Being Particularly : The Syntax of Sentence-Final Particles in Singaporean English

Lan Yingjie Emmanuel College

8th June 2017

This dissertation is submitted for the degree of Master of Philosophy.

Department of Theoretical and Applied Linguistics University of Cambridge Declaration

This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration except where specifically indicated in the text.

This dissertation contains 23 183 words, of which 20 000 words are counted towards the official word limit.

1 Acknowledgements

This thesis being my third in as many academic years, one might expect it to have been a little easier. Unfortunately, that was not quite how things panned out: I was stretched and challenged to exceed whatever I had previ- ously done, as it should be. To be able to complete this dissertation on my dearly beloved Singaporean English (SgE) therefore has been only possible because of all the people who have supported me throughout this endeav- our: my professors, my friends, family and loved ones, and my heavenly Father.

To Dr Theresa Biberauer, my thesis supervisor, I owe an incredible debt of gratitude that I fear I can never repay. A powerhouse of intellectual energy and linguistic enthusiasm, Theresa pushed me to go beyond my own intellectual limits again and again. I am especially grateful for her generosity of time and energy, especially she was meant to focus on research instead of teaching this year. Despite her rigorous standards, she also knew when I needed to ease off, and I was very much touched by her care and concern the entire year. I also want to express my deepest thanks for her confidence in my abilities and the way she treated me, not merely as a student, but a younger lin- guist to be mentored. She will always be a role model for me to aspire towards.

I also want to express my thanks to my Singaporean friends who served as my native speaker informants for this study: Gwyneth Teo, Charles Nicholas Lim, Justin Daniel Pereira, Derek Lim, Ludwig Tan, Goh Hui Min, Christine Yong, Leonard Yip, Teo Min Xun, Cheryl Tan and Natalie Ng. In particular, the cartographic approach to syntax requires an exhaustive search of syntactic possibilities, and I ended up bombarding Gwen and Charles with so many

2 test questions at strange hours. Had I not been conscientiously verifying my data with the other respondents, this thesis might have been better titled A Microvariational Syntax of the SgE of Gwen and Charles. Thank you for being so available, even if you were eight timezones away. To Lee Junwen, whose analysis of lah and ah formed the starting point for my exploration of pitch contours in SgE, thank you for always being an excellent sounding board for my ideas. I am also grateful to Dr Michael Yoshitaka Erlewine, for his time, ideas and the many useful suggestions he generously gave.

Many of the ideas here were also developed at the Department of Theoreti- cal and Applied Linguistics MPhil seminar Topics in Syntax: Fine Structures, ably led by Dr Craig Sailor and Dr Jamie Douglas. To them, as well as my fellow seminar mates, be they, postgraduate, undergraduate (special mention goes to Ollie Sayeed and Samuel Andersson), or Sten Vikner, I am grateful for your company and suggestions.

As many linguistics undergraduates will attest, syntax can be incredibly intimidating, especially at the beginning. While my continued explorations have only shown it to be more complex and challenging than I first imagined it to be, I want to thank Professor Kenichi Namai of Waseda University for helping me fall in love with that complexity. He showed me how beautiful elegant linguistic argumentation in syntax could be, and how a logical mind could begin to tease apart the mysteries of language. Without his encourage- ment, I might have given up after that first grueling class, much less go on to complete a dissertation under his watchful mentorship in my last year of undergraduate study. I would not be where I am today without him.

To my family: Mama, Papa, and Yingli, who have put up with my very nomadic education over the past six years, thank you for the love and care. I might not always be around for many family events, having found myself in a different country each year for the last four years, but I have always known that you were supporting me and cheering me on from home to finish what I had started. To my aunts Pah Yen Sze and Ho Jia Yin, and my grandaunt Ho

3 Siew Bee (Yˇım`a), thank you for your care and putting up with my incessant questions about Hokkien.

I also owe my thanks to my teammates at the Cambridge University Fencing Club, especially to my teammates on the Men’s First Team. My life in Cambridge would have been unbearably unbalanced had I only restricted myself to linguistics: you guys gave me something to look forward to at train- ings, and I cherish the memories we have together (especially our triumph over the Dark Blue Menace at the Varsity Match). I shall eternally remain, very proudly, a Light Blue fencer.

To Judith, who has encouraged my dreams for the past two years, even if it meant that we would be apart for what felt like an eternity, thank you for the love, care, and prayers you have lavished on me. It has been a challenging two years, but I am very aware of the sacrifices you made for us.

And last but not least, I give thanks to God for His grace that has brought me through. May my work continue to uncover the exquisite intricacies of His hand.

4 Abstract

This thesis looks at sentence-final particles (SFPs) in Singaporean English. By combining existing descriptions with new insights, I address existing issues of SFPs and their pitch contour variants. Subsequently, through a cartographic approach, the linear ordering of SFPs relative to each other is mapped and their relative structural heights determined. The differences in structural position then motivate a proposal for a discourse domain above the CP as postulated by Wiltschko (2016). I then find evidence for an additional functional projection in this domain beyond what has been proposed in the literature. Finally, I examine the apparent violations of the Final-over-Final Condition by the head-final SFPs in SgE that dominate a head-initial clause.

5 Contents

1 Introduction 9

2 Singaporean English and Sentence-Final Particles 12 2.1 The Origins of SFPs in SgE ...... 12 2.2 Tone in SgE SFPs ...... 15 2.2.1 Tone with lah, ah and what?...... 15 2.2.1.1 lah ...... 15 2.2.1.2 ah ...... 17 2.2.1.3 what ...... 19 2.2.1.4 Interim Conclusions ...... 19 2.2.2 Tone and the “Cantonese set” of SFPs ...... 19

2.2.2.1 lor33 ...... 20

2.2.2.2 meh55 ...... 21

2.2.2.3 ma22 ...... 22 2.2.2.4 hor: A Problem? ...... 23 2.2.2.5 Another Problem: leh ...... 26 2.2.3 Conclusion ...... 30 2.3 Other SFPs and Potential Candidates ...... 30 2.3.1 Other SFPs ...... 30 2.3.1.1 one ...... 30 2.3.2 SFP-like Items in SgE ...... 32 2.3.2.1 Sentence-final already ...... 33 2.3.2.2 Sentence-final only ...... 34 2.3.2.3 Sentence-final already and only as SFPs in SgE 35

6 2.4 Boundary Tone Revisited ...... 36 2.5 Conclusion ...... 40

3 Syntactic Analysis of SgE SFPs 41 3.1 Introduction ...... 41 3.2 SFPs in Varieties of Chinese ...... 41 3.3 Substratum Transfer and SgE SFPs ...... 47 3.3.1 Transferred Properties of SgE SFPs ...... 47 3.3.2 SFP Linear Ordering in SgE ...... 48 3.3.3 The Low/High Divide in SgE SFPs ...... 56 3.3.3.1 Embedded and Matrix Clause Modification . 56 3.3.3.2 SFP Positions with Question Tags ...... 58 3.3.4 SgE SFP Structural Heights ...... 64 3.3.5 Rethinking the Low/High Divide ...... 71 3.3.6 Situating The Groups ...... 73

4 SFPs in the Discourse Domain 75 4.1 Motivating the Discourse Domain ...... 75 4.2 The Complex Speech Act Structure ...... 76 4.3 SgE SFPs and the Discourse Domain ...... 80

4.4 The Case of hor24 ...... 88 4.5 Conclusion ...... 90

5 SgE and The Final-over-Final Condition 91 5.1 The Final-over-Final Condition ...... 91 5.2 Particles and FOFC ...... 93 5.2.1 FOFC and Chinese SFPs ...... 93 5.2.2 Types of Particle Acategoriality ...... 95 5.2.3 Acategoriality and Borrowing ...... 95 5.2.4 Acategoriality, FOFC and SFPs ...... 97 5.3 Resolving FOFC and SgE SFPs ...... 97 5.3.1 Intonational SFPs ...... 98 5.3.1.1 The Low and Middle SFPs ...... 98 5.3.2 Pitch Contour SFPs ...... 103

7 5.4 Summing Up ...... 104

6 Conclusion 105

8 Chapter 1

Introduction

‘Singlish also has an array of words that [...] dramatically alter the tone of what you’re saying when tacked on to the end of a sentence. “I got the cat lah”, is an assurance that you have the cat. “I got the cat meh?” is the puzzled realisation that you may have lost it.’

Tessa Wong, The Rise of Singlish1

Singaporean English is a variety of English spoken on the former British colony of Singapore. When the British established a colony in Singapore in 1819, English was put in place as the language of administration. Subse- quently, it was retained upon independence in 1965 as the working language of the country and continues to be widely used in most spheres, public and private, developing into what is referred to as Standard Singaporean English (SSE). SSE has also been observed to not differ meaningfully from standard varieties of English spoken elsewhere, and hence is considered an inner-circle variety in Kachru’s (1982, 1985) three-circle model of World Englishes (Gupta, 1994).

