SOUTHERN VIETNAMESE REDUPLICATION:
AN OPTIMALITY THEORY ANALYSIS

By

Nghi Lam

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Linguistics—Master of Arts

2025

ABSTRACT

This thesis offers a unified account of reduplication in Southern Vietnamese, concentrating

on two under-documented polysyllabic patterns — triplication and AABB — and on the

avoidance effects that block certain base–reduplicant combinations. Earlier studies describe

these facts, but either restrict attention to disyllables or cannot capture how tone, avoidance,

and reduplicant shape interact.

This study addresses these gaps through three main contributions. First, it provides

substantial new data on previously under-documented reduplication patterns that (1) en-

hance Vietnamese linguistic documentation and (2) provide a robust empirical foundation

for formal analysis. Second, it offers a description of the identical patterns avoided across

reduplication types, previously sparsely covered in the literature. Third, I offer a formal anal-

ysis of the polysyllabic reduplication types — triplication and AABB — using tools from

Optimality Theory (OT). Although the analysis draws on OT for formalization, it remains

framework-neutral and broadly applicable.

A key claim of this thesis is that the so-called triplication in Vietnamese results from the

attachment of two distinct reduplicative morphemes, each with its own allomorphs, and not

recursive reduplication. In this view, Vietnamese reduplication is best understood as the use

of lexically specified reduplicant morphemes. Specifically, morphology selects an allomorph

according to phonological properties of the base (notably tone); phonology then merely

enforces its placement. High-ranking alignment constraints and Max-IO constraints keep

the selected allomorph at its designated edge and preserve its lexical tone. Avoidance patterns

therefore emerge from the selectional restrictions imposed by lexically specified allomorphs,

thus avoiding theoretical issues such as ranking paradoxes.

Rather than invoking multiple constraint rankings, all attested patterns can be captured

within a single unified grammar in which lexically specified reduplicants select appropriate

stems. This allomorphy-based analysis provides a principled account of complex reduplication

patterns in Southern Vietnamese, addressing and resolving previous theoretical challenges.

Copyright by
NGHI LAM
2025

ACKNOWLEDGEMENTS

I am profoundly grateful to Professor Karthik Durvasula, my advisor, for his guidance,

encouragement, and insightful feedback on every chapter of this thesis. Whenever self-doubt

crept in, his reassurance kept me moving forward. I also thank my committee members,

Professor Yen-Hwei Lin and Professor Silvina Bongiovanni, whose constructive comments

and questions sharpened the analysis of Vietnamese triplication.

This work benefited from lively discussions with members of the Phonology Reading

Group and my peers in the Department of Linguistics. I am also grateful to the linguistics

faculty, whose courses have reminded me time and again how interesting linguistics is. Special

thanks to Pam Withers, Ben Lampe, and Melissa French for patiently handling my many

administrative questions and miscellaneous requests.

Last but not least, I owe deepest gratitude to my parents for their unwavering support

through every decisions that I made. I thank my boyfriend, Dang, for his love and constant

encouragement from miles away during the hardest stretches of this journey. I am equally

grateful for my friends in Vietnam, especially Phuong, Lu, Thu, Van, and Lam, whose humor

and support lift me up throughout the periods when everything feels impossible.

iv

TABLE OF CONTENTS

CHAPTER 1

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 2

BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Tone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Final consonants
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Syllable structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 Word structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6 Word stress
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interim summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7

CHAPTER 3

DATA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1 Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Exclamation (AABB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Pejorative ([-i@k]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4
Intensification (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Intensification (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5
3.6 Lexicalized reduplicatives . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interim summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.7

CHAPTER 4

ANALYSIS REVIEW . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 P.-P. Nguyen (1998)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Hoang et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Vu (2007)
4.4 Chengdu Mandarin AABB and tone sandhi (Lin, 2015) . . . . . . . . . .
4.5 Taiwanese (Zhang & Lai, 2007; Hsiao, 2018) . . . . . . . . . . . . . . . .
Interim summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6

CHAPTER 5

ANALYSIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 Correspondence Theory . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Prosodic hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Phonotactics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Pejorative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6
Intensification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7 Exclamation (AABB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.8 Triplication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interim summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.9

CHAPTER 6

CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . .

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

APPENDIX A

GENERAL STRUCTURE OF REDUPLICATION . . . . . . . .

1

4
4
7
7
9
9
11
12

13
13
16
19
25
27
31
33

34
34
36
37
39
46
51

53
53
55
55
57
61
65
67
71
82

84

86

92

v

APPENDIX B

ADDITIONAL DATA . . . . . . . . . . . . . . . . . . . . . . . .

96

vi

CHAPTER 1

INTRODUCTION

Reduplication is a productive morphophonological method in Vietnamese which helps form

a large amount of new words (Hoang et al., 2008). Consequently, it is one of the most exten-

sively studied topics in Vietnamese linguistics (Hoang, 1985; Thompson and O’Harrow, 1988;

Hoang, 1998; Cao, 2007; Vu, 2007; Hoang et al., 2008; inter alia). Most existing work consists

of descriptive analyses of reduplication (Ngo, 1984; Hoang, 1985; Thompson & O’Harrow,

1988; D.-H. Nguyen, 1997; Srichampa, 2002). Other studies cover experiments and acoustic

analyses of Vietnamese reduplication (T. Nguyen & Ingram, 2006; T. Nguyen & Ingram,

2007; A.-H. Pham & Pham, 2020). Formal analysis of Vietnamese reduplication, however, is

limited (Agbayani, 1997; Inkelas & Zoll, 2005; Vu, 2007).

Among existing studies, disyllabic reduplication predominates in the literature, as it

constitutes the majority of reduplicative words in Vietnamese. It embodies the key phonetic

and semantic characteristics of Vietnamese reduplication, such as tone harmony and the

encoding nuances of intensification and attenuation (Diep & Hoang, 1996). However, this

has led to a research gap between disyllabic reduplication and other reduplication forms,

such as triplication and AABB, even though they are also productive and have patterns

that are similar to those of disyllabic reduplication.1 Of these polysyllabic forms, triplication

poses challenges in Vietnamese linguistics (P.-P. Nguyen, 1998; Hoang et al., 2008). However,

little formal work has been done to analyze these challenges under a theoretical framework.

An important recurring pattern across Vietnamese reduplication involves identity avoid-

ance, where speakers systematically avoid adjacent identical syllables. Despite its prevalence

in Vietnamese and many world languages (Kennard, 2004; An, 2012; Wivell et al., 2024

among others), identity avoidance has not been adequately formalized in previous studies on

Vietnamese reduplication, receiving only brief mention in Vu (2007).

This thesis makes three primary contributions. First, it provides substantial new empirical

1The notion of productive will be explained in Section 2.5.2.

1

data on previously under-documented reduplication patterns — triplication and AABB,

used in tandem with the data and example sentences drawn from prior work by native

speakers Ngo (1984), Hoang (1998), and Vu (2007). The novel data are obtained from my

fieldwork, where informants provided acceptability judgments on various reduplicative forms

and participated in an elicitation task in which informants were given a real word in the

language with no known occurrence of triplication or AABB. This new dataset not only

enhances the Vietnamese linguistic documentation but also provides the empirical foundation

necessary for testing theoretical claims about reduplication processes in this thesis. Any

instances of such novel data will be explicitly marked with (Novel data) throughout the

thesis.

Second, this thesis proposes a formalization of Vietnamese reduplication patterns within

Optimality Theory (OT), introducing a novel theoretical perspective: Vietnamese triplica-

tion is shown to result from the attachment of two distinct reduplicative morphemes, each

with its own set of allomorphs, rather than repeated reduplication as previously suggested

in the literature. Morphology chooses which allomorph surfaces, triggered by phonological

properties of the base (notably tone), while phonology only enforces that choice through

high-ranking Align and Max-IO that keep the allomorph at its stored edge and preserve

its lexical tone.

Third, identity avoidance effects are extensively described, and are argued to result from

ordinary selectional restrictions: an allomorph that would duplicate the base’s rhyme is

simply absent from the lexicon. No special anti-identity constraint, and no ranking paradox,

is required. In other words, certain reduplicant allomorphs simply do not select stems of

particular phonological shapes, and this lexical restriction superficially resembles identity

avoidance. This approach removes the necessity of positing explicit phonological constraints

specifically designed to avoid identity.

Existing analyses rely on cophonologies (Orgun, 1996; Inkelas et al., 1997; Anttila, 2002;

Inkelas & Zoll, 2007) and Morphological Doubling Theory, both of which assume different

2

reduplication types have separate phonological grammars. In contrast, I propose a unified

phonology where Vietnamese reduplication processes are best understood through lexically

specified reduplicant morphemes and allomorph selection. Note that while the analysis is

presented within the OT framework, it does not depend exclusively on it; rather, OT serves

as a formalization tool for my proposal.

The rest of this thesis is organized into five chapters. Chapter 2 covers the relevant mor-

phological and phonological aspects of Vietnamese linguistics. Descriptions of Vietnamese

reduplication processes are given in Chapter 3. Chapter 4 addresses some existing approaches

to analyses of triplication and AABB in Vietnamese and cross-linguistically. Chapter 5 goes

through formal analyses provided for each reduplication process in Vietnamese. The thesis

concludes with Chapter 6.

3

CHAPTER 2

BACKGROUND

This chapter discusses aspects of Vietnamese phonology and morphology relevant to the

current analysis. Since the author and informants are Vietnamese speakers of the South-

ern variety, background information (except the syllable structure, word stress, and word

structure) and phonetic transcriptions in the subsequent sections are of the Southern vari-

ety as well, unless mentioned otherwise. This is why the title explicitly refers to Southern

Vietnamese reduplication.

This section is organized as follows: Section 2.1 covers the tone system of Southern

Vietnamese, the final consonant and vowel inventories are briefly introduced in Section 2.2

and Section 2.3 respectively, Section 2.4 covers the syllable structure, Section 2.5 outlines

different word structures in Vietnamese, and Section 2.6 closes out this section with a brief

discussion on word stress in Vietnamese.

2.1 Tone

There are five tones in Southern Vietnamese.1 The following description is from Brunelle

(2009) accompanied by a graph in Figure 2.1. A1 is a level tone that starts around mid F0

range and has no creakiness nor breathiness throughout its production.2 A2 is also a level

tone, but it starts lower than A1 and is produced with a falling pitch. B1 is a rising tone

but, similar to A1, has no creakiness nor breathiness throughout its production. B2 is a low

falling-rising tone, and unlike its Northern counterpart, there is no final glottalization. C1 is

a falling-rising tone with no laryngealization, similar to its Northern counterpart. Note that,

unlike the Northern variety, there is no C2 in Southern Vietnamese since this is merged with

C1.3

1There are different arguments for whether Vietnamese has six or eight tones, see A.-H. Pham (2001)

and A. H. Pham (2004) for the discussion. For this analysis, the six-tone system is adopted.

2Alphanumerical labels for tones are adapted from Michaud (2005).
3See Brunelle (2009) for a detailed discussion of the production and perception of C1 and C2 in Northern

and Southern Vietnamese.

4

Figure 2.1 Average F0 contours of Northern and
Southern Vietnamese tones (Brunelle, 2009)

A summary of the tone system is presented in Table 2.1. The organization is from Cao

(2003); the contour description is adopted from A.-H. Pham (2001)’s formal structure of

Vietnamese tones. Tones are organized in terms of their contours and registers. Register

here refers to the vocal range where the tone is produced (Vo, 1997). There are two registers

— high (+H) and low (−H). This means that tones A1, B1, and C1/C2 are produced at a

higher vocal range. Tones A2 and B2 on the other hand are produced at a lower vocal range.

Contour refers to the tonal shape; there are two general patterns — flat and non-flat (Vo,

1997; A.-H. Pham, 2001). Flat refers to level tones, with their overall shape not changing

much throughout the tone. Non-flat, on the other hand, refers to tones that change their

shape over the production time.

Contour

Flat Non-flat

Register

+H A1 B1 C1

−H A2

B2

Table 2.1 Southern Vietnamese tones (A.-H. Pham, 2001; Cao, 2003)

5

Following A.-H. Pham (2001), the tone contour can be captured through the use of what

the author refers to as points, denoted by a dot •. These points indicate whether there is

movement during the tone’s time course. A single point indicates no movement during the

tone. If there is movement during the tone, more than one points are used to indicate the

contour. The specification for tone contour of each tone is presented in (1).

The model in (1) has some of the following indications. During the time course of flat

tone (i.e., A tone), there is no movement, hence one point under the C(ontour) node. There

is movement in the non-flat tones (i.e., B and C tones); hence, there are two to three points

under the C node. The difference between the number of points in the C node for either non-

flat tone is irrelevant in the case of reduplication and thus will not be discussed in further

detail.4

(1) Specification for tone contour
A1/A2

C

•

B1/B2

C

•

•

C1/C2

C

•

•

•

Based on their structural complexity and occurrence in Vietnamese, tone markedness is

identified by register and contour (Vo, 1997; A.-H. Pham, 2001; A. H. Pham, 2004; Hoang

et al., 2008). In terms of register, low tones are more marked than high tones. In terms of

contour, the tone markedness is evaluated in pairs between flat and non-flat tones. C1-C2

are more marked than B1-B2 tones, which are more marked than A1-A2 tones. Thus, A1 is

the most unmarked tone, while C1/C2 is the most marked tone.5. The markedness of tones

is summarized in (2)–(3), where ‘<’ indicates the left-hand tone is less marked than the

right-hand one.

(2) Tone markedness in terms of register:

(3) Tone markedness in terms of contour:

A1, B1, C2 < A2, B2, (C2)

A1-A2 < B1-B2 < C1(-C2)

4See A.-H. Pham (2001) for a more in-depth analysis of Vietnamese tone.
5Under an eight-tone system, in terms of contour, B1-B2 in checked syllables (i.e., D1 and D2 ) are the

most marked; consequently, D2 is the most marked in Vietnamese (A.-H. Pham, 2001).

6

2.1.1 Tone-coda co-occurrence restriction

In East and Southeast Asian languages, there is a prevalent constraint on possible tone

and coda combinations (Hyman, 2014). In the case of Vietnamese, this can be observed when

the syllable has an obstruent coda. Specifically, B1 and B2 tones are allowed to appear in

syllables with obstruent codas /p, t, k/ (e.g., (4a)) but A and C tones are not (e.g., (4b)).

(4)

a. thích [th1tB1] ‘to like’

b.

* hãc [ha:kC2]

2.2 Final consonants

In Vietnamese, only six consonants from the whole Vietnamese consonant inventory are

allowed as coda. Note that /w/ is not usually counted as a coda consonant because it is

considered to be an off-glide in diphthongs. The permissible codas are given in the following

Table 2.2.

Place

Labial Alveolar Palatal Velar

t

n

p

m

w

k

N

j

Plosive

Manner

Nasal

Approximant
(Semivowels)

Table 2.2 Vietnamese codas

Other permissible codas are the labial-velars [

>
kp, >Nm]. These are doubly articulated con-

sonants that only occur after rounded vowels in Vietnamese, hence are allophones of /k, N/

when following rounded vowels.

2.3 Vowels

There are eleven monophthongs in Vietnamese, which are presented in Table 2.3 below.

Note that /i/ has an allophone [1] that occurs with alveolar codas.

7

Frontness

Front Central

Back

Height

High

Mid

Low

/i/

/e/

/E/

/W/, /u/

/7/, /7:/

/a/, /a:/

/o/

/O/

Table 2.3 Vietnamese monophthongs (Thompson & O’Harrow, 1988)

Vowel (and consonant) features are given based on the feature geometric model proposed

by Clements and Hume (1995) to account for the place specifications of consonants and

vowels. The revised model of the V-place node from Parker (2018) is used instead of the

Aperture node to distinguish the difference between central and back vowels. The revised

model is given in Figure 2.2 below.

Oral Cavity

C-place

Vocalic

V-place

Aperture

[Labial]

[Coronal]

[Dorsal]

[open]

[± back]

Figure 2.2 Feature geometric model (Clements & Hume, 1995)
and Revised model of the V-place (Parker, 2018)

Labial, coronal, and dorsal features are distinctive, in which they are not specified for

a value {+, −, 0}. Front vowels /i, e, E/ are defined as [coronal]. Central and back vowels

are both specified as [dorsal]. To distinguish between central and back vowels, central vowels

are defined as [−back], while back vowels are defined as [+back]. Among the back vowels,

the rounded ones /u, o, O/ are, in addition, specified with [labial]. For the analysis at hand,

however, I will employ the binary feature nodes [±high] and [±low] instead of [±open1, 2, 3] of

the Aperture node to represent the vowel height. Thus, the high vowels will have the height

nodes [+high, −low]; mid vowels will have [−high, −low], low vowels will have [−high, +low].

8

2.4 Syllable structure

The syllable structure in Vietnamese is given in Figure 2.3. In this structure, onset (C1)

and nucleus (V) are obligatory; on-glide (G1), coda (C2), and off-glide (G2) are optional.

Tone is realized on the rhyme (A.-H. Pham, 2001). There are arguments around the status

of the on-glide, but I’ll assume that the on-glide is a segment of its own and is a part of the

rhyme (Cu et al., 1977; V.-L. Nguyen & Edmondson, 1997; A.-H. Pham, 2009).

Onset

Rhyme

Glide6 Nucleus

Coda

C1

G1

V

C2/G2

7

Figure 2.3 Vietnamese syllable structure

2.5 Word structure

The organization of different word structures here mainly follows Ngo (1984), D.-H.

Nguyen (1997), and Vu (2007). The smallest meaningful constituent possible is a simple

word.8 It is a word that contains only one morpheme that has a lexical or grammatical

meaning, which can be divided further into monosyllabic (e.g., (5a)) and polysyllabic (e.g.,

(5b)) (D.-H. Nguyen, 1997).

(5)

a. Monosyllabic word: [ña:A2] ‘house(s)’, [kW@C1] ‘door(s)’, . . .

b. Polysyllabic word: [áuA2.ñ1nA2] ‘scarecrow’, [lu >NmA1.l1nA2] ‘sparkling’, . . .

Any words that are made up of two or more morphemes are complex words, which are

divided into two groups — compounds (Section 2.5.1) and reduplicatives (Section 2.5.2).

2.5.1 Compounds

A compound is composed of two simple words and can be divided into two types —

coordinate and subordinate (D.-H. Nguyen, 1997). Coordinate compound is composed of two

6Also called on-glide to differentiate from off-glide which occurs in coda, or pre-vocalic /w/ since only

/w/ occurs here.

7Also called off-glide.
8Any labels used here for different word structures are from D.-H. Nguyen (1997).

9

words and has a broader meaning than that of its components. For instance, the word for

clothes in Vietnamese (6a) is composed of two simple words: one for shirts and one for pants.

Subordinate compound is composed of two words as well, but its meaning is narrower than

that of its components, specifically, the head of the compound (the leftmost syllable). For

instance, the word for bicycle in Vietnamese (6b) is composed of two simple words: one for

vehicle and one for pedal ; the composition of these two words results in a compound with a

meaning that is narrower than that of the head of the compound, i.e., the word for vehicle.

(6)

a.

[Pa:wB1] ‘shirt(s)’ + [kw7NA2] ‘pants’ = [Pa:wB1.kw7NA2] ‘clothes’

b.

[sEA1] ‘vehicle’ + [âa:pB2] ‘pedal’ = [sEA1.âa:pB2] ‘bicycle’

2.5.2 Reduplicatives

Reduplicatives in Vietnamese can be split into two groups: productive reduplicatives (redu-

plicatives, henceforth) and lexicalized reduplicatives (P.-P. Nguyen, 1998; Vu, 2007; A.-H.

Pham & Pham, 2020). Productive reduplicatives are outputs of a reduplication process, and

thus the main data of the analysis. On the other hand, lexicalized reduplicatives are not

results of a reduplication process. They are listed in dictionaries, while productive ones are

not because there is a large amount of them in the language and their constructions can be

subsumed under a set of formation rules (P.-P. Nguyen, 1998).

A reduplicative has two components — a base, which carries the core meaning, and a redu-

plicant, which is the copy of the base that is affixed to it. To be specific, reduplicant here the

actual phonological projection of a phonologically-unspecified lexical entry Red (McCarthy

& Prince, 1994). An example is given below in (7). In this example, the reduplicant Red

precedes the base. More reduplicative data will be presented in Chapter 3.

(7)

[c7mA2
Red
‘Slightly slow’

c7mB2]
slow

Note that in the case of lexicalized reduplicatives, there are forms that can be broken

down to obtain its base (and hence its structure is somewhat similar to that of a subordinate

10

compound) (e.g., (8a); the supposed base is boldface); however, there are other cases in which

they can’t be broken down any further (and hence similar to polysyllabic simple word) (e.g.,

(8b)). More examples of this reduplication form and explanations are offered in Section 3.6

and Appendix B.

(8)

a.

[lu@NA1.lu@NA1]
‘always’

b.

[âuA1.âuC1]
‘papaya’

Additionally, reduplicatives are not counted as compounds because if we strictly follow the

definition for compounds in this case, they are composed of simple words, while reduplicatives

are not made up of simple words, but rather of morphemes (which happens to have simple

words in the mix).

2.6 Word stress

In Vietnamese, there’s little evidence for word stress. There is past literature that de-

scribe word stress in Vietnamese as final stress (i.e., word final); however, this can also be

accounted as phrase-final lengthening and not word stress per se (Brunelle, 2017). However,

experimental work on word stress in reduplication pointed out that the second syllable in

disyllabic reduplication is more acoustically prominent (T. Nguyen & Ingram, 2006). There’s

also no acoustic evidence to support the claim of contrastive stress patterns between com-

pounds and phrases. Thus, under forced elicitation condition, there’s a juncture between

constituents in phrases, but none in compounds (T. Nguyen & Ingram, 2007). As a conse-

quence, each constituent in phrases bears stress, and constituents are produced continuously

with the rightmost syllable stressed (T. Nguyen & Ingram, 2007; A.-H. Pham, 2009). For

this analysis, I will assume that in Vietnamese, the rightmost syllable in the word will get

the prominence stress, following T. Nguyen and Ingram (2007), Vu (2007), and A.-H. Pham

(2009).

11

2.7

Interim summary

In this chapter, the descriptive groundwork was laid out for analyzing and understanding

Vietnamese reduplication. The section started with examining the tonal system, coda con-

sonants, and vowels, which are central to understanding the primary alternations between

syllables in reduplicatives. Next, syllable structure was discussed to pinpoint where such

alternations would occur. I discussed the word structure to clarify how the type of base in-

fluences and is influenced by the reduplication process. The chapter closes with word stress

in Vietnamese to show its role in distinguishing the base from the reduplicant in a prosodic

structure of the reduplicative. With these foundational elements in place, the next chapter

will present and describe the reduplication data.

12

CHAPTER 3

DATA

This study focuses on productive reduplication processes. Thus, the following sections present

data from reduplicatives derived from clearly replicable processes, including nonce forms. As

mentioned in the introduction, all of the data, except for novel words from my informants,

is from Ngo (1984), Hoang (1998), Vu (2007), and myself. Since the transcriptions were

given for Standard (Northern) Vietnamese, I re-stated and re-described the data here with

Southern transcriptions instead to align with the variety of Vietnamese used here.

Instead of categorizing reduplication forms by the number of syllables or the shape, I

organized them by their morpho-semantic function: Section 3.1 for attenuation, Section 3.2

for exclamation, Section 3.3 for pejorative, Section 3.4 and Section 3.5 for intensification

of different base and reduplication form, and a brief discussion of lexicalized reduplication

forms in Section 3.6. To make the data presentation clear, the leftmost word in each example

or data point is the base, the rightmost word is the reduplicative.

3.1 Attenuation

The predominant function of disyllabic reduplication is attenuation. This function is only

applicable to adjectives and verbs.1 In disyllabic reduplication, the reduplicant’s form can

either be a full or partial copy of its base. One thing to note is that every base can undergo

both full and partial reduplication. However, the choice and frequency of each type depend

on speaker preference.2 In addition, attenuation reduplication limits to monosyllabic bases.

If a speaker wants to attenuate the meaning of an adjective or verb that has more than one

syllable (e.g., (1a) & (1b) respectively), the meaning has to be paraphrased.

(1)

a.

[lu >NmA1.l1nA2]
sparkling
‘sparkling’

̸→

b.

[NiA1
suspect
‘to suspect’

N7:A2]
suspect

̸→

1Note, however, that some sources refer to the meaning of verbal (full) reduplication as verbs with

iterative marking (Thompson & O’Harrow, 1988).

2See A.-H. Pham and Pham (2020) for SVN speakers’ full vs. partial reduplication preference.

13

In the full reduplication case, the reduplicant is an exact copy of the base (e.g., (2a)).3

In the partial reduplication case, the reduplicant copies all but the base’s tone contour (and

coda) (e.g., (2b) & (2c)).

(2)

a.

b.

c.

