CHAPTER TWO

SEGMENTAL INVENTORY: A REVIEW

2.1 Introduction

In investigating a language, one of the first tasks of a linguist is to

establish the phonemic inventory of the language. Dzongkha has been studied
by a number of linguists, though not many, for example, by Mazaudon (1985),
Mazaudon & Michailovsky (1988) Van Driem (1992) and Watters (2002). The
phonemic inventory of the language differs (slightly) in all the four linguists.
Though all the four works put forward their inventories for standard Dzongkha,
their language informants are speakers of other regional varieties. Mazaudon
(1985) and Mazaudon & Michailovsky (1988) have their informants from
Chapcha 7, Watters's (2002) informant is from Pasakha 8 and Van Driem (1992)
does not mention about the background of his informants, but his work, being a
grammar on Dzongkha, is presumed to be on the standard variety spoken in
Thimphu. So except Van Driem (1992), all others describe the regional dialects
of Chukha District.

Therefore there is a need, firstly, to establish the segmental inventory of

the dialect of Paro, which is our subject of study.

In this chapter, we first review the phonemic inventory of Dzongkha

put forward by the four linguists who have worked on the language, and
propose a revised one with illustrative examples showing the phonemic
contrasts. We review the consonant inventories of the four linguists in table
2.1, 2.2, 2.3 and 2.4, and then propose a consolidated one. (We retain the
original phoneme symbols by the authors in the review section and use
appropriate (revised) IPA symbols in our consolidated proposal).

7 Chapcha is a village under Chukha District (southern Bhutan). but the place is approximately
50 kilometers south ofThimphu.
8 Pasakha is a village in the extreme south of Chukha District. bordering West Bengal, India.

15
 9
2.2 Consonant phonemes

In this section we review the consonant phoneme inventories by the

four linguists mentioned above. It is noteworthy to mention here that there is
no detailed phonetic and phonological study done on phonemes of the
language. The focus of all the four linguists has been on other aspects of the
language. Mazaudon (1985) is a study on Dzongkha number systems;
Mazaudon and Michailovsky (1988) is a diachronic study of Dzongkha level
tone and falling melodic contour in the perspective of their historical origin;
Van Driem (1992) is a descriptive grammar of Dzongkha, with a pedagogical
goal and Watters (2002) is a study of sounds and tones of Dzongkha, but in
combination with four other languages (Lhomi, Sherpa, Dolpo Tibetan and
Mugom Tibetan). He has just one inventory for all the five languages; besides,
his subject of study in Dzongkha is the regional dialect of Pasakha (a place in
southern Bhutan - Indo-Bhutan border).

Thus, there is not yet a comprehensive study done in the segmental

aspects of Dzongkha variety spoken in Paro.

We first present the consonant inventories of all the four linguists in table
2.1,2.2,2.3 and 2.4.10

Table 2.1: Consonant phonemes of Dzongkha after Mazaudon (1985)

Affricate PJ bJ Is dz
Ish dzh
Fricative s z z
zh
" zh
Nasal m n J1 I)

Approximant w hi r

<) Until we propose our new obstruent phonation ty-pes as: voiceless tense (VL T*). voiceless lax

(VLL). voiceless aspirated (VLAS) and voiced (V D), we continue to use the existing term
'devoiced' for the VLL consonants. We also maintain the phonemic symbols used by the
respective authors.
lOWe postpone our discussion on the effect of consonant and prosodic structure until chapter 3
and 4.

16
 As shown in table 2.1, Mazaudon (1985) has 43 consonants: 20 stops
with a four-way phonation distinction for stops and affricates across five places
(bilabial. dental, retroflex, palatal and velar) and a three-way phonation
distinction for fricatives across two places (dental and palatal). He does not
have the voiceless glottal fricative Ihl in his inventory. In his inventory the

palatal affricates are analyzed as palatal stops. There is no voicing contrast in

all four nasals and the dental lateral has a voiced - voiceless (I - hI) distinction.

The fourth phonation type (devoiced) in stops and affricates (bh, dh, c\.h, gh,

bjh, and dzh) and the third phonation type in fricatives (zh) and (:;;h) are not

studied thoroughly, thus we do not have any phonetic infonnation about their
contrast with their counterparts. He analyzes them either as plain voiced with
low tone merging with the voiced series or as voiceless aspirates like the
regular voiceless aspirated series but occurring with the low tone. However, he
mentions dialectal variation as the source of confusion for the phonation types.

Table 2.2: Consonant phonemes of Dzongkha after Mazaudon and

Michailovsky (1988)11

Affricate p. bj Is dz c j
tsh dzh ch
Fricative s z C :;; h
zh :;;h
Nasal hm m hn n
Approximant hi r y w

Mazaudon and Machailovsky (1988) have 46 consonants in their

inventory: 16 stops and 12 affricates across five places (bilabial, dental,
retroflex, palatal and velar) with four phonation types (voiceless, aspirated.