16th August 2015, BBC News. Retrieved from http://www.bbc.co.uk/news/magazine- 33809914, 31st May 2017.

9 However, shortly after colonisation, an influx of migrants from southern China and India joined the existing Malay and Peranakan2 communities on the island, leading to an explosion of ethnic diversity. Accordingly, English was used alongside varieties of Chinese such as Hokkien3, Teochew, Cantonese, Hakka and Mandarin, as well as varieties of Malay from the native Malay and Peranakan inhabitants, and various South Asian languages including Tamil as well. Under such intense conditions of language contact, vernacular Singaporean English, affectionately known as Singlish but referred to here as SgE, emerged as a local creole variety of English distinct from SSE. Linguists have therefore proposed that SgE and SSE exist in a diglossic situation (Bao and Hong, 2006), with SSE taking the place of the high-prestige variety and SgE the low-prestige variety. Others have instead suggested that a continuum model is more suitable, so as to account for the possibility of incorporating SgE features into what would be a standard SSE sentence (Platt, 1975).

Either way, SgE continues to be an unusual creole with syntactic features rarely found in a variety of English. This thesis investigates one such fea- ture: the sentence-final particle (SFP) system in SgE. Often described as one of SgE’s most characteristic features, these particles are also renowned for being fiendishly difficult for non-native speakers to pick up. This system appears to have developed as the result of linguistic influence from varieties of Chinese and possibly Malay (Lim, 2007, 2011), and consists of various particles that appear at the end of sentences in SgE to mark a variety of functions ranging from tense/aspect and focus, to speaker attitudes and other pragmatics-related discourse functions. Since these SFPs also appear to be head-final items dominating an otherwise head-initial clause, they also seem to violate the Final-over-Final Condition (FOFC), an otherwise robust syntactic universal proposed by Biberauer, Holmberg and Roberts (2008, 2014). As the

2The Peranakans, also known as Straits-born Chinese, are the descendants of Chinese immigrants to the Malay archipelago. Arriving between the 15th and 17th century before the British colonisation in 1819, they developed their own creole language that is closely- related to Malay known as Baba Malay. 3Also known as Southern Min, though the Singaporean variety is somewhat distinct from the well-studied Taiwanese variety usually referred to as Taiwanese Southern Min.

10 scholarship in recent years has mainly focused on Chinese SFPs (which also famously appear to violate FOFC), little attention has been paid to exploring the apparent FOFC violations in SgE.

This thesis therefore studies the SFP system in SgE from a syntactic perspective. Chapter 2 surveys the diversity of these particles and their uses, attempting to resolve several outstanding issues, including the nature of their tonal variation and underlying forms. Chapter 3 then goes on to explore the ability of these SFPs to “stack”, where they combine with each other to form a sentence-final SFP cluster. This stacking is constrained: not all combinations are possible (either due to semantic or syntactic reasons). By applying a cartographic approach to such combinations, it is possible to map out a field of where each SFP attaches in the syntactic structure, allowing separation of these SFPs into distributional classes that attach at different structural heights. Following this separation, Chapter 4 explores the nature of the SFPs that attach high up in the structure, finding that they show consistent characteristics to support placement in a domain that is distinct from the CP, dedicated to discourse information, such as speech act structure. Chapter 5 then returns to the apparent violation of the Final-over-Final Condition by SFPs in SgE, arguing that SFPs in SgE actually do not violate FOFC due to their syntactic features which exempt them from it. Finally, Chapter 6 concludes the thesis.

11 Chapter 2

Singaporean English and Sentence-Final Particles

2.1 The Origins of SFPs in SgE

Sentence-final particles (SFPs) are one of the most well-known and well- documented features of SgE and appear frequently in discourse, the famous lah being one of them (Gupta, 1992b; Kwan-Terry, 1991; Ler, 2006; Lim, 2007; Platt, 1987; Wee, 2004). Often punctuating speech heard in Singapore, SFPs serve a variety of discourse marking functions and are notoriously difficult to characterise.

As noted in the introduction, SgE emerged from language contact between English, locally-spoken Sinitic vernaculars (Hokkien, Teochew, Cantonese, Hakka and Mandarin), local varieties of Malay (Bazaar Malay and Baba Malay), and to a lesser extent, Tamil (Sato and Kim, 2012). As such, while ostensibly English-lexified, SgE exhibits syntactic features that would be unusual for a variety of English. These include radical pro drop (where any pronominal argument can be omitted) (Sato and Kim, 2012; Tan, 2007), a lack of overt wh-movement (Lan, 2016), and, of course, SFPs.

These SFPs, as their name suggests, come at the end of a sentence and

12 fulfill various discourse functions. For example, it is possible to turn a declarative sentence into a question by adding the interrogative SFP ah at the end of the sentence, as shown below in (1):

(1) a. You going to school.

‘You are going to school.’

b. You going to school ah? sfp

‘Are you going to school?’

Such syntactic features are characteristic of the Chinese varieties from which SgE draws much of its grammar. For example, Mandarin SFPs do the same, as presented in (2):

(2) a. z`ai ch¯ı wˇanc¯an. MandarinNˇı You prog eating dinner

‘You are eating dinner (now).’

b. z`ai ch¯ı wˇanc¯an ma? MandarinNˇı You prog eating dinner sfp

‘Are you eating dinner (now)?’

Phonologically, Chinese SFPs have been observed to be prosodically inte- grated into the sentences that they combine with, with no pause intonation separating them, unlike question particle “tags” in other languages (Matthews and Yip, 1994; Simpson, 2014). This property is also shared by SgE SFPs and also makes for a very useful diagnostic of SFP-hood.

Accordingly, it would follow that these features have their origin in sub- strate influence from these Chinese varieties, making existing linguistic treat- ments of such phenomena in these related languages a logical starting point for the analysis of SgE.

13 Along this vein, SFPs in SgE have been recognised as having originated largely from varieties of Chinese (Gupta, 1992b; Platt, 1987). In a com- prehensive historical survey, Lim (2007) traces the origins of the various SFPs, concluding that some of the earlier SFPs like lah, ah and what 1 came from Bazaar Malay and Hokkien. She concludes that the bulk of the SFPs came at a later age from Cantonese: lor, hor, leh, meh and ma. Tracing the historical linguistic ecology in Singapore, she proposes three eras characterised by different linguistic ecologies:

1. pre-1800–mid-1970s, where Hokkien and vernacular forms of Malay dominated in the early colonial settlement, and English was mainly limited to the colonial elite and in the administration;

2. mid-1970s–late 1980s, where national policy on official languages led to an increased prominence of English as lingua franca, while Mandarin, Malay and Tamil were spoken at home;

3. and late 1980s–present, where English started becoming the language spoken at home as well. However, due to the burgeoning trend of Cantopop culture exported from Hong Kong, Cantonese also rose in presence and prestige. (note: While plausible as a cultural force in the 1980s, it must be noted that this trend seems to have waned since the mid-2000s, possibly due to the diminishing role of non-Mandarin dialects2 at home, leading to a younger generation that is not fluent in Cantonese.)

Lim notes that her proposed eras indicate approximate points in time, and the divisions do not suggest that the eras are discrete and change abruptly, but that the linguistic situations in each era morph into the next. Accordingly, she proposes that the first group of SFPs arose in the first era, establishing a

1While superficially similar to the English interrogative pronoun what, this what is distinct in SgE. See Section 2.2.1.3 for details. 2Language policy aimed at reducing divisions among the various Chinese dialect groups and preparing a bilingual population able to trade with China led to the government describing all non-Mandarin varieties of Chinese as dialects, despite their generally agreed upon status among linguists as mutually unintelligible languages.

14 functional category in the SgE grammar that facilitated the wholesale import of the second group from Cantonese with their tonal features intact. It seems plausible therefore, to conclude that there might be different groups of SFPs with similar feature bundles (such as tone) due to their common time and place of import. We will return to this idea in Section 2.3.2, when we examine other possible lexical items in SgE that show very similar syntactic behaviour to the “classic” SFPs often discussed in the scholarship of SgE. Such lexical items include the sentence-final adverbs already and only, the SFP status of which are still under debate.

2.2 Tone in SgE SFPs

Since it has long been generally accepted in the literature that SgE SFPs mostly originate from southern varieties of Chinese (Gupta, 1992b; Lim, 2007; Platt, 1987), the presence of tone on some SgE SFPs is not surprising. Following Lim’s (2007) notation, I represent tone in this discussion using pitch level numbers, where 5 represents a high tone and 1 a low tone, with the sequence of digits showing starting and ending (as well as intermediate, if any) pitches: lor33 therefore represents flat mid-level tone.