[ka:wA1]
tall
‘tall’

[sa:kB1]
close.to
‘close to’

[ñOC1]
small
‘small’

ka:wA1]
[ka:wA1
Red
tall
‘slightly tall’

/ *[ka:wA2.ka:wA1]

/ *[sa:kA1.sa:kB1]

[sa:NA1
Red
‘slightly close to’

sa:kB1]
close.to

ñOC1]
small

[ñOA1
Red
‘slightly small’

/ [ñOC1
Red

ñOC1]
small

However, unlike adjectives, verbs can only take full reduplication. Consider the data given

in (3). Partial reduplication in (3a) being unacceptable is expected because the base has a

flat tone (i.e., A1), meaning regardless of the form, the reduplicant’s tone is always a full

copy of that of the base. The same partial form in (3b) is unacceptable as well, even though

it obeys the tone alternation for partial reduplication and the tone-coda restriction.

(3)

a.

[NiA1]
suspect
‘to suspect’

/ *[NiA2.NiA1]

NiA1]
[NiA1
Red
suspect
‘to suspect(att)’4

b.

[th1tB1]
like
‘to like’

th1tB1]
like

[th1tB1
Red
‘to like(att)’

/ *[th1nA1.th1tB1]

The degree of the verb is attenuated in the sense that the verb is less direct or blunt

(e.g., (4a) with (3a) verb, where the base form appears in the sentence, compared to the

corresponding (4b), where the reduplicative is used; base and reduplicant are in boldface).

Similarly, attenuation also conveys uncertainty (e.g., (5a) with (3b) verb, compared to the

(5b) reduplicative counterpart). However, not all verbs are compatible with attenuation, as

3In cases of full reduplication, the base and the reduplicant are identical in form, making it impossible

to systematically distinguish between them. Thus, gloss annotations are assigned arbitrarily.

4att = attenuation

14

it is unusual to attenuate a dynamic verb like eat in (6), where (a) contains the base form,

and (b) the reduplicative. Regardless, a discussion of the syntactic, semantic, and pragmatic

properties of different verb types is beyond the scope of the current study.

(4)

(5)

(6)

a. ta:wA1
1sg
‘I’ve suspected him/her/it for a while.’

NiA1
suspect

rojA2
already

âa:C2
past

nOB1
3sg

b. ta:wA1
1sg
‘I’ve kinda suspected him/her/it for a while.’

NiA1
suspect

rojA2
already

âa:C2
past

nOB1
3sg

NiA1
Red

th1tB1
nOB1
a. ta:wA1
1sg
3sg
like
‘I like him/her/it.’

PaNA1
a. Pa:NA1
An
eat
‘An has a meal.’

k7mA1
rice

b. ta:wA1
1sg
‘I kinda like him/her/it.’

th1tB1
like

th1tB1
Red

nOB1
3sg

b. # Pa:NA1
An
‘An kinda has a meal.’

PaNA1
Red

PaNA1
eat

k7mA1
rice

3.1.1 Tone pattern

The tone harmony pattern in partial reduplication is same-register tone harmony, in

which the reduplicant’s tone contour is always A with its register copied from the base. One

explanation for why it is an A tone is because of tone neutralization; the reduplicant might

have been a total copy of the base, but the tone is neutralized to an unmarked tone (i.e.,

a flat (A) tone) (Hoang et al., 2008). Due to the tone-coda co-occurrence restriction, if the

base has an obstruent coda (e.g., (2b)), the reduplicant will not take the base’s coda but will

have the homorganic nasal instead.

Another possible explanation, that is of interest of the analysis in Chapter 5, is fixed

flat tone in the lexical representation of the reduplicant. This means that the reduplication

morpheme has a fixed flat tone, with the function of attenuation in the language.

A consequence of the fixed A tone contour and full copy of the base’s tone register is that

if the base has a flat tone (e.g., the unacceptable form in (2a)), the reduplicant is always

an exact copy of the base (cf. the acceptable and the unacceptable form in (2a)). These two

15

tone harmony patterns are considered to be the representative tone pattern in Vietnamese

reduplication as a whole (Diep & Hoang, 1996; Cao, 2007).

3.2 Exclamation (AABB)

In past literature, AABB (in this analysis, exclamation) reduplication has been described

to be used for repetition (for adjectives and verbs) (Srichampa, 2002) and emphatic (Vu,

2007). (7) shows examples of how the base form and its reduplication for exclamation are in

sentences.5 Compared to (7a), sentence (7b) contains the reduplicative. This form emphasizes

the large amount of clothes piling up. Consequently, the sentence as a whole functions as an

exclamation..

Unlike attenuation and pejorative reduplication, exclamation can only take disyllabic

inputs. If we have a monosyllabic base (e.g., [Pa:wB1] ‘shirt(s)’) like (7c), we cannot have a

sentence like (7d) that has the reduplication of [Pa:wB1] ‘shirt(s)’. Instead, we can re-phrase

the sentence like (7e), which is the same as (7c) with an additional question particle (q)

[v7jB2], changing the sentence to a rhetorical question that expresses exclamation.

(7)

a. úO >NmA1
in
‘In A’s room, clothes pile up like a mountain.’

kw7NA2
pants

fO >NmA2
room

Pa:wB1
shirt

c7kB1
stack

A
A

thanA2
into

nujB1
mountain

b. úO >NmA1
in
‘In A’s room, clothes pile up like a mountain. (emph)’6

kw7NA2
pants

fO >NmA2
room

Pa:wB1
shirt

Pa:wB1
shirt

kw7NA2
pants

A
A

c7kB1
stack

thanA2
into

nujB1
mountain

c. úO >NmA1
Pa:wB1
in
shirt
‘In A’s room, shirts pile up like a mountain.’

fO >NmA2
room

thanA2
into

c7kB1
stack

A
A

nujB1
mountain

d.

* úO >NmA1

fO >NmA2
room

A
A

Pa:wB1
shirt

Pa:wB1
shirt

c7kB1
stack

thanA2
into

nujB1
mountain

in
‘In A’s room, clothes pile up like a mountain. (emph)’

e. úO >NmA1
Pa:wB1
in
shirt
‘Why are there so many shirts in A’s room?’

fO >NmA2
room

thanA2
into

c7kB1
stack

A
A

nujB1
mountain

v7jB2
q

5(7) is novel data.
6emph = emphasis

16

Similar to pejorative reduplication, this form applies to items across all lexical categories.

For adjectives, there is an alternation between this reduplication and the [-a:] reduplication

to express emphasis or intensification of the adjective’s degree.

A common feature among all lexical categories is that the input must take the form of a

coordinate compound (e.g., (8a), (9a), and (10b)). For adjectives, lexicalized reduplicatives

are also possible inputs (e.g., (10a)). Additional novel data (see Appendix B) further confirms

that lexicalized reduplicatives can serve as valid inputs across categories (e.g., (8b), (9b), and

(10c)).

By contrast, inputs with other structures, such as subordinate compounds, reduplicatives,

and mono- or polysyllabic words, are not possible. This includes nominal and adjectival

subordinate compounds (e.g., (8d), (10d)), nominal and verbal lexicalized reduplicatives (e.g.,

(8d), (9c)), nominal polysyllabic words (e.g., (8b)), verb phrases (e.g., (9d)), and adjectival

productive reduplicatives (e.g., (10d)).

(8) Nouns:

a.

b.

[Pa:wB1
shirt
‘clothes’

kw7NA2]
pant

[Pa:wB1
shirt
‘clothes (emph)’

Pa:wB1
shirt

kw7NA2
pant

kw7NA2]
pant

Pi@kB1]
[Pa:wB1
shirt
Ri@k
‘shirts (pej)’

Pi@kB1
Pa:wB1
[Pa:wB1
shirt
Ri@k
shirt
‘shirts (pej, emph)’

Pi@kB1]
Ri@k

(Novel data)

c.

[ka:A2.fEA1]
‘coffee’

* [ka:A2.ka:A2.fEA1.fEA1]
Intended : ‘coffee (pej)’

d.

[sEA1.âa:pB2]
‘bicycle’

* [sEA1.sEA1.âa:pB2.âa:pB2]
Intended : ‘bicycle (pej)’

e.

[cu@NA2.cu@NA2]
‘dragonfly’

* [cu@NA2.cu@NA2.cu@NA2.cu@NA2]

Intended : ‘dragonfly (pej)’

17

(9) Verbs:

a.

b.

[kW@jA2
laugh
‘to laugh and talk’

nOjB1]
talk

nOjB1]
[kW@jA2
laugh
talk
‘to laugh and talk (emph)’

kW@jA2
laugh

nOjB1
talk

[PaNA1
Pi@kB1]
Ri@k
eat
‘eat (pej)’

[PaNA1
PaNA1
Pi@kB1
Ri@k
eat
eat
‘eat (pej, emph)’

Pi@kB1]
Ri@k

(Novel data)

c.

[NuC1.NeA1]
‘sleep’

* [NuC1.NuC1.NeA1.NeA1]

Intended : ‘sleep (pej, emph)’

d.

[la:mA2.áa:jA2]
‘to study’

* [la:mA2.la:mA2.áa:jA2.áa:jA2]

Intended : ‘to study (pej, emph)’

(10) Adjectives:

a.

b.

c.

[m7A1.ma:NA2]
dreamy
‘dreamy’

âEpB2]
pretty

[s1nA1
pretty
‘pretty’

[sopB1
si@kB1]
fluffy
Ri@k
‘fluffy (pej)’

m7A1
dream

[m7A1
dream
‘dreamy (emph)’

ma:NA2
-

ma:NA2]
-7

s1nA1
pretty

[s1nA1
pretty
‘pretty (emph)’

âEpB2
pretty

âEpB2]
pretty

sopB1
[sopB1
fluffy
fluffy
‘fluffy (pej, emph)’

si@kB1
Ri@k

si@kB1]
Ri@k

(Novel data)

d.

[va:NA2.va:NA2]
‘slightly yellow’

* [va:NA2.va:NA2.va:NA2.va:NA2]

Intended : ‘slightly yellow (emph)’

e.

[sanA1.la:B1]
‘green’

* [sanA1.sanA1.la:B1.la:B1]
Intended : ‘green (emph)’

7[ma:NA2] doesn’t have meaning in Vietnamese, but it’s an [-a:NA] reduplicant of an Adj, an unproductive

reduplication process

18

Similar to pejorative reduplication, this reduplication form also exhibits separability ef-

fect, shown in (11).8 The phrase [v@jB1.ca:C1] ‘also neg’ can be used as the “separator”.

Same conclusion can be made from the separability effect for exclamation form, that is the

base and Red in exclamation reduplication belongs to separate prosodic structures, i.e., each

is a prosodic word in the same phonological phrase.

(11) Pa:wB1

Pa:wB1
shirt

v@jB1
with

ca:C1
neg

kw7NA2
pants

kw7NA2
pants

mu@A1
buy

jiA2
what

ma:A2
then

ñi@wA2
many

shirt
v7jB2
q
‘You bought so many clothes, and for what? (pej, emph)’

The general form of the reduplicative is AABB, where each syllable of the input AB is

reduplicated. Because the copies of A and B are segmentally identical to A and B themselves,

it is difficult to determine which instance is the Red. There is no empirical evidence at this

point to definitively identify which element is the reduplicant in AABB forms, since both

copies are segmentally identical. I have chosen to analyze the leftmost syllable of each pair

as the reduplicant, making them prefixes, primarily for consistency with other reduplication

processes in Vietnamese, particularly the attenuation forms, which are generally analyzed as

prefixal. This assumption allows for a more uniform treatment across different reduplication

types in the language.

Exclamation reduplication bears a resemblance to the AABB reduplication found in

Chengdu (Lin, 2015) and Taiwanese (Hsiao, 2018), particularly regarding its form and the

way it intensifies meaning. For this reason, I will extend their analyses of alignment in AABB

reduplication to the Vietnamese data.

3.3 Pejorative ([-i@k])

Pejorative reduplication is highly productive; this form can be applied to any lexical

item, regardless of structures or number of syllables, in Vietnamese. Unlike attenuation, in

pejorative reduplication, the reduplicant always follows the base. For example, in (12a), the

8(11) is novel data.

19

reduplicant [si@kB1] is preceded by its base [sEA1] ‘vehicle’ and not the other way around

(i.e., *[si@kB1.sEA1]).

(12)

a.

[sEA1]
vehicle
‘vehicle(s)’

/ *[sEA1.si@kB2] / *[si@kB1.sEA1]

si@kB1]
Ri@k

[sEA1
vehicle
‘vehicles (pej)’9

b.

c.

[v7:C1]
notebook
‘notebook(s)’

[v7:C1
notebook
‘notebooks (pej)’

vi@kB1]
Ri@k

/ *[v7:C1.vi@kB2]

[áa:NB2]
friend
‘friend(s)’

[áa:NB2 ái@kB2] / *[ái@kB1/B2.áa:NB2]

[áa:NB2
friend
‘friends (pej)’

ái@kB1]
Ri@k

The reduplicant’s rhyme is always [-i@k]; it never copies the base’s rhyme. If the base has

two syllables or more (e.g., (13a)–(g)), the number of syllables in Ri@k has to be the same as

that of the base.10 The rightmost syllable then will be the one that has the [-i@k] rhyme; any

syllables preceding Ri@k in the reduplicative will copy their correspondents in the base. For

instance, in (13a), given the base [ka:A2.fEA1] ‘coffee’, Ri@k copies both syllables but changes

the rhyme of the rightmost syllable to [-i@k]. Hence, we have the reduplicant [ka:A2.fi@kB1].

(13)

a.

[ka:A2.fEA1]
coffee
‘coffee(s)’

[ka:A2.fEA1
coffee
‘coffees (pej)’

ka:A2.fi@kB1 / *B2]
Ri@k

b.

* ki@kB2 fEA1

d.

* ki@kB2 fEA1 ka:A2 fEA1

(number of σ’s in the reduplicative

(Ri@k in the first syllable)

̸= number of σ’s in the base)

e.

* ka:A2 fEA1 ki@kB2 fEA1

c.

* ka:A2 fi@kB1 ka:A2 fEA1

(Ri@k in the third syllable)11

(Ri@k in the second syllable)

9pej = pejorative
10Ri@k = reduplicant in pejorative process
11These violations are the same inputs of any lexical categories.

20

f.

[Pa:wB1
shirt
‘clothes’

kw7NA2]
pant

kw7NA2
[Pa:wB1
shirt
pant
‘clothes (pej)’

Pa:wB1
shirt

ki@kB1/B2]
Ri@k

g.

[k7wA1.lakB2.áoB2]
club
‘club(s)’

[k7wA1.lakB2.áoB2
club
k7wA1.lakB2.ái@kB2/B1]
Ri@k
‘clubs (pej)’

(Novel data)

According to Vu (2007), pejorative reduplication is only possible for bases with at most

two syllables. This is false, as shown by cases like (13g) in which the base has three syllables

and the process is still applicable, as long as characteristics of pejorative reduplication (fixed

rhyme, tone contour, etc.) are obeyed.

One characteristic of pejorative reduplication that will be crucial for the analysis of the

prosodic structure of the reduplicative in the complete work is the syntactic separability effect

(Vu, 2007). This means that additional syntactic constituents can separate the base from

Ri@k (e.g., (14)–(15), where the base and reduplicant are in boldface; for more examples,

see Appendix B).12 Specifically, in (14), (a) contains the base, while (b)–(c) contain the

reduplicative. In (15), (a)–(b) contain the base, while (c)–(f) contain the reduplicative. This

effect also occurs in coordinate compounds and lexicalized reduplicatives. The fact that the

base and Ri@k can be separated suggests that each belongs to a separate prosodic structure,

i.e., each is a prosodic word within the same phonological phrase (Vu, 2007).

(14) Monosyllabic base: Separating constituent (sep) = [xo >NmA1] ‘no’

xo >NmA1
neg

ciwB2
a. EmA1
2sg
agree
‘You never sleep or anything at all!’

NuC1
sleep

jiA2
what

h7tB1
all

xo >NmA1
neg

b. EmA1
2sg
‘You never sleep or anything at all! (pej)’

Ni@kB1
Ri@k

NuC1
sleep

jiA2
what

h7tB1
all

12Syntactic structures of the constituents used as the “separator” are left open for future research as they

are beyond the scope of this analysis.

21

c. X sep X

xo >NmA1
neg

EmA1
2sg
‘You never sleep or anything at all! (pej)’

Ni@kB1
Ri@k

NuC1
sleep

xo >NmA1
neg

jiA2
what

h7tB1
all

(15) Polysyllabic base (coordinate compound): sep = [v@jB1.ca:C1] ‘also neg’

a. kW@jA2
laugh
‘What are you laughing and talking about?’

(ka:jB1)
(clf)

nOjB1
talk

âOB1
that

jiA2
what

b. X sep Y

nOjB1
kW@jA2
talk
laugh
‘What are you laughing and talking about? (pej)’

(ka:jB1)
(clf)

v@jB1
with

ca:C1
neg

âOB1
that

jiA2
what

c. kW@jA2
jiA2
laugh
what
‘What are you laughing and talking about? (pej)’

kW@jA2
laugh

ni@kB1
Ri@k

(ka:jB1)
(clf)

nOjB1
talk

âOB1
that

d. X Y sep X Y

ca:C1
kW@jA2
neg
laugh
‘What are you laughing and talking about? (pej)’

kW@jA2
laugh

ni@kB1
Ri@k

nOjB1
talk

v@jB1
with

(ka:jB1)
(clf)

jiA2
what

âOB1
that

e.

* X sep Y X Y

kW@jA2
laugh
Intended : ‘What are you laughing and talking about? (pej)’

kW@jA2
laugh

ni@kB1
Ri@k

(ka:jB1)
(clf)

nOjB1
talk

v@jB1
with

ca:C1
neg

jiA2
what

f.

* X Y X sep Y

kW@jA2
v@jB1
laugh
with
Intended : ‘What are you laughing and talking about? (pej)’

kW@jA2
laugh

ni@kB1
Ri@k

(ka:jB1)
(clf)

nOjB1
talk

ca:C1
neg

jiA2
what

âOB1
that

âOB1
that

3.3.1 Tone pattern

Since the reduplicant has a fixed [-i@k] rhyme and the tone-coda restriction (see Sec-

tion 2.1), there is only one possible tone for the reduplicant, that is B tones. As for the tone

register, it is said to be a same-register tone harmony for Southern (and Central) speakers,

unlike the Northern variety in which Ri@k’s tone register is always +H (i.e., the tone is always

22

B1). However, additional novel data (see Appendix B) shows that, for SVN speakers, there’s

an alternation between the use of B1 and B2 in Ri@k, i.e., both B1 and B2 are possible in

the reduplicant with any bases, and Ri@k does not have to copy the tone register of the base,

regardless of the number of syllables in the base (e.g., (12c) and (13f)). There is a preference

for B1 however, which would make it similar to the Northern variety’s.

However, there is no tone alternation when the tone register of the base is +H; Ri@k’s tone

register has to be +H too (e.g., (12a)–(b), and (13a)). The description given in Vu (2007)

is at least true for SVN when the base has two or more syllables; in such cases, B1–B2

alternation still exists, but there is a preference for both the base and Ri@k to have the same

tone register. One possible explanation is that since +H tones are unmarked, there is no need

for further alternation; any changes will result in a more marked tone. One problem with

this is that there are no attested processes in Vietnamese that would favor high tones over

non-high tones, and among reduplication processes themselves, there is no such a pattern.

For this study, I will provide an analysis for both B1 and B2 in monosyllabic bases, and tone

register copying in polysyllabic bases; the preference between B1 and B2 is a probable future

research topic as it is not within the scope of the current study.

3.3.2 Avoidance pattern

When the base’s rhyme is [-i@k], the process is blocked. In that case, the pejorative

meaning of the base can be expressed periphrastically or a replacement allomorph can be used

instead. For instance, in (16b)–(a), a replacement allomorph, which carries the pejorative

marking of the base, is used to express the pejorative meaning of the base.

(16)

a.

[si@kB1]
circus
‘circus(es)’

b.

[ti@kB2]
party
‘party(s)’

[si@kB1.su: >NmC1]
circus-pej
‘circuses (pej)’

[ti@kB2.tu: >NmA2]
party-pej
‘parties (pej)’

/ * [si@kB1

circus

si@kB1]
Ri@k

/ * [ti@kB2
party

ti@kB2]
Ri@k

/ *[ti@kB2.ti@kB1]

23

In cases where the base has the rhyme [-i@k] with tone B1, as in (16a), speakers reject

a form like *[si@kB1.si@kB1]. One plausible reason is homophony avoidance in which such a

string is identical to the obsolete productive full nominal reduplication pattern that means

‘every circus’. To avoid that, speakers resort to a periphrastic form shown in (16a).

The same story cannot explain [-i@kB2] bases (16b). Even if the reduplicant switched

to B1, *[ti@kB2.ti@kB1], the string is not homophonous with any existing word, yet it is

still rejected. What matters is the phonological (partial) identity of the rhyme, in which the

patterns avoided are not completely identical to each other (only the rhyme [-i@k] (and the

tone contour but the tone register). Vietnamese therefore instantiates a more general ban

on adjacent identical rhymes, rather than simple homophony avoidance.

Partial-identity avoidance of this sort is cross-linguistically common. French noun-forming

-iste /ist/ blocks /ist/ after a base that already ends in [-is] or [-iz] (e.g., /bodis/ “Baudis

(name)” + /-ist/ → [bodist], *[bodisist]; see Corbin and Plénat (1992) and de Lacy (1999)).

Nisgha shows a similar restriction: the third-person singular suffix -t /t/ cannot follow stems

ending in [s], [t], or [ì] (Russell, 1995; de Lacy, 1999). Moreover, the morphological identity

effects surveyed in Yip (1995) — English plural -s + possessive ’s and Chinese Mandarin

double -le (perfective marker -le + sentence-final particle conveying current state change

-le) — also block adjacent identical material, though those languages repair the violation

by deletion of either copy of the affix. Another way to consider this is that there is a null

allomorph in the lexicon that emerges when there are adjacent identical segments (i.e.,

English) or affixes (i.e., Chinese Mandarin).

By contrast, no repair is necessary in Vietnamese because there is simply no null allo-

morph in the lexicon, unlike cases in Chinese Mandarin and English.. The process is ineffable

when there are adjacent identical segments (i.e., [-i@k] base + [-i@k] reduplicant). Morphology

therefore selects a periphrastic alternative as the reduplicant, and phonology merely enforces

that choice. The avoidance effect is thus morphological, a matter of allomorph inventory, not

a phonological identity constraint. Treating it this way avoids the ranking paradoxes that

24

an explicit anti-identity constraint would create (see Section 5.5).

3.4

Intensification (1)

In Vietnamese, there are two reduplication processes that can be used to express inten-

sification of a polysyllabic verb or an adjective — one form involves a fixed rhyme of [-a:]

and the other form is triplication.13 This section discusses the first process.

[-a:] reduplication is only applicable to disyllabic verbal and adjectival reduplicatives.

Existing data shows that this form is observed with lexicalized reduplicatives (e.g., (17));

however, novel data shows that reduplicatives from productive processes, pejorative to be

specific, are also possible (e.g., (18a)–(c)). As a result, this process is possible for nouns as

well (e.g., (18a)).

(17)

a.

[kaNA2.ñaNA2]
‘complain’

[kaNA2.ña:A2.kaNA2.ñaNA2]
‘constantly complain’

b.

[l1nA1.t1nA1]
‘nonsensical’

[l1nA1.ta:A1.l1nA1.t1nA1]
‘nonsensical (emph)’

c.

[s7:NB1.sa:kB1]
‘ill-mannered’

[s7:NB1.sa:A1.s7:NB1.sa:kB1]
‘ill-mannered (emph)’

(18)

a.

ñi@kB1]
Ri@k

[ña:A2
house
‘houses (pej)’

b.

c.

[GEpB1
Gi@kB1]
couple
Ri@k
‘to couple (pej)’

[va:NA2
yellow
‘yellow (pej)’

vi@kB1]
Ri@k

ña:A1
Ra:

[ña:A2
house
‘houses (pej, emph)’

ña:A2
house

ñi@kB1]
Ri@k

Ga:A1
Ra:

[GEpB1
couple
‘to couple (pej, emph)’

GEpB1
couple

Gi@kB1]
Ri@k

va:NA2
va:A1
[va:NA2
yellow
Ra:
yellow
‘yellow (pej, emph)’

vi@kB1]
Ri@k

(Novel data)

(Novel data)

(Novel data)

13To better distinguish the two intensification processes, I will refer to the former process as [-a:] redu-

plication and the latter as triplication, utilizing their main segmental difference.

25

Similar to the pejorative process, this reduplication also involves a fixed rhyme and tone

contour, [-a:] and A tone contour, respectively. The tone harmony pattern is also of the same

register, similar to reduplication processes described above.

The general form is ABAB in which B is Ra: with fixed segments, and A is Red of A,

B for B.14 Thus, this form resembles attenuation and exclamation reduplication, in that the

reduplicant is a prefix.

This reduplication also exhibits syntactic separability effect, presented in (19).15 Here,

the reduplicative [kaNA2.ña:A2.kaNA2.ñaNA2] ‘constantly complaining’ is split by the word

[hajA1] ‘or’.