II Mazaudon and Michailovsky (1988) do not specify the place of aniculalion for the
consonanls. The places specified in Ihe table are assumed by the author.

17
voiced and devoiced) and seven fricatives across three places (dental, palatal
and velar) with three phonation distinctions (voiceless, voiced and devoiced),

except for the voiceless glottal fricative Ihi. Of the four places for nasals two of

them (bilabial 1m! and dental In/) make voicing contrasts, and among

approximants the lateral II/ has a voicing contrast. Consonants are not specified

for place of articulation in the inventory.

The voiceless and aspirated series are associated with the high register,
and the voiced and devoiced with the low.

Table 2.3: Consonant phonemes of Dzongkha after Van Driem (1992) "

Stop p b t d tr dr k g
bO th dO thr drO kh
Affricate pc bj ts dz c j
tsh ch
Fricative s z sh zh h
ZO zho

Approximant hi w
hr r

Van Driem (1992) identifies 44 consonants: 16 stops and II affricates

across six places (bilabial, alveolar, dental, retroflex, palatal and velar) with
four phonation distinctions (voiceless, voiceless aspirated, voiced and
devoiced) except the alveolar series of affricates, which is devoid of the
devoiced phonation, and seven fricatives across three places (alveolar, palatal
and velar) with three phonation distinctions (voiceless, voiced and devoiced),

except for the voiceless glottal fricative Ihi. According to Van Driem (1992)

the 'devoiced' consonants are called so because historically they are derived
from voiced consonants. In articulatory terms, the devoiced consonants are
unvoiced, but in contrast to the voiceless consonants. and they are followed by
a murmured or breathy voiced vowel with the low register tone. He also

" The diacritic (0) is used forthe devoiced phonation type in Van Driem ( 1992)
18
observes that for speakers other than from Western Bhutan (Thimphu, Paro,
Punakha, Haa, Wangdi Phodrang, Chukha) there is no distinction between the
devoiced and voiced. This can be put otherwise. For all other speakers of
Dzongkha, there is always a first language; Dzongkha is a second language for
them, while for speakers from western Bhutan, Dzongkha is their first
language.

Voiceless and voiceless aspirated obstruents occur in the high register

tone while voiced and devoiced occur in the low. Sonorants can occur in both
the registers.

Table 2.4: Consonant phonemes after Watters (2002) 11

Stop P b d tr dr k g
ph th th thr d'r kh kh
Affricate ptS ptS ts dz tS d3
tsh ts h
Fricative s z
Nasal m hn n Tt I)
Approximant hi I w
hr r

Waters (2002) has 43 consonants: 16 stops and 12 affricates with a four-

way phonation distinction (voiceless, voiceless aspirated, devoiced and voiced)
across five places (bilabial, alveolar, post-alveolar, retroflex and velar) of
articulation. Waters (2002) is the only linguist who investigates the segmental
properties of consonants in Dzongkha. He describes the phonetic properties of
the four phonation types for stops as:

• Voiceless> voiceless without aspiration

• Devoiced> voiceless sometimes with a small degree of aspiration 14

followed by usually breathy voice (devoiced) characterized by a slightly
longer VOT than the voiceless unaspirated series

13 The elevated I"! symbol is used by Watters (2002) for the devoiced consonants to distinguish
it from the regular aspirated Ihl consonants

19
• Aspirated> voiceless with heavy aspiration by contrast with the devoiced
series, and

• Voiced > voiceless preceded by voicing l5 .

Voiceless and voiceless aspirated obstruents occur with the high register
tone while voiced and de voiced occur in the low. Sonorants occur in both the
registers.

In the next section, we look at the differences in the phonemic inventories

of consonants.

2.2.1 Differences in consonant inventories

In Tables 2.1, 2.2, 2.3 and 2.4, a number of differences can be seen vis-
a-vIs: I. number of phonemes; 2. place of articulation and 3. manner of
articulation.

As we can see in tables 2.1, 2.2, 2.3 and 2.4, the inventories differ with
regard to the number of phonemes: Mazaudon (1985) and Watters (2002) have
43 each, Van Oriem (1992) has 44, and Mazaudon & Michailovsky (1988) has
46 phonemes. The best way to present their differences is, perhaps, in terms of
the manner of articulation.

2.2.2 Stops

As shown in tables 2.1, 2.2, 2.3 and 2.4, except for Mazaudon (1985),
all others make a four way place distinction with a four way phonation
distinction. Here again, there is a difference in place for the stops; in Watters
(2002), stops are distributed over bilabial, alveolar. retroflex and velar. while

" Waters (2002) observes that the aspiration is not consistently articulated by all speakers. and
it is not phonemic.
15 Waters's (2002) observes that the voiced stops are not completely voiced during closure. His
description of the 'voiced' phonation is little awkward because irrespective of whether the
closure duration of a stop is completely or partially voiced the sound is a voiced one.