Based on her earlier work (2007) discussed above, Lim (2011) argues that the earlier set, comprising lah, ah and what, either came via Bazaar/Baba Malay, both non-tonal languages, and hence were transferred without tone, or came via Chinese varieties but lost their tonal features over the longer period of time (relative to the later set comprising lor, hor, leh, meh and ma). In contrast, the later set—referred to as the “Cantonese set” for convenience— retained their original Sinitic tone in SgE.

2.2.1 Tone with lah, ah and what?

2.2.1.1 lah

At this point, the question naturally arises: can the earlier set of particles still exhibit tone (in the form of boundary tone) even if they do not possess

15 lexical tone as Lim (2007, 2011) claims? Lah for instance, can be used in at least three different ways3, differentiated by pitch contour, in response to the question in (3):

(3) A: Do you want to eat durian?

(4) a. Speaker wants to persuade addressee “without excessive pressure” (Wong, 2004, p. 773).

B: No lah24. I’m full from lunch. lah24

‘I want to persuade you that I can’t eat durian because I’m full.’ b. Speaker wants to “impose an idea or a view” on addressee (Wong, 2004, p. 765).

B: No lah21. I’m full from lunch. lah21

‘I want you to know that I can’t eat durian because I’m full.’ c. Speaker wants to show that “the idea is something self-evident or obvious” (Wong, 2004, p. 768).

B: No lah51! I’m already trying not to puke! lah51

‘It is known/should be obvious to you that I hate durian.’

The question naturally arises: does the existence of the three types of lah differentiated by pitch contour therefore mean that contrary to Lim (2007, 2011), lah (and possibly ah and what) could have tone? Ironically, Lim (2007) acknowledges these variants of lah, as well as similar variants of ah, but does not address how these fit in with her theory that these three SFPs have come to lack tone currently.

3 There is also another form: lah55. Lah55 is enumerative and used when listing, such as “cows lah55, chickens lah55, etc.” Since it can also appear sentence-medially, I do not consider it an SFP and leave it out of the discussion. I note however, that it does seem to be an edge particle, and hope to return to it in future research.

16 To deal with the three variants, Wong (2004) adopts a homophonous

approach, treating each variant as a separate particle, referring to the lah24 of 4 (4a) as “persuasive la” , the lah21 of (4b) as “impositional la” and the lah51 of (4c) as “propositional la”. However, this approach suggests that these SFPs therefore carry lexical tone, a stance which Wong indeed takes (2004, p. 762). Accepting such a homophonous account however, loses the striking distinction that Lim (2007, 2011) makes between the earlier and the later set of the SFPs. Coincidentally, we can observe that only the earlier set has such apparent homophony, since the later “Cantonese set” as specified by Lim (2007) only has one tonal variant for each particle.

To make sense of this, we have to consider another possibility: that boundary tone is at work with the earlier set of SFPs: lah, ah and what. Indeed, Lee (2016b) adopts such a unitary approach in his analysis of lah. By

applying a compositional analysis approach to lah21 and lah51, Lee proposes that lah has an intrinsic meaning that combines with the different pitch contours to produce the pragmatic effects. The meaning of lah(p) is therefore given as “the marked proposition p is entailed by the conjunction of all factors relevant to the truth of p that are true in the actual world” (Lee, 2016b, p. 13). Accordingly, pitch contour 51 indicates that the speaker expects no dispute about p. Lee observes that pitch contour 51 is suspiciously similar to Sag and Liberman’s (1975) surprise-redundancy contour. Pitch contour 21 indicates instead that there is no expectation that the speaker expects dispute about p, which Lee suggests is the “neutral declarative intonation” in Standard American/ described by Pierrehumbert and Hirschberg (1990).

2.2.1.2 ah

While such analysis has only been done for lah so far, it seems likely to apply to ah as well (what does not require it, since no variants have been observed).

4Owing to the largely spoken nature of SgE, there exists some variation in how lah is written. While Wong (2004) uses la, I prefer to adopt the more common spelling of lah. These spelling variations are common throughout the set of SgE SFPs, including the spelling of ah/a, among others. To minimise confusion, they will be pointed out where relevant.

17 Two types of pitch contour have been observed with ah: ah21 and ah24 (Lim, 2007), though with three types of meaning. Examples are given below in (5):

(5) a. Speaker requires a confirmatory response from addressee.

A: You don’t like durian ah21? ah21

‘So you don’t like durian?’

b. Speaker asks a question that requires a response from addressee.

A: Where got sell good durians ah24? ah24

‘Where do they sell good durians?’

c. Speaker is softening an imperative to addressee.

A: Come eat durians later ah24? ah24

‘Come and eat durians later, okay?’

I propose that like lah, ah also combines with different pitch contours to produce these pragmatic effects. I suggest ah indicates that information concerning the proposition p it marks is sought from the addressee by the speaker. This can also combine with the illocutionary force of the sentence to produce different pragmatic outcomes. In (5a), the pitch contour 21 indicates that the speaker expects p to be true. Therefore, felicitous responses by the addressee would either be confirmation, or denying p with an explanation of why p might appear true but not be true. It would be infelicitous for the addressee to respond with a negative answer and not explain why.

However, (5b) and (5c) present a trickier situation: both have the same pitch contour applied on ah, but seem to mean different things! Before continuing however, I wish to draw attention to this pitch contour 24, that we earlier saw on lah, which was mentioned but not discussed by Lee (2016b).

The use of lah24 was described by Wong (2004) to be persuading the addressee gently, which in turn appears as part of (5b) and (5c)! If we take pitch contour 24 to be that of persuasion, then it is possible to analyse (5b) as using

18 persuasion to gently inquire (since leaving the ah would make the question rather forceful). (5c) can then be analysed as using gentle questioning to soften an imperative. Therefore, we see that the pragmatics of ah can be worked out compositionally through the illocutionary force, pitch contour and base meaning of the SFP itself.

2.2.1.3 what

Finally, we turn our attention to what. Lim (2007) suggests that what has no pitch variants, hence we need not apply the same analysis as earlier (though we will return to this at the end of the chapter). Its usage is presented in (6):

(6) A and B are buying gifts for friends whose birthdays are in the following month.

a: Eh you got buy present for C or not? ‘Did you get a gift for C?’ b: No? His birthday is not next month what! ‘No? His birthday isn’t next month after all!’

2.2.1.4 Interim Conclusions

Having considered lah and ah, I suggest therefore, as per Lee (2016a,b), that the tone exhibited on the SFPs of Lim’s (2007, 2011) earlier set is boundary tone and not intrinsic lexical tone. This also means that syntactic treatments for lah and ah need only deal with the base form, instead of the possible variants differentiated by lexical tones, allowing a far cleaner approach.

2.2.2 Tone and the “Cantonese set” of SFPs

Having thus concluded that the earlier set of SFPs possess boundary tone, we turn our attention to the “Cantonese set”. Lim (2007) lists them as lor, hor, leh, meh and ma. I describe them below.

19 2.2.2.1 lor33

The SFP lor33, pronounced with a mid tone, indicates obviousness (Platt and Ho, 1989). This is shown in (7), from the Singaporean (ICE-SIN) component of the International Corpus of English (ICE)5.

(7) A and B are discussing what B plans to cook for an upcoming celebration.

a: So what are you going to cook? ‘What are you going to cook?’ b: Told you wat. Either chicken wing or– ‘I’ve already told you. It’s either chicken wings, or–’ a: Chicken wing, bee hoon. ‘Chicken wings, vermicelli.’ b: Ya lah. Bee hoon or fried rice lor, chicken wing lor. ‘Yeah. Vermicelli or fried rice, and chicken wings.’

(ICE-SIN S1A-007)

In (7), since B has already told A the planned dishes, B responds to A’s question with the SFP wat (usually spelled what). Lor is then used in the list to mark the dishes as presupposed information, since these were previously given.

Additionally, in negative contexts, lor can further convey a sense of resignation (Lim, 2007; Wee, 2002), as in (8):

(8) A and B are discussing A’s recent job applications.

a: I’ve been shortlisted with another girl. She said she will give us a written test ‘I’ve been shortedlisted with another girl. (The inter- viewer) said she will give us a written test.’

5Dialogue presented from ICE-SIN is based on transcriptions of recorded natural speech of SgE native speakers.

20 b: What’s entailed in the essay uh? ‘What will be on that written test?’ a: I don’t know. They refuse to say. They say it won’t be fair. Okay lor. Just go there and shock. ‘I don’t know, they won’t tell me. They say doing so will not be fair (to other candidates). Oh well, I’ll just got there and get a shock (when I take it) then.’

