ/L dominates IDENT-T, and if an input has a low tone on a glottal stop as in /CVʔL/, the input would not surface faithfully. The faithfulness constraint IDENT-T in (19c) does not preserve tones on glottal stops and the markedness constraint *ʔ/L bans low tone on glottal stops. So, a hypothetical input /Caʔ/ will surface with high tone [Ca ʔ].

Conversely, when IDENT-T dominates *ʔ/L, an input tone associated with a moraic or syllabic TBU surfaces faithfully. In this situation, an input with a low tone glottal stop /Caʔ/ will surface with an output with a high tone glottal stop due to *ʔ/L and the lack of faithfulness constraints. However, the input tone on a mora or a syllable should not undergo changes in the output as in [Caʔ] due to the higher ranked faithfulness constraints that target only moraic TBU’s. Draft

So, the neutralization of tones in syllables with a glottal stop in Burmese is due to a ranking in which *ʔ/L dominates IDENT-T. The tone neutralizes to high tone due to the markedness constraint that bans low tone on a syllable with a glottal stop. An input with low tone /khaʔ/ becomes [khaʔ] in the output as in (20).

(20) *ʔ/L >> IDENT-T /khàʔ/ *ʔ/L IDENT-T a. σH

μ | [kha ʔ]

*

b. σL μ | [kha ʔ]

W* L

The markedness constraint *ʔ/L that prefers high tone on a glottal stop is

perhaps the opposite to what might be expected. Phonetic studies including the one on Burmese by Watkins (2000) show that the fundamental frequency of syllables with a glottal stop is phonetically realized with pitch lowering. The pitch lowering in glottal stops is due to the creaky phonation on its syllable. As shown in a study on Arabic, however, final glottal stops raise the pitch of the preceding vowel (Hombert et al. 1979). In this paper, glottal stops will be regarded as a pitch raiser, while creaky phonation will be seen as a pitch lowerer.

At least in Burmese, there is supporting phonetic evidence that glottal stop syllables in fact have high tone. The F0 measured by Watkins (2000) demonstrates that the onset of the measured F0 in glottal stop syllables is higher than the high tone, showing that glottal stops require a high tone. The drastic drop of F0 in glottal stop syllables, however, is a phonetic effect caused by creaky phonation of the glottal stop.

3.2 Neutralization to high tone in creaky phonation In addition to glottal stop syllables, I propose that the neutralization of tones also occurs in syllables with creaky phonation. As discussed in section 2.5, syllables with creaky phonation and syllables with a glottal stop show phonetic similarities in fundamental frequency and closed quotient.8

8 Michaud (2004: 137-139) provides an insightful overview of phonetic distinctions among glottal stop, glottal constriction, and creaky voice. While a glottal stop is “a gesture of closure that has less coarticulatory effects on the voice quality of the surrounding segments”, glottal constriction is realized throughout the whole syllable rhyme. Creaky voice is “the irregular, low-frequency

Draft

In syllables with creaky phonation, the pitch will become lower as attested in previous phonetic studies. Even so, I propose that low tone is banned in syllables with creaky phonation *V/L.

The neutralization in syllables with creaky phonation results from constraint interaction between IDENT-T and *V/L. In Burmese, *V/L dominates IDENT-T. So, a creaky syllable with a low tone in the input /kha/ will surface with a high tone [kha ]. The creaky phonation on the vowel will make a drastic fall of F0 as shown in (15a).

There are supporting data that show neutralization of tones in syllables with creaky phonation. Syllables with low or high tone in Burmese become creaky when they form a word with the following syllable. Okell dubs these low and high tone syllables as ‘creakable’ syllables.

The examples in (21) show the environment in which creakable syllables acquire creaky phonation. Creaky (or creakable) syllables are bold. Low tone syllables in (21a-e) and high tone syllables in (21f-h) become creaky when they are followed by another syllable. Syllables that follow creakable syllables can be any syllable or tonal type observed in Burmese.