20
Van Driem (1992) has bilabial. dental, retroflex and velar as the places.
Mazaudon (1985) has identified five places (bilabial, dental, retroflex, palatal
and velar), and four places in Mazauzon & Michailovsky (1988) are not
specified. However, all the four linguists have four phonation types; voiceless
unaspirated, voiceless aspirated, voiced and devoiced 16. Mazaudon (1985)
analyses the palatal series of affricates as stops.

2.2.3 Affricates

Van Driem (1992), Watters (2002), and Mazaudon & Michailovsky

(1988) recognize three series of affricates, while Mazaudon (1985) has only
two series (he identifies the palatal affricate series as palatal stops as mentioned
in section 1.1.1). Like the stops, affricates too have been placed over different
places; Van Driem (1992) has bilabial, alveolar and palatal places, while
Watters has bilabial, alveolar and post-alveolar, and Mazaudon (1985) has
bilabial and dental as the places. The places in Mazaudon & Michailovsky
(1988) are again not specified. All four have four types of phonations for
affricates (voiceless unaspirated, voiceless aspirated, voiced and devoiced) as
in stops, except for the alveolar series in Van Driem (1992), which has only
three. He recognizes the fact that the alveolar series has only three phonation
types: voiceless unaspirated, voiceless aspirated and voiced.

2.2.4 Fricatives

As shown in tables 2.1. 2.2, 2.3 and 2.4, all the linguists have two series
of fricatives at different places; Mazaudon (1985) has dental and palatal places
while Van Driem (1992) has alveolar and palatal, and Watters (2002) has
alveolar and post-alveolar places for the fricatives. Mazaudon & Michailovsky
(1988) do not specify the places. Except Watters (2002), who has only two
phonation types (voiceless and voiced), all others have a three way phonation

'6 The phonetic details of the so called 'devoiced' stop have never been studied in depth. The
term is assigned to the stop based on historical ground; historically this stop is derived from the
proto Tibetan voiced stop.

21
distinction (voiceless, voiced and devoiced). Waters (2002), perhaps, analyses
the devoiced phonation type as a segment that differs only in pitch (tone) with
the voiceless one (though he does not mention this explicitly in his analysis).

2,2.5 Nasals

Van Driem (1992) and Mazaudon (1985) have identified four nasals
(lm/, In!, IJlI & IIJl) while Watters (2002) has five (with one more voiceless

Ihn!) and Mazaudon and Michailovsky (1988) have six (with two more
voiceless !hml & !hn!) nasals. Watters (2002) has alveolar place for nasal In!
while Van Driem (1992) and Mazaudon (1985) have dental as the place. The
place for Inl in Mazaudon & Michailovsky (1988) is not specified. The other
places are bilabial, palatal and velar. In Van Driem (1992) and Mazaudon
(1985), all series of nasals are (simply) voiced, while the other two linguists
distinguish voiced and voiceless nasals. Watters (2002) make a distinction
between voiced and voiceless nasal for the alveolar place, and Mazaudon &
Michailovsky (1988) has the same distinction for 1m! and In!.

2.2.6 Approximants

In all the inventories, the glides IjI or Iyl and Iwl do not differ; Ijl or Iyl

is a voiced palatal glide and Iwl is voiced labiovelar glide. But sonorants (Ill &

Ir/) differ in both place and phonation types. Van Driem (1992) and Watters

(2002) have four (two voiced and two voiceless) across the same places
(alveolar). Mazaudon's (1985) inventory has three; a voiced Irl and lateral III

with a voicing distinction, while Mazaudon & Michailovsky (1988) have three
glides with a voicing distinction in the lateral.

In the next section we propose our consolidated consonant inventory of

Dzongkha.

22
2.3 Consonant Phonemes: the consolidated inventoryl7

In this section, we bring together all the four consonant inventories of

the four linguists along with our analysis and propose a consolidated consonant
inventory; additional examples (minimal pairs) are also provided. We have the
consolidated consonant phonemes in table 2.5.

Table 2.5: Dzongkha consonant phonemes 18

VLT p* t* t* k*
VLL P t k [7]"
-
0
0
r /l VLAS ph th th kh
VO b d <t 9

-......= VLT pc* t,* t.*

-::I

'"
,.Q
0
VLL
VLAS
p•
p.h t,h
tc
tch
VO b~ dl d*
VLT s* c*
VLL s c h
VO z *

VL IJ1 I}

VO m n J1 I)

-=
E
Q
VL
VO r
r
=
Q
r/l

VO j w
"
:-g
r.:;;

17 The phonemic symbols are from IPA (revised to 2005) except the voiceless tense symbols.
" In absence of a symbol for the feature ··tense"' in the IPA, we take the liberty of borrowing
the voiceless tense diacritic (C*) from Korean after Kim & Duanmu (2004), Silva (2006), Kim
(2004) and others.
"We do not specifY the phonation type of tile '1': we analyze '7' as a phonemic (or rather sub-
phonemic) segment as it contrasts with other stops and nasals (only) in the final position. thus
in brackets (we postpone further discussion on the glottal stop until chapter 5).