(19) xo >Nm
Neg
‘Stop complaining whatsoever!’

kaNA2.ña:A2
Red

hajA1
or

kaNA2.ñaNA2
complain

jiA2
what

h7tB1
all

3.4.1 Avoidance pattern

When the base’s rhyme is [-a:] and has an A tone, the process is blocked. Having either

the rhyme [-a:] or the flat tone alone does not block the reduplication, as shown by (20) with

the rhyme [-a:]. But when the two patterns simultaneously occur in the base, the process

is blocked, as shown in (21) (Vu, 2007). In such cases, the intensification meaning can be

expressed periphrastically or the speaker can use a replacement form instead. For instance,

given a base like [l7wA1.la:A1] ‘(of time) long’ in’ (21), the base (the rightmost syllable, to

be specific) has [-a:] as its rhyme and an A tone, so this reduplication process cannot be

applied; a replacement form is used instead. In this case, the replacement form is obtained

through the exclamation reduplication process (Vu, 2007).

(20)

[á7jB2.áa:B2]
‘inappropriate’

[á7jB2.áa:A2.á7jB2.áa:B2]
‘very inappropriate’

(Novel data)

14Since the reduplicant’s rhyme is always [-a:], for convenience, I will refer to the reduplicant in intensifi-

cation (1) as Ra:.

15(19) is novel data.

26

(21)

[l7wA1.la:A1]
long
‘(of time) long’

l7wA1
long

la:A1]
[l7wA1
Red
-16
‘(of time) long (emph)’

la:A1
Red

*

la:A1
Ra:

[l7wA1
l7wA1.la:A1]
long
long
‘Intended : ‘(of time) long (emph)’

The same allomorph-selection logic from Section 3.3 and Section 5.5 applies here, which

will be discussed in Section 5.6. This also parallels with English and Chinese Mandarin

identity avoidance (Yip, 1995).

3.5

Intensification (2)

Triplication is a productive process in Vietnamese, yet it is documented in a limited

amount of literature (Hoang et al., 2008). It has been described and generalized in only a

few studies (see Appendix B; cf. Hoang, 1985, 1998; P.-P. Nguyen, 1998; Srichampa, 2002;

Hoang et al., 2008). Moreover, triplication, with a similar intensifying function, can also

be found in various languages or dialects, such as Taiwanese (Zhang & Lai, 2007) and the

Chengdu variety of Mandarin (Lin, 2015).

Based on these available data (a sample list of data given in (22)), triplication in Viet-

namese applies to only monosyllabic adjectives. In the reduplicative, the first syllable is

assumed to be the base because the segments and the tones of the monosyllabic input are

preserved here (Hoang et al., 2008). This syllable is also said to carry the primary stress of

the word (Hoang et al., 2008); (P.-P. Nguyen, 1998) extended the stress carrier to include

the third syllable as well. The second syllable consistently bears an A2 tone and, if the

base ends in an obstruent coda, it also has a homorganic nasal coda, due to the tone-coda

co-occurrence constraint.

For the third syllable, there are two sub-patterns: if the base has an B1 tone and is a

checked syllable (i.e., one that has an obstruent coda), then the third syllable has B2 (cf.

(22a)–(b) and (22c)); otherwise, the reduplicant will have an A1 tone (and homorganic nasal

16Since the syllable does not have a meaning, a dash - is used in the gloss.

27

coda if there is a coda in the base). In general, the third syllable and the first syllable (i.e.,

the supposed base) harmonize in terms of either contour (i.e., flat versus non-flat tones)

tones or register (i.e., high versus low tones) (Hoang et al., 2008).

(22)

a.

[sa:kB1]
‘close to’

[sa:kB1.sa:NA2.sa:kB2]
‘very close to’

[s7wB1]
‘ugly’

[s7wB1.s7wA2.s7wA1]
‘very ugly’

c.

e.

b.

d.

[P1tB1]
‘little’

[P1tB1.P1nA2.P1tB2]
‘very little’17

[satB2]
‘clean’

[satB2.sanA2.sanA1]
‘very clean’

[tWNC1]
‘crazy’

[tWNC1.tWNA2.tWNA1]
‘very crazy’

f.

[lwa:NC2]
‘watery’

[lwa:NC2.lwa:NA2.lwa:NA1]
‘very watery’

Besides the triplication pattern presented in (22), another pattern of form XYY that

has three syllables is also documented, shown in (23). In these cases, the input is an ad-

jectival phrase of the form XY; the adverb (i.e., the rightmost syllable Y) is the base for

reduplication. The tone harmony pattern here is similar to that in attenuation. However,

they function differently — while the same pattern in attenuation is for degree minimizing,

it is for intensification here.

(23)

a.

>
kpB1]

[xoA1.xo
‘very dry’

[xoA1.xo >NmA1.xo
‘very very dry’

>
kpB1]

b.

[th7:mA1.fWkB1]
‘very aromatic’

[th7:mA1.fWNA1.fWkB1]
‘very very aromatic’

c.

[satB2.áatB1]
‘very clean’

[satB2.áanA1.áatB1]
‘very very clean’

In the current analysis, the focus on triplication is of the type in (22), in which there are

two copies of the one base, and not the type in (23) in which there is only one copy of the

base. The tone pattern of XYY can then be subsumed under that of partial reduplication

presented in Section 3.1.

17Novel data

28

A number of triplication forms have a truncated disyllabic form, in which the second

syllable in the triplication form is omitted from the disyllabic one (e.g. (24); for more exam-

ples, see Hoang et al. (2008)). For instance, (22a)/(24a) can be truncated to [sa:kB1.sa:kB2]

‘very close to’ (i.e., (24e)); both forms yield the same meaning, but the triplication form has

slightly more emphasis. However, this truncation is not possible in all cases of triplication,

as shown by the illicit forms in (24g)–(h).

(24)

a.

[sa:kB1.sa:NA2.sa:kB2]
‘very close to (emph)’

b.

c.

d.

[jWNC1.jWNA2.jWNA1]
‘indifferent (emph)’

[hOkB1.hONA2.hOkB2]
‘very hot’

[tWNC1.tWNA2.tWNA1]
‘very crazy’

e.

f.

[sa:kB1.sa:kB2]
‘very close to’

[jWNC1.jWNA1]
‘indifferent’

g.

* [hOkB1.hOkB2]

‘very hot’

h.

* [tWNC1.tWNA1]

‘very crazy’

3.5.1 Tone pattern

In general, there are five tonal schema (base-reduplicant alignment asides) observed in

triplication: (1) B1.A2.A1, (2) B1.A2.B2, (3) B2.A2.A1, (4) C1.A2.A1, and (5) C2.A2.A1,

in which each tone counts for a syllable in the reduplicative. Of these patterns, (2), (3), and

(5) do not follow tone harmony within the same register if we consider the first syllable and

the third syllable to be the base and the reduplicant respectively.

Consider (25a) for schema (2) and (25b) for schema (3). In (25a), the base and the second

reduplicant only differ in the tone, where both tones B1 and B2 are non-flat. However, they

belong to different registers (i.e., B1 being +H and B2 being −H) and thus do not follow

the same-register harmony pattern. If both of the reduplicants and the base follow the

tone pattern in partial reduplication (see Section 3.1), the first reduplicant (i.e., the second

syllable) should have been [sa:NA1] and [sa:NA1] for the second one (i.e., the third syllable)

as well. However, these two forms will (1) result in identical syllables in a sequence, and (2)

29

changing the fixed A2 in the second syllable, which is illicit in Vietnamese triplication (see

Section 5.8.1 for the discussion).

In (25b), assuming the first syllable [satB2] to be the base, its relations with both the first

reduplicant [sanA2] and the second reduplicant [sanA1] are different from each other. The

second reduplicant does not follow the harmony pattern; it is not in the same register nor has

the same contour as that of the base (i.e., A1 in the reduplicant and B2 in the base). If this

reduplicant and the base were to follow the tone pattern in partial reduplication, the second

reduplicant should also be [sanA2], i.e., having an A2 tone, which is the flat counterpart of

B2, but that would result identical syllables in a sequence.

(25)

a.

sa:NA2
[sa:kB1
Red
close
‘very close to’

sa:kB2]
Red

b.

sanA2
[satB2
Red
clean
‘very clean’

sanA1]
Red

In conclusion, both tonal harmony patterns in examples of triplication in (25) are different

from each other and different from the general pattern of tonal harmony in partial redupli-

cation. This pattern also leads to another problem in triplication — base and reduplicant

alignment, described in the following section.

3.5.2 Avoidance pattern

Triplication is blocked when the base tone is A2. In such cases, the intensification meaning

has to be conveyed by paraphrase or through replacement forms. For instance, in (26), an

existing word in the language that means ‘very naughty’ is used in place of the triplicated

form of [liA2] ‘naughty’.

(26)

[liA2]
‘naughty’

[liA2.l7mB2] / *[liA2.liA2.liA1/B2]
‘very naughty’

(Novel data)

If A2 bases were to be allowed, given the existing tone pattern in triplication, there would

be two possible schema: (a) A2.A2.A1 and (b) A2.A2.B2. In the schema (b), which has B2 in

the third syllable, the base may or may not have an obstruent coda, but it has A2 tone and

not B1; thus, this schema is impossible. The schema (a) is not possible as well. Section 5.8

30

will go into the details of why a form like schema (a) is not possible, in which this is also

due to allomorphy selection, similar to pejorative and intensification (1).

3.5.3 Reduplicant placement

When the base’s tone is A1, the base is no longer the first syllable, but is now the third

syllable instead, shown in the boldface syllable in (27). The difference in the placement of

the base leads to the shift of the morphemes RA2 and Red. The reduplicants follow the

base for all cases, in which RA2 immediately follows the base and then Red. The order is in

reverse for A1 base; RA2 precedes the base and them the Red precedes RA2. Additionally,

the tone in the reduplicant is also changed. In (27), the reduplicant has an C1 tone instead

of A1. Keeping the base in the first syllable and changing the tone of the reduplicant in the

third syllable from A1 to C1 is not possible too.

(27)

[n7wA1]
brown
‘brown’

n7wA2
[n7wC1
Red
RA2
‘very brown’

n7wA1]
brown

(Novel data)

*

n7wA2
[n7wA1
n7wA1]
Red
RA2
brown
Intended : ‘very brown’

*

n7wA2
[n7wA1
n7wC1]
Red
brown
RA2
Intended : ‘very brown’

Note that there is a difference between the base in (28a)–(b) and that in (27). The analysis

for this mismatch in alignment will be presented Section 5.8, in which I will argue for these

two reduplicants being two separate morphemes, each having different allomorphs that are

subject to their own alignment constraints.

(28)

a.

[jWNC1.jWNA2.jWNA1]
‘very indifferent’

b.

[kONC1.kONA2.kONA1]

‘very small’

3.6 Lexicalized reduplicatives

Lexicalized reduplicatives here are forms that are not produced by a productive word

formation process in Vietnamese (P.-P. Nguyen, 1998; Vu, 2007). Example data are shown

in (29)–(34). By the nature of the studies, most past literature listed and analyzed these

31

forms along with the productive reduplicatives and did not make a distinction between the

two (Hoang, 1985; D.-H. Nguyen, 1997; Hoang et al., 2008). The “base” (if there’s any)

is determined by whether either component bears a lexical category in Vietnamese (e.g.,

(29)–(34)), but there are other forms that neither components of the reduplicatives bear any

meaning (e.g., (35)).

Some forms, however, may appear like a result of a productive reduplication process. For

instance, in the forms in (31), both constituents of the reduplicative are similar all but the

onset, which can be considered as an example of partial reduplication. However, the onset

alternation cannot be explained phonological nor phonetically. Additionally, this process

is not observed elsewhere in the language, and it cannot be applied to other words as a

word formation like the productive reduplication processes described in previous sections.

Or words in (35a)–(b), they resemble outputs of partial reduplication (i.e., same-register tone

harmony). However, they are not productive reduplicatives, and neither of the components

can be the base. One explanation might be that these words historically derived from a

productive process but that process nor the form itself is productive (Vu, 2007).

(29)

(Full) reduplication of nouns (distributive and/or plurality marking Thompson and

O’Harrow, 1988; Vu, 2007)

a.

a:jA1]
[a:jA1
who
who
‘everyone’

(30)

(Full) reduplication of adverbs

a.

[lu@NA1
always
‘always’

lu@NA1]
always

(31) Onset replacement

a.

[lanA1
clever
‘hasty’

canA1]
-18

b.

[NWjA2
person
‘everybody’

NWjA2]
person

b.

[h7:jA1
slightly
‘slightly’

h7:jA1]
slightly

b.

[lu@NB1
-
‘in a hurry’

ku@NB1]
panic

18Since the syllable does not have a meaning, a dash - is used in the gloss.

32

(32) Onset and tone replacement

a.

á7:jA2]
[c7:jA1
play
-
‘to play around’

(33) Rhyme replacement

a.

kujB1]
[kamB1
plant
-
‘to concentrate’

(34) Rhyme and tone replacement

a.

mEC1]
[ma:kB1
-
cool
‘(of weather) cool’

b.

áatB1]
-

[satB2
clean
‘completely clean’

b.

[áWkB2
upset
‘upset’

áojB2]
-

b.

tawB1]
-

[t1nC1
awake
‘refreshed’

(35) Cases in which the base and the reduplicant are indistinguishable

a.

[th1nA1.th1tB1]
‘silently’

b.

[áiA2.áOmC2]
‘water sound’

3.7

Interim summary

In this chapter, a description for each of the productive reduplication processes is pre-

sented. In terms of shape, there are two main patterns: full and partial. Observations reveal

that some reduplication processes but the exclamation one block forms with certain rhyme-

tone combinations (e.g., [-i@k] is blocked in the pejorative process or [-a:Aα] is blocked in

the exclamation process). In contrast, attenuation and exclamation, allow identical syllables.

These are discovered through novel data and comparative examples across processes, show-

ing systematic gaps. Moreover, since the scope of this thesis specifically targets the Southern

variety rather than the Northern or Standard variety, I have clarified and re-described pre-

vious accounts of data where necessary. Preliminary analyses of the data have also been

proposed and will be explored in detail in Chapter 5. In the following chapter, previous the-

oretical analyses of Vietnamese reduplication and similar processes in other languages will

be reviewed.

33

CHAPTER 4

ANALYSIS REVIEW

Description and formal analyses of Vietnamese reduplication are provided in a number of

works — Ngo (1984), Hoang (1985), Thompson and O’Harrow (1988), Agbayani (1997),

D.-H. Nguyen (1997), Srichampa (2002), and Vu (2007). In this study, three representative

work will be reviewed — descriptive (P.-P. Nguyen, 1998; Hoang et al., 2008) and formal

analyses (Vu, 2007).

In addition, because the analysis of triplication and exclamation reduplication will draw

on similar processes in other languages, a review of triplication in Chengdu Mandarin (Lin,

2015) and Taiwanese (Zhang & Lai, 2007), and AABB in Chengdu Mandarin (Lin, 2015)

and Taiwanese (Hsiao, 2018) will be briefly discussed as well.

4.1 P.-P. Nguyen (1998)

P.-P. Nguyen (1998) discusses the morphological aspect of Vietnamese through two pro-

cesses, reduplication and affixation. Description of reduplication is given from two approaches:

the reduplicative’s meaning and the reduplicating pattern.

In terms of meaning, there are two types. One is reduplication as a grammatical process in

which the derived meaning is predicable and can be grouped into two types, attenuation and

intensification. The other is a lexical process in which the derived meaning is not predictable

from the reduplicating pattern. In general, this is similar to the classification shown in

Chapter 3.

Pattern-wise, there are two types: onset alternation and tone alternation. Onset alterna-

tion refers to cases where the onsets in both syllables are changed; either or both the rhyme

and the tone are preserved. For example, (1) shows [P-] changing to [ñ-] in the onset and (2)

shows [P-] to [t-] and [s-] (examples from Hoang et al. (2008)).

(1)

a.

[PeB1]
‘lonely’

[PeB1.ñeB2]
‘completely lonely’

b.

[PiwB1]
‘weak’

[PiwB1.ñiwB2]
‘very weak’

34

(2)

a.

[Po >NmA1.to >NmA2]
‘soft spoken’

b.

[PWmA1.sWmA2]
noisy

Note that there has been no satisfactory explanation for the alternation between any of

the onset pairs, and no fixed meaning is associated with the changes. Thus, this is classified

as the lexical type because there is no predictable alternation pattern and the meaning

associated to the pattern.

Tone alternation refers to cases where the tones of the syllables in the word can have

either the same contour and register or just the register. There are two patterns — full and

partial, shown in (3a)–(b) respectively. The derived meaning can be split into two groups,

attenuation and intensification, shown in (3b)–(c) respectively. Note that in each of the

examples, the base of the reduplication process is identifiable.

(3)

a.

[ka:wA1]
tall
‘tall’

ka:wA1]
[ka:wA1
Red
tall
‘slightly tall’

b.

[sa:kB1]
close.to
‘close to’

[sa:NA1
Red
‘slightly close to’

sa:kB1]
close.to

c.

[sa:kB1]
‘close to’

[sa:kB1.sa:NA2.sa:kB2]
‘very close to’

However, there are also cases where the meaning is lexicalized. For example, consider

[âuA1.âuC1] ‘papaya’ and [á1mA2.á1pB2] ‘coucal’. These two words have the tone harmony

pattern of partial reduplication, but the base is not identifiable, unlike cases in (3a)–(c). In

addition, neither of them have an attenuating or intensifying meaning. Thus, they are of the

lexical type of reduplication.

Affixation covers the [-7p] affix with examples shown below. This is a prefix that signifies

movement or instability of the verb it attaches to.

(4)

a.

[fo >NmA2]
‘inflate’

[f7pB1.fo >NmA2]
‘inflate and deflate’

b.

[NWNA2]
‘stop’

[N7pB2.NWNA2]
‘hesitant’

It is similar to the pejorative process with the [-i@k] rhyme discussed here in the sense

that both have a fixed rhyme, takes the onset of the stem that it attaches to, and presents

some tone harmony pattern. However, it is not a productive pattern like [-i@k].

35

P.-P. Nguyen (1998) only gives a description of the tone harmony pattern in reduplica-

tion and affixation and how they are used to form new words in Vietnamese. The author

does not give a formal analysis; however, they are one of the limited number of works that

acknowledges triplication.

4.2 Hoang et al. (2008)

Hoang et al. (2008) discusses several morphological aspects of Vietnamese, including

reduplication. Similar to P.-P. Nguyen (1998), they describe it as both a word formation

process and a word type, but provide greater detail and cover more patterns. Reduplication

is classified based on syllable count, derived meaning, and harmony pattern.

Of particular relevance here are the two hypotheses proposed by Hoang et al. (2008) for

triplication: (1) repetition of a disyllabic base and (2) repetition of a monosyllabic base.

The first treats triplication as reduplication of a disyllabic form, motivated by examples like

[sa:kB1.sa:NA2.sa:kB2] ‘very close to (emph)’ in (5a), where the second reduplicant (A2) can

be omitted, resulting in a disyllabic form with less emphasis (5b).

(5)

a.

sa:NA2
[sa:kB1
Red
close.to
‘very close to (emph)’

sa:kB2]
Red

b.

sa:NA2
[sa:kB1
Red
close.to
‘very close to’

sa:kB2]
Red

This leads to the question of whether triplication is derived by reduplicating a disyllabic

base. If so, the process can be schematized as follows:

(6) X.X → X.XA2.X

However, this hypothesis faces issues, which are many trisyllabic forms lack disyllabic

counterparts, and the base–reduplicant alignment varies. Sometimes the base precedes the

reduplicant (as in 5b), while in other cases it follows (28a)–(b). Additionally, no explanation

is offered for why the reduplicant has A2. Due to these inconsistencies, Hoang et al. (2008)

ultimately favors the second hypothesis.

In the alternative view, triplication begins with a monosyllabic base that can function

as a word of its own. There are two structural possibilities: either the base and reduplicants

36

are contiguous (e.g., (7)) or not (e.g., (8)).

(7)

a. X → X.XA2

(8)

a. X → X.X

b. X.XA2 → X.XA2.X

b. X.X → X.XA2.X

Hoang et al. (2008) posits that tone assignment follows arbitrary rules, which are the

second syllable has fixed A2 tone, and the third syllable harmonizes with the base either by

the tone contour or the tone register. Yet, this account mirrors the disyllabic-base hypothesis.

Both suffer from inconsistent base placement and tone harmony. If this pattern can be

proposed for the base and the third syllable in the trisyllabic form, the same thing can be

proposed for the disyllabic-base hypothesis as well.

A notable difference between Hoang et al. (2008) and earlier works like P.-P. Nguyen

(1998) and Vu (2007) (and this thesis) lies in lexicalized reduplication. The latter treat

reduplication with onset/rhyme alternation as lexicalized; Hoang et al. (2008), however, sees

them as compounds only if both syllables are meaningful. If only one or neither is meaningful,

they are reduplicatives, even if non-productive, as long as there is a “harmony” between the

two syllables in the word. An example is the [P-]–[ñ-] onset pair in (1); there is no reason for

why this occurs or if this would make the syllables more harmonious (P.-P. Nguyen, 1998).

While both perspectives have merit, this thesis restricts its focus to productive redu-

plication patterns — those described in P.-P. Nguyen (1998) and Vu (2007) — to analyze

patterns that are predictable. The scope also extends to triplication and AABB patterns,

for which formal analyses will be proposed.

4.3 Vu (2007)

Vu (2007) provides a unified OT analysis for four productive reduplication processes:

full, partial, [-i@k], and [-a:] reduplication. The different processes are differentiated by their

tone/segmental pattern: full, Emergence of the Enmarked (TETU) (i.e., partial) (McCarthy

& Prince, 1995), Melodic Overwriting (i.e., [-i@k] and [-a:] reduplication) (Alderete et al.,

1999). Partial reduplication is TETU because it is the emergence of the unmarked tone —

37

flat tones. [-i@k] and [-a:] reduplication are melodic overwriting because neither rhymes can

be phonologically derived.

The analysis involves conflicting constraint rankings within a language; to account for

that, the author opts for Cophonology (Orgun, 1996; Arto Anttila, 1997; Inkelas and Zoll,

2007 among others). This allows distinct co-existing phonological systems, or cophonologies,

in a language. There’s a Master Ranking which is the partial ranking of constraints that all

individual cophonologies must obey.

The author proposes two cophonologies that the four reduplication processes follow: (1)

*Marked ≫ Ident-BR and (2) Ident-BR ≫ *Marked in which *Marked is the set

markedness constraints for the V(owel)-place. The two cophonologies share other constraints

that are dominant in all reduplication processes: Faith-IO, PhonoA for tone markedness,

alignment constraints, and FtBin. Note that alignment constraints refer to a set of alignment

constraints that govern the placement of the base and the reduplicant in reduplicative; there

are two different alignment constraints for the reduplicants, one [-i@k] and one for all the

remaining reduplicants, because only [-i@k] is a suffixal reduplicant but others are prefixes.

Three reduplication processes but the partial one belong to the Ident-BR ≫ *Marked

cophonology. *Marked ≫ Ident-BR is the TETU ranking schema, and thus the partial

reduplication belongs to this one for the unmarked tone (i.e., flat tones) to emerge.

Vu (2007) is one of the few that formalized the reduplication processes in Vietnamese.

The analysis accounts for the different tone patterns in the processes as well as the prosodic

structure of the base and the reduplicatives using cophonologies. However, Vu (2007) does not

discuss AABB and triplication, even though they are also productive reduplication processes.

In addition, there is no analysis given for the ineffability of [-i@k] and [-a:Aα] bases when

they have the same rhyme as the fixed rhyme in the reduplicant of the pejorative and

intensification processes. The author also provides that reduplicatives of the intensification

process do not have the syntactic separability effect; however, novel data from my field work

show that there is such an effect. That poses a possible problem for the analysis in Vu (2007)

38

such that one might have to propose an additional Align constraint, i.e., Align-L(Ra:, ϕ),

which rules that for every reduplicant Ra:, there is a phonological phrase such that the left

edge of Ra: coincides with the left edge of ϕ. This thesis will address and fill in the gaps

listed here.

4.4 Chengdu Mandarin AABB and tone sandhi (Lin, 2015)

Lin (2015) examines four reduplication patterns in Chengdu Mandarin, AA, AAB, ABB,

and AABB, and proposes an OT analysis to account for the reduplicant size and placement,

and the tone sandhi of these patterns. For the purpose of this analysis, the focus will be

solely on AABB, as exemplified below.

(9)

a. s145.u@n31
‘gentle’

s145.s1551.u@n31.u@n31
‘very gentle’

b. ta13.faN44<45
‘generous’2

ta13.ta55.faN44<45.faN44<45
‘very generous’

There are four possible ways to derive the AABB form from a disyllabic base AB, as

shown below.

1. A1A2B1B2

2. A1A2B1B2

3. A1A2B1B2

4. A1A2B1B2

Forms (3) and (4) are ruled out because the tone in the initial position is always identical

to that in the input, suggesting that the initial syllable must be the base. The forms in (1)

and (2) differ by how much of the reduplicant or the base intervenes one another. In (1), the

base is intervened by the reduplicant, while in (2), not only is the base intervened by part

of the reduplicant (A2), the reduplicant is also intervened by part of the base (B1). Thus,

in terms of Contiguity (McCarthy & Prince, 1995), (1) violates only Contig-IB in (10),

while (2) violates both Contig-IB and Contig-IR in (11). Thus, the desired form is (1),

A1A2B1B2, in which the reduplicant appears to be disyllabic like its base and infixal.