(ICE-SIN S1A-004)

Here, A uses lor to indicate her sense of resignation to not being able to know what the written test will comprise. She then makes her negative expectations clear by mentioning that she expects to be unpleasantly surprised when she takes the test.

Using a grammaticalisation analysis based on Traugott’s approach (1982, 1988, 1989; Traugott and K¨onig,1991), Wee (2002) proposed that the use of lor to indicate resignation developed from the original meaning where lor is used to stress obviousness. He argues that despite Gupta’s (1992b) claim that lor is used to mark a suggestion, her examples actually indicate resignation.

2.2.2.2 meh55

The particle meh55, pronounced with a high tone, signifies negative presup- position from the speaker. This achieves the pragmatic effect of skepticism (Lim, 2007; Wee, 2004).

(9) A and B attended a concert where the singer failed to show up due to having fainted backstage. A and B were not informed of this. While C did not attend, C had heard of the fainting and subsequent non- performance (though C is not clear at which point of time the fainting happened).

c: Didn’t he come out and sing at all? ‘Did he not come out to sing at all?’

21 a: He didn’t.

b: He didn’t.

c: I thought he came out and perform then he fainted? ‘I thought he came out to perform and then he fainted?’ b: No, nonsense.

c: He fainted even before the concert, so flop! ‘So he must have fainted before the concert even began, such a flop!’ a: He fainted meh? ‘Did he really faint?’ c: He fainted. ‘Oh yeah, he fainted.’ a: Really?

(ICE-SIN S1A-068)

In (9), A does not know why the singer did not appear on stage and is skeptical of C’s assertion that the singer had fainted. This skepticism is made clear in the closing line of the dialogue, where A expresses doubt again at C’s statement.

2.2.2.3 ma22

The particle ma22 has been described as presenting a piece of information or advice as being obvious to the addressee (Kwan-Terry, 1991; Wee, 2004). Lim (2011) presents (10) in support of such an analysis.

(10) A wants to know why B made a phone call to him/her.

22 a: How come you call me? ‘Why did you call me?’ b: You page for me ma. ‘You paged for me, after all (so naturally I’m returning your call).’

However, such an analysis of ma runs into problems in situations like (11), where the proposition before ma cannot be obvious to the addressee:

(11) A has discovered B is dating a boy whom A considers not very good- looking and cannot fathom why.

a: Why you go steady with him?! ‘Why are you dating him!’ b: When he proposed to me, he read poems until I very moved ma. ‘Well, I was very moved when he read poetry to me while asking me to be his girlfriend, so I naturally agreed.’

Here, B does not expect A to know the circumstances of the proposal, hence the reason B gives cannot be obvious to A. Yet, ma is used. Instead, I propose that ma marks naturalness; ma(p) indicates that p is a natural course of action. If we adopt such an explanation, then both (10) and (11) make sense.

2.2.2.4 hor: A Problem?

At this point, we stumble upon the first potential problem in Lim’s (2007) classification: the particle hor. While Lim (2007) only covers hor24, it turns out hor21 also exists. Since Cantonese only has ho2, this jeopardises her claim that this set of SFPs were imported from Cantonese with lexical tone intact.

Lim cites Wee (2004) in her description of hor24 as both asserting the proposition it is attached to, and trying to garner support for it (2007). She

23 further notes that since hor24 requires that a speaker assert a proposition, it is only found with questions with declarative form, such as in (12) and never with polar or wh-interrogatives.

(12) hor24 (Gan, 2000, p. 37)

a: I bought a Prada wallet yesterday.

b: You are very rich hor24. ‘You are very rich, aren’t you?’

a: No lah21. Got sale wat21. ‘No, I’m not really. It was a sale (as you know).’

However, hor24 is not the only form hor takes: in addition to the previously discussed form, Marie (1988) describes a form of hor that appears at the end

of an imperative, given in (13). I identify this as hor21.

(13) hor21 (Marie, 1988, p. 45) In a photocopying room, the person-in-charge instructs the students to form a queue.

a: Stand in line hor21. ‘Stand in line, please.’

For her, hor is ambiguous and can be either a question marker soliciting agreement, or a politeness device used to soften the tone of the order in the imperative, depending on whether hor is attached to a declarative or an imperative respectively. Gan (2000) refers to these as “Marked Polar Question hor” and “Request hor” respectively. He also notes two more uses of hor: “Filler hor” and “Repair hor”. However, both forms take the same form as

“Marked Polar Question hor”: hor24. Therefore, since his distinctions appear to be based on their intended conversational effects, I treat them as being syntactically identical to “Marked Polar Question hor”.

24 Kim and Wee (2009) rightly note, however, that there are situations where hor21 is not a question marker soliciting agreement even when attached to a declarative, such as in (14):

(14) hor21 with strong illocutionary force (Kim and Wee, 2009, p. 249)

a: This is my pencil hor21. Don’t touch it. ‘This is MY pencil. Don’t touch it.’ b: Okay.

Instead, they propose despite earlier claims to the contrary that hor weak- ens illocutionary force (Gupta, 1992b), (14) suggests hor can also be used with strong illocutionary force. Instead, they suggest hor has no inherent illocutionary force modifying function and acquires its influence only through discourse contexts. They further propose a speaker who uses hor therefore is asserting a proposition in a context where authority is asymmetrically distributed, and thus confirming that the hearer shares the same mental representation (Kim and Wee, 2009, p. 252).

Using the compositional approach we applied to lah, Lee (2016b) suggests that hor(p) has the meaning of p being in common ground. hor21(p) then asserts that p being in common ground, while hor24(p) (he refers to it as hor35) questions if p is in common ground.

By now, the paradox becomes clear: Lim (2007) claims hor came from Cantonese with its lexical tone intact, yet we see two different types differen- tiated by pitch contour, like the earlier set. Therefore, while hor did come from Cantonese, it seems that it has actually lost its lexical tone and become underspecified. This underspecification then allows it to combine with the pitch contours available in SgE.

At this point, I also note that hor24 can also appear sentence-initially, as seen in (15):

25 (15) A is trying to convince B to let her host a group dinner at her place.

a: (to B) Everyone says my cooking is the best, hor24? ‘Everyone says that my cooking is the best, right?’

a: (turns to C) Hor24? My cooking is the best, right? ‘Right? My cooking is the best, eh?’

Therefore, hor24 seems rather unusual, since unlike other SFPs, it can be separated from the main utterance by an unusually long break in prosody. I note that these two characteristics are associated with tag questions, suggest- ing that hor24 may therefore act like a tag question. This suggestion will be explored in Chapter 4.

2.2.2.5 Another Problem: leh

Finally, we turn our attention to the last SFP in the set: leh. Lim (2007) describes it as having a high level pitch highlighting a comparison. It has also been described as possessing meaning similar to “what about?” (Gupta,

1992b; Platt, 1987). For example, (16) illustrates how leh55 can be used to question in comparison:

(16) leh55 (Lim, 2007, p. 451)

a: My parents very old fashion ah21? Then your parents

leh55? ‘Are you saying that my parents are old-fashioned? But what about your parents?’

At this point, the notion of contrastive topic comes in very handy in de- scribing leh55. Contrastive topic marking indicates that an utterance addresses only one question/issue, while leaving other (implied) topics undiscussed (Con- stant, 2014). Building on earlier work on contrastive topic in English (B¨uring, 2003) and the Japanese contrastive wa (Tomioka, 2010), Constant (2011) identifies three diagnostics for contrastive topic (CT):

26 Diagnostics for Contrastive Topic (Constant, 2011, p. 18)

1. CT marks non-maximal elements like most, but resists maximal elements like all. 2. CT marks partial answers, but resists direct and complete resolution answers. 3. CT marks contrasting sub-questions of a larger issue, but resists simple out-of-the-blue questions.

All of these apply to leh55 in SgE, as shown in (17), (18), (19) and (20):

(17) Leh55 marks non-maximal elements, as in (a) but resists maximal elements, as in (b).

a. Most of the things you want hard to do leh55.

‘It’s hard to do most of the things you’ve asked for.’

b. (??)All the things you want hard to do leh55.

‘It’s hard to do all of the things you’ve asked for.’

(18) Leh55 marks partial answers, as in (a) but resists direct and complete resolution answers, as in (b). a. (Is Charles dieting?)

He say he dieting leh55... (but he’s eating fried chicken.)

‘He said he is dieting... (but he’s eating fried chicken.)’

b. (How did you know Charles is dieting?)

(??)He told me leh55.

‘He told me.’

(19) Leh55 marks contrasting sub-questions of a larger issue. Here, someone knocks on the door and A yells from inside.