23
 As shown in table 2.5, Ozongkha consonants are distributed across five
places of articulation (labial, dental, retroflex, palatal, and velar) and six
manners of articulation (stop, affricate, fricative, nasal, liquid and glide). All
phonemes, stops (labial, dental. retroflex and velar). affricates (labial. dental
and palatal), fricatives (dental and palatal), nasals (labial, dental. palatal and
velar), liquids (dental and retroflex) and glides (labiovelar and palatal) occur in
more than one phonation type. Obstruents make a contrast between voiceless
tense, voiceless lax, voiceless aspirated and voiced (henceforth, VLT, VLL,
VLAS and VO respectively). A three-way voiceless stop phonation type is, so
far, believed to be attested only in Korean (Kim & Ouanmu 2004. Silva 2006,
Kim 2004 and others). We shall postpone our in-depth discussion on the VL T
and VLL distinction until chapter 3. For now, the VLT type occurs in the high
register and VLL in low.

A two-way register tone system in Ozongkha is a well-established fact.

All linguists who have worked in the language, viz. Mazaudon (1985),
Mazaudon & Michailovsky (1988), Van Oriem (1992) and Watters (2002)
have analyzed the tone system in Ozongkha as a register tone system. The VL T
and VLAS obstruents occur in the high register and VLL and VO in the low.
All sonorants occur in both the registers.

Next section is our analysis of stop consonants in Ozongkha.

2.3.1 Stops

We analyse stops with four phonation types, three voiceless (VL T,

VLL, and VLAS) as in Korean and an additional voiced (VO) one, as
mentioned in table 2.5. We analyse stop places as Mazaudon (1985) and
Mazaudon & Machailovsky (1988) do, with labial, dental. retroflex and velar.
The VL T and VLAS occur in the high register and VLL and VO occur in the
low register (see chapter 5 [on Ozongkha tone] for a detailed discussion). All
stops occur initially and medially but the occurrence of stops in the final

24
 position is restricted to voiceless (tense or lax not specified) velar Ikl and labial

/pPO. Word medially and in intervocalic positions, the VLL stops become

voiced. The phonemic contrasts of the stops across all places are given in I. In
all the examples, the low tone is not marked and high tone is marked by the
high diacritic ( ').

(I) Labial p* 'a slice of cooked meat'

p pan a 'in the middle'

ph pha? 'pig (astrology),

b ba? 'bamboo mat'

Dental t* t*a? 'tiger'

t ta? 'flag, ceremonial scarf

th tha? 'loom'

d da? 'lick (V)'

Retroflex t* t*um 'to spread around'

t til 'to make into slices'

th thU 'pattern'

ct ctll "enemy, san1eness'

Velar k* k*e 'voice (of timbre)

k ke 'line, queue'

kh khete 'clever, intelligent'

9 gep . remuneration for monk'

20 Ik/ occurs in some regional varieties instead of /pi, and sometimes both occur in free
variation.

25
2.3.2 Affricates

Dzongkha has eleven affricates spread across three places. Affricates

are spread across labial, dental and palatal places of articulation. All series of
affricates have four phonation contrasts: VL T, VLL, VLAS and VD, except the
dental series, which has only three phonation types (VL T, VLAS and VD).The
palatal affricates occur in four phonation types as in stops while the VLL is not
attested in the dental place as Van Driem (1992) correctly observes; this fact
could have been an oversight in the other authors' analyses 21 • As already
mentioned in section l.l. I. the labio-palatal series in Mazaudon (1985) are
analysed as palatal stops, we analyse them as affricates as analysed by others.
All affricates occur in initial and medial positions and, like the stops, affricates
too are voiced word medially and in intervocalic position. None of the
affricates are attested in the final position. The phonemic contrasts of the
affricates are given in (2). VLT and VLAS occur in the high register and VLL
and VD in the low (see chapter 5 for a detailed discussion).

(2) Labial p~* p~*1i 'monkey'

p~ p~1i ·thin, slim'

p~h p~hll 'bird'

b'f; b'f;ll 'cow shed'

Dental ts tslim 'fodder'

tsh tshlim 'boundary, niece'

dz dZllm 'pot, chop (V)'

Palatal t~* te*1i 'hair'

te t~ll 'tea'

t~h tchli 'pair'

d'f; d'f;ll 'hundred'

" We have no data of Dzongkha with VLL dental affricates.