1T = special sandhi tone
2T < T = sandhi tone < citation tone

39

(10) Contiguity-IB (Contig-IB): The portion of the base standing in correspondence

to the input forms a contiguous string. (Fitzgerald, 2000)

(11) Contiguity-IR (Contig-IR): The portion of the reduplicant standing in corre-

spondence to the input forms a contiguous string. (Fitzgerald, 2000)

According to Lin (2015), reduplicants in different patterns are considered as allomorphs

of the same functional affix, and the alternation in the reduplicant size and placement can

be accounted for by Morphoprosodic alignment (MPA) model by Kennedy (2008). For the

reduplicant size, it is proposed that the different sizes are the result of two separate patterns

of reduplicants — stem-internal and stem-external — shown below.

(12) Stem-internal reduplicant: [Red + B]stem

(13) Stem-external reduplicant: [Red # [B]stem]

Each of the patterns is accounted for by two alignment constraints, Align + Syll and

Align # Foot, respectively. Based on MPA model, the disyllabic reduplicant in AABB is

stem-external; thus, only Align # Foot is shown in (14).

(14) Align # Foot : Stem-external morphological boundaries occur at foot boundaries.

Although the reduplicant in AABB appears infixal on the surface, it is actually suffixal

in nature.3 Based on (14), there are two possible reduplicant alignment: (a) A.B. # A.B and

(b) A. # A.B # .B. The desired form is (b), which is the same to the form proposed for

AABB. It is triggered by a set of linearity constraints, Lin(M) in (15) and Lin(S) in (16),

that requires the input segmental sequence to be preserved in the output.

(15) Lin(M): The linear order of the morphemes in the input is kept in the output.

(16) Lin(S): The linear order of the segments in the input is kept in the output.

The proposed ranking is in (17) with the input ta13.faN44<45 ‘generous’. The tone is

omitted from the tableau for simplicity. Candidate (c) is eliminated because it is not a foot

3This is similar to the other reduplication patterns in Chengdu Mandarin.

40

and hence violates the alignment constraint.4 Candidate (a) of the form A.B. # A.B is

eliminated because it violates the linear order of segments, in which ta precedes faN in the

input but it is not always the case in this output candidate (e.g., ta follows faN). This leaves

candidate (b) of the form A. # A.B # .B to be the optimal output at the expense of violating

the linear order of the base and the reduplicant. This output does not violate Lin(S) because

the segment ta preceding faN in the input is preserved here.

(17) Lin(S), Align # Foot ≫ Lin(M)

ta.faN # Red

Lin(S) Align # Foot Lin(M)

a.

ta.faN. # (ta.faN)

b. ☞ ta. # (ta.faN) # .faN

c.

ta.faN. # .faN

∗!

∗!

∗

∗!

Having discussed the alignment of the base and the reduplicant, we now turn to the tone

sandhi. The general tone sandhi rules are listed in Table 4.1.

Input

Output

Example

a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
l.

45.45
31.45
13.45
55.45
45.55
31.55
13.55
55.55
45.13
31.13
13.13
55.13

45.44
31.44
13.44
55.44
45.53
31.53
13.53
55.53
45.11
31.11
13.11
55.11

tCin thiEn ‘today’
tso thiEn ‘yesterday’
miEn pau ‘bread’
tshau ku ‘straw mushroom’
tshau s@u ‘wonton’
thAN ko ‘candy’
fu mu ‘parents’
tso s@u ‘left hand’
tsoN tCiau ‘religion’
Ci kuan ‘habit’
s1tCiEi ‘world’
tsau fan ‘breakfast’

Key: tones are separated by ‘.’; tones that undergo general tone sandhi are in boldface.

Table 4.1 Tonal combinations that undergo tone sandhi in Chengdu (Lin, 2015)

4There is a dominant FtBin constraint which requires the foot to be minimally disyllabic.

41

These combinations can be summarized as rules in (18). Specifically, the tone 45 changes

to tone 44 when preceded by another tone; the tone 55 shifts to tone 53 when in the phrase-

final position; and the tone 13 changes to 11 when preceded by another tone. Notice that

environment that triggers the change in tone 45 and 13 are similar to each other and hence

the shaded rows in Table 4.1.

(18) General tone sandhi (TS) rules:

1. 45 [Hr, lh] → 44 [Hr, l]/T 5

2. 55 [Hr, h] → 53 [Hr, hl]/

]

3. 13 [Lr, lh] → 11 [Lr, l]/T

In A1A2B1B2 reduplication,6 A2 always bears the fixed tone 55, while B1 and B2 undergo

regular tone sandhi as shown in Table 4.1 and encoded in (18). For instance, the word for

‘tidy’, ts@n55.tChi31, has the form ts@n55.ts@n55.tChi31.tChi31 ‘very tidy’ when reduplicated.

(19)

ts@n55.tChi31 ‘tidy’ → ts@n55.ts@n55.tChi31.tChi31 ‘very tidy’

This fixed 55 tone on the reduplicant is assumed to be a floating tone introduced by the

morphological reduplication process. When the reduplicant is disyllabic, this floating H tone

docks onto its left edge, a behavior governed by the constraints *Float, MaxFloat, and

Align-HR-L, as defined below.

(20) *Float: No floating tones. (Wolf, 2005)

(21) MaxFloat: All autosegments that are floating in the input have output correspon-

dents. (Wolf, 2005)

(22) Align(H, L, Red, L) (abbr. Align-HR-L): A high tone must be associated with the

leftmost tone bearer of a reduplicant. (Carleton & Myers, 1994)

The desired ranking is in (23). Notice that since the lack of the realization of an input

floating tone in the output would violate either *Float or MaxFloat, as in candidate (a).

5T = sandhi tone
6The subscript numbers are used for ease of reference.

42

Candidate (b) is eliminated because the input floating tone is realized, but it is associated

with the rightmost tone bearer of a reduplicant instead.

(23) T = floating tone in the input; T = output tones that correspond to an input

floating tone

45-Red, 55

*Float MaxFloat Align-HR-L

a.

b.

45-337

∗!

∗!

55 -33

∗!

c. ☞ 45- 55

Based on the general tone sandhi rules in (18), the allotone pairings 45∼44, 55∼53,

and 13∼11 all preserve tone register, with faithful preservation of the left tone segment

from input to output. This supports the dominance of the constraints Ident-IO-Reg and

Ident-IO-L-T, shown in (24) and (25), respectively.

(24)

Ident-IO-Reg— The register of an input tone must be preserved in the output.

(25)

Ident-IO-L-T — The left tone segments of an input must be preserved in the output.

The alternations from 45 and 13 to 44 and 11 in post-tonal or non-initial positions can

be explained as leveling of rising contours and deletion of the left tone segment. This is

captured by the constraints Max-IO-T-L (no deletion at left edge) and *Rise (ban on

rising contours), whose ranking is given in (26). This ensures that the change from 45, 13 to

44, 11 occur in non-initial position only and not when 45 and 13 occur in isolation.

(26) Max-IO-T-L ≫ *Rise

The shift from 55 to 53 in utterance-final position is attributed to a boundary low tone

that incorporates with the base 55 tone at the end of the utterance. This is supported by the

7It is unclear why 33 is here. It might have been a typo in the original text because according to the

general tone sandhi in Table 4.1, it should have been 44.

43

constraint Boundary Low, which enforces the parsing of boundary L tones at the surface

(Chen, 2000). The ranking in (27) ensures that this alternation results from boundary tone

incorporation rather than insertion of new tone segments.

(27) Boundary Low ≫ Dep-IO-T

Further evidence for this comes from how isolated tones 45 and 13 interact with boundary

L tones. In such cases, the boundary L can surface either as a complex tone (i.e., 454[Hr,

lhl] and 131[Lr, lhl]), in which the input tone is preserved, or a level tone (i.e., 44 [Hr,

ll] and 11[Lr, ll]), in which the right input tone segment is deleted for the boundary L to

surface.8 The constraint *ComCon (ban on complex contours) is needed to prevent the

former scenario, and its dominance over Boundary Low is shown in (28). Additionally,

the change of 45 and 13 to 44 and 11 in the post-tonal/non-initial position, indicating that

*Rise, which triggers the alternation, dominates Max-IO-T, which forbids the change.

(28)

*ComCon, Max-IO-T ≫ Boundary Low

When combining the general and special sandhi processes in A1A2B1B2 sequences, two

cases arise: (1) when the base has 55 (i.e., A1.55) and (2) when the base has 45 or 13

(i.e., A1.45 or A1.13). When the base has 55, it is copied onto the reduplicant B1. the B1

in the input, which is now B2 in the output, has 55 in the utterance’s final position and

thus is changed to 53 (i.e., A1.A2.55.53) due to the boundary L. The change, however, is

not reflected in the reduplicant, violating Ident-BR but obeying Boundary Low. If this

change were copied, it would instead violate Ident-IR, and hence, the ranking proposed in

(29). When the base has 45 or 13, the change to a level tone 44 or 11 is also applied to

the reduplicant B1, which means the constraint *Rise that drives this change dominates

Ident-IR.

8l = boundary L tone

44

(29)

Ident-IR, Boundary Low ≫ Ident-BR

X.55 # Red

Ident-IR Boundary Low Ident-BR

a.

b.

X-X.53-53

∗!

X-X.55-55

∗!

c. ☞ X-X.55-53

∗

Finally, a special tone sandhi condition arises when the reduplicant follows a base tone

55. In this context, the reduplicant never surfaces with any of the expected general sandhi

tones (55, 13, 45, 11, 44, 53), but instead with 31. Tone 55 is avoided because the language

militates against 55. 55 , in which 55 corresponds to the floating H tone, violating (30).

Tones 13 and 45, both having a rising contour LH, are universally disfavored due to *Rise.

Tones 11 and 44 are ruled out by ¬BR-Con (31), since they share contours with 55 (i.e., 11:

L, 44: L, and 55: H). Lastly, tone 53 is excluded because it shares a register with 55, which

is High, violating ¬BR-Reg (32).9

(30)

*55. 55 : 55. 55 is prohibited.

(31) ¬BR-Con: The reduplicant and the base should have different contours.

(32) ¬BR-Reg: The reduplicant and the base should have different registers.

All of this comes to the ranking in (33). However, in the case of fixed 55 in A2 in

A1A2B1B2, the tone of B1 does not change to 31 This means that Ident-BR is preserved

and hence the ranking in (34).

(33)

*Rise, *55- 55 ≫ ¬BR-Con, ¬BR-Reg

(34)

Ident-BR ≫ ¬BR-Con, ¬BR-Reg10

9Note that the anti-faithfulness constraints (31)–(32) are controversial due to limited cross-linguistic

support and their data-driven formulation. A full evaluation lies beyond the scope of this summary.

10*Rise, *55- 55 dominate Ident-BR because *Rise ≫ Ident-IR and *55- 55 ≫ *Float, MaxFloat

≫ Ident-IR in other processes.

45

Lin (2015) offers separate analyses for reduplicant alignment and size on one hand, and

tone sandhi on the other. As a result, there is no interaction between the two shown; neither

constraint set appears to influence the rankings or outcomes of the other. While Contig-IB

and Contig-IR are invoked to rule out unattested forms, their rankings relative to the

tonal or alignment constraints are not explored in their paper. Additionally, the constraints

in (31)–(32) are not easy to motivate beyond the data at hand. It is important to provide an

analysis that is consistent between the reduplication patterns and the tone sandhi patterns

to show that the overall analysis works; a challenge that I will pursue in this thesis (see

Chapter 5).

4.5 Taiwanese (Zhang & Lai, 2007; Hsiao, 2018)

4.5.1 Taiwanese AABB (Hsiao, 2018)

Taiwanese AABB is exemplified below.

(35)

a. ki33

kuai21
weird

strange
‘strange’

ki33
strange
‘very strange’

ki33
strange

kuai53
-

kuai21
weird

b. khin33
light
‘relaxed’

sang55
loose

khin33
light

khin33
light
‘very relaxed’

sang33
-

sang55
loose

Note that Hsiao (2018) only discusses the base-reduplicant alignment and the redupli-

cant’s shape; there is no mention of the phonological material. Given the form AABB, there

are four logically possible alignments between the base and the reduplicant, similar to the

Chengdu Mandarin case. Each form corresponds to a specific constraint ranking involving

stem-alignment and reduplicant-alignment constraints, defined in (36)–(39).

(36) AlignRed-R: For every Red, assign one violation mark for every syllable that in-

tervenes between the right edge of that Red and the right edge of the Wd.

(37) AlignRed-L: For every Red, assign one violation mark for every syllable that in-

tervenes between the left edge of that Red and the left edge of the Wd.

46

(38) AlignStem-R: For every stem, assign one violation mark for every syllable that

intervenes between the right edge of that stem and the right edge of the Wd.

(39) AlignStem-L: For every stem, assign one violation mark for every syllable that

intervenes between the left edge of that stem and the left edge of the Wd.

The proposed constraint ranking is (40), in which the reduplicant is a suffix.

(40) A1A2B1B2: AlignStem-L, AlignRed-R ≫ AlignStem-R, AlignRed-L

There is both cross-linguistic and internal evidence supporting this ranking. One cross-

linguistic observation is that full reduplication in most Sinitic languages patterns as suf-

fixation. In Pingnan Baihua, for instance, the expressive suffix kuai appears in the second

syllable position where the reduplicative would be (i.e., A kuai BB), suggesting that the

reduplicant occupies a suffixal position in the form AABB. The evidence from within Tai-

wanese comes from literary Taiwanese, in which the emphatic marker tsai appears in the

second and fourth syllables of the construction X.X.Y.Y, again pointing to the interpretation

that the first and third syllables are the base, and the second and fourth are the reduplicant,

located where a regular suffix would be.

To further refine the analysis, a set of Contiguity constraints is added to the alignment

ranking, yielding a revised constraint ranking in (41).

(41) Revised (40):

A1A2B1B2: AlignStem-L, AlignRed-R ≫ AlignStem-R, AlignRed-L, Con-

tigRed, ContigStem

While both Hsiao (2018) and the Chengdu Mandarin paper make use of Contig con-

straints, they do so in opposite ways. In Chengdu Mandarin, Contig is invoked to explain

why A1A2B1B2 surfaces as optimal, whereas in the Taiwanese case, Contig is ranked lower

to allow the A1A2B1B2 pattern to win, despite its violation of contiguity.

Because AABB in Taiwanese is a case of full reduplication, the focus of this analysis is

on the alignment between base and reduplicant, rather than tone sandhi. As such, it is not

47

clear what influences, implications, or interactions of the alignment constraints and Contig

have in other reduplication processes or tone sandhi in Taiwanese.

AABB in Taiwanese bears a closer resemblance to Vietnamese AABB reduplication than

to that of Chengdu Mandarin. In both Taiwanese and Vietnamese, the process is full redupli-

cation. However, while Red behaves as a suffix in Taiwanese (i.e., A1A2B1B2), in Vietnamese,

it appears to be a prefix (i.e., X.X.Y.Y).11

4.5.2 Taiwanese triplication and tone sandhi (Zhang & Lai, 2007)

Taiwanese tone sandhi, like other Southern Min dialects, is positionally conditioned.

Specifically, tones undergo sandhi in non-XP-final positions, while the tone of the XP-final

syllable remains unchanged. This positional dependency is illustrated in the tone sandhi

pattern in (42).12

(42) Taiwanese tone sandhi in non-XP-final positions:

33

24

51

55

21

Because tone sandhi applies only in non-final positions, the analysis assumes that in

reduplication, the final syllable is the base while the preceding syllables are reduplicants.

In a trisyllabic reduplicative form, the second syllable undergoes general tone sandhi. In

contrast, the first syllable behaves differently depending on the base tone. It follows the

general sandhi pattern only when the base tone is 21 or 51. When the base tone is 55, 33, or

24, however, the first syllable surfaces with a rising tone 35. This distribution is summarized

in (43).

11I used XY notation for Vietnamese to avoid confusion with A and B tones.
12The five tones considered here are for unchecked syllables, which are syllables that are either open or

closed with a sonorant coda.

48

(43) Tone patterns in Taiwanese reduplication

Monosyllabic base Single reduplication Double reduplication13

21
51
55
33
24

51-21
55-51
33-55
21-33
33-24

51-51-21
55-55-51
35-33-55
35-21-33
35-33-24

To derive these patterns, the authors propose two main mechanisms: (1) the docking

of a floating High (H) tone and (2) the selection of appropriate tonal allomorphs for non-

final syllables. The floating H tone is key to understanding the tone on the first syllable of

trisyllabic reduplication. The sandhi tone on the second syllable is first determined through

allomorph selection, and the tone on the first syllable is then derived by docking a floating H

tone to the sandhi tone. If the sandhi tone starts with a H tone (i.e., 51 or 55), the floating

H docks to the left edge of the syllable. If the sandhi tone starts with a mid tone (i.e., 33 or

21), the floating H docks to the right edge.

This analysis is supported by a set of ranked constraints.

(44) Realize(Float): A floating tone must be realized in the output.

(45) Align(Float, L, Wd, L) (abbr. Align-L(Float, Wd)): The left (L) edge of a floating

tone must be aligned with the left edge of a word (Wd).

(46)

Ident-RR(Tone, L) (abbr. Id-RR(T, L)): The left edges of the tones of two redupli-

cants derived from the same base must be identical.

(47)

Ident-RR(Tone) (abbr. Id-RR(T)): The tones of two reduplicants derived from the

same base must be identical.

Realize(Float) is undominated since all outputs in the dataset realize the floating H

tone. Id-RR(Tone, L) is ranked above Align-L(Float, Wd), favoring outputs in which the

left edge of the two reduplicants derived from the same base are identical. Id-RR(Tone) is

13Triplication is called double reduplication in Zhang and Lai (2007).

49

ranked lowest because identity between the tones of the two reduplicants is not observed

except in cases where the base tone is 21 or 51. This ranking is formalized in (48).

(48) Realize(Float), Id-RR(Tone, L) ≫ Align-L(Float, Wd) ≫ Id-RR(Tone)

The tone sandhi in non-final syllables is treated as an allomorph selection process rather

than a productive phonological rule, especially since such alternations fail to generalize

to novel words (Tsay & Myers, 1996). This is captured by the constraint Appropriate-

Allomorph (Allmph), which selects surface tonal allomorphs from the lexicon.

(49) Appropriate-Allomorph (abbr. Allmph): For an existing syllable, select its sur-

face tonal allomorph as follows:

UR

21
51
55
33
24

Surface allomorph

XP-final

Non-XP-final

21
51
55
33
24

51
55
33
21
33

The revised ranking is in (50). Realize(Float) and Id-RR(Tone, L) rank above the

Allmph constraint, ensuring that (1) no cases where the two reduplicant tones are similar

to each other but does not realize the floating H tone in the output, and (2) the floating H

being on the right edge of the first syllable when the base has a sandhi tone that does not

have an initial H. In turn, Allmph is ranked above Align-L(Float, Wd). For instance, with

a base tone 55, an RR sequence like 35-33 and 53-53 are possible outputs, 53-53 does not

violate Align-L(Float, Wd) but the supposed output 35-33 does, so Allmph must dominate

to rule out cases like 53-53 because while 35-33 violates that once (for the 35 in the first

syllable instead of 33), 53-53 violates that twice.

50

(50) Revised (48):

Realize(Float), Id-RR(Tone, L) ≫ Allmph ≫ Align-L(Float, Wd)

≫ Id-RR(Tone)

Zhang and Lai (2007) assume a fixed alignment between base and reduplicant and does

not examine possible interactions between alignment constraints and the tonal constraints

discussed. Comparing this analysis with that of the Chengdu Mandarin, both authors deal

with floating H tones but propose different constraint sets. The Chengdu analysis uses

*Float and MaxFloat to ban unassociated tones and penalize their deletion, respec-

tively. In contrast, Zhang and Lai (2007) uses Realize(Float), which stems from Kurisu

(2001)’s Realize Morpheme constraint, requiring every underlying morpheme to receive

some phonological exponent.

In addition, both Zhang and Lai (2007) and Lin (2015) analyze the pattern of single

reduplication differently. In Chengdu Mandarin, since this pattern is productive, Lin (2015)

analyzes this as a docking of floating H tone in the reduplicant. As for Taiwanese, since this

pattern is not productive, Zhang and Lai (2007) proposes the constraint Allmph following

Tsay and Myers (1996).

Note that although the tone sandhi pattern in single reduplication is not productive,

speakers still produce the first-syllable tone in double reduplication presented in (43) even

when they get the second-syllable tone that is not presented in (43) or not following the

pattern shown in (42) (Zhang & Lai, 2007). There seems to be a split in productivity: one

reduplicant (the first syllable) is derived via a productive process involving the docking of

the floating H tone, while the other reduplicant (the second syllable) reflects a lexicalized,

non-productive process of allomorphy selection.

4.6

Interim summary

This section summarizes how triplication and AABB patterns have been described and

analyzed both cross-linguistically and within Vietnamese. While triplication and redupli-

cation mechanisms more broadly have been extensively discussed in Vietnamese literature,

51

three representative works are selected here: the descriptive accounts of P.-P. Nguyen (1998)

and Hoang et al. (2008), and the formal analysis of Vu (2007). Two salient features of triplica-

tion are the apparent identity avoidance and a tone pattern that diverges from expectations.

However, one possible approach that has been largely overlooked is the idea that triplica-

tion may not involve multiple rounds of base copying. Instead, it could involve allomorph

selection. From this perspective, what looks like reduplication may reflect the presence of

two separate morphemes, each contributing to the phonological representation and interpre-

tation of the overall reduplicative form. This alternative analysis will be developed further

in Section 5.8.

Since the analysis of triplication and AABB reduplication will make reference to the

same processes in other languages, a review of these processes in Chengdu Mandarin and

Taiwanese were discussed as well (Zhang & Lai, 2007; Lin, 2015; Hsiao, 2018). One noticeable

observation is that the analyses provided by these works are either of alignment or tone

sandhi only, or both but separately. Hence, they do not show how the two analyses might

interact with each other. The analysis in this thesis will attempt to close that gap in a similar

analysis for Vietnamese reduplication; it is crucial to show how the two patterns of alignment

and tone interact with each other for the sake of a uniform analysis of reduplication. Such a

holistic approach contains possible analyses for tones; for instance, one cannot say something

is a default tone if it has incorrect consequences on other tone patterns.

52

CHAPTER 5

ANALYSIS

Although the core of this study centers on tone patterns and base-reduplicant alignment

in triplication and AABB reduplication, re-analyzing other productive reduplication pat-

terns in Vietnamese, especially those previously treated in OT (Vu, 2007), offers valuable

insight toward a unified OT account. Such an approach makes it possible to evaluate how

tone interacts with alignment constraints across the full range of productive reduplication,

contributing to a broader understanding of Vietnamese morphophonology.

This chapter is arranged as follows. The relevant theoretical models that the analysis

utilizes are briefly discussed in Section 5.1 and Section 5.2. From Section 5.4 to Section 5.8,

the analysis of each reduplication process — attenuation, pejorative, intensification, excla-

mation, and triplication — is presented.

5.1 Correspondence Theory

This analysis is presented in the Optimality Theory (OT) framework, in which languages

differ by how constraints are ranked against each other (Prince & Smolensky, 2002). Under

this theoretical framework, given some input, an infinite set of output candidates are gener-

ated by the function Gen and evaluated by the function Eval for well-formedness against a

set of ranked and violable constraints. The optimal candidate is selected as the one that vi-

olates the lowest-ranked constraints and violates the dominant constraints the least number

of times.

The model in OT that is of interest here is Correspondence Theory (McCarthy & Prince,

1995) which involves correspondence relation between stem and base, between base and

reduplicant, and between stem and reduplicant. The correspondence relation is defined in

(1).

(1) Correspondence (McCarthy & Prince, 1995): Given two strings S1 and S2 , cor-

respondence is a relation R between the elements of S1 and those of S2. Elements

53

α ∈ S1 and β ∈ S2 are correspondents of one another when α R β.

The system of relations is schematized as follows,

(2) Full model

Input:

/AfRed +

Base/

IR–Faithfulness

IB–Faithfulness

Output:

R

B

BR–Identity

For this analysis presented in this chapter, the basic version of the model above will be

used instead since it is sufficient enough. The basic model is shown below.

(3) Basic model

Input:

/AfRed +

Base/

IB–Faithfulness

Output:

R

B

BR–Identity

Constraints must govern the correspondence and identity of correspondent elements.

There are separate constraints for each I(nput)-O(utput) and B(ase)-R(eduplicant) corre-

spondence relation. The following are three of the constraint families that will play a role in

the analysis with its corresponding domain-specific instances:

(4) Max constraint family: Every segment of string S1 has a correspondent in string S2.

a. Max-IO: Every segment of the input has a correspondent in the output.

b. Max-BR: Every segment of the base has a correspondent in the reduplicant.

(5) Dep constraint family: Every segment of string S2 has a correspondent in string S1.

a. Dep-IO: Every segment of the output has a correspondent in the input.

b. Dep-BR: Every segment of the reduplicant has a correspondent in the base.

(6)

Ident constraint family: S2 correspondents of S1’s [γF] segment are also [γF].

a. Ident-IO: Output correspondents of an input [γF] segment are also [γF].

b. Ident-BR: Reduplicant correspondents of the base [γF] segment are also [γF].