(21) Induced creaky tone (Okell 1969: 18-20) Low tone syllable becomes creaky before a. lade (he) came ladelu low the man who came b. θounze thirty θounzeŋá high thirty-five c. θu he θṵjḛ creaky phonation his d. ŋabei five feet ŋabei ʧauʔleʔma glottal stop five feet six inches e. pjo be happy mǝpjotǝbjo toneless not very happy High tone syllable becomes creaky before f. sǝka word sǝɡaɡou low even a word g. phǝja pagoda phǝjabeí high the side of the pagoda h. mín you minǝsa toneless instead of you

Syllables with a glottal stop coda are not creakable. Because syllables with glottal stop codas neutralize to high tone, they cannot neutralize to low tone as expected in syllables with creaky phonation. Thus, the exemption of syllables with glottal stop from being creakable can be accounted for by the tonal neutralization process. Okell states that syllables with creaky phonation are not creakable, either. Considering that creaky syllables already have creaky phonation, the neutralization to low tone accompanying creaky phonation does not show any effect. The non-application of creaky phonation can also be explained if syllables with creaky phonation cannot become more creaky. vibration of the vocal folds”. For example, in Hanoi Vietnamese, glottal constriction and creaky voice are significant in distinguishing some tones.

Draft

4 Tone and syllable types in Burmese This section discusses the relationship between tone and syllable types, which has been observed in many Southeast Asian languages such as Vietnamese or Hmong (Chen 2000: 5-6). In these languages, it is common that syllables with obstruent codas (so-called checked syllables) have fewer tones than open syllables or syllables with a sonorant coda (so-called smooth syllables). Tonal contrasts regarding syllable types have been reconstructed for Proto-tone systems in Chinese, Thai, Vietnamese and Hmong (Sagart 1988). In Burmese, syllables with a glottal stop or syllables with creaky phonation neutralize the two-way tonal contrast observed in open syllables or syllables with a nasal coda.

Theories proposed to account for the asymmetry of the number of tones in syllables with an obstruent coda include two-system theory, subset theory, and underspecification theory (from Hyman 2007).

The two-system theory would contend that Burmese has four tones: three tones (low, high, creaky) in smooth syllables and one tone in glottal stop syllables (checked syllables). On the other hand, the subset theory would regard the tone on a glottal stop syllable as an allotone of other tones. The underspecification theory based on Weidert (1987: 368) would describe the Burmese tone system with a three-way tonal contrast, while glottal stop syllables are not specified with tones.

In contrast, this paper argues that the effect of glottal stop coda in Burmese is a result of tonal neutralization. Burmese has two tones: high and low. Syllables with glottal stop force neutralization to high tone, and syllables with creaky phonation force neutralization to low tone. Thus, this analysis follows the tentative conclusion by Hyman (2007), which suggests the validity of the subset theory in languages that show effects of coda consonants on tone. 5 Conclusion This paper has shown that Burmese has a glottal stop coda based on arguments such as place assimilation, blocking of obstruent voicing, phonotactic restrictions on clusters in the coda, and vowel quality in closed syllables. Moreover, phonetic experiments support this claim by showing that syllables with a glottal stop have shorter duration due to the non-moraic coda. Burmese has two tones: high and low. The two-way tonal contrast neutralizes to high tone in syllables with a glottal stop coda due to the markedness constraint *ʔ/L outranking relevant faithfulness constraints. In syllables with creaky phonation, the tonal contrast neutralizes to high tone because of the markedness constraint *V/L. The neutralization is achieved by the current proposal, in which consonants are TBU’s and no faithfulness constraints preserve underlying tone on non-moraic TBU’s such as consonantal TBU’s. Draft

References Bao, Z. 1999. The Structure of Tone. Oxford University Press. Beckman, J. 1997. Positional faithfulness, positional neutralization, and Shona vowel harmony.