27 a: Charles ah21 /*leh55? ‘Is that you, Charles?’

(20) But leh55 resists simple out-of-the-blue questions.

a: Do you want to eat oysters tonight? ‘Do you want to eat oysters tonight?’ b: Nah, too unhealthy.

a: Fish soup leh55? ‘What about fish soup?’

Since all three diagnostics are met, we conclude leh55 does indeed mark contrastive topic in SgE.

Beyond leh55, at least one other form of leh exists, as presented in (21). I identify this as leh21.

(21) A and B are discussing what to get for a friend.

a: Why not get him books?

b: He hates reading leh21 / leh55. ‘He hates reading, though.’ / ‘But he hates reading!’

In (21), using leh21 allows B to convey that the proposition it marks is ei- ther new information, or, reminding A of something he/she hasn’t considered.

Interestingly, it would also be felicitous to use leh55 as well. The reading would then express that the earlier proposition contradicts what B knows about the situation.

Surprisingly, we can have both types of leh in the same sentence:

28 (22) A is comparing John, who didn’t go to school but is rich, with Joe, who studied hard but is poor.

John leh55/*21, never go school but now very rich leh21/55. sfp sfp

‘John, on the other hand, didn’t go to school, but, let me tell you, is very rich now.’ / ‘John, on the other hand, didn’t go to school, but, in contrast, is very rich now.’

Like (21), both leh-s can mark clauses: leh21 marks the sentential clause as informative, while leh55 marks it as contrastive topic. However, leh55 can also mark items smaller than that, such as nominals (such as John).

This points to an underspecification analysis for leh: it has some seman- tics of its own, but derives its full meaning from the relations it enters into following merge. It seems the base meaning of leh could be described as approximating “but”: where either a comparison (as in leh55) would contradict the statement responded to, or new information provided/existing information reminded (with pitch contour 21) would allow further modification. Merging leh with a nominal gives it the contrastive topic reading, while merging it with a clause gives the informative reading. In (22), John in John leh is nominal. Therefore, only the contrastive topic reading of leh55 is available, since nominals do not provide enough information to generate an informative reading. In never go school but now very rich leh though, I suggest the optionality comes from two possible merger sites that are linearised at the edge of the clause. One of these would be clausal (the proposition that John is now very rich), giving the contrastive topic reading, but the other seems propositional, allowing the informative reading from leh21. The possibility of having two merger sites also hints at additional structure present, which we will discuss in Chapter 4.

Adopting the underspecification analysis, like with hor, I suggest leh has lost its lexical tone, allowing it to combine with pitch contours. In leh55’s

29 case, I suggest that high level pitch (55) is not lexical tone, but the end of a comparison intonation, which affects the whole utterance, not just the SFP. This is clear if we leave leh out of the utterance: there would still be a rise in pitch over then your parents? in (16). We will revisit the difference between pitch contour and intonation in Section 2.4.

2.2.3 Conclusion

At this point, I conclude that while Lim’s (2007) system is generally on the right track, it requires some modification to address minor details otherwise unaccounted for. SgE is a dynamic and well-used language, and the recency with which the “Cantonese set” of SFPs entered SgE (1970s up to the late 1980s) suggests that there would still be ample time since her proposal for SgE to further develop. I therefore proceed to examine the syntactic properties of the SFPs discussed above on the basis that they possess basic forms that could be modified by pitch contours, explaining the variants possible.

2.3 Other SFPs and Potential Candidates

Beyond the SFPs described by Lim (2007), there are also other SFPs that have received less attention in the literature, and several SFP-like lexical items on which there is no clear consensus over their status. I briefly cover them here.

2.3.1 Other SFPs

2.3.1.1 one

A common but understudied SFP in SgE is one. Interestingly, while there also exists nominaliser one and pronoun one in SgE, these forms of one behave as nominals (Wong, 2005) and need not be sentence-final. Consider (23):

30 (23) a. Pronoun one:

Don’t buy the green one (on the table).

‘Don’t buy the green one (on the table).’

b. Nominaliser one:

I will take can eat one (from the cupboard).

‘I’ll take edibles (from the cupboard).’

c. SFP one:

He is very smart one (*because he studies).

‘It is the case that he is very smart.’

I therefore only focus on SFP one for now, though we will return to the other uses of one in Chapter 5. Wong (2005) describes SFP one as “...expressing definitiveness and the speaker’s commitment that the given proposition is the one and only way of thinking something...”, noting that one can be used to try and influence the addressee’s way of thinking, such as in (24):

(24) A and B are waiting for a regularly late friend.

a: Almost 2pm already. Where’s he? ‘It’s almost 2pm, where is he?’ b: He said he’s on his way already, probably nearby!

a: Knowing him, sure late one. ‘Knowing him, it is the case that he’ll be late.’

The literature attributes the origins of one to Chinese, like many of the other SFPs, despite its English form (Gupta, 1992a; Wong, 1994). Building on earlier scholarship (Alsagoff and Ho, 1998), Wong (2005) links one in SgE with the particle de in Mandarin. Consider (25) and (26):

31 (25) A is convinced he will fail an exam he just took, and B is reassuring him.

a: I confirm fail this exam lah21. So hard leh21! ‘I am definitely going to fail this exam. It was so hard!’

b: Won’t fail one lah21. You studied so hard! ‘That’s probably not going to happen! You studied so hard (for it)!’

(26) Mandarin equivalent of B’s words in (25):

hu`ı b`uj´ıg´e de la. MandarinB´u not will not-pass sfp sfp Nˇı ni`an l`e zh`em`edu¯o! you study prf this-much

‘That’s probably not going to happen! You studied so hard (for it)!’

In Mandarin, de has also been described as an SFP, as we can see in the gloss of (27) provided by Simpson (2014, p. 158):

(27) Mandarin de as an SFP (Chao, 1968, in Simpson, 2014, p. 158):

¯a b´u hu`ı d¯a-y`ıng nˇı de le ba. MandarinT he not will promise you sfp sfp sfp

‘He won’t promise you now, don’t you think?’

Therefore, one is likely to be a calque that plays the same role that sentence-final de plays in Chinese.

2.3.2 SFP-like Items in SgE

Beyond traditional SFPs in SgE, which tend to be monosyllabic particles that appear at the end of the sentence, there also exist other SFP-like items whose potential status as SFPs are debated. Sentence-final already and sentence-final only are two cases in point (Cheong, 2016).

32 2.3.2.1 Sentence-final already

In SgE, sentence-final already marks perfective and inchoative aspect (Bao and Hong, 2006), as shown in (28) and (29) respectively.

(28) Perfective already (Bao and Hong, 2006): a. I bought a place already. b. They have all moved towards disposables already.

(29) Inchoative already (Bao and Hong, 2006):

Are you spring-cleaning already?

‘Have you started spring-cleaning?’

Parallels have been observed between sentential le in Mandarin and SgE’s sentence-final already (Bao, 2005; Hiramoto, 2015), citing the possibility of substrate transfer from the local Sinitic varieties as an explanation for sentence-final already in SgE. In Mandarin, Soh and Gao (2006) found that sentence-final le gave rise to a perfective reading when the sentence de- noted a telic situation, and an inchoative reading when the situation was atelic. Therefore, they proposed a unification of sentence-final le’s perfectivity and inchoativity as a transition marker. As we have seen in (28) and (29), sentence-final already also has perfective and inchoative properties, hence Cheong (2016) extends their approach to SgE and suggests a similar account for already.

This is similar to Soh’s (2012) analysis of the Malay SFP dah, often referred to as the colloquial counterpart of the aspectual auxiliary sudah in Standard Formal Malay (and often translated into English as already). While sudah only appears pre-verbally (Koh, 1990, p. 204), dah can appear either pre-verbally, post-verbally or sentence-finally. She then argues that pre-verbal dah is semantically distinct from post-verbal and sentence-final dah. While the post-verbal already is unavailable in SgE, it is possible to have already

33 occur pre-verbally and sentence-finally. Soh’s proposal then dovetails with Hiramoto’s (2015) suggestion that the development of sentence-final already in SgE could also have been influenced by Malay, since there is clearly a distinction between pre-verbal and sentence-final forms of dah.

2.3.2.2 Sentence-final only

In Standard English, only is a focus-sensitive operator possibly ambiguous depending on the item being focused. For example, I only introduced A to B could mean B met one person (A) as a result of me (but I also introduced A to others), or that I also introduced B to others, but A only met one person as a result of me. Alternatively, if introduced is stressed, such as in I only introduced A to B, it could mean that all I did was to introduce A to B.