54

Among these constraint families, Max will be used most intensively. One reason is that

the difference the input and the output, mainly involves no deletion of segments or features

that are specified in the input for either the base or the reduplicant.

5.2 Prosodic hierarchy

Following Vu (2007), the prosodic hierarchy will be used in this analysis for the descrip-

tion of stress patterns and syntactic separability in Vietnamese. In the prosodic hierarchy,

sentences can be a structure such that its categories can be organized hierarchically such that

in each domain, various phonological and morphological processes can take place (Selkirk,

1978; Hayes, 1989, et seq.).

Prosodic hierarchy
Utt

Category
Utterance

IP

ϕ

ω

Ft

σ

µ

Intonational phrase

Phonological phrase

Prosodic word

Foot

Syllable

Mora

5.3 Phonotactics

I posit the constraints listed below, along with Max-IO, to be undominated in the

language since no inputs nor outputs violate them. The tone-coda co-occurrence restriction

is not the main focus of this analysis; thus, an undominated arbitrary markedness constraint,

Phono, will be posited. The unpacking of this constraint into more motivated constraints is

left to future work. Because no outputs in the language violate the phonotactic constraint,

it is assumed to be dominant. This ensures that any sequence of obstruent coda with A or

C tones is barred from the inventory of the language as a whole, and hence will never have

a surface representation.

55

Following Vu (2007), I assume that every foot is binary and hence FtBin (σ), where

each foot must consist of two syllables, in Vietnamese. The constraint’s definition is given

in (7).

(7) FtBin (σ): Feet are binary under syllabic analysis. (Prince and Smolensky, 2002).

This is enforced by ParseSyll, which requires that every syllable be parsed into a foot.

Stress assignment is handled by Align-R(Ft, H(Ft)), which demands that the right edge

of a foot align with that of its head, ensuring that stress consistently falls on the rightmost

syllable.

(8) ParseSyll: All syllable σ must be parsed by a foot Ft.

(9) Align-R(Ft, H(Ft)): For every foot Ft, there’s a head foot H(Ft) such that the right

edge of Ft coincides with the right edge of H(Ft).

To prevent empty syllables in outputs, a Fill constraint is assumed to be undominated.

This constraint penalizes any syllabic position that is not filled by segmental material. In

the context of Base-Reduplicant Correspondence Theory (BRCT), the reduplicant Red is

a morpheme that is phonologically empty in the input (McCarthy & Prince, 1994). Can-

didates that insert the reduplicant without filling it will violate Fill and are eliminated.

Note that satisfying Fill in the output does not violate Max-IO, since the reduplicant is

phonologically empty in the input and thus has no input correspondents. It is also assumed

that the shape of the reduplicant is minimally a syllable. Previous work in OT analyses

of reduplication has proposed various ways to derive this through constraint interactions.1

However, since this issue is not within the scope of the present analysis, I will assume a

minimal syllabic shape for the reduplicant.

(10) Fill: Syllable positions are filled with segmental elements. (Prince and Smolensky,

2002)

1One consequence of template derivation is that a Realize Morpheme constraint (Kurisu, 2001) could

be used in place of Fill, but since no derivation was attempted, Fill is used instead.

56

5.4 Attenuation

In attenuation, there are two main patterns: full reduplication, where the reduplicant is

an exact copy of the base, and partial reduplication, where the reduplicant exhibits a fixed

flat tone (i.e., A tone) while the remaining segmental material is copied from the base. Note

that attenuation reduplicatives are not subject to the syntactic separability effect. Thus, an

assumption is made, which is attenuation reduplicatives are of a Pr(osodic) W(or)d (ω) and

therefore also a foot. Each of the base and reduplicant is then a syllable in size. The following

alignment constraints are adopted to regulate the positions of both the base and reduplicant

within the PrWd. Ultimately, we need specific Align for different reduplicant allomorphs,

but in the interest of brevity, the constraints are presented as such.

(11) Align-L(Base, ω) (Base-L(ω)): For every base, there’s a PrWd ω such that the left

edge of the of the base coincides with the left edge of ω.

(12) Align-R(Base, ω) (Base-R(ω)): For every base, there’s a PrWd ω such that the

right edge of the of the base coincides with the right edge of ω.

(13) Align-L(Red, ω) (Red-L(ω)): For every Red, there’s an ω such that the left edge

of Red coincides with the left edge of the PrWd.

(14) Align-R(Red, ω) (Red-R(ω)): For every Red, there’s an ω such that the right edge

of Red coincides with the right edge of the PrWd.

In attenuation, the reduplicant surfaces as a prefix. This prefixal position satisfies Align-

L(Red, ω) and Align-R(Base, ω) but violates Align-L(Base, ω), as the base is no longer

aligned with the left edge of the PrWwd due to the insertion of the reduplicant. The alignment

ranking that captures this is presented below with the corresponding tableau. Along with the

undominated constraints that govern Vietnamese phonotactics, these alignment constraints

are dominant. For simplicity, only relevant alignment constraints present in the tableau.

57

(15) Align-L(Red, ω), Align-R(Base, ω) ≫ Align-R(Red, ω), Align-L(Base, ω)

B + R

Red-L(ω) Base-R(ω) Red-R(ω) Base-L(ω)

a. ☞ [(σR.σB)]

∗

∗

b.

[(σB.σR)]

∗!

∗!

5.4.1 Full reduplication

In full reduplication, the reduplicant is a total copy of the base. Hence, both Max-IO

and Max-BR are satisfied; the input base is preserved in the output, and there is no deletion

of segments nor mismatch between the base and its reduplicant. Markedness constraints (M)

are dominated and therefore do not influence the choice of output. The optimal candidate

is the one that fully satisfies faithfulness to both input and base, provided it also meets the

undominated phonotactic constraint and is not empty segmentally. Thus, the desired ranking

is in (16).

(16) Phono, Fill, Align-L(Red, ω), Align-R(Ft, H(Ft)), Align-R(Base, ω), Pars-

eSyll, FtBin, Max-IO, Max-BR ≫ M, Align-R(Red, ω), Align-L(Base, ω)

Consider the example tableau below for the input /sa:NB1/ ‘bright’. For simplicity and

relevance of empty the reduplicant, of the dominant constraints besides Max-IO and Max-

BR, only Fill will be in the tableau. Candidate (b) and (c) are ruled out because they each

violates Max-BR and Max-IO respectively. Candidate (d) is ruled out because it violates

the undominated Fill for having an empty syllable. This leaves candidate (a) to be the

optimal output.

58

(17)

Input: /sa:NB1/ ‘bright’

R + sa:NB1

Fill Max-IO Max-BR

a. ☞ sa:NB1.sa:NB1

b.

c.

d.

sa:NA1.sa:NB1

∗!

sa:NA1.sa:NA1

∗!

R.sa:NB1

∗!

5.4.2 Partial reduplication

Since the reduplicant’s tone has a fixed A contour and copies the base tone register, we can

either approach this as the fixed tone contour arising through different constraint rankings

(i.e., A tone is phonologically derived) or it being lexically specified in the reduplicant input

(i.e., an allomorph of the reduplicant).

A set of markedness constraints can be imposed to opt for the flat tone to occur in the

reduplicant at the expense of an unfaithful mapping between the base and reduplicant. For

the purpose of the analysis, a modified version of NoContour in Pulleybank (1997), which

prohibits more than one tone to be linked to a vowel, is proposed. Recall from Section 2.1

that in their tone structure model, flat tones only have one point linked to its contour node.

Thus, this constraint will rule out non-flat tone, which has more than one point under their

contour node.

(18) NoContour: No more than one point may be linked to the contour node.

The proposed ranking, shown in (19), will follow the model of The Emergence of The

Unmarked (TETU), in which Faith-IO ≫ M ≫ Faith-BR (McCarthy & Prince, 1994).

Under this schema, the inactive markedness constraint M (due to being dominated by Max-

IO) becomes active in BR mappings where IO faithfulness is irrelevant. The ranking points

out that any tone and not necessarily flat tones is permitted in the language, but because

59

NoContour ≫ Max-BR, the reduplicant will not copy the tone from the base and has to

have a flat tone.

(19) Max-IO ≫ NoContour ≫ Max-BR

Similarly, another way to approach fixed segmentism here is to consider the tone contour

A and tone register α being specified in the reduplicant as a requirement of the lexicon.

If that’s the case, the input will be something like Base.redAα in which Base represents

the phonological material of the base and Aα subscript represents the specified segment in

the reduplicant. With the reduplicant being specified with a tone contour and tone register

already, the only thing that is left to do is filling in any of the empty positions, in this case,

onset, nucleus, and coda and thus the ranking below.

(20) Fill, Max-IO ≫ Max-BR

Note that this is the same ranking for full reduplication in (16). Previously, Phono and

Fill have been established to be undominated in the constraint ranking to (1) rule out any

input or output that violates the phonotactic constraint on tone and coda, and (2) no empty

syllable can have a surface representation. On the contrary to the case of fixed tone through

constraint ranking, NoContour is not needed here. Max-IO dominates Max-BR to rule

out any candidate that alters the phonological content of the base or the reduplicant as an

effort to obey Max-BR.

Consider the input /sa:kB1/ ‘to be close to’ in the tableau below. Candidate (c) is

eliminated for violating the highly ranked Fill. Between candidates (a) and (b), (a) is the

winner because (b) violates Max-IO for deleting the flat contour to preserve the identity

relation between the base and reduplicant. Candidate (a) is the optimal output at the expense

of deleting the non-flat contour in the reduplicant.

60

(21)

Input: /sa:kB1/ ‘to be close to’

RAα + sa:kB1

Fill Max-IO Max-BR

a. ☞ sa:NA1.sa:kB1

b.

c.

sa:kB1.sa:kB1

∗!

RAα.sa:kB1

∗!

∗

∗

Comparing the two analytical options, the phonological-derivation approach requires an

additional markedness constraint, NoContour, which complicates the constraint hierarchy.

In contrast, the allomorphy case relies on a simpler ranking that only necessitates tone

register identity. This analytical choice becomes particularly significant when we consider

other reduplication processes — specifically exclamation/AABB forms — which only permit

the full reduplication ranking shown in (16). For the purpose of developing a unified OT

analysis and avoiding conflicting constraint rankings within the same prosodic domain, I

will adopt the allomorphy approach for the attenuation process, as presented below. This

decision eliminates the need for multiple competing constraint hierarchies within Vietnamese,

providing a more coherent overall account of the language’s reduplication.

(22) Revised (16)

Phono, Fill, Align-L(Red, ω), Align-R(Ft, H(Ft)), Align-R(Base, ω), Pars-

eSyll, FtBin, Max-IO ≫ Max-BR, Align-R(Red, ω), Align-L(Base, ω)

5.5 Pejorative

Recall from Section 3.3 that the reduplicant in the pejorative process is either [-i@kB1] or [-

i@kB2]. The process disallows any input that has the same rhyme [-i@k] as the reduplicant’s.

In Southern Vietnamese, there is a diphthong reduction process in which the rime /-i@k/

surfaces as [-i:k]. SVN speakers tend to omit the second vowel and lengthen the first as

a compensatory strategy.2 This reduction can be explained within the framework of OT

2This process of diphthong reduction is observed in all three diphthongs [i@, W@, u@] in SVN.

61

using a universally ranked family of markedness constraints on place features: *Pl-Dor,

*Pl-Lab ≫ *Pl-Cor, where coronals are the least marked compared to dorsals and labials

(Lombardi, 1997; Prince & Smolensky, 2002). The nucleus [-i:] then emerges in the output

due to a ranking *Pl-Dor, *Pl-Lab ≫ *Pl-Cor ≫ Max-BR. However, this still does

not explain why the coda surfaces as [-k], a dorsal segment, despite dorsals being the most

marked under this ranking.

Since the rime contains a marked structure that cannot be explained by phonological

constraints alone, it is assumed to be specified lexically. In other words, the specification

[-i@k] in Ri@k in pejorative reduplication is lexically provided rather than derivable solely

from phonology.

Regarding the tone, there are three possible underlying representations of the reduplicant

that can be considered: full specification /Ri@k B1/, partial specification /Ri@k Bα/, and no

specification /Ri@k/. Given the fixed B1, the full specification /Ri@k B1/ needs a markedness

constraint on the reduplicant’s tone to dominate Max-IO in cases where the base with a

low tone (i.e., tone register = 2) wants the reduplicant to have a low tone too at the expense

of the output being unfaithful to the input. However, a consequence of this would be that if

M ≫ Max-IO, Max-BR, the segment or structure militated against by M (in this case, a

high tone) is barred from the inventory of the whole language.

Another possibility is the partial specification of the tone contour /Ri@k Bα/. However,

this appears redundant because phonotactic constraints Phono, Max-BR, and possibly

other markedness constraints already favor B1 or B2 as appropriate.

The last possibility is no specification /Ri@k/. In this case, phonotactic constraints gener-

ate B tones, and Max-BR allows either B1 or B2 depending on the tone of the base. When

the tone register of the reduplicant matches that of the base, Max-BR will ensure that. In

contrast, when the tone freely alternates between B1 and B2, other tone-related markedness

constraints might dominate Max-BR to yield the correct output, as will be discussed later.

This is the best and least problematic option.

62

Turning to alignment constraints, the reduplicant Ri@k must be properly aligned with the

phonological phrase (ϕ) (Vu, 2007). This is motivated by the observation that there can be

additional syntactic constituents separating the base from the reduplicant in the pejorative

process (see Section 3.3 for examples).

Monosyllabic base X (14):
→ ϕ

ϕ

Polysyllabic base X.Y (15):

ϕ

→

ϕ

ω

Ft

σ

ω

ω

Ft

Ft

σ

σ

ω

ω

ω

ω

Ft

Ft

Ft

Ft

σ

σ

σσ

σσ

The constraint Align-R(Ri@k, ϕ) (Ri@k-R(ϕ)) ensures that the right edge of Ri@k coincides

with the right edge of the phonological phrase. This is the opposite of attenuation redupli-

cation, in which the reduplicant is prefixed. Furthermore, inserting Ri@k as a suffix violates

the base counterpart of the alignment constraint, Align-R(Base, ϕ) (Base-R(ϕ)).

A second alignment constraint, Align-L(Ri@k, ϕ) (Ri@k-L(ϕ)), is the leftward equivalent

of the one above. It is always violated in pejorative reduplication because Ri@k always follows

the base, and hence its left edge will never align with the left edge of the phonological phrase.

For the stem, which is assumed to be a PrWd, two alignment constraints are posit:

Align-L(Base, ϕ) (Base-L(ϕ)) and Align-R(Base, ϕ) (Base-R(ϕ)), ensuring that both

edges of the base align with the phonological phrase. Because Ri@k follows the base, its right

edge must align with that of ϕ, which means Align-R(Ri@k, ϕ) and Align-L(Base, ϕ) are

satisfied, while Align-L(Ri@k, ϕ) and Align-R(Base, ϕ) are violated.

This results in the following alignment ranking,

(23) FtBin, Align-R(Ri@k, ϕ), Align-L(Base, ϕ) ≫ Align-L(Ri@k, ϕ), Align-R(Base,

ϕ)

The lower-ranked constraints Align-L(Ri@k, ϕ) and Align-R(Base, ϕ) are not ranked

against Fill but should dominate Max-IO altogether, similar to the attenuation process.

63

In tableaux for attenuation reduplication, FtBin (and Align-R(Ft, H(Ft))) were not

shown because the reduplicant is always a disyllabic PrWd, and hence these are dominant.

However, in the case of pejorative reduplication, some inputs have stems composed of two

PrWds (e.g., coordinate compounds). Hence, FtBin is necessary in this analysis for com-

pleteness as to eliminate candidates that (1) retain the prosodic structure of the base or (2)

have a reduplicant composed of two separate PrWds.

(24) Alignment tableau for Base and Ri@k

B + Ri@k

FtBin Ri@k-R(ϕ) Base-L(ϕ) Ri@k-L(ϕ) Base-R(ϕ)

a. ☞ [(σ.σ)]B [(σ.σ)]R

b.

c.

[(σ)]B [(σ)]B [(σ.σ)]R

[(σ.σ)]B [(σ)]R [(σ)]R

∗!∗

∗!∗

∗!

∗!

∗

∗

∗∗

∗

∗∗

∗

In pejorative reduplication, because the rime [-i@k] is lexically specified, the only parts of

the reduplicant that need to be filled are the onset and tone positions. The onset is copied

faithfully from the base. The tone, however, may mismatch if the base has tone A or C, since

these are phonotactically incompatible with the coda [-k]. Thus, the tone of the reduplicant

must be adjusted, and the following ranking emerges,

(25) Phono, Fill ≫ Max-IO ≫ Max-BR.

Since there is no markedness constraint between Max-IO and Max-BR, the most opti-

mal output will be the one that preserves the segmental content of the input and only changes

the tone contour if necessary due to phonotactic restrictions. Novel data also suggest that

the reduplicant tends to adopt the tone register of the base. In such cases, a markedness

constraint like *L, which bans low tones, must dominate Max-BR in order for B1 to surface

when the base has a low tone.

(26)

*L: Low tones are prohibited. (Bickmore, 2000)

Thus, the revised ranking is as follows,

64

(27) Fill, Max-IO ≫ *L ≫ Max-BR

The ranking in (27) only emerges when the base has a low tone. However, this ranking

will be in conflict with the ranking in (25) because there cannot be two conflicting rankings

for one morpheme. It follows then the ranking between *L and Max-BR is variable; there

is no strict order between them, allowing either constraint to dominate the other in different

bases. However, this is not within the scope of the analysis, so I will leave this for future

research. In the meantime, the no-alternation case is assumed for this analysis. The full

ranking is presented below.

(28) Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO, Align-R(Ri@k,

ϕ), Align-L(Base, ϕ) ≫ Align-L(Ri@k, ϕ), Align-R(Base, ϕ), Max-BR

The ranking works as follows. Consider the input /lanA2/ ‘mild’ in the tableau below.

Candidate (c) and (d) are eliminated because each violates Fill and Max-IO. Between

candidates (a) and (b), candidate (a) is the optimal output because it has less violation in

Max-BR than candidate (b) does.

(29)

Input: /lanA2/ ‘mild’

lanA2 + Ri@k

Fill Max-IO Max-BR

a. ☞ lanA2.li@kB2

b.

c.

d.

lanA2.li@kB1

lanA2.lanA2

∗!∗∗

lanA2.Ri@k

∗!

∗∗∗

∗∗∗

∗∗∗∗!

5.6

Intensification

Recall from Section 3.4 that the reduplicant in the [-a:] reduplication always has a [-

a:] rhyme and a flat tone with its register copied from that of the base. Similar to the

pejorative process, this also avoids inputs that have the same [-a:] rhyme as the reduplicant

65

does. Regarding the alignment of the base and reduplicant, [-a:] reduplication has a prefixal

reduplicant similar to that in the attenuation process. One noticeable difference is that since

there is separability effect observed in the reduplicative, each of the base and the reduplicant

is a Prwd and hence also a Ft. As a result, the whole reduplicative is a phonological phrase,

mirroring the prosodic structure of the reduplicative from the pejorative process, illustrated

below.

Polysyllabic base X.Y (19):

ϕ

→

ϕ

ω

ω

ω

ω

Ft

Ft

Ft

Ft

σ

σ

σσ

σσ

This results in the alignment ranking in (30). It is similar to that of the pejorative

reduplication as both share a similar prosodic domain governed by the same alignment

constraints. The only difference is that in this case, the left-edge Align for the reduplicant

and right-edge counterpart for the base dominate the right-edge Align of the reduplicant and

left-edge Align of the base, which makes the reduplicant a prefix, similar to the reduplicant

of the attenuation process.

(30) FtBin, Align-L(Ra:, ϕ), Align-R(Base, ϕ) ≫ Align-R(Ri@k, ϕ), Align-L(Base,

ϕ)

Regarding the phonological material of the reduplicant, there are two possible cases to

consider: /R-a:/ or /R-a:Aα/. In the first case, the tone in the reduplicant is phonologically

derived. The possible ranking is Max-IO ≫ NoContour ≫ Max-BR.

The markedness constraint NoContour motivates the choice of an A tone in the redu-

plicant, while Max-BR ensures that the tone register is preserved. This latter ranking re-

sembles that of the TETU model, which as a result inferring that the flat tone is unmarked.

66

This is also parallel to the partial specification in the reduplicant of the attenuation and

pejorative processes.

In the second case, the reduplicant is specified with both the rhyme and the tone contour

[-a:A]; the tone register is derived phonologically. Here, a simple ranking of Max-IO ≫ Max-

BR is sufficient to derive the correct output, selecting the tone register α that matches the

base. This is similar to the ranking for full specification of the reduplicant in the attenuation

and pejorative reduplication in (16) and (25) respectively.

Comparing the two options, the /R-a:/ case requires the addition of the markedness con-

straint NoContour, which introduces additional candidates and complexity. In contrast,

the /R-a:Aα/ case relies on a simpler ranking that only necessitates tone register iden-

tity and involves fewer competing candidates. More importantly, the former ranking with

NoContour would create a direct conflict with the rankings established for full reduplica-

tion processes (attenuation, pejorative, and later, AABB) within the same prosodic domain,

as these processes do not require a flat tone in the reduplicant. Since a unified analysis

should avoid contradictory constraint rankings within the same language system, this anal-

ysis adopts the /R-a:Aα/ case, which maintains consistency with the constraint hierarchy

established for other reduplicative processes in Vietnamese.

In the interest of space, I will not provide example tableaux for the constraint ranking

of the intensification process in (31). However, since the rhyme and the tone contour are

specified in the input, the reduplicant only needs to copy the onset and the tone register of

its base. One thing to note is that, given the form X1Y1X2Y2, the specification of [-a:Aα] is

for the second syllable of the reduplicant, Y1; X1 remains a total copy of X2.

(31) Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO, Align-L(Ra:,

ϕ), Align-R(Base, ϕ) ≫ Align-R(Ri@k, ϕ), Align-L(Base, ϕ), Max-BR

5.7 Exclamation (AABB)

The alignment analysis for both the base and the reduplicant, as well as the positioning

of the reduplicant within the output of the exclamation process, follows the work of Lin

67

(2015) and Hsiao (2018). Both of these works analyze reduplication processes in Chengdu

and Taiwanese, respectively, that involve full reduplication of the form AABB, which is

structurally similar to the reduplication seen in this exclamation process.

This type of reduplication involves total copying, meaning the reduplicant is a faithful

reproduction of the base. Consequently, the phonological exponent of the reduplicant does

not pose significant challenges, as long as the faithfulness constraints Max-IO and Max-BR

dominate all markedness constraints, except those necessary for phonotactics and alignment.

Thus, the ranking will be similar to that of full reduplication and specification in attenuation,

pejorative, and intensification.

Now, let us turn to alignment. Two constraints are adopted from Hsiao (2018) that govern

the positioning of the base within the prosodic word.3 First, Align-L(Base, ω) ensures that

the left edge of the base aligns with the left edge of the prosodic word. Second, Align-

R(Base, ω) requires that the right edge of the base coincide with the right edge of the

PrWd. Similarly, alignment constraints for the reduplicant are posited: Align-L(Red, ω)

and Align-R(Red, ω) require that the reduplicant align with the prosodic word at its

respective edges (Hsiao, 2018).

To further capture the mapping between the reduplicant and the base, I also include con-

tiguity constraints. Following Lin (2015) and Hsiao (2018), ContigBase and ContigRed

are adopted, as defined below,

(32) ContigBase: Let base = s1s2. . .sn and ω = w1w2. . .wm

Assign one violation mark (*) for every wy if there’s no sx where sx R wy (Golston,

1996).

(33) ContigRed: Let Red = r1r2. . .rn and ω = w1w2. . .wm

Assign one violation mark (*) for every wy if there’s no rx where rx R wy (McCarthy

& Prince, 1995).

3Hsiao (2018) uses stem instead of base in the constraint; for the purpose of the analysis, I change all

mentions of “stem” to “base” in the alignment constraints here.

68

In both attenuation and pejorative reduplication, the base and reduplicant appear to

form a contiguous string. However, in the current process, since we do not yet know how the

reduplicant is arranged with respect to the base, there is the possibility that contiguity is

disrupted, as is potentially the case in candidate outputs like (1), (3), and (6).

This uncertainty arises because we cannot definitively determine which syllables are part

of the base and which are part of the reduplicant. Thus, there are four possible arrange-

ments of the reduplicant within the reduplicative, each corresponding to a distinct ranking

of alignment and contiguity constraints:

1. X.X.Y.Y

Align-L(Base, ω), Align-R(Base, ω), ContigRed ≫ Align-L(Red, ω), Align-

R(Red, ω), ContigBase

The reduplicant X.Y infixes between the base X.Y.

2. X.X.Y.Y

Align-L(Red, ω), Align-R(Red, ω), ContigBase ≫ Align-L(Base, ω), Align-

R(Base, ω), ContigRed

The reduplicant is split into two, the 1st half X is a prefix, and the 2nd half Y is a

suffix.

3. X.X.Y.Y

Align-L(Base, ω), Align-R(Red, ω) ≫ Align-R(Base, ω), Align-L(Red, ω),

ContigBase, ContigRed

Both base and reduplicant are split — the 1st half X of reduplicant follows its base X,

and the 2nd half Y of the reduplicant follows its base Y.