26
2.3.3 Fricatives

Fricatives occur in three phonation types across three places. In terms

of place of articulation we analyse fricatives as alveolar (ls*l, lsi and Iv),

palatal (/,,*1, 1,,1 and 1.,1) and glottal (/h/) with three phonation types: VLT, VLL

and VO for dental and palatal. and a voiceless glottal Ihi following Mazaudon
and Michailovsky (1988), Van Oriem (1992 and Mazaudon (1985). The VL T
series occurs in the high register and VLL and VO occur in the low register.
Watters (2002) suggests that fricatives occur as alveolar and post-alveolar with
voicing distinction (voiceless and voiced) as shown in table 2.1; he probably
assumes tonal contrast for the two voiceless fricative). However, we analyze
fricatives as three phonemes (VL T*, VLL and VO) following other three
linguists (also see chapter 3 for the acoustic cues for the voiceless fricatives).
All fricatives occur in word initial and medial position. However, as in the
stops and affricates, the VLL series get voiced in word-medial and intervocalic
positions. All fricatives occur in the initial position and medial position. The
phonemic contrasts of the fricatives are shown in (3).

(3) Alveolar s* s*u "soil'

s se "eat (V)'

z ze "day (calendar),

Palatal ,,* ,,*a "quarrel, partition wall'

"night'
"., "a

.,a "keep (V)'

Velar h hem "blue, green

2.3.4 Nasals

Unlike all the authors in table 2.1, 2.2. 2.3. and 2.4. we will propose a
two way phonatory contrast (voiceless and voiced) across all four places
(labial, dental, palatal and velar). The phonemic contrasts in (4) support our

27
proposal for voiceless - voiced contrast. The voiceless nasals occur in the high
register and voiced ones can occur in both high and low registers (see chapter 4
for detailed discussion).

(4) mil 'lock'

mllP 'husband'

rp,lltll 'low'

rna: 'butter'

rna: 'war

'lui: 'a kind of bamboo'

n-lJ neni 'to get sick'

nllp 'black'

IJIlP 'snot'

nil 'promise (N),

IJIl 'curse

)1- ~ Jlll 'fish'

JlIl~i 'yoke'

~IlP 'reach for something (with hand)'

Jle 'sleep'

Jle 'site of pilgrimage'

~e 'injure (in an accident),

28
 IJ - 9 IJa 'a kind of fasting'

IJa 'a kind of healing chant'

'pillow'
93

IJIl 'I'

IJe 'five, drum'

ge 'early'

2.3.5 Liq uids

Our analysis of liquids conform to the analysis of Watters (2002) and

22
Van Driem (1992) in both place (dental for III and retroflex for Ir/) and
phonation type (voiceless and voiced), the other two linguists do not include
the voiceless retroflex Irl in their inventories. The voiceless - voiced phonemic
contrast of the liquid consonants are presented in (5).

(5) I-I III "mountain'

Ie 'payment, wage'

Ie 'god, deity'

lu 'sheep'

hi 'pour'

Iii 'dismantle'

r- r rum 'dismantle (of stone fence) (lNTR)'

" Articulatorily. the tongue touches both teeth and alveolar ridge, but we choose to specify
'dental' as the place feature.

29
 plm 'dismantle (of stone fence) (TR)'

rup 'help (V)'

'tea mixer / churner"

2.3.6 Glides

The two glides (labio-velar /w/ and palatal /jI) are the only consonants
that correspond across all the four inventories. The phonemic contrasts of
glides are given in (6). Both the glides occur in both high and low Register.

(6) w wal) 'name of a place'

'blessing from a monk'

j jl1 'one of a pair'

jti 'itch'

We now turn to vowels.

2.4 Vowel Phonemes

In this section, we review the vowel phonemes by the four linguists.

The vowel inventories do not differ as much as the consonants do; yet, each
inventory differs from the other in the analysis of vowel phonemes. We shall
first present the four inventories by the four linguists and identify the areas in
which they differ. As in the consonant inventories, we shall again retain the
respective authors' original vowel symbols in the review section and then use
IPA symbols in our consolidated inventory. The four vowel inventories are
presented below after their respective authors.

30
 Figure 2.1: Mazaudon (1985)

y u

e
e o
a

In figure 2.1, Mazaudon (1985) has eight vowels (three front unrounded

Iii, Iy I and lei; two front rounded Iyl and lei; two back rounded lui and 101 and

an open unrounded Ia!). And there is no further discussion on the vowels by the
author.

Figure 2.2: Mazaudon & Michailovsky (1988)

ue u

e oe o

In figure 2.2, Mazaudon & Michailovsky (1988) have eight vowels

(three front unrounded Iii, lei and lEI; two front rounded luel and loe/; two back

rounded lui and 101 and an open vowel laI). They claim that there is no

opposition between the front unrounded lei and front unrounded IE I in short,
open monosyllables, and suggest that the realization is rather open. They also
claim that vowel length is phonologically distinctive, but there are no examples
in support of the contrasts.