In SgE, sentence-final only can associate with various elements in the predicate, such as the indirect object (30a), direct object (30b), verb (30c), entire predicate (30d), or even the subject (30e):

(30) Association of only in SgE (Cheong, 2016)

a. Clara lend the book to [Jing]F only, (never lend to other people).

‘Clara lent the book only to [Jing]F, (not to anyone else.)’

b. Clara lend [the book]F to Jing only, (never lend other things).

‘Clara lent [the book]F only to Jing, (not anything else.)’

c. Clara [lend]F the book to Jing only, (never give to her).

‘Clara [lent]F the book only to Jing, (she didn’t give it to her).’

d. Clara [lend the book to Jing]F only, (never do other things).

‘Clara only [lent the book to Jing]F, (she didn’t do anything else).’

e. [Clara]F lend the book to Jing only, (other people didn’t lend to her).

‘Only [Clara]F lent the book to Jing, (no one else lent their theirs to Jing).’

34 Interestingly, as Cheong (2016) observes, the uses in (30a-30d) are congru- ent with analyses for the Cantonese sentence-final zaa (’only’) (Law, 2004) and Mandarin’s sentence-final ´eryˇı (’only’) (Erlewine, 2010; Tang, 1998). (30e) however, is found in Law’s (2004) account of Cantonese sentence-final zaa6, but not in what Erlewine (2010, 2016) found in Mandarin’s sentence-final ´eryˇı, where association with the subject is ungrammatical.

Hiramoto’s (2015) research on sentence-final adverbs in Asian Englishes indicates that sentence-final only (13.1%, or 39 sentence-final occurrences out of a total of 298 occurrences of only) is used at a far higher frequency in SgE than in other varieties of English, such as Canadian English (1.7%, or 2 of 120 occurrences) or British English (0.6%, or 1 out of 173 occurrences). She argues that this is due to substrate transfer from the local languages like Malay and the Chinese varieties spoken in Singapore.

2.3.2.3 Sentence-final already and only as SFPs in SgE

The extant literature is mixed on the status of sentence-final already and sentence-final only in SgE. While Gupta (1992b) classifies already as SFPs, Hiramoto (2015) only refers to them as sentence-final adverbs, but notes that they behave like SFPs. Bao (1995) also notes that there may be particles that are regular words in Standard English but perform grammatical and pragmatic functions akin to SgE SFPs, such as already and only.

More importantly, analogous particles fulfilling the same function in Can- tonese and Mandarin have long been considered to be SFPs. Since there is no other equivalent in SgE, it seems reasonable to conclude that already and only play the equivalent roles in the SgE system. Their surface similarities to their Standard English adverbs should not disqualify them from such a designation, especially considering the sheer amount of calquing that we observe in SgE.

Therefore, I consider sentence-final already and only in SgE to be SFPs, as

6Law (2004, p. 29) disagrees with Tang’s (1998) judgement that sentence-final zaa cannot associate with the subject. See footnote 7 of Cheong (2016) for more details.

35 part of the limited set of functional morphemes appearing at the right edge of the clause. While it is also possible to use already and only clause-internally as they tend to be used in Standard English, already and only in such contexts do not behave like SFPs and will not be discussed here.

2.4 Boundary Tone Revisited

Up to this point, I have presented a general overview of SFPs in SgE largely based on the existing scholarship in the field. The SFPs discussed so far are summarised in Table 2.1.

36 Table 2.1: Summary of SFPs in SgE presented so far. SFP Form Description Example(s)

lah lah24 Speaker aims to persuade addressee gently. (4a) lah21 Speaker wants to impose idea or view on ad- (4b) dressee. lah51 Speaker wants to show that proposition is self- (4c) evident or obvious.

ah ah21 Speaker requires confirmatory response from ad- (5a) dressee. ah24 Speaker asks a question that requires addressee (5b), (5c) response. what what Speaker indicates that information is obvious and (6) contradicts something previously asserted.

lor lor33 Speaker indicates information is obvious. (7), (8)

meh meh55 Speaker indicates skepticism. (9)

ma ma22 Speaker indicates that the proposition preceding (10), (11) is a natural course of action.

hor hor24 Seeks confirmation, can be used as a tag question. (12), (15) hor21 Speaker wishes to assert a proposition in common (13), (14) ground with addressee.

leh leh55 Speaker wishes to question in comparison with (16) something earlier discussed. leh21 Speaker indicates that proposition is either new (21) information, or information that addressee hasn’t considered. one one Speaker expresses definitiveness and commitment (24), (25) that the given proposition is the one and only way of thinking something. already already Marks perfective and inchoative aspect. (28), (29) only only Focus marker that can associate with various (30) elements in the predicate.

We have also seen that some SFPs, such as lah, ah and leh, come in several variants. These variants are compositional: the SFP first modifies the

37 proposition p, then the pitch contour is applied to it, audibly manifesting on the SFP. There are three pitch contours available in SgE, summarised in Table 2.2.

Table 2.2: Pitch contours available in SgE. Pitch Contour Meaning 21 (small fall) The speaker does not expect dispute about p. 24 (small rise) The speaker wants to soften the force of the proposition p. 51 (sharp fall) The speaker will accept no dispute about p.

A question naturally follows: can the other SFPs also take these pitch contours? By systematically testing this possibility with each SFP, we obtain the results in Table 2.3.

Table 2.3: Possible pitch contours on SgE SFPs. SFP Contour 21 Contour 24 Contour 51 Q Intonation lah XXX ah XXX what X lor meh X ma X hor XXXX leh XX one X already X only X

The answer is, surprisingly, yes. We are even able to describe variants of SFPs that have not yet been described with pitch contours in the literature, but nonetheless are completely acceptable to native speakers of SgE, such as ma21, hor51 and leh24, the uses of which are shown in (31), (32) and (33) respectively.

38 (31) A is asked how she was talent-scouted to become a basketballer.

a: The coach thought I was quite tall ma21. So he ask me play lor. ‘The coach thought I was quite tall, naturally he asked me to play.’

(32) In a photocopying room, the person-in-charge earlier instructed the students to form a queue, but was ignored.

a: OI! STAND IN LINE HOR51! ‘STAND IN LINE!’

(33) A and B are vice-captain and captain of a sports team respectively. They are deciding the starting line-up for the season, and there is one place left, with two players left competing for it. However, B does not like either player.

a: “I think this guy good.” ‘I think this guy is good.’ b: But I don’t really like him... ‘I don’t quite like him.’

a: The other guy leh55? ‘What about the other guy?’

b: I also don’t really like him leh24. ‘I also don’t quite like him either.’

Beyond the three SFP variants, it also seems that the usual form of the

SFP what actually takes the form of what21 too. We also observe that certain SFPs can take a rising intonation: it is unsurprising that these are SFPs which are compatible with questions in SgE.

39 However, even though the SFP system in SgE appears to be compositional, we must still account for the absence of certain combinations, such as *ma51. I suggest that these combinations are not found because of incompatible semantics: *ma51 would require a very strident imposition of the speaker’s opinion on the naturalness of p while explaining p to the addressee. There- fore, the availability of a base SFP/pitch contour is dependent on semantic compatibility.

Finally, we return to the difference between pitch contours and intonation. Pitch contours tend to manifest themselves mainly at the end of the sentence. In contrast, intonation tends to affect the entire utterance: it is possible to detect a rise in pitch throughout on the sentence You eat dinner? This rising intonation (indicated in Table 2.3 as “Q Intonation”) in SgE, rather unsurprisingly, marks questions: the same utterance of You eat dinner without the rising intonation is interpreted as a declarative. Therefore, it can be found on sentences with SFPs if the SFPs are compatible with question semantics. Intonation is not just limited to questions, we have also seen that SgE has a comparison intonation that manifests in leh55.

2.5 Conclusion

In this chapter, I have described common SFPs in SgE, beginning with those discussed in the literature and then moving on to others that have not received much attention so far. I have also showed that the apparent homophony in certain SFPs is instead a system of base SFPs with their own meaning that combine with pitch contours (which express speaker-addressee attitudes) in a compositional manner to produce complex semantics. In the next chapter, I will explore how these SFPs and pitch contours are positioned in the syntactic structure.

40 Chapter 3

Syntactic Analysis of SgE SFPs

3.1 Introduction

Having described a selection of SFPs in SgE, we now turn to examining the syntactic interactions of these SFPs with the SgE clause. Since evidence suggests that much of the SFP system in SgE originated from the Chinese varieties (with some additional influence from Baba Malay/Bazaar Malay), we first begin by looking at SFPs in these languages, the syntactic status of which have long been actively debated in Chinese linguistics (see Erlewine (2016) for discussion and references).