4. X.X.Y.Y

Align-R(Base, ω), Align-L(Red, ω) ≫ Align-L(Base, ω), Align-R(Red, ω),

ContigBase, ContigRed

Both base and reduplicant are split — the 1st half X of reduplicant precedes its base X,

and the 2nd half Y of the reduplicant precedes its base Y.

69

In all of the possible rankings, two contiguity constraints ContigBase and ContigRed

are always dominated because either the reduplicant is split, the base is split, or both are

split in the possible arrangements of B and R. For the purpose of this analysis, I assume that

(4) is the optimal arrangement as there is no empirical evidence at this point to definitively

identify which element is the reduplicant in AABB forms, since both copies are segmentally

identical. This configuration reflects full reduplication of each syllable in the base, consistent

with the general Vietnamese pattern where the reduplicant precedes the base (cf. attenuation

processes and Hoang (1985, 1998) and Cao (2007)).

Accordingly, the complete ranking for the AABB reduplication pattern is as follows,

(34) Phono, Fill, Align-R(Base, ω), Align-L(Red, ω) ≫ Align-L(Base, ω), Align-

R(Red, ω), ContigBase, ContigRed ≫ Max-IO, Max-BR

In this ranking, each reduplicant is a faithful copy of its base, provided that the optimal

output satisfies (1) the undominated alignment constraints, (2) the requirement to fill in

phonological material (e.g., onset, tone contour, and tone register), (3) faithfulness to the

segmental content of the input, and (4) base-reduplicant identity.

The addition of the constraints ContigBase and ContigRed to the ranking creates

an important distinction between AABB reduplication and previously discussed processes.

In AABB forms, these contiguity constraints must be dominated by alignment constraints,

allowing the base or reduplicant to be split. In contrast, for all previously discussed pro-

cesses (attenuation, pejorative, and intensification), both the base and reduplicant always

form contiguous strings. Therefore, in those cases, ContigBase and ContigRed can be

undominated without affecting the outputs, as shown in the revised ranking (35). This rank-

ing difference captures the structural distinction between AABB reduplication and other

reduplication types, while maintaining consistency in the overall constraint ranking.

(35) Revised (22), with alignment and contiguity constraints added

Phono, Fill, Align-L(Red, ω), Align-R(Ft, H(Ft)), Align-R(Base, ω), Pars-

eSyll, FtBin, Max-IO, ContigBase, ContigRed ≫ Max-BR, Align-R(Red,

70

ω), Align-L(Base, ω)

5.8 Triplication

Recall from Section 3.5 that Vietnamese triplication typically follows specific patterns

based on the tone of the base syllable. The general pattern is Base.RA2.RA1, B2

4, with the

base normally being the first syllable (P.-P. Nguyen, 1998; Hoang et al., 2008). However,

several systematic variations occur:

1. When the base has tone B1: The pattern is Base.RA2.RB2, where RB2 has tone B2.

2. When the base has tone A1: The pattern is RC1.RA2.Base, where RC1 has tone C1

(note the reversed order).

3. For all other valid base tones: The pattern is Base.RA2.RA1, where RA1 has tone A1.

Importantly, this process does not permit bases with tone A2. There are two distinct

problems regarding triplication — formation and allomorphy selection, which were discussed

in Section 3.5. They can be subsumed under one analysis: allomorphy selection.

5.8.1 Two separate morphemes

Another hypothesis that has been overlooked is the possibility of having two separate mor-

phemes that make up triplication. Consequently, this makes triplication different from simply

multiple applications of one reduplication process. If this is the case, the tone and avoidance

patterns discussed in Section 3.5 can be formally captured. Crucially, then, Vietnamese trip-

lication is a morphological, such that two distinct reduplicative morphemes are concatenated,

each with its own lexical entry. Phonology’s role is strictly enforcement: high-ranking Align

and faithfulness constraints keep each chosen allomorph at its lexically specified edge and

tone, but do not decide those properties themselves.

The main evidence for two separate reduplicants is the existence of trisyllabic reduplica-

tives with corresponding disyllabic forms. Recall (5), repeated below. There are two main

differences between (36a) and (36b): (1) their meaning, in which (36a) conveys an emphatic

4RA1, B2 = reduplicant in triplication with A1 or B2

71

meaning of (36b), and (2) the addition of [sa:NA2] from (36b) to (36a). I suggest that this

[sa:NA2] element is a reduplicant with a fixed A2 that contributes the intensification. The

reason this reduplicant consistently bears an A2 tone is due to it being specified in the lexi-

con. This fixed-tone behavior is attested in Vietnamese, as seen in other examples in Hoang

et al. (2008, pp. 78 & 90).

(36) Repeating (5)

a.

sa:NA2
[sa:kB1
Red
close.to
‘very close to (emph)’

sa:kB2]
Red

b.

sa:kB2]
[sa:kB1
Red
close.to
‘very close to’

Having established that the intensification effect arises from the addition of a fixed-A2

reduplicant, consider a similar pattern in (27), repeated below. Unlike the previous example,

the disyllabic form in (37a) is illicit, and only the trisyllabic form in (37b) yields the intended

intensification. This implies that in such cases, two separate reduplicants are at work: one is

the fixed-A2 reduplicant RA2 from earlier examples, and the other is an abstract reduplicant

Red.5

(37)

a.

* [n7wC1
Red

n7wA1]
brown
* [n7wA1
n7wC1]
Red
brown
Intended : ‘very brown’

b.

n7wA2
[n7wC1
Red
RA2
‘very brown’

n7wA1]
brown

(Novel data)

If there are indeed two reduplicants at play, why do we not observe forms where the base

combines with RA2 alone, or with RA1, B2, C1 alone? A possible explanation is that RA2 can

only attach to a specific type of base, one that already includes a reduplicative morpheme, as

seen in (37b), (36a), and (28a)–(b). In this view, RA2 requires a base that already possesses

a reduplicative feature, which RA1, B2, C1 supplies as in (37b) and (36a), or a reduplicant

supplies as in (28a)–(b).

This selection restriction aligns with Fabb (1988), who shows that suffixes often have

constraints on which bases or affixes they can attach to. For instance, in English, the suffix

5The form is irrelevant for now but will be discussed later.

72

-ary selects the suffix -ion to create a compound suffix -ion-ary that will then select a word

to attach to. Since -ary is bounded by -ion, there are no words in English that has only

-ary (e.g., revolution and revolutionary but *revolu-nary). A consequence of this is that the

selectional properties of -ion will transfer to -ion-ary and hence both apply to the same type

of words (i.e., verbs) (Fabb, 1988). Coming back to Vietnamese, we do not observe forms

where the base combines with RA1, B2, C1 alone because it is bound by RA2. In return, RA2

combines with the base satisfies the selectional property of RA1, B2, C1 that it only selects

bases that are disyllabic and already has a reduplicative morpheme.

In conclusion, triplication is not merely the result of iterative reduplication applied to a

monosyllabic base. Instead, I suggest that there are two distinct structural patterns: (1) a

monosyllabic base accompanied by two distinct reduplicants — RA1, B2, C1 and RA2— and

(2) a disyllabic lexicalized base combined with the intensifying morpheme RA2. In the first

pattern, intensification results from the coordinated contribution of both morphemes. In the

second, the lexicalized reduplicative already satisfies the reduplicative condition required by

RA2.

The idea that triplication involves two separate morphemes, rather than repeated appli-

cations of the same reduplication process, opens the door to a deeper morphological analysis.

In particular, it raises the question of how these two morphemes interact with the base and

with each other phonologically and morphologically. One key consequence of this hypothe-

sis is that each morpheme may have its own set of allomorphs, with distinct phonological

properties and selectional restrictions. This observation helps explain patterns of identity

avoidance and tonal alternation that are otherwise difficult to capture. The next subsection

builds on this hypothesis by formalizing the role of allomorphy in shaping the surface forms

of triplication, particularly how RA2 and RA1, B2, C1 select their forms depending on the base

tone and structure.

73

5.8.2 Allomorphy selection

The pattern of identity avoidancce and tone harmony in triplication can be explained

through the lens of allomorphy selection. Each of the two reduplicant morphemes, RA2 and

RA1, B2, C1, has its own allomorphs and selectional restrictions. These restrictions determine

which combinations of base and reduplicant are permitted.

Specifically, I propose that the intensifying morpheme RA2 has two allomorphs: a prefixal

allomorph that attaches to A1 bases, and a suffixal allomorph that occurs with all other

bases. Prefix-suffix allomorphy in a single morpheme is not unusual, as seen in languages

such as Nakanai (an Austronesian language in Papua New Guinea)’s nominalizer and Biak

(an Austronesian language in Indonesia)’s third-person subject marking (Kalin, 2022). For

each allomorph, there is a distinct Align that they follow, presented below. In other words,

morphology chooses the appropriate allomorph (prefix versus suffix), and once that choice

is made, the phonological Align constraints place it at the correct edge.

Similar to other reduplication processes, these alignment constraints will be dominant.

(38)

a. RA2 prefix: Align-L(RA2, ω), Align-R(Base, ω)

b. RA2 suffix: Align-R(RA2, ω), Align-L(Base, ω)

For ineffable A2 bases, RA2 selects the base of the structure base + reduplicant. However,

it would create an identity relationship where both the RA2 syllable and the base syllable have

A2 tone. This identity relationship is systematically avoided in the language (see Section 3.5

for the avoidance pattern in triplication).

Consequently, each allomorph of RA2 leads to different placements of RA1, B2, C1. For RA2

prefix, only RC1 associates with it, in which this RA1, B2, C1 is also a prefix as a consequence.

However, between RA2 and RC1, RA2 is closer to the base than RC1 is. Thus, RC1 is the

leftmost syllable in a trisyllabic reduplicative, and the base is the rightmost one; the Align

that governs the placement of RC1 (i.e., (39)) will dominate that of RA2, as shown in (40)

with an accompanied tableau.

(39) RC1 prefix: Align-L(RC1, ω), Align-R(Base, ω)

74

(40) Align-R(Base, ω), Align-L(RC1, ω) ≫ Align-L(RA2, ω)

Base + RA2 + RC1 Base-R(ω) RC1-L(ω) RA2-L(ω)

a. ☞ C1.A2.Base

b.

c.

d.

Base.A2.C1

∗!∗

A2.C1.Base

A2.Base.C1

∗!

∗!∗

∗!

∗!∗

∗

∗

Similar argument can be made for RA1 and RB2, such that RA2 suffix leads to having

RA1 and RB2 suffixes. Since RA2 is suffixal with regard to the base, RA1 and RB2 are the

rightmost syllable in the word, while the base being the leftmost. The Align that governs

the placement of RA1 and RB2 (i.e., (41a)–(b)) will dominate that of RA2, as shown in (42)

with an accompanied tableau.6

(41)

a. RA1 suffix: Align-R(RA1, ω), Align-L(Base, ω)

b. RB2 suffix: Align-R(RB2, ω), Align-L(Base, ω)

(42) Align-L(Base, ω), Align-R(RA1, ω), Align-R(RB2, ω) ≫ Align-L(RA2, ω)

Base + RA2 + RA1 RA1-R(ω) Base-L(ω) RA2-R(ω)

a. ☞ Base.A2.A1

b.

c.

d.

A1.A2.Base

Base.A1.A2

A1.Base.A2

∗!∗

∗!

∗!∗

∗!∗

∗!

∗

∗

Regarding RA1, B2, C1, it alternates between A1, B2, and C1. For these varying tones, the

tone specification in the reduplicant can fall into three levels: (1) no specification (phonolog-

6RA1 is used in the tableau, but the alignment of RB2 should also be the same.

75

ically derived), (2) partial specification (e.g., only contour or register is fixed), and (3) full

specification (both tone and segments fixed, forming an allomorph).

5.8.2.1 A1

When there is no specification, the tone is fully derived in the phonology. This derivation

can follow two possible rankings:

(43) Max-IO ≫ *L/Ocp (T), NoContour ≫ Max-BR

Both ranking use NoContour for the flat tone to emerge. The first ranking uses *L

to penalize low tones (i.e., −H tone) and allows a high tone, in this case A1, to emerge

(Bickmore, 2000). Another way to approach this is to have Ocp (T) instead of *L. This

constraint avoids adjacent identical tones, ruling out A2 due to identity with the tone RA2.

Ocp (T) is based on the Ocp constraint proposed by Golston (1996), which militates against

adjacent identical elements. This constraint, in turn, is derived from the Obligatory Contour

Principle (OCP) introduced in Leben (1973), which prohibits adjacent identical tones.

When there is partial specification of the contour (denoted Aα), the ranking still involves

either *L or Ocp (T). The NoContour constraint is not needed because the flat contour

is already specified in the input.

(44) Max-IO ≫ *L/Ocp (T) ≫ Max-BR

When the tone is partially specified for register only (e.g., T1 for +H), there is only one

approach, using NoContour ensures a flat tone surfaces. There is no need for *L nor Ocp

because if the reduplicant is already specified with a H register, it inherently satisfies *L and

Ocp.

(45) Max-IO ≫ NoContour ≫ Max-BR

When fully specified as A1, Max-IO simply dominates Max-BR to enforce complete

faithfulness to the input, making the tone part of the reduplicant’s lexical representation.

Thus, the result is a constraint ranking similar to that of full reduplication in the attenuation

process (i.e., (16))

76

(46) Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO ≫ Max-BR

Among these three patterns — no specification, partial specification, and full specifica-

tion — the full specification approach offers the most theoretically consistent analysis. While

both the no specification and partial specification approaches require additional markedness

constraints like *L, Ocp (T), or NoContour to derive the correct tone patterns, the full

specification approach relies simply on Max-IO dominating Max-BR (46). Crucially, this

ranking parallels the constraint rankings already established for attenuation, pejorative, in-

tensification, and AABB reduplication within the language. Adopting the full specification

analysis therefore maintains a unified constraint ranking across all reduplicative processes

in the same prosodic domain. In contrast, the other approaches would introduce ranking

paradoxes where certain markedness constraints must sometimes be dominated (in full redu-

plication) and sometimes dominate (when phonologically deriving the reduplicant’s tone).

By treating the A1 tone as lexically specified in the reduplicant, we eliminate these conflicts

and provide a more coherent overall analysis of Vietnamese prosodic morphology.

5.8.2.2 B2

For B2, a similar set of derivations can be proposed. With no specification, there are two

possible rankings account for the surface form.

(47) Max-IO ≫ Ocp (T)/*H ≫ Max-BR

In the first ranking, adjacent identical tones are avoided. However, for Ocp (T) to work

here, we have to change its scope from identical tones in adjacent syllables in a prosodic

word to those in any pairing of syllables in the prosodic word. For instance, if Ocp works

for first syllable-second syllable and second syllable-third syllable in a linear order, it should

now be applicable to first syllable-third syllable as well. Note that the change of scope in

Ocp (T) does not affect the ranking for A1 in (43) because the base and the reduplicant in

question never have a matching tone.

A second account appeals to the markedness constraint *H, which penalizes high tones

77

(Bickmore, 2000). The base has B1, and with Max-IO ≫ *H, it remains unaltered. In the

case of B2, its base already has B1, meaning *H will always be violated at most once; if

the reduplicant has B1 instead of B2, it will have an additional violation. Although A2

is a possible outcome under this ranking, tone-coda co-occurrence constraints make it less

optimal, as A2 would require a homorganic nasal coda in the reduplicant, further violating

Max-BR.

In partial specification cases with only a contour (Bα), register assignment follows the

same logic as in the no-specification scenario and hence has the same constraint ranking.

When only register is specified (−H), surface forms like A2, B2, and C2 are possible. However,

phonotactic constraints again rule out A2 and C2 due to required coda alternations. Since

B2 can surface without phonotactic violations, it incurs fewer Max-BR violations, making

it the optimal form. Fully specified B2 reduplicants are protected by the high-ranking Max-

IO constraint. This means that both latter specifications have the same constraint ranking

as that of full reduplication in attenuation and AABB.

(48) Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO ≫ Max-BR

As with the A1 reduplicant analysis, the full specification approach for B2 offers consid-

erable theoretical advantages. While the no-specification approach requires either extending

the scope of Ocp (T) to non-adjacent syllables or introducing *H constraint, and the partial

specification approach involves complexities with register assignment and phonotactic in-

teractions, the full specification approach simply relies on the established ranking Max-IO

≫ Max-BR (48). This ranking aligns with the constraint rankings already established for

attenuation and AABB reduplication processes. The consistency of this pattern across dif-

ferent reduplicative processes suggests that treating B2 as fully specified in the reduplicant’s

lexical representation may be the most parsimonious approach.

5.8.2.3 C1

For C1, both full-specified and contour-specified reduplicants share the same constraint

ranking: Max-IO ≫ Max-BR. In full specification, the argument mirrors that of A1 and

78

B2. In the contour-specified case, C2 is a competitor, but it deletes not only the base’s

contour but also the base’s register in the reduplicant, violating Max-BR twice. Hence, C1

is preferred.

When tones are unspecified or only register-specified, the following ranking accounts for

the outcome.

(49) Max-IO ≫ Ocp (T), Ident-BR ([αCg]) ≫ Max-BR

Ocp (T) eliminates repeated tones, leaving only A1 and A2 as possibilities. B2 and

C2 are excluded because they mismatch the base in both register and contour. To further

ensure that C1 is chosen over B1 (which shares the same contour), we can introduce the

ad-hoc constraint Ident-BR ([αCg]), which favors base-reduplicant pairs that agree in the

α value of the [Constricted Glottis] feature. Since A1 and C1 lack [Cg], i.e., [−Cg], while

B1 possesses it (Brunelle, 2009), this constraint favors C1.

The ranking works as follows, as exemplified by the input /n7wA1/ ‘brown’ with an

unspecified reduplicant Red. The optimal output is candidate (a) with C1 in the reduplicant,

which deletes the base’s tone contour in the reduplicant. Candidate (b) also violates Max-

BR as much as the optimal output does, is eliminated since it violates other markedness

constraints that dominate Max-BR. Candidate (c) with the identical base and reduplicant

is eliminated because they have a pair of identical tones in the output. Candidate (d) is

eliminated for violating the dominant Max-IO.

(50)

Input: /n7wA1/ ‘brown’

79

Red + RA2 + n7wA1

Max-IO Ocp (T) Ident-BR ([αCg]) Max-BR

a. ☞ n7wC1.n7wA2.n7wA1

b.

c.

d.

n7wB1.n7wA2.n7wA1

∗!

n7wA1.n7wA2.n7wA1

∗!

n7wC1.n7wA2.n7wC1

∗!

∗

∗

The preceding analyses of the three tonal allomorphs (RA1, RB2, and RC1) show that the

selectional restrictions on RA1, B2, C1 demonstrate that morphology consults phonology, i.e.,

the tone of the base, when the allomorph chooses the appropriate base. The high-ranking

Max-IO and Align constraints show that phonology cannot freely alter a lexically specified

allomorph once selected.

The relationship between morphology and phonology reveals a noticeable ranking para-

dox, which lies in Ocp. Ocp operates across the tone alternations in triplication, and there is

no separability effect in triplication, meaning the trisyllabic form behaves as a single prosodic

word (PrWd). This means that Ocp is operating in the domain of PrWd. Compare to AABB

and the AB base in that process (see Section 3.2), there is separability effect which indicates

that A and B in AB and AA, BB in AABB form are all separate PrWds. This raises a

paradox: if Ocp is active within PrWds in triplication, it should also be active in AABB

reduplication, where the domain is similarly defined. However, identical adjacent syllables

are allowed in AABB, contradicting Ocp’s supposed influence. This paradox implies a rank-

ing conflict, which is Ocp dominates Max-BR in triplication but is dominated in AABB.

These ranking contradictions highlight the inadequacy of underspecified representations for

reduplicants.

In addition to resolving internal ranking issues, the analysis of these three reduplicant

patterns reveals a significant advantage of the full specification approach over alternatives.

Both no-specification and partial-specification analyses require introducing markedness con-

80

straints that would create ranking paradoxes when considered alongside other reduplication

processes in Vietnamese. These constraints would need to dominate Max-BR in triplication

to derive the correct tonal patterns, yet must be dominated by Max-BR in processes like full

reduplication. The full specification approach avoids these ranking conflicts entirely by treat-

ing each reduplicant tone (A1, B2, and C2) as lexically specified allomorphs of the RA1, B2, C1

morpheme. This approach maintains a uniform constraint ranking across all reduplicative

processes within the same prosodic domain: Max-IO consistently dominates Max-BR.

Thus, understanding triplication as involving tone-specific allomorphs rather than phono-

logical derivation provides the most coherent analysis of Vietnamese tonal reduplication as

a whole.

This allomorphy-based analysis not only avoids ranking paradoxes but also aligns with

typological observations: true triplication is rare cross-linguistically and often reducible to

multiple affixation or allomorph selection. Vietnamese triplication, then, exemplifies a mor-

phological strategy encoded in the lexicon rather than a product of recursive reduplication.

The desired ranking for each allomorph is presented below. One thing to note is that each

allomorph is subject to their own set of alignment constraints; phonological-material wise,

it is lexically specified.

(51) A1:

Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO, Align-L(Base,

ω), Align-R(RA1, ω), Align-R(RB2, ω), ContigBase, ContigRed, ≫ Align-

L(RA2, ω), Max-BR

(52) B2:

Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO, Align-L(Base,

ω), Align-R(RA1, ω), Align-R(RB2, ω), ContigBase, ContigRed ≫ Align-

L(RA2, ω), Max-BR

81

(53) C1:

Phono, Fill, Align-R(Ft, H(Ft)), ParseSyll, FtBin, Max-IO, Align-R(Base,

ω), Align-L(RC1, ω), ContigBase, ContigRed ≫ Align-L(RA2, ω), Max-BR

Triplication may be better understood not as a base repeating twice but as two redu-

plicative allomorphs attached to a single base. If it was base being copied n times, the form

of the reduplicant is phonologically derived. However, under an allomorphy-based analysis,

the number of possible reduplicants is inherently limited by the inventory of available allo-

morphs in the lexicon, each associated with a specific morphological or semantic function.

This allomorphy-based approach accounts for tone alternations, identity avoidance, and the

distribution of reduplicants in triplication, tying together the phonological and morphological

aspects of Vietnamese reduplication.

5.9

Interim summary

This chapter has provided an OT analysis of several reduplication patterns in Vietnamese,

each serving distinct functions: attenuation, intensification of polysyllabic adjectives ([-a:]

reduplication), exclamation (AABB), and intensification of monosyllabic adjectives (tripli-

cation). While each pattern could be analyzed in isolation and as a phonological deriva-

tion through constraint interactions, doing so risks positing multiple independent constraint

rankings within a single language, which is theoretically undesirable. Processes involving full

reduplication, like AABB, should be analyzed in tandem with their tone patterns, as tone

provides critical evidence for how these forms are derived.

I suggest a unified ranking across all reduplication types by adopting an allomorphy

approach. Each reduplicant is treated as having a set of lexically specified allomorphs that

select the base or base-reduplicant combination they attach to (Fabb, 1988) and hence is

not phonologically derived. This perspective helps avoid complications that arise under a

reduplicative analysis, especially those involving tone identity or alternation that cannot be

easily explained if we assume multiple rounds of phonological copying.

While I draw on tools from OT to formalize the constraints and interactions that govern

82

these patterns, my argument is not theory-internal to OT. Instead, OT serves here as a

framework to express broader generalizations about morphophonological behavior, particu-

larly the role of tone in morphophonological processes like reduplication and allomorphy.

83

CHAPTER 6

CONCLUSION

This thesis has developed a unified, OT-based analysis of Vietnamese reduplication, focusing

on previously under-analyzed data from Southern Vietnamese. The analysis offers an account

of various reduplicative patterns — from well-documented disyllabic forms to complex cases

such as triplication and AABB — while addressing tone behavior and avoidance effects across

processes.

Three main contributions have emerged from this research. First, this thesis provides

substantial new empirical data on previously under-documented reduplication patterns, no-

tably triplication and AABB. This novel data, gathered through rigorous fieldwork involving

native speakers, fills an important gap in Vietnamese linguistic documentation and provides

the empirical foundation necessary for robust theoretical analysis.

Second, this thesis proposes a formalization of Vietnamese reduplication under OT, ar-

guing that the so-called triplication in Vietnamese emerges not from repeated reduplication

but from attaching two distinct reduplicative morphemes, each possessing its own set of

allomorphs. In other words, Vietnamese triplication is morphological, such that (1) mor-

phology chooses the allomorph, which is triggered by phonological properties of the base

(especially tone), and (2) phonology merely enforces the choice through high-ranking Align

and Max-IO keep the selected allomorph at its lexically stored edge and preserve its tone.

As a whole, Vietnamese reduplication is best understood through the selection of lexically

specified reduplicant morphemes rather than phonological derivation. This perspective offers

a unified explanation for tone harmony, avoidance effects, and prosodic alignment across

various reduplication processes.

Third, avoidance patterns are offered a description, and they can be understood as a

consequence of the allomorph selecting a particular base to attach to, which may resemble

identity avoidance on the surface. The analysis utilizes the OT framework to formalize the

argument; however, it is not dependent on the theory.