Figure 2.3: Van Driem (1992)

ii u

e o

a a

31
 In figure 2.3, Van Driem (1992) has eight vowels: two front unrounded
Iii and lei; two front rounded lilI and 16/; two back rounded lui and 101 and two

low open lal and Ia!. He analyses the length of five vowels as phonologically

distinctive, the vowels are la!, lei, Iii, 101 and lui, while lilI, 161 and Ia! as

always long.

(7) Short

a - a: map 'husband' ma:p 'red'

e .. e: sep 'stallion' se:p 'yellow'

i - i: tsip 'stone wall' tsi:p 'astrologer'

0 .. 0: phop 'put something down' pho:p 'cup'

u .. u: ku 'body part (Hon), ku: 'scoop'

However, he notes that the difference between the long and the short
vowels are of timbre and vowel quality, besides length.

Figure 2.4: Watters (2002)

y u

[e] e (0)

[( E) 1 (~)

(re) [u]

In figure 2.4, in Watters' (2002) inventory, the front and central vowels
which are in square brackets [ ] are problematic vowels. He does not identify if

the difference is of vowel duration or quality in both the pairs (lei vs / E/ and

131 vs lui). The front and back vowels which are in parenthesis ( ) are supposed

to be having allophonic relationships, but he does not specify which is an

32
 allophone of which in both the pairs (front and back). This leaves the number
of vowels in his inventory uncertain; however he calls for further phonetic
study on the vowels, especially the front vowels to further establish the vowel
contrasts.

In the next section. we present our consolidated inventory of Dzongkha

vowels.

2.5 Vowel phonemes: the consolidated inventory23

Dzongkha has, perhaps, one of the richest vowel inventories among the
Tibetan dialects. Besides ten monophthongs (pure vowels) it has seven
diphthongs. The monophthongs further have secondary features - length and
nasalization contrasts.

In this section, we review all the previous vowel inventories and then
propose a consolidated one with acoustic evidence. The formant (F I, F2 and
F3) matching experiment is a well-established phonetic paradigm in research
on the distinction of vowel quality. We shall establish vowel contrast with
vowel formant values. First, we shall present formant readings of a male
speaker to support our analysis of the vowel contrasts in Dzongkha. Then
further vowel qualities such as length and nasalization will also be discussed in
this section. We first present the vowel formants.

2.5.1 Vowel formants: facts about vowel formants and vowel quality

Johnson (2000) suggests that the distinctive features of vowels, in other

words the vowel contrasts are better analyzed with acoustic properties. rather
than articulatory properties. We have the summary of the correlation between

2' All phoneme symbols are from IPA (revised to 2005)

33
 the two fonnant frequencies (F I and F2)24 and the vowel quality (Ladefoged
1981) listed as:

I. The vowel distinction is best noticeable in the frequencies of F I and F2

11. Vowel height correlates with the frequency of the first formant (F I), i.e.
lower vowels have higher F I, and higher vowels have lower F I
111. Vowel frontness correlates with the second fonnant (F2), i.e. front vowels
have higher F2 and back vowels have lower F2.

2.5.2 Formant plot experiment

Ten words, which are minimal pairs. of a male speaker are recorded
(three tokens) with all the ten vowels preceded by the voiceless palatal fricative

consonant IQ/. Then FI, F2 and F3 values of all the three tokens are measured

and their averages listed in table 2.6.

Table 2.6: Average formant (Fl, F2 and F3) (Hz) for Dzongkha vowels

i Qi 'lice' 289 2044 2721

e 2 Qe 'know' 344 1861 2611

E 3 QS 'strength' 435 1916 2574

Y 4 QY 'reminiscence' 289 1755 2117

e 5 ~e 'part of plough' 326 1624 2117

11 6 QII 'meat' 472 1331 2172

a 7 Qa 'quarrel' 600 1112 2245

0 8 QO ~dice ' 399 765 2373

:l 9 Q:J 'come' 472 783 2355

u 10 QU 'force, strength' 326 801 2154

24 We chose to neglect the values of F3 for our analysis since our main concern is with the
vowel 'height" and 'frontness'

34
 The mean formant values for the first two formants (F I and F2) are
plotted in figure 2.5 using JPlotformant software (Billerey - Mosier 2002) for
the male speaker and we get the results in figure 2.5. (For lack of IPA symbol
options in the software, vowel numbers are used instead of vowel symbols, [see
table 2.6 for vowel equivalents D.

Figure 2.5: Fl - F2 Plot of Dzongkha vowels

, ,
F2
, ,
2000 1500 1000 500
J I I I I I I I I I I I I 1.....200

0 0 -300
1 40 0
0
S 10
2
0 -401)
0 9
3 0 0
6 8 -500
F1
0 -600
7

Looking at figure 2.5 and the vowel contrasts in table 2.6, it is apparent
that Dzongkha has ten vowel contrasts, and the vowel inventory quite well
shapes with the IPA vowel plotting. We have three front unrounded (Iii, lei and

I E I), three back rounded (lui, 101 and hI), two front rounded Iyl and lei, a low

unrounded central vowel/TJI and a central open unrounded la/.