3.2 SFPs in Varieties of Chinese

Given the Sinitic origins of the SFP system in SgE, it should come as no surprise that SFPs in Chinese varieties are also used to communicate a range of discourse-sensitive meaning concerning speaker attitudes and “emotional colouring” (Matthews and Yip, 1994). Like SFPs in SgE, SFPs in Chinese varieties are phonologically light elements, often monosyllabic, and they typi- cally (in most cases, obligatorily) occur in a sentence-final position (Simpson, 2014). However, notable exceptions include l´aizhe and ´eryˇı, which are disyl- labic. SFPs therefore need not be always monosyllabic, strengthening the case for only (which arguably plays the same role in SgE that ´eryˇı plays in

41 Mandarin) and already in SgE to be considered SFPs.

Following Zhu (1982), Mandarin SFPs have traditionally been divided into three distributional classes with fixed relative orders (Erlewine, 2016;

Paul, 2014; Simpson, 2014). Today, these are referred to as SFP1, SFP2 and

SFP3 (Paul, 2015, p. 360). Some representative members of each class given in Table 3.1 (Erlewine, 2016).

Table 3.1: Distributional Classes of Chinese SFPs. Class SFP Description

SFP1 le Li and Thompson’s (1981) “currently relevant state” marker l´aizhe Recent past (Paul, 2015, pp. 258-260) ne Durative aspect (Constant, 2011) ´eryˇı Exclusive ‘only’ (Erlewine, 2010)

SFP2 ma Polar question ba Imperative ne Contrastive topic (Constant, 2011, 2014) or follow-up and constituent questions (Cheng, 1991)

SFP3 ou Impatience a Softening ei Gentle reminder

Members of SFP1 surface at the rightmost boundary of the verb/predicate phrase and express aspect, while those in SFP2 come next to perform clause- typing functions. Finally, the outermost class of SFP, SFP3, follows and pertains to the speaker’s attitude or feelings. Only one SFP from each distribution class may be found in a sentence, and where there are multiple SFPs from different distribution classes, they adhere to the order outlined above, giving us the sentences found in (34) and (35):

42 (34) SFP1 le comes before SFP2 ma (Paul, 2015, p. 252) a. ¯a b`u ch¯ouy¯an le ma? MandarinT s/he neg smoke prf q

‘Does s/he no longer smoke?’

b. *T¯a b`u ch¯ouy¯an ma le? Mandarin s/he neg smoke q prf

(35) SFP2 ba comes before SFP3 ou (Zhu, 1982, p. 212) a. b’ou! MandarinJ`ınl´ai

enter sfp2.sfp3

‘Hurry, come in!’

b. *J`ınl´ai ou ba! Mandarin

enter sfp3

Paul (2014, 2015) took the standard view that all Mandarin SFPs are head-final heads in the clausal spine (Cheng, 1991; Lee, 1986; Tang, 1998) and proposed that each class corresponds to different heads in an adjusted version of Rizzi’s (1997) split CP hypothesis, shown in Figure 3.1.

CAttitudeP

CForceP Attitude SFP3 ClowP Force SFP2 TP Clow SFP1

Figure 3.1: Paul’s (2014, 2015) proposal for Mandarin SFPs in a split CP approach.

Paul proposed that SFP2 corresponds to the traditional Force head, while

SFP3 is a split C head above Force labelled “Attitude”. Since SFP1 does

43 not directly correspond to any heads within Rizzi’s system, she proposes that SFP1 is a head labelled “Clow”. These facts on ordering and structure generally hold for other varieties of Chinese such as Cantonese, though there have been proposals to split the Force head in Cantonese (Lam, 2014).

Within the three distributional classes traditionally accepted, there is also strong evidence for a split between the low (SFP1) and the high (SFP2 and

SFP3) classes, argued for in Mandarin (Erlewine, 2016; Paul, 2014, 2015) and Cantonese (Tang, 1998). For example, low SFPs may be available in either the matrix or the embedded clause, as per (36), but high SFPs can only occur in matrix clauses, as per (37).

(36) SFP1 ´eryˇı ‘only’ is available in either the matrix or embedded clause (Erlewine, 2012).

g`aos`u-le y¯ı g`e h´aizˇı t¯a k´eyˇı ch¯ı y¯ı g`e MandarinWˇo I told-prf one clf child s/he can eat one clf d`ang¯ao ´eryˇı. cake only

‘I told one child [that s/he can only eat [one cake]F].’

‘I told [one child]F [that s/he can only eat one cake].’

(37) SFP2 ma can only modify the matrix clause (Li and Thompson, 1981, p. 557).

b`u zh¯ıd`ao t¯a l´ai ma. MandarinNˇı You neg know s/he come q

‘Do you not know that s/he’s coming?’ *‘You don’t know whether s/he’s coming.’

Erlewine (2016) suggested that this difference follows from Paul’s (2014, 2015) Rizzian split CP proposal. Since embedded clauses have been proposed

44 to have a “truncated” left periphery (Haegeman, 2006), matrix clause-only distribution of the high SFPs can be explained by hypothesizing that em- bedded clauses are truncated to include the Clow head for SFP1 but not the Force and Attitude heads.

Another difference that has been observed between low and high SFPs is in their position with respect to tag questions (Constant, 2011). This is illustrated in (38) and (39).

(38) Durative aspect ne (SFP1) comes before the tag question (m´eiyˇou) (Constant, 2011, p. 22).

d`ai-zhe (ne) m´eiyˇou (#ne)? MandarinY`aosh´ı

key carry-dur sfp1 not-have sfp1

‘Do you have the keys?’ Literally: ‘Are you carrying the keys or not?’

(39) Contrastive topic ne (SFP2) comes after the tag question (m´eiyˇou) (Constant, 2011, p. 22).

¯angs¯an q`u-gu`o r`ıbˇen. MandarinZh Zhangsan go-exp Japan Nˇı q`u-gu`o (#ne) m´eiyˇou (ne)?

you go-exp sfp2 not-have sfp2

‘Zhangsan has been to Japan. Have you?’ Literally: ‘Zhangsan has been to Japan. Have you or not?’

It seems that in Mandarin at least, low SFPs must come before the neg- ative tag question (in this case, m´eiyˇou), while high SFPs come after the tag question. Using this approach, Constant (2011) argued for a separation of durative ne and contrastive topic ne. Erlewine (2016) then used this to

45 argue for a lower structural height just above the vP for low SFPs, since a position just above the vP while the other SFPs remained up high in the C layer would easily accommodate the position of the negative tag question in the structure below the high SFPs.

Structural height-wise however, contra Paul (2014, 2015), Erlewine (2016) showed that the position of the low SFPs is actually far lower than the split-C position proposed. Using semantic scope evidence from negation, interaction with modals and alternative question disjunction, he demonstrated clearly that low SFPs actually attach above the vP but below the TP, while analyz- ing SFP2 as C and following Paul (2014, 2015) in locating SFP3 in a high neo-performative head labelled Attitude. His proposed structure is given in Figure 3.2.

AttitudeP

CP Attitude SFP3 TP C SFP2 subject T ...

SFP1P

SFP ... 1

vP

Figure 3.2: Erlewine’s (2016) proposal for Mandarin SFPs.

With this understanding of the syntactic organisation of SFPs in Chinese varieties, we now turn our attention to SFPs in SgE.

46 3.3 Substratum Transfer and SgE SFPs

At this point, it is useful to briefly consider the nature of the influence of Chinese varieties in SgE. As earlier discussed, Lim (2007) sketched a linguis- tic history of the SFPs and their origins on an individual SFP basis. She argued that while the original, once-prominent substrates of Bazaar Malay and Hokkien were later replaced as the lingua franca, the earlier particles were not lost and remained robust features of SgE due to the Founder Princi- ple (Mufwene, 1996, 2004). As the original substrates faded from common interethnic use, SgE became a language with increasing numbers of native speakers using it as their dominant language, providing the environment for crystallisation and consolidation of the SgE grammatical system that was compatible with SFPs.

Taking a broader approach, Bao (2015) looked at the aspectual system of SgE, arguing that it showed split loyalty in pattern with the Chinese system in interpretation, but with English in its morphosyntax. Therefore, he argued that SgE derived its grammar from Chinese, but this grammar had to fit English morphosyntax in the transfer process. Hence, only parts of the Chinese system that were compatible with the surface requirements of English morphosyntax were transferred into SgE. Therefore, my approach to characterising the syntax of SFPs in SgE therefore is to test for the existence of properties that might have made the transfer from Chinese, and then branch out into the specifics unique to SgE.

3.3.1 Transferred Properties of SgE SFPs

We earlier saw that one of the most characteristic features of the Chinese SFP system is their division into distributional classes with a fixed relative linear order, with the possibility of multiple SFPs in a single sentence. The examples in (40) attest to the possibility of having multiple SFPs in an SgE sentence.