84

This analysis opens several directions for future research. Particularly, a more robust

morphophonological account may be needed. In verb attenuation, the lack of observed allo-

morphy, in contrast to the adjectival case, suggests that distinct morphemes may underlie the

reduplication of verbs and adjectives. The use of a placeholder constraint Phono to simplify

Vietnamese phonotactics invites a more detailed exploration of tone–coda co-occurrence re-

strictions. The preference for B1 reduplicants in pejorative constructions, even with non-B1

bases, may reflect a tonal markedness pattern and warrants empirical testing. The sepa-

rability effect and internal movement of reduplicants in AABB reduplication suggest the

possibility of two distinct reduplication processes at work. If this is the case, it could help

explain both the separability effect observed in AABB reduplicatives and the apparent in-

ternal movement of reduplicants. Importantly, this movement cannot simply be attributed

to compound structure, as compounds block internal movement, ruling out any analysis in

which a portion of the compound (i.e., the reduplicant) is moved. Finally, the separability

seen in AABB, pejorative, and intensification constructions remains insufficiently addressed,

especially with respect to stress placement and prosodic structure, and presents an oppor-

tunity for more detailed analysis.

85

BIBLIOGRAPHY

Agbayani, B. (1997). Phonological Constraints and Base-Reduplicant Non-Identity: The Case

of Vietnamese [Master’s thesis, University of California, Irvine].

Alderete, J., Beckman, J., Benua, L., Gnanadesikan, A., McCarthy, J., & Urbanczyk, S.
(1999). Reduplication with Fixed Segmentism. Linguistic Inquiry, 30 (3), 327–364.
https://doi.org/10.1162/002438999554101

An, Y.-r. (2012). The gradient OCP: Evidence from Korean reduplication. In Rebeka
Campos-Astorkiza & Jon Franco (Eds.), Papers in linguistics by the BIDE gener-
ation (pp. 33–53, Vol. 46). UPV/EHU Press.

Anttila, A. (2002). Morphologically conditioned phonological alternations. Natural Language
and Linguistic Theory, 20 (1), 1–42. https://doi.org/10.1023/A:1014245408622

Arto Anttila. (1997). Deriving Variation from Grammar. In Frans L. Hinskens, Roeland van
Hout, & W. Leo Wetzels (Eds.), Current Issues in Linguistic Theory (pp. 35–68,
Vol. 146). John Benjamins Publishing Company. https://roa.rutgers.edu/files/63-0
000/63-0000-ANTTILA-0-0.PDF

Bickmore, L. S. (2000). Downstep and fusion in Namwanga. Phonology, 17 (3), 297–331.

https://doi.org/10.1017/S0952675701003943

Brunelle, M. (2009). Tone perception in Northern and Southern Vietnamese. Journal of

Phonetics, 37 (1), 79–96. https://doi.org/10.1016/j.wocn.2008.09.003

Brunelle, M. (2017). Stress and phrasal prominence in tone languages: The case of Southern
Vietnamese [Edition: 2017/02/16 Publisher: Cambridge University Press]. Journal of
the International Phonetic Association, 47 (3), 283–320. https://doi.org/10.1017/S0
025100316000402

Cao, X.-H. (2003). Tiếng Việt: Mấy vấn đề ngữ âm, ngữ pháp, ngữ nghĩa (Vietnamese: A few
phonetic, grammatical and semantic problems. The Ministry of Education; Training.

Cao, X.-H. (2007). Ngữ âm tiếng Việt (Vietnamese Phonetics). VNU.

Carleton, T., & Myers, S. (1994). On tonal transfer. https://roa.rutgers.edu/files/16-0594/

16-0594-CARLETON-0-0.PDF

Chen, M. Y. (2000). Tone sandhi: Patterns across Chinese dialects (Vol. 92). Cambridge

University Press.

Clements, G. N., & Hume, E. V. (1995). The internal organization of speech sounds. In J. A.

Goldsmith (Ed.), The Handbook of Phonological Theory (pp. 245–306). Blackwell.

86

Corbin, D., & Plénat, M. (1992). NOTE SUR L’HAPLOLOGIE DES MOTS CONSTRUITS
[Publisher: Armand Colin]. Langue Fran¸caise, (96), 101–112. http://www.jstor.org/
stable/41558606

Cu, D.-T., Hoang, V.-T., & Nguyen, N.-T. (1977). Ngữ âm học tiếng Việt hiện đại (Phonology

of the modern Vietnamese). The Ministry of Education; Training.

de Lacy, P. (1999). Morphological haplology and correspondence. In University of Mas-
sachusetts Occasional Papers: Papers from the 25th Reunion (pp. 51–88, Vol. 24).
GLSA Publications. https://roa.rutgers.edu/files/298-0299/roa-298-lacy-1.pdf

Diep, Q.-B., & Hoang, V.-T. (1996). Ngữ pháp tiếng Việt (Vietnamese Grammar) (Vol. 1).

The Ministry of Education; Training.

Fabb, N. (1988). English Suffixation Is Constrained Only by Selectional Restrictions [Pub-
lisher: Springer]. Natural Language & Linguistic Theory, 6 (4), 527–539. http://www.
jstor.org/stable/4047592

Fitzgerald, C. M. (2000). Vowel Hiatus and Faithfulness In Tohono O’Odham Reduplication.
Linguistic Inquiry, 31 (4), 713–722. https://doi.org/10.1162/002438900554532

Golston, C. (1996). Direct Optimality Theory: Representation as Pure Markedness. Lan-

guage, 72 (4), 713–748. https://doi.org/10.2307/416100

Hayes, B. (1989). The prosodic hierarchy in meter. In P. Kiparsky & G. Youmans (Eds.),

Phonetics and Phonology (pp. 201–260, Vol. 1). Academic Press.

Hoang, V.-H. (1985). Từ láy trong Tiếng Việt (Vietnamese Reduplication). Social Sciences

Publishing House.

Hoang, V.-H. (1998). Từ điển từ láy tiếng Việt (Vietnamese Reduplication Dictionary). Social

Sciences Publishing House.

Hoang, V.-H., Ha, Q.-N., & Nguyen, V.-K. (2008). Từ tiếng Việt (Vietnamese word). Social

Sciences Publishing House.

Hsiao, Y. E. (2018). An Analysis Of Taiwan Southern Min Tetrasyllabic Adjective Redupli-
cation. Journal of Chinese Linguistics, 46 (1), 123–148. https://doi.org/10.1353/jcl.
2018.0004

Hyman, L. M. (2014). Coda constraints on tone. In S. Bendjaballah, N. Faust, M. Lahrouchi,
& N. Lampitelli (Eds.), Language Faculty and Beyond (pp. 103–122, Vol. 12). John
Benjamins Publishing Company. https://doi.org/10.1075/lfab.12.09hym

Inkelas, S., Orgun, C. O., & Zoll, C. (1997). The Implications of Lexical Exceptions for

87

the Nature of Grammar. In I. Roca (Ed.), Derivations and Constraints in Phonology
(pp. 393–418). Clarendon Press. https://doi.org/10.1093/oso/9780198236894.003.00
13

Inkelas, S., & Zoll, C. (2005). Reduplication: Doubling in Morphology (1st ed.). Cambridge

University Press. https://doi.org/10.1017/CBO9780511627712

Inkelas, S., & Zoll, C. (2007). Is grammar dependence real? A comparison between cophono-
logical and indexed constraint approaches to morphologically conditioned phonology.
Linguistics, 45 (1). https://doi.org/10.1515/LING.2007.004

Kalin, L. (2022). Infixes really are (underlyingly) prefixes/suffixes: Evidence from allomorphy
on the fine timing of infixation. Language, 98 (4), 641–682. https://doi.org/10.1353/
lan.2022.0017

Kennard, C. H. (2004). Copy but don’t repeat: The conflict of dissimilation and reduplication
in the Tawala durative. Phonology, 21 (3), 303–323. https://doi.org/10.1017/S09526
75704000296

Kennedy, R. (2008). Evidence for Morphoprosodic Alignment in Reduplication. Linguistic

Inquiry, 39 (4), 589–614. https://doi.org/10.1162/ling.2008.39.4.589

Kurisu, k. (2001). The Phonology of Morpheme Realization. https://doi.org/10.7282/T3R4

9PKM

Leben, W. (1973). Suprasegmental Phonology [PhD Thesis]. MIT. http://www.ai.mit.edu/p

rojects/dm/theses/leben73.pdf

Lin, H.-s. (2015). Chengdu Reduplication: An Optimality Theoretic Analysis. Language and

Linguistics, 16 (6), 843–872. https://doi.org/10.1177/1606822X15602613

Lombardi, L. (1997). Coronal sonorants and markedness. In Maryland Working Papers in

Linguistics (pp. 156–175, Vol. 5). University of Maryland.

McCarthy, J. J., & Prince, A. (1994). The emergence of the unmarked: Optimality in prosodic
morphology. In Proceedings of the North East Linguistics Society 24 (Vol. 18). GLSA,
UMASS. https://scholarworks.umass.edu/linguist_faculty_pubs/18

McCarthy, J. J., & Prince, A. (1995). Faithfulness and reduplicative identity. In J. Beckman,
L. W. Dickey, & S. Urbanczyk (Eds.), Papers in Optimality Theory (pp. 249–384,
Vol. 10). GLSA, UMASS. https://scholarworks.umass.edu/linguist_faculty_pubs/10

Michaud, A. (2005). Final consonants and glottalization: New perspectives from Hanoi Viet-

namese. Phonetica, 61 (2-3), 119–146. https://doi.org/10.1159/000082560

88

Ngo, T.-N. (1984). The syllabeme and patterns of word formation in Vietnamese. New York

University.

Nguyen, D.-H. (1997). Vietnamese. John Benjamins.

Nguyen, P.-P. (1998). Các phương diện hình thái học tiếng Việt (Aspects of Vietnamese
Morphology). Hội thảo Quốc tế Việt Nam học (Internal Conference on Vietnamology),
1, 139–150. https://repository.vnu.edu.vn/bitstream/VNU_123/20295/1/018.pdf

Nguyen, T., & Ingram, J. (2007). Acoustic and perceptual cues for compound-phrasal
contrasts in Vietnamese. The Journal of the Acoustical Society of America, 122.
https://doi.org/10.1121/1.2747169

Nguyen, T., & Ingram, J. C. L. (2006). Reduplication and word stress in Vietnamese.

Nguyen, V.-L., & Edmondson, J. A. (1997). Thanh điệu và chất giọng trong tiếng Việt hiện
đại (phương ngữ Bắc Bộ): Khảo sát thực nghiệm (The tones and voice quality in
modern northern Vietnamese: Instrumental case studies). Ngôn ngữ (Language), 1,
1–16.

Orgun, C. O. (1996). Sign-Based Morphology and Phonology: With special attention to Op-
timality Theory [PhD Thesis]. University of California, Berkeley. https://roa.rutger
s.edu/files/171-0197/171-0197-ORGUN-0-0.PDF

Parker, S. (2018). Central vs. Back Vowels. Work Papers of the Summer Institute of Lin-
guistics, University of North Dakota Session, 44 (1). https://doi.org/10.31356/silwp
.vol44.05

Pham, A. H. (2004). Vietnamese Tone: A New Analysis (L. Horn, Ed.; 1st ed.). Routledge.

https://doi.org/10.4324/9780203500088

Pham, A.-H. (2001). Vietnamese tone: Tone is not pitch [Section: University of Toronto
Linguistics PhD students’ Dissertations]. Toronto Working Papers in Linguistics. Re-
trieved October 28, 2024, from https://twpl.library.utoronto.ca/index.php/twpl/arti
cle/view/6507

Pham, A.-H. (2009). The identity of non-identified sounds: Glottal stop, prevocalic /w/ and
triphthongs in Vietnamese [Section: Articles]. Toronto Working Papers in Linguistics,
34. Retrieved October 28, 2024, from https://twpl.library.utoronto.ca/index.php/t
wpl/article/view/6803

Pham, A.-H., & Pham, A.-A. (2020). Productive Reduplication in Southern Vietnamese
[i]. Journal of the Southeast Asian Linguistics Society, 13 (1). Retrieved October 28,
2024, from https://link.gale.com/apps/doc/A651817246/AONE?u=anon~e7c7873
d&sid=googleScholar&xid=afce68dd

89

Prince, A., & Smolensky, P. (2002). Optimality Theory: Constraint Interaction in Generative

Grammar. https://doi.org/https://doi.org/10.7282/T34M92MV

Pulleybank, D. (1997). Optimality Theory and Features. In Optimality theory: An overview
(pp. 59–101, Vol. 59). Blackwell. https://www.researchgate.net/profile/Douglas-Pul
leyblank/publication/247074150_Optimality_theory_and_features/links/540d4d
590cf2f2b29a383763/Optimality-theory-and-features.pdf

Russell, K. (1995). Morphemes and candidates in Optimality Theor. https://roa.rutgers.ed

u/files/44-0195/44-0195-RUSSELL-0-0.PDF

Selkirk, E. O. (1978). On prosodic structure and its relation to syntactic structure. In T.

Fretheim (Ed.), Nordic Prosody II (pp. 111–140). Tapir.

Srichampa, S. (2002). Vietnamese verbal reduplication. In Collected papers on Southeast
Asian and Pacific languages (pp. 37–47). Pacific Linguistics. https://openresearch-r
epository.anu.edu.au/server/api/core/bitstreams/ebd2dcac-aee0-491d-971f-1e3bec9
7cc1a/content

Thompson, L. C., & O’Harrow, S. D. (1988). A Vietnamese Reference Grammar. University

of Hawaii Press. https://books.google.com/books?id=XN7SHNeZ_ksC

Tsay, J., & Myers, J. (1996). Taiwanese Tone Sandhi as Allomorph Selection. Annual Meeting
of the Berkeley Linguistics Society, 22 (1), 395. https://doi.org/10.3765/bls.v22i1.13
47

Vo, X.-H. (1997). Nghiên cứu về chức năng của thanh điệu tiếng Việt (theo phương pháp định
lượng) (Quantitative Studies on Vietnamese Tones) [PhD Thesis]. Hanoi University
of Education.

Vu, S. X. (2007). A unified analysis of some Vietnamese reduplication forms. In M. Alves,
P. Sidwell, & D. Gil (Eds.), SEALS VIII Papers from the 8th Annual Meeting of
the Southeast A (pp. 165–191). Pacific Linguistics, Research School of Pacific; Asian
Studie. http://sealang.net/sala/archives/pdf8/vu1998unified.pdf#page=16.00

Wivell, G., Miatto, V., Karaka¸s, A., Kostyszyn, K., & Repetti, L. (2024). All about ablaut:
A typology of ablaut reduplicative structures. Linguistic Typology, 28 (3), 505–536.
https://doi.org/10.1515/lingty-2023-0018

Wolf, M. (2005). For an Autosegmental Theory of Mutation. Proceedings of the West Coast

Conference on Formal Linguistics. https://doi.org/10.7282/T3CZ3B3S

Yip, M. (1995). Identity Avoidance in Phonology and Morphology. Proceedings of the Con-
ference on Morphology and its relation to Syntax and Phonology. https://roa.rutger
s.edu/files/82-0000/roa-82-yip-3.pdf#page=20.73

90

Zhang, J., & Lai, Y. (2007). Two aspects of productivity in Taiwanese Double Reduplication.

Kansas Working Papers in Linguistics. https://doi.org/10.17161/KWPL.1808.1786

91

APPENDIX A

GENERAL STRUCTURE OF REDUPLICATION

Legend

X

X

N

T

α

Base

Red

Nasal

red

blue

Verb

Adjective

purple

Verb + Adjective

Tone contour

black

Noun + Verb + Adjective

Tone register

Novel words

green

Noun

Type

Base

Base + Reduplicant

Attenuation

Pejorative

X

X

X

X.Y

X-T1

X.X

X.X

XAα.X (α matches that of X)

X.Xi@kB1/B2

X.Y.X.Yi@kB1/B2

X.Xi@kBα

X.Y T1

X.Y.X.Yi@kBα

X(.Y)i@k

∅

X

X.Y

X.Xi@kBα (less alternation)

X.Y.X.Yi@kBα (less alternation)

X-T1

X.Xi@kBα

X(.Y)i@k

∅

92

Exclamation

X.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X

∅

X.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.XECα

X.Xa:Aα.X.XECα

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.Xa:Aα.X.Xa:NAα

X.Xa:NAα

X.Xa:NBα

X.Xa:Aα.X.Xa:NBα

X.Xa:pB2

X.Xi@kBα

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.Xa:A2.X.Xa:pB2

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

93

X.X

∅

Intensification (1)

X.YTα

X.Ya:Aα.X.YTα

X.Xi@kB1

X.X-a:A1.X.Xi@kB1

X.YTα

X.Ya:Aα.X.YTα

X.Xi@kB1

X.X-a:A1.X.Xi@kB1

X.X.Y.Y

X.X.Y.Y

X.Xa:NAα

X.X.Y.Y

X.X.Y.Y

X.Xa:Aα.X.Xa:NAα

X.X.Y.Y

X.X.Y.Y

X.XaNAα

X.X.Y.Y

X.X.Y.Y

X.Xa:Aα.X.XaNAα

X.X.Y.Y

X.X.Y.Y

X.Xa:pBα

X.X.Y.Y

X.XE:Cα

X.X.Y.Y

X.Xa:Aα.X.Xa:pBα

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.X.Y.Y

X.Xa:Aα.X.X.XE:Cα

94

Intensification

(2)/

Triplication

X

XB1

X.Y

X

XB1

X.XA2.XA1

XB1.XA2.XB2 (X has an obstruent coda)

X.YAα.Y

X.XA2.XA1

XB1.XA2.XB2 (X has an obstruent coda)

95

APPENDIX B

ADDITIONAL DATA

B.1 Attenuation

Verb

(1)

a.

[NiA1]

‘to suspect’

[NiA1.NiA1]

‘suspect (att)’

*[NiA2.NiA1]

b.

[ñ1nA2]

‘to look’

[ñ1nA2.ñ1nA2]

‘look (att)’

c.

[th1tB1]

‘to like’

[th1tB1.th1tB1]

‘like (att)’

*[ñ1nA1.ñ1nA2]

d.

[li@kB1]

‘to glance’

[li@kB1.li@kB1]

‘glance (att)’

*[th1nA1.th1tB1]

*[li@NA1.li@kB1]

Adjective

(2) A1 words:

a.

[ka:wA1]

‘tall’

[ka:wA1.ka:wA1]

‘slightly tall’

*[ka:wA2.ka:wA1]

b.

[âENA1]

‘black’

[âENA1.âENA1]

‘slightly black’

(3) A2 words:

*[âENA2.âENA1]

a.

[va:NA2]

‘yellow’

[va:NA2.va:NA2]

‘slightly yellow’

*[va:NA1.va:NA2]

b.

[ja:A2]

‘old’

[ja:A2.ja:A2]

‘slightly old’

*[ja:A1.ja:A2]

96

(4) B1 words:

a.

[t1mB1]

‘purple’

[t1mB1.t1mB1]

‘slightly purple’

[t1mA1.t1mB1]

b.

[sopB1]

‘porous; fluffy’

[sopB1.sopB1]

‘slightly fluffy’

[somA1.sopB1]

c.

[sa:kB1]

‘close to’

[sa:kB1.sa:kB1]

‘slightly close to’

[sa:NA1.sa:kB1]

*[sa:kB1.sa:NA1]

(R is the rightmost σ)

*[sa:kA1.sa:kB1]

(tone-coda co-occurrence)

(5) B2 words:

a.

[âEpB2]

‘pretty’

[âEpB2.âEpB2]

‘slightly pretty’

[âEmA2.âEpB2]

*[âEpA2.âEpB2]

(tone-coda co-occurrence)

b.

[satB2]

‘clean’

[satB2.satB2]

‘slightly clean’

[sanA2.satB2]

c.

[tW@B2]

‘similar to’

[tW@B2/A2.tW@B2]

‘slightly similar to’

(6) C1 words:

a.

b.

c.

[âOC1]

[ñOC1]

‘red’

[âOC1/A1.âOC1]

‘slightly red’

‘small’

[ñOC1/A1.ñOC1]

‘slightly small’

[xwEC1/A2]

‘healthy’

[xwEC1.xwEC1]

‘slightly healthy’

(7) C2 words:

a.

b.

c.

d.

[lwa:NC2]

‘watery’

[lwa:NC2/A1/A21.lwa:NC2]

‘slightly watery’

[ña:wC2]

‘mushy’

[ña:wC2/A1.ña:wC2]

‘slightly mushy’

[jEC2]

[kiC2]

‘easy’

[jEC2/A1.jEC2]

‘slightly easy’

‘careful’

[kiC2/A1.kiC2]

‘slightly careful’

1Note that there’s an A1 option here as well because there is a preference for A1.C2 combination rather

than A2.C2 one, but both are possible regardless, see A.-H. Pham and Pham (2020) for the discussion.

97

B.2 Pejorative

Monosyllabic base

Noun

(8)

a.

[sEA1]

‘car(s)’

[sEA1.si@kB1]

‘cars (pej)’

*[sEA1.si@kB2]

b.

c.

[ña:A2]

‘house(s)’

[ña:A2.ñi@kB1/B2]

[Pa:wB1]

‘shirt(s)’

[Pa:wB1.Pi@kB1]

‘houses (pej)’

‘shirts (pej)’

d.

[si@kB1]

‘circus(es)’

alternative:

*[Pa:wB1.Pi@kB2]

*[si@kB1.si@kB1]
[si@kB1.su: >NmC1]

e.

f.

[áa:NB2]

‘friend(s)’

[áa:NB2.ái@kB1/B2]

[ti@kB2]

‘party(s)’

alternative:

*[ti@kB2.ti@kB1/B2]
[ti@kB2.tu: >NmA2]

‘circusses (pej)’

‘friends (pej)’

‘parties (pej)’

g.

[v7:C1]

‘notebook(s)’

[v7:C1.vi@kB1]

‘notebooks (pej)’

h.

[muC2]

‘hat(s)’

[muC2.mi@kB1/B2]

‘hats (pej)’

*[v7:C1.vi@kB2]

Verb

(9)

a.

[PaNA1]

‘to eat’

[PaNA1.Pi@kB1]

‘to eat (pej)’

*[PaNA1.Pi@kB2]

b.

[ñ7:A2]

‘to ask for

[ñ7:A2.ñi@kB1/B2]

‘to ask for a favor

a favor’

(pej)’

c.

[GEpB1]

‘to couple’

[GEpB1.Gi@kB1]

‘to couple (pej)’

d.

[li@kB1]

‘to side-eye’

*[li@kB1.li@kB1]

‘to side-eye (pej)’

*[GEpB1.Gi@kB2]

*[li@kB1.li@kB2]

98

e.

f.

[lwi@NB2]

‘to practice’

[lwi@NB2.li@kB1/B2]

‘to practice (pej)’

[ji@kB2]

‘to destroy’

*[ji@kB2.ji@kB1]

‘to destroy (pej)’

*[ji@kB2.ji@kB2]

g.

[NuC1]

‘to sleep’

[NuC1.Ni@kB1]

‘to sleep (pej)’

h.

[Na:C2]

‘to fall’

[Na:C2.Ni@kB1/B2]

‘to fall (pej)’

*[NuC1.Ni@kB2]

Adjective

(10)

a.

[ka:wA1]

‘tall’

[ka:wA1.ki@kB1]

‘tall (pej)’

*[ka:wA1.ki@kB2]

b.

c.

d.

e.

[va:NA2]

‘yellow’

[va:NA2.vi@kB1/B2]

‘yellow (pej)’

[sopB1]

‘porous; fluffy’

[sopB1.si@kB1]

‘fluffy (pej)’

*[sopB1.si@kB2]

[satB2]

‘clean’

[satB2.si@kB1/B2]

‘clean (pej)’

[ñOC1]

‘small’

[ñOC1.ñi@kB1]

‘small (pej)’

f.

[lwa:NC2]

‘watery’

[lwa:NC2.li@kB1/B2]

‘watery (pej)’

*[ñOC1.ñi@kB2]

Novel words

(11) Noun

a.

b.

c.

d.

[hwa:A1]

‘flower(s)’

[hwa:A1.hwi@kB1/B2]

‘flowers (pej)’

[kwa:A2]

‘present(s)’

[kwa:A2.k(w)i@kB1/B2]

‘presents (pej)’

[áenB2]

‘sick’

[áenB2.ái@kB1/B2]

‘sick (pej)’

[miC2]

‘America’

[miC2.mi@kB1/B2]

‘America (pej)’

(12) Verb

a.

b.

[m7:A1]

‘to dream’

[m7:A1.mi@kB1]

‘to dream (pej)’

[s7:A2]

‘to touch’

[s7:A2.si@kB1/B2]

‘to touch (pej)’

99

c.

[ñ7:B1]

‘to miss/remember’

[ñ7:B1.ñi@kB1]

‘to miss/remember

(pej)’

d.

e.

f.

[cajB2]

‘to run’

[cajB2.ci@kB1/B2]

‘to run (pej)’

[c7:C1]

‘to transport’

[c7:C1.ci@kB1/B2]

‘to transport (pej)’

[vEC2]

‘to draw’

[vEC2.vi@kB1/B2]

‘to draw (pej)’

(13) Adjective

a.

b.

c.

d.

e.

f.

g.