Phonology 14.1-46. Bernot, D. 1963. Esquisse d'une description phonologique du birman. Bulletin de la Société de

Linguistique de Paris 58.164-224. Bradley, D. 1982. Register in Burmese. In Papers in South-East Asian Linguistics 8: Tonation, ed.

by D. Bradley, 117-132. Canberra: Australia National University. Broselow, E. 2001. Uh-oh: Glottal stops and syllable organization in Sulawesi. In Surface syllable

structure and segment sequencing, eds. by E. Hume et al., 77-90. Leiden: Holland Institute of Generative Linguistics.

Brunelle, M. 2005. Register in Eastern Cham: Phonetic, Phonological and Sociolinguistic approaches, Ph. D. Dissertation, Cornell University.

Chen, M. 2000. Tone Sandhi: Patterns across Chinese Dialects. Cambridge University Press de Lacy, P. 2006. Markedness: Reduction and preservation in phonology. Cambridge University

Press. DiCanio, C. 2006. The Phonetics of Consonantal Pitch Perturbation in a Complex Tone System.

Paper from Conference in Language Complexity, Dynamique du Language, Lyon, France. Green, A. D. 2005. Word, foot, and syllable structure in Burmese. In Studies in Burmese

linguistics, ed. by J. Watkins, 1-25. Pacific Linguistics. Hombert, J.-M., J. J. Ohala, & W. G. Ewan 1979. Phonetic explanations for the development of

tones. Language 55.37-58. Hubbard, K. 1994. Duration in moraic theory, Ph. D. Dissertation, UC Berkeley. Hyman, L. 2007. Coda Constraints on Tone. Paper from Workshop on Segments and Tone,

Amsterdam, the Netherlands. Itô, J. 1989. A prosodic theory of epenthesis. Natural Language and Linguistic Theory 7.217-259. Maddieson, I. 1978. The frequency of tones. UCLA Working Papers in Phonetics 41.43-52. Maran, L. R. 1971. Burmese and Jingpho: A study of tonal linguistic Processes. Urbana, U. of IL. Matisoff, J. A. 2003. Lahu. In The Sino-Tibetan languages, eds. by G. Thurgood & R. J. LaPolla,

208-221. London & New York: Routledge. Mehnert, D. & E. Richter. 1976. Untersuchungen zur Phonetik und Phonologie des modernen

Burmesischen: Teil 3. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikations-forschung, 29.145-166.

Michaud, A. 2004. Final Consonants and Glottalization: New Perspectives from Hanoi Vietnamese. Phonetica 61.119-146.

Okell, J. 1969. A reference grammar of colloquial Burmese. London: Oxford University Press. Okell, J. 1994. Burmese: An Introduction to the Script. NIU Center for Southeast Asian Studies. Prince, A, and P. Smolensky. 1993/2004. Optimality Theory: Constraint interaction in generative

grammar. Oxford: Basil Blackwell. Prince, A. 2000. Comparative tableaux. Ms., New Brunswick, NJ: Rutgers University. Sagart, L. 1988. Glottalised Tones in China and South-East Asia. In Prosodic Analysis and Asian

Linguistics: To Honour R.K. Sprigg, eds. by D. Bradley, E. J. A. Henderson and M. Mazaudon, 83-93. Canberra: Australian National University.

Thein Tun, U. 1982. Some acoustic properties of tones in Burmese. In Papers in South-East Asian Linguistics 8: Tonation, ed. D. Bradley, 77-116. Australian National University.

Watkins, J. 2000. Notes on creaky and killed tone in Burmese. SOAS Working Papers in Linguistics 10.139-149.

Weidert, A. 1987. Tibeto-Burman Tonology. Amsterdam: John Benjamins. Yip, M. 1995. Tone in East Asian languages. In The Handbook of Phonological Theory, ed. by J.

A. Goldsmith, 476-494. Cambridge: Blackwell. Yip, M. 2002. Tone. Cambridge University Press. Draft