The front open unrounded Iii and rounded Iyl have exactly the same FI

value (289 Hz), while the back rounded lui is slightly lower (by 37Hz) than its

front counterparts. The high mid front rounded lei is slightly higher (by 18 Hz)

than its front unrounded counterpart lei, and the front unrounded lei is lower

than lei by approximately 100 Hz. The front unrounded low vowel [~l occurs

as an allophone of lei. The back rounded vowels lui, 101 and hi are more or less

consistently spaced in the vowel continuum.

35
 The low open vowel lal is lower than the central vowel lei, and it is

closer to back vowels than front vowels, and thus, we may analyze it as a back
vowel.

Thus, Dzongkha vowel inventory can be classified as shown in table 2.7.

Table 2.7: Classification of Dzongkha vowels

i High Front Unrounded

e mid Front Unrounded

e Front

y High Front Rounded

e Higher mid Front Rounded

11 Lower mid Centre

a Low

0 Rounded

:) Lower Back Rounded

u High Back Rounded

In the next section, we look at the secondary features such as vowel

length and nasalization contrasts.

2.6 Vowel length and nasalization

2.6.1 Vowel length

Vowel length in Dzongkha is contrastive, except the central vowel lei;

this vowel is always short. Difference in vowel length also has implication on
prosodic structure in Dzongkha. We will, however, postpone our discussion on
'vowel length and prosody' until chapter five on Dzongkha Tones. The
minimal pairs in table 2.8 and 2.9 show vowel length contrast in Dzongkha.

36
 Table 2.8: Dzongkha vowel length contrast in open monosyllabic words

e pe 'example' pe: (V)'

e CE 'strength' CE: 'wander (V)'

y phy 'offering to phy: 'push (V)'

deity'

e tcha ·you' tcha: 'go mad'

a pha 'pig (astrology), pM: 'there'

phS 'yeast' phS: 'stomach'

"
0 to 'food' to: 'release (V)'

u cll cll:

Vowel length is more apparent in open monosyllabic words than in

closed syllables. However, there are examples of length contrasts in closed
syllables as well. Few examples are given in table 2.9,

Table 2.9: Dzongkha vowel length contrast in closed monosyllabic words

dzim 'tongs dzi:m 'eyebrow

phop 'put down' pho:p 'bowl, cup'

sem 'mind, soul' se:m 'pea'

thap 'forehead' tha:p 'absolute, complete'

gep 'remuneration for monks' g e :P 'king'

37
2.6.2 Nasalized vowels

Nasalization is distinctive. at least with four vowels, three back vowels

ii, 6 and a and the front high unrounded i in Dzongkha. Nasalized vowels are

attested only in open (both short and long) syllables. Table 2.10 lists minimal
pairs of nasalized vowels.

Table 2.10: Dzongkha vowel nasalization contrast

i el 'lice' ei 'tree, firewood'

pelt6 'tip, terminal' pei: 'glue

ri •caste , ri 'price, long'

u su 'incense' sii ·tell, say (Hon)'

eu '\vater' eii: 'basket'

phu 'rub (V)' phd 'stores (N)'

0 kh6 'he' kh6 'freeze (V)'

s6 'teeth' s6 'go (IMP)'

t6 'support' t6 'penalty'

a pii? 'meat' pa: 'grass lawn'

ga: 'tent' ga: 'hill'

thii? 'loom' th~: 'play ground'

Length in nasalized vowels is also contrastive. This is shown in table

2.11.

Table 2.11: length contrast in nasalized vowels

dzI 'indifferences' dzI: 'pond, fight (V)'

sii 'story' sii: 'say, tell (Hon)

kh6 'gravy (of curry)' kh6: 'they'

sa 'incense leaves' sa: 'balance (N),

38
2.7 Diphthongs

Along with the ten monophthongs, there are seven diphthongs as well.
Diphthongs in Dzongkha occur only in open syllable. Dzongkha diphthongs
are listed in table 2.12 with examples.

Table 2.12: Dzongkha diphthongs

'hom, fence'
mu
au au 'thief
'stick'
khau
iu tiu 'pigeon'
'bullet, stupid'
diu
eu teu 'navel'
'hailstorm '
seu
ell thw 'ash'
'forehead'
pew
ou thou 'hammer'
'hide'
kou
Ili Ili 'mother
'yuck!'
khlli

Thus, Dzongkha has 34 vowels: 10 short monophthongs; 9 long (:)

monophthongs; 4 short nasalized n monophthongs; 4 long nasalized (:)

monophthongs and 7 diphthongs. All 34 vowels in Dzongkha are listed in table

2.13.