[sanA1]

‘blue/green’

[sanA1.si@kB1]

‘blue/green (pej)’

[hoNA2]

‘pink’

[hoNA2.hi@kB1/B2]

[la:wB1]

‘rude’

[la:wB1.li@kB1]

‘pink (pej)’

‘rude (pej)’

[â7mB2]

‘strong/deep’

[â7mB2.âi@kB1/B2]

‘strong/deep (pej)’

[hwa:NC1]

‘in a panic’

[hwa:NC1.hi@kB1/B2]

‘in a panic (pej)’

[úeC2]

[rOC2]

‘late’

‘clear’

[úeC2.úi@kB1/B2]

[rOC2.ri@kB1]

‘late (pej)’

‘clear (pej)’

Polysyllabic base

Noun

(14)

a.

[ka:A2.fEA1]
coffee
‘coffee(s)’

[ka:A2.fEA1
coffee
‘coffees (pej)’

ka:A2.fi@kB1 / *B2]
Ri@k

b.

* ki@kB2 fEA1

d.

* ki@kB2 fEA1 ka:A2 fEA1

(number of σ’s in the reduplicative

(Ri@k in the first syllable)

̸= number of σ’s in the base)

e.

* ka:A2 fEA1 ki@kB2 fEA1

c.

* ka:A2 fi@kB1 ka:A2 fEA1

(Ri@k in the third syllable)2

(Ri@k in the second syllable)

f. sEA1

âa:pB2
pedal

vehicle
‘bicycle(s)’

âa:pB2
pedal

sEA1
vehicle
‘bicycles (pej)’

sEA1
vehicle

âi@kB2/B1
Ri@k

2These violations are the same inputs of any lexical categories.

100

g.

[Pa:wB1
shirt
‘clothes’

kw7NA2]
pant

kw7NA2
[Pa:wB1
shirt
pant
‘clothes (pej)’

Pa:wB1
shirt

ki@kB1/B2]
Ri@k

h. cu@NA2.cu@NA2

dragonfly
‘dragonfly(s)’

cu@NA2.cu@NA2
dragonfly
‘dragonflies (pej)’

cu@NA2.ci@kB1/B2
Ri@k

i.

[k7wA1.lakB2.áoB2]
club
‘club(s)’

[k7wA1.lakB2.áoB2
club
k7wA1.lakB2.ái@kB2/B1]
Ri@k
‘clubs (pej)’

(Novel data)

Verb

(15)

a.

áa:jA2
la:mA2
do
HW
‘to do HW’

la:mA2
la:mA2
áa:jA2
do
do
HW
‘to do HW (pej)’

ái@kB2/B1
Ri@k

b.

* la:mA2 ái@kB2 la:mA2 áa:jA2

d.

* la:mA2 áa:jA2 li@kB2 áa:jA2

(R in 2nd syllable)

(R in 3rd syllable)

c.

* li@kB2 áa:jA2 la:mA2 áa:jA2

(R in 1st syllable)

nOjB1
e. kW@jA2
laugh
talk
‘to laugh and talk’3

kW@jA2
laugh
‘to laugh and talk (pej)’

kW@jA2
laugh

nOjB1
talk

ni@kB1/*B2
Ri@k

f. NuC1.NeA1

sleep
‘to sleep (emph)’

NuC1.Ni@kB1/*B2
NuC1.NeA1
sleep
Ri@k
‘to sleep (pej)’

Adjective

(16)

a. va:NA2
Red
‘slightly yellow’

va:NA2
yellow

va:NA2
va:NA2
Red
yellow
‘slightly yellow (pej)’

va:NA2.vi@kB1/B2
Ri@k

3in a collective sense

101

b. m7:A1.ma:NA2

dreamy
‘dreamy4’

m7:A1.ma:NA2
dreamy
‘dreamy (pej)’

m7:A1.mi@kB2/B1
Ri@k

c. sanA1

blue/green
‘green’

la:B1
leaf

sanA1
blue/green
‘green (pej)’

la:B1
leaf

sanA1.li@kB1
Ri@k

d. s1nA1
pretty
‘pretty’

âEpB2
pretty

âEpB2
s1nA1
pretty
pretty
‘pretty (pej)’

s1nA1.âi@kB1
Ri@k

Novel words

(17) Noun

a. hONA2
clf
‘rock(s)’

âa:B1
rock

b. kONA1
clf
‘dog(s)’

cOB1
dog

c. sEA1

vehicle
‘vehicles’

koB2
vehicle

hONA2
âa:B1
clf
rock
‘rocks (pej)’

hONA2
clf

âi@kB1/B2
Ri@k

kONA1
kOB1
clf
dog
‘dogs (pej)’

cONA1
clf

ci@kB1/B2
Ri@k

koB2
sEA1
vehicles
vehicle
‘vehicles (pej)’

sEA1
vehicle

ki@kB2/B1
Ri@k

kOC1
d. kajA1
grass
tree
‘vegetation’

kOC1
kajA1
grass
tree
‘vegetation (pej)’

kajA1
tree

ki@kB1
Ri@k

(18) Verb

ñajC1
a. cajB2
run
jump
‘to run and jump’

ñajC1
jump

ñi@kB1/B2
cajB2
run
Ri@k
‘to run and jump (pej)’

cajB2
run

4[ma:NA2] doesn’t have meaning in Vietnamese, but it’s an [-a:NA] reduplicant of an Adj, an unproductive

reduplication process

102

(19) Adjective

a. menA1.moNA1

vast
‘vast’

menA1.moNA1
vast
‘vast (pej)’

menA1.mi@kB1
Ri@k

b. vojB2.va:NA2

hasty
‘hasty’

c. ro >NmB2.ra:NA2

bustling
‘bustling’

vojB2.va:NA2
hasty
‘hasty (pej)’

vojB2.vi@kB2/B1
Ri@k

ro >NmB2.ra:NA2
bustling
‘bustling (pej)’

ro >NmB2.ri@kB2/B1
Ri@k

B.3 Exclamation

Noun

(20)

a. Pa:wB1
shirt
‘clothes’

kw@NA2
pant

Pa:wB1
shirt

Pa:wB1
shirt
‘clothes (emph)’

kw@NA2
pant

kw@NA2
pant

b.

* Pa:wB1 kw@NA2 Pa:wB1 kw@NA2

(reduplicative in the form ABAB)

c. ka:A2.fEA1
‘coffee’

d. sEA1.âa:pB2
‘bicycle’

* ka:A2.ka:A2.fEA1.fEA1

Intended : ‘coffee (emph)’

* sEA1.sEA1.âa:pB2.âa:pB2
Intended : ‘bicycle (emph)’

e. cu@NA2.cu@NA2
‘dragonfly’

* cu@NA2.cu@NA2.cu@NA2.cu@NA2
Intended : ‘dragonfly (emph)’

Verb

(21)

nOjB1
a. kW@jA2
talk
laugh
‘to laugh and talk’

nOjB1
kW@jA2
talk
laugh
‘to laugh and talk (emph)’

kW@jA2
laugh

nOjB1
talk

b.

* kW@jA2 nOjB1 kW@jA2 nOjB1

(reduplicative in the form ABAB)

c. NuC1.NeA1
‘sleep’

* NuC1.NuC1.NeA1.NeA1

Intended : ‘sleep (emph)’

103

d.

la:mA2.áa:jA2
‘to study’

* la:mA2.la:mA2.áa:jA2.áa:jA2
Intended : ‘to study (emph)’

Adjective

(22)

a. m7A1.ma:NA2

dreamy
‘dreamy’

b. s1nA1
pretty
‘pretty’

âEpB2
pretty

m7A1
dream

ma:NA2
m7A1
-
dream
Intended : ‘dreamy (emph)’

ma:NA2
-

s1nA1
pretty

âEpB2
s1nA1
pretty
pretty
Intended : ‘pretty (emph)’

âEpB2
pretty

c. va:NA2.va:NA2
‘yellow (att)’

* va:NA2.va:NA2.va:NA2.va:NA2

Intended : ‘yellow (att) (emph)’

d. sanA1.la:B1
‘green’

* sanA1.sanA1.la:B1.la:B1
Intended : ‘green (emph)’

(23)

a. ái@NB1
lazy
‘lazy’

ña:kB1
lazy

ái@NB1
lazy

ái@NB1
lazy
‘lazy (emph)’

ña:kB1
lazy

ña:kB1
lazy

b. hamA1.h7:C1

‘eager’

hamA1.hamA1.h7:C1.h7:C1
‘eager (emph)’

c. N7:A1.Na:kB1

‘dazed’

N7:A1.N7:A1.Na:kB1.Na:kB1
‘dazed (emph)’

d. th7:A1.th7NC1
‘distracted’

th7:A1.th7:A1.th7NC1.th7NC1
‘distracted (emph)’

e.

lamA2.liA2
‘reticent’

lamA2.lamA2.liA2.liA2
‘reticent (emph)’

f. úu >NmA2.âi@pB2
‘expansive’

úu >NmA2.úu >NmA2.âi@pB2.âi@pB2
‘expansive (emph)’

104

g. vojB2.va:NA2

‘hasty’

vojB2.vojB2.va:NA2.va:NA2
‘hasty (emph)’

Novel words

(24) Noun

a.

fW@NA2
district

foB1
street
‘city(s)’

foB1
foB1
street
street
‘city(s) (emph)’

fW@NA2
district

fW@NA2
district

Pu@NB1
b. PaNA1
eat
drink
‘eating and drinking’

su@jB1
c. soNA1
stream
river
‘rivers and streams’

PaNA1
eat

PaNA1
eat
‘eating and drinking (emph)’

Pu@NB1
drink

Pu@NB1
drink

soNA1
river

su@jB1
soNA1
stream
river
‘rivers and streams (emph)’

su@jB1
stream

d. âW@NA2
road
‘roads’

sa:B1
road

âW@NA2
road
‘roads (emph)’

âW@NA2
road

sa:B1
road

sa:B1
road

e. ña:A2
house
‘house(s)’

kW@C1
door

f. hOjC1
ask
‘Q&A’

âa:pB1
answer

kW@C1
ña:A2
ña:A2
house
door
house
‘house(s) (emph)’

kW@C1
door

hOjC1
hOjC1
ask
ask
‘Q&A (emph)’

âa:pB1
answer

âa:pB1
answer

(25) Verb

a. âWNB1
stand
‘to be on edge’

NojA2
sit

b. GakB2
reap
‘to reap; to gain’

ha:jB1
pluck

âWNB1
âWNB1
stand
stand
‘to be on edge (emph)’

NojA2
sit

NojA2
sit

GakB2
reap

GakB2
reap
‘to reap; to gain (emph)’

ha:jB1
pluck

ha:jB1
pluck

105

Pi@kB1
c. PaNA1
eat
Ri@k
‘eat (pej)’

Pi@kB1
PaNA1
PaNA1
eat
Ri@k
eat
‘eat (pej, emph)’

Pi@kB1
Ri@k

(26) Adjective

a. ma:kB1.mEC1

cool
‘(of weather) cool’5

b. âEpB2.âEC2
beautiful
‘beautiful’6

c. menA1.moNA1

vast
‘vast’

d. xwEC1.xwaNB1

lively
‘lively’7

e. c7mB2.ca:pB2

slowly
‘slowly’8

f. xoA1
dry
‘dry/dull’

xa:NA1
hoarse

ma:kB1
cool
‘(of weather) cool (emph)’

ma:kB1
cool

mEC1
-

mEC1
-

âa:A1
âEpB2
pretty
Ra:
‘beautiful (emph)’

âEpB2.âEC2
beautiful

ma:A1
Ra:
menA1
vast

menA1.moNA1
vast

moNA1
-

moNA1
-

menA1
vast

menA1
vast
‘vast (emph)’

xwEC1
healthy
‘lively (emph)’

xa:A1
Ra:

xwEC1.xwaNB1
lively

c7mB2
slow

ca:A2
Ra:
c7mB2
slow

c7mB2
slow
‘slowly (emph)

c7mB2.ca:pB2
slowly

ca:pB2
-

ca:pB2
-

xoA1
dry

xoA1
dry

xa:NA1
hoarse

xa:NA1
hoarse

xoA1
dry
‘dry/dull (emph)’
5[mEC1] doesn’t have meaning in Vietnamese, but it’s an [-EC] reduplicant of an Adj, an unproductive

xa:NA1
hoarse

xa:A1
Ra:

xoA1
dry

reduplication process

6[âEC2] doesn’t have meaning in Vietnamese, but it’s an [-EC] reduplicant of an Adj, an unproductive

reduplication process

7[xwaNB1] doesn’t have meaning in Vietnamese, but it’s an [-aNB] reduplicant of an Adj, an unproductive

reduplication process

8[ca:pB2] doesn’t have meaning in Vietnamese, but it’s an [-a:pB] reduplicant of an Adj, an unproductive

reduplication process

106

g.

jeC2.ja:NA2
easily
‘easily’9

jeC2
easy

jeC2
easy

ja:NA2
-

ja:NA2
-

ja:A2
jeC2
Ra:
easy
‘easily (emph)’

jeC2.ja:NA2
easily

h. sopB1
fluffy
‘fluffy (pej)’

si@kB1
Ri@k

sopB1
sopB1
fluffy
fluffy
‘fluffy (pej, emph)’

si@kB1
Ri@k

si@kB1
Ri@k

B.4

Intensification (1)

Verb

(27)

a. kaNA2.ñaNA2
‘complain’

b. kamB1.kujB1
‘concentrate’

kaNA2.ña:A2.kaNA2.ñaNA2
‘constantly complain’

kamB1.ka:A1.kamB1.kujB1
‘constantly concentrate’

c.

lwa:NB2.cwa:NB2
‘totter’

lwa:NB2.ca:A2.lwa:NB2.cwa:NB2
‘constantly totter’

d.

la:mC1.ña:mC1
‘blabber’

la:mC1.ña:A1.la:mC1.ña:mC1
‘constantly blabber’

Adjective

(28)

a. m7:A1.ma:NA2

dreamy
‘dreamy’

ma:A2
m7:A1
dream
Ra:
‘dreamy (emph)’

m7:A1.ma:NA2
dreamy

b.

* m7:A1 ma:NA2 m7:A1 ma:A2

d.

* va:NA2 va:A2 va:NA2 va:NA2

‘dreamy (emph)’

‘sligthly yellow (emph)’

c.

* s1nA1 âa:A2 s1nA1 âEpB2

e.

* sanA1 la:A1 sanA1 la:B1

‘pretty (emph)’

‘green (emph)’

(29)

a.

9[ja:NA2] doesn’t have meaning in Vietnamese, but it’s an [-a:NA] reduplicant of an Adj, an unproductive

reduplication process

107

l1nA1.t1nA1
‘nonsensical’

l1nA1.ta:A1.l1nA1.t1nA1
‘nonsensical (emph)’

b. s7:NB1.sa:kB1
‘ill-mannered’

s7:NB1.sa:A1.s7:NB1.sa:kB1
‘ill-mannered (emph)’

c. ñiB1.ñanC1
‘cutesy’

ñiB1.ñaA1.ñiB1.ñanC1
‘cutesy (emph)’

d. áiA2.áOmC2
‘water sound’

áiA2.áa:A2.áiA2.áOmC2
‘water sound (emph)’

Novel words

(30) Noun

Pi@kB1
a. Pa:wB1
shirt
Ri@k
‘shirts (pej)’

Pa:wB1
Pa:A1
Pa:wB1
shirt
shirt
Ra:
‘shirts (pej, emph)’

Pi@kB1
Ri@k

b. ña:A2
house
‘houses (pej)’

ñi@kB1
Ri@k

(31) Verb

a. NuC1.NeA1

sleep
‘to sleep’

b. ñ7:A2

ñi@kB1
ask.favor
Ri@k
‘ask for a favor (pej)’

c. GEpB1
Gi@kB1
Ri@k
couple
‘to couple (pej)’

ña:A1
Ra:

ña:A2
house
‘houses (pej, emph)’

ña:A2
house

ñi@kB1
Ri@k

Na:A1
NuC1
NuC1.NeA1
sleep
Ra:
sleep
‘to sleep (emph)’

ñi@kB1
ñ7:A2
ask.favor
Ri@k
‘ask for a favor (pej, emph)’

ñ7:A2
ask.favor

ña:A1
Ra:

Ga:A1
Ra:

GEpB1
Gi@kB1
Ri@k
couple
‘to couple (pej, emph)’

GEpB1
couple

108

(32) Adjective

a. ma:kB1.mEC1

cool
‘(of weather) cool’

b. âEpB2.âEC2
beautiful
‘beautiful’

c. menA1.moNA1

vast
‘vast’

d. xwEC1.xwaNB1

lively
‘lively’

e. c7mB2.ca:pB2

slowly
‘slowly’

f. xoA1
dry
‘dry/dull’

xa:NA1
hoarse

g.

jeC2.ja:NA2
easily
‘easily’10

ma:kB1
cool
‘(of weather) cool (emph)’

ma:kB1
cool

mEC1
-

mEC1
-

âEpB2
âa:A1
Ra:
pretty
‘beautiful (emph)’

âEpB2.âEC2
beautiful

ma:A1
Ra:
menA1
vast

menA1.moNA1
vast

moNA1
-

moNA1
-

menA1
vast

menA1
vast
‘vast (emph)’

xwEC1
healthy
‘lively (emph)’

xa:A1
Ra:

xwEC1.xwaNB1
lively

c7mB2
slow

ca:A2
Ra:
c7mB2
slow

c7mB2
slow
‘slowly (emph)

c7mB2.ca:pB2
slowly

ca:pB2
-

ca:pB2
-

xoA1
dry

xoA1
dry

xa:NA1
hoarse

xa:NA1
hoarse

xa:A1
Ra:

xoA1
dry
‘dry/dull (emph)’

xoA1
dry

xa:NA1
hoarse

jeC2
easy

jeC2
easy

ja:NA2
-

ja:NA2
-

ja:A2
jeC2
easy
Ra:
‘easily (emph)’

jeC2.ja:NA2
easily

10[ja:NA2] doesn’t have meaning in Vietnamese, but it’s an [-a:NA] reduplicant of an Adj, an unproductive

reduplication process

109

h. va:NA2
yellow
‘yellow (pej)’

vi@kB1
Ri@k

va:NA2
va:A1
va:NA2
yellow
yellow
Ra:
‘yellow (pej, emph)’

vi@kB1
Ri@k

B.5

Intensification (2) (Triplication)

Monosyllabic base

Adjective

(33)

a.

liA2
‘naughty’

b. mW@kB1
‘smooth’

c. sa:kB1

‘smooth’

liA2.l7mB2 / *liA2.liA2.liA1/B2
‘naughty (emph)’

mW@kB1.mW@NA2.mW@kB2
‘smooth (emph)’

sa:kB1.sa:NA2.sa:kB2
‘close to (emph)’

d. ca:kB1

‘slapping sound’

ca:kB1.ca:NA2.ca:kB2
‘slapping sound (emph)’

e. sopB1
‘fluffy’

f. tu@kB1

‘everything’

g. t7kB1

‘everything’

h. satB2
‘clean’

sopB1.somA2.sopB2
‘fluffy (emph)’

tu@kB1.tu@NA2.tu@kB2
‘everything & anything’

t7kB1.t7NA2.t7kB2
‘everything & anything’

satB2.sanA2.sanA1
‘clean (emph)’

i. tEwB2

‘a tiny bit’

tEwB2.tEwA2.tEwA1
‘small (emph)’11

11used in compounds

110

j. majC1

‘a small portion’

majC1.majA2.majA1
‘(of chance) small (emph)’

k. tWNC1
‘crazy’

tWNC1.tWNA2.tWNA1
‘crazy (emph)’

Novel words

(34)

a. n7wA1
‘brown’

n7wC1.n7wA2.n7wA1
Intended: ‘brown (emph)’

b. sanA1

‘blue/green’

sanC1.sanA2.sanA1
Intended: ‘blue/green (emph)’

c. hOkB1
‘hot’

d. s7wB1
‘ugly’

e. P1tB1
‘little’

f. âEpB2
‘pretty’

g. áa:kB2
‘silver’

h. âOC1
‘red’

i. ñOC1

‘small’

j. ña:wC2
‘mushy’

k.

hOkB1.hONA2.hOkB2
Intended: ‘hot (emph)’

s7wB1.s7wA2.s7wA1
Intended: ‘ugly (emph)’

P1tB1.P1nA2.P1tB2
Intended: ‘little (emph)’

âEpB2.âEmA2.âEmA1
Intended: ‘pretty (emph)’

áa:kB2.áa:NA2.áa:NA1
Intended: ‘silver (emph)’

âOC1.âOA2.âOA1
Intended: ‘red (emph)’

ñOC1.ñOA2.ñOA1
Intended: ‘small (emph)’

ña:wC2.ña:wA2.ña:wA1
Intended: ‘mushy (emph)’

111

lwa:NC2
‘watery’

lwa:NC2.lwa:NA2.lwa:NA1
Intended: ‘watery (emph)’

Polysyallbic base

Adjective

(35)

>
kpB1

a. xoA1.xo:
‘very dry’

xoA1.xo: >NmA1.xo:
‘very dry (emph)’

>
kpB1

b. th7:mA1.fWkB1
‘very aromatic’

th7:mA1.fWNA1.fWkB1
‘very aromatic (emph)’

c. satB2.áatB1
‘very clean’

satB2.áanA1.áatB1
‘very clean (emph)’

d. thW@A1.th7:kB1
‘very sparse’

thW@A1.th7:NA1.th7:kB1
‘very sparse (emph)’

e. nO >NmB1.áWNA2

‘flaming hot; feverish’

nO >NmB1.áWNA2.áWNA2
‘feverish (emph)’

f.

>
kpB1

laNA1.lo
‘rolling around’

laNA1.lo >NmA1.lo
‘rolling around and around’

>
kpB1

g. âENA1.siA2
‘very black’

âENA1.siA2.liA2
‘very black (emph)’

h. th7mA1.siA2

‘very discolored’

th7mA1.siA2.liA2
‘very discolored (emph)’

i. âENA1.thujA1
‘burnt black’

âENA1.thujA2.lujA2
‘burnt black (emph)’

j. t1mB1.l1mB2
‘very purple’

t1mB1.s1mA2.l1mB2
‘very purple (emph)’

112

k.

luA1.áuA1
‘busy’

luA1.suA1.áuA1
‘very busy (emph)’

l.

l7:A1.m7A1
‘half-asleep’

l7:A1.t7A1.m7A1
‘half-asleep (emph)’

B.6 Lexicalized reduplicatives

(36)

(Full) reduplication of nouns (distributive and/or plurality marking

(Thompson & O’Harrow, 1988; Vu, 2007))

a.

b.

c.

d.

e.

[a:jA1]

‘who’

[a:jA1.a:jA1]

‘everyone’

[NWjA2]

‘person’

[NWjA2.NWjA2]

‘everbody’

[NajA2]

‘day’

[NajA2.NajA2]

‘every day’

[sa:NA1]

‘morning’

[sa:NA1.sa:NA1]

‘every morning’

[ciwA2]

‘afternoon’

[ciwA2.ciwA2]

‘every afternoon’

(37)

(Full) reduplication of adverbs

a.

b.

c.

d.

e.

[lu@NA1]

‘always; continually’

[lu@NA1.lu@NA1]

‘always; continually’

[h7:jA1]

‘slightly, somewhat’

[h7:jA1.h7:jA1]

‘slightly, somewhat’

[âewA2]

‘equal; regular’

[âewA2.âewA2]

‘equally; regularly’

[thW@NA2]

‘often’

[thW@NA2.thW@NA2]

‘usually’

[ma:jC2]

‘for a long time’

[ma:jC2.ma:jC2]

‘always’

(38) Onset replacement

a.

b.

c.

[lanA1]

‘clever’

[lanA1.canA1]

‘hasty’

[ñaNA2]

‘eat with front teeth’

[kaNA2.ñaNA2]

‘complain’

[ku@NB1]

‘panic’

[lu@NB1.ku@NB1]

‘in a hurry’

(39) Onset and tone replacement

a.

b.

[c7:jA1]

‘play’

[c7:jA1.á7:jA2]

‘to play around’

[satB2]

‘clean’

[satB2.áatB1]

‘completely clean’

113

(40) Rhyme replacement

a.

b.

c.

[kamB1]

‘plant’

[kamB1.kujB1]

‘concentrate’

[áWkB2]

‘upset’

[áWkB2.áojB2]

‘upset’

[c7mB2]

‘slow’

[c7mB2.ca:pB2]

‘slowly’

(41) Rhyme and tone replacement

a.

b.

c.

d.

[ma:kB1]

‘cool’

[ma:kB1.mEC1]

‘cool (weather)’

[âEpB2]

‘beautiful’

[âEpB2.âEC2]

‘beautiful’

[t1nC1]

‘awake’

[t1nC1.tawB1]

‘awake; refreshed’

[jeC2]

‘easy’

[jeC2.ja:NA2]

‘easily’

(42) Cases in which the base and the reduplicant are indistinguishable

a.

b.

c.

d.

e.

[l1nA1.t1nA1]

‘nonsensical’

[th1nA1.th1tB2]

‘silently’

[áiA2.áOmC2]

‘water sound’

[s7:NB1.sa:kB1]

‘ill-mannered’

[lWNC2.thWNC2]

‘walk slowly and aimlessly’

114