39
 Table 2.13: Dzongkha vowel phonemes

Short
Long

Short nasalized

Long nasalized

Diphthongs

2.8 On the size of Dzongkha segmental inventory

It is worthwhile reviewing the segmental size of Dzongkha before we

conclude this chapter, because the size (83 phonemes) does draw our attention.
Including length distinction in vowels (both non-nasalised and nasalised) we
have claimed that Dzongkha has 83 phonemes, which includes 49 consonants.
The question that arises now is, are these many, 83 phonemes, usual in a
language? The answer to this question necessitates an areal survey.

The number of segments in languages seems to vary widely. According

to the ULeA UPSID 25 survey, the smallest inventories included only II
segments as in Rotokas 26 and Mura27 , and the largest has 141 segments as in
28
!XU (Maddieson 1984). However, the study also claims that the typical size
of an inventory lies between 20 and 37 segments, but only 7% of the languages
in the survey fall within these limits. The mean number of segments per
language is believed to be little over 31 according to the survey.

Then, the number of segments in Dzongkha seems to be little more than

usually expected in languages in general; it exceeds more than double the size
of the average number of segments in languages. However, a quick survey on
the number of phonemes in peripheral Tibeto-Burman languages saves us from
our disbelief. The size of phonemic inventories tends to vary considerably

lS (University of California, Los Angeles) Phonological Segment Inventory Database

26 Rotokas (East Papuan) is spoken in Vougainville, an island in Papua New Guinea
27 Mura (Amazonas) is spoken in Brazil
28 !Xu (Khoisan) is a language spoken in Southern Africa

40
across Tibeto-Bunuan languages. Ao (Chungli)29 (Temsunungsang 2009) has
just 19 segments (4 vowels and 15 consonants) at one extreme, on the other,
Zhongu Tibetan JO (Sun 2003) too, like Dzongkha, has as many as 84 phonemes
(42 simple consonants, 33 complex consonants, and 9 vowels), which is two
more than number of segments in Dzongkha. In fact, quite a number of dialects
of Tibetan seem to be bigger-sized in tenus of their segmental inventory; Kham
Tibetan l ] (Olson n.d) has 64 segments (57 consonants and 7 vowels), and
Rgyalthang Tibetan J2 (Xiaosong 1996) has 73 segments (42 consonants and 31
vowels, including diphthongs and nasa Ii sed vowels). Thus, the segment size of
Dzongkha (83 phonemes) seems to be quite usual by Tibetan standard.

2.9 Conclusion

In this chapter we began by recognizing the need to revIew the

phonemic inventory of Dzongkha; we reviewed the phonemic inventories of
four linguists: Mazaudon and Michailovsky (1988), Watters (2002), Van
Driem (1992) and Mazaudon (1985). We have shown that Dzongkha has 83
phonemes (including vowel length and nasalization): 48 consonants across five
places (labial, dental, retroflex, palatal and velar) with four phonatory
distinctions (VLT, VLL VLAS and VD) for obstruents (except fricatives); 10
monophthongs (with length distinction on 9 vowels and nasalization and length
distinction on four) and 7 diphthongs.

In section 2.2, we have compared the consonant phonemes of the earlier

researchers and proposed a consolidated consonant inventory in section 2.3.
The significant alteration we brought about in this section is the reclassification
of earlier devoiced series of consonants as a new feature Voiceless Lax. This
makes the claim that in fact, Dzongkha has three voiceless stops and affricates:

" Ao (Tibeto-Burman) spoken by Ao tribe in North-East India

,10 Zhongu (Tibeto-Burman) is a variety of Tibetan spoken in the Zhongu valley of Songpan

County in Northern Sichuan. a borderland between Tibetan and Qiang

,11 Kham Tibetan (Tibeto-Bunnan) is a variety of Tibetan spoken in Tibet (Sikang. Szechwan

and Yunnan provinces of China)

," Ryalthang Tibetan is spoken in Zhongdian County, Diqing Tibetan autonomous prefecture in
China

41
voiceless tense, voiceless lax and voiceless aspirated in addition to voiced, and
three types of fricatives: voiceless lax, voiceless tense and voiced.

In section 2.4, we compared the vowel inventories of the earlier

researchers, and proposed a consolidated inventory in section 2.5 through
section 2.7. We have also supported our claim with acoustic evidence. The
average F I and F2 formant values of the male speaker further maintains our ten
vowel contrasts. In addition to 7 diphthongs and 10 monophthongs we have
also demonstrated the length and nasalisation distinction in monophthongs.

Then, in section 2.8, we did a quick survey on the number of segments

in Dzongkha in comparison with other Tibeto-Burman languages, including
three other dialects of Tibetan, and found the segment size of Dzongkha fairly
usual by the standard of Tibetan dialects.

In the next chapter, we start with our phonation study of obstruents in

Dzongkha: a four-way phonation for stops.

42