47 (40) a. You eat dinner already ah21? sfp sfp

‘Have you had dinner?’ b. Speaker is scolding someone who has been secretly helping himself to the speaker’s food in a shared pantry.

You think I don’t know you been taking my food lah21 hor21! sfp sfp

‘I know you’ve been taking my food!’

The property of relative linear ordering when there are multiple SFPs present also exists in SgE, as shown by the examples in (41), when contrasted with those in (40).

(41) a. *You eat dinner ah21 already? sfp sfp

b. *You think I don’t know you been taking my food hor21 lah21! sfp sfp

Our next task therefore is to establish the linear ordering of SFPs in SgE.

3.3.2 SFP Linear Ordering in SgE

The linear ordering seen in (40) and (41) clearly suggests that already must precede ah, and lah must precede hor. By using ordered pairs of SFPs to test for grammaticality, it is possible to determine the relative ordering, and hence the distributional classes that they fall into. For instance, it is possible to say the following sentence in (42).

48 (42) A mobile phone salesperson is making a pitch to a customer who still uses a feature phone (as opposed to a smartphone).

Please lah21! Nowadays, everyone use smartphone only already sfp sfp

one lah21 hor51! You use this phone, super outdated leh21. sfp sfp sfp

‘Come on. These days, everyone only uses smartphones! If you use this phone, you’re really outdated.’

Since (42) had five SFPs in order, it follows that there must be at least five distributional classes of SFPs in SgE, with all the SFPs coming from different distributional classes. In addition, since they can only come in the order presented in (42) (changing the order would have made the utterance ungrammatical), we can surmise that the order that the SFPs appear indicate their relative positions (but not heights) in the structure. This gives us the tentative field presented in Table 3.2.

Table 3.2: SFP Linear Ordering in SgE: A First Pass. Distributional Classes (Left to Right) only already one lah hor

Other SFPs do seem to fall under similar categories: (43a) indicates for instance, that what must come in a distribution class occuring linearly after one but before hor.

(43) The speaker is with a friend awaiting a habitually late friend. The friend is annoyed and bemoans the lateness. The speaker utters the following to express how he isn’t surprised by the lateness.

a. He always late one what, hor24? sfp sfp sfp

‘Well, he’s always late anyway right? No surprises there.’

49 b. He always late one lah21, hor24? sfp sfp sfp

‘Well, he’s always late anyway, right?’

c. He always late one (*lah21) what (*lah21), hor24? sfp sfp sfp sfp sfp

‘Well, he’s always late anyway, right?’

(43b) shows that lah can also occur between one and hor. (43c) then shows that lah and what cannot co-occur, suggesting that they belong to the same distributional class.

(44a) then suggests that ah is in the same category as hor, since it comes after what.

(44) The speaker is eating an unusually large amount of food for lunch and his colleagues express surprise. He then utters the following, pausing at the ellipsis to turn and utter the last SFP to another colleague who is nearby and knows that he didn’t have breakfast. a. I never eat breakfast what... sfp

...ah24? sfp

‘Well, I didn’t have breakfast after all, didn’t I?’

b. I never eat breakfast what... sfp

...hor24? sfp

‘Well, in my defense, I didn’t have breakfast, right?’

50 c. I never eat breakfast what... sfp

...(*hor24) ah24 (*hor24)? sfp sfp sfp

‘Well, in my defense, I didn’t have breakfast, didn’t I?’

Again, (44b) shows that hor can occur in the same surface position as ah in (44a). (44c) then shows that ah falls in the same distribution class as hor, since they cannot appear together.

We therefore update Table 3.2 to include what and ah, shown in Table 3.3:

Table 3.3: SFP Linear Ordering in SgE: Second Pass. Distributional Classes (Left to Right) only already one lah hor what ah

What therefore remains is to achieve a classification of the remaining SFPs: ma, lor, meh and leh. These are set out in (45), (46), (47) and (48) respectively.

(45) A, B and C are having a conversation where B is talking about how he was scolded by his boss at work for not following earlier instructions, but the instructions were contradictory. A asks: So why didn’t you do what he wanted? B then utters the following to explain his dilemma, pausing to turn to C when he says ah for sympathy:

51 a. I didn’t know what he want one ma... sfp sfp

...ah24? sfp

‘Well, it was the case that I didn’t know what he wanted, so naturally I didn’t do anything, right?’ b. I didn’t know what he want one what... sfp sfp

...ah24? sfp

‘Well, it was the case that I didn’t know what he wanted, right?’

c. I didn’t know what he want one (*what) ma (*what)... sfp sfp sfp sfp

...ah24? sfp

‘Well, it was the case that I didn’t know what he wanted, so naturally I didn’t do anything, right?’

The sentence in (45a) clearly indicates that ma must therefore go between the class that contains one and the class that contains ah. (45b) shows that what (belonging to the proposed class) can be in that position. (45c) then shows that it cannot occur with SFPs of that class.

Similarly, (46a) allows us to pinpoint the linear position of lor: it must go between the classes that hold one and ah.

(46) A, B and C are talking about A and C’s colleague, who is very selfish. B asks A Why do you put up with his behaviour? A then utters the following in reply, pausing to turn to C when she says ah for agreement:

52 a. He’s the super selfish kind of person one lor33... sfp sfp

...ah24? sfp

‘He’s just the super selfish kind of person, nothing I can do, right?’

b. He’s the super selfish kind of person one ma... sfp sfp

...ah24? sfp

‘He’s just the super selfish kind of person, hence nothing I can do, right?’

c. He’s the super selfish kind of person one (*ma) lor33 (*ma)... sfp sfp sfp sfp

...ah24? sfp

‘He’s just the super selfish kind of person, nothing I can do, right?’

(46b) shows that ma (from the proposed class) can occur in that positions as well, then (46c) shows us that it also cannot co-occur with SFPs of its proposed class.

We then find meh’s position in the linear order from (47):

(47) a. A and B are talking about a friend who has decided to live healthy, but still drinks beer daily. B is incredulous and replies to A:

Want to be healthy can still drink beer one meh, hor24? sfp sfp sfp

‘Is it really the case that you can drink beer and still be healthy? I assume you think that’s a valid question too?’

53 b. A replies to B’s response:

Even super healthy people also got drink beer one what, sfp sfp

hor24? sfp

‘Well, even super healthy people do drink beer, right?’

c. Identical context to (a):

Want to be healthy can still drink beer one (*what) meh sfp sfp sfp

(*what), hor24? sfp sfp

‘Is it really the case that you can drink beer and still be healthy? I assume you think that’s a valid question too?’

(47a) shows that meh must go after the distributional class that contains one and before the class containing hor. (47b) then shows that what (from the proposed class) can also occur in that position. (47c) then shows meh cannot co-occur with SFPs from its proposed class.

Finally, we turn to leh. While we earlier proposed in Chapter 2 that leh has two possible positions, only the informative reading position is strictly sentence-final. To avoid conflating them, we restrict ourselves to sentences where leh has the informative reading. This restricts us to examples with

leh21.

54 (48) Charles and Leanne meet a friend, who asks if Charles would like to work as a barista one day. Leanne expresses her certainty that Charles would like that, uttering the following and turning to Charles

for agreement when she says hor24:

Of course! He likes to drink coffee one leh21, hor24? sfp sfp sfp

‘Of course! You know, it’s the case that he likes to drink coffee, right?’

(48) therefore indicates that leh21 must come between one and hor24, suggesting it might be in the same class as lah, what, ma, lor and meh. (49) also indicates that leh21 cannot co-occur with any of these SFPs, offering support to co-classification.

(49) a. *You like to drink coffee (*lah) leh21 (*lah). sfp sfp sfp

b. *You like to drink coffee (*what) leh21 (*what). sfp sfp sfp

c. *You like to drink coffee (*ma) leh21 (*ma). sfp sfp sfp

d. *You like to drink coffee (*lor) leh21 (*lor). sfp sfp sfp

e. *You like to drink coffee (*meh) leh21 (*meh). sfp sfp sfp

However, unlike previously, I propose that leh21’s inability to co-occur with these SFPs is not due to it being in the same class, but due to semantic conflict. For reasons that will become clear in the next chapter, I propose that leh21 instead belongs to a distributional class by itself between the class containing one and the class containing what.

We can then update Table 3.3 to Table 3.4, as well as give tentative identities (a-f, to avoid confusion with SFP classes in Chinese varieties) to

55 each distribution class for clarity. However, this field only indicates the relative ordering, with the exact position of each distributional class within the clause yet to be fleshed out. Using this field, we will now proceed to probe the structural height of each distributional class within the clause.

Table 3.4: SFP Linear Ordering in SgE: Third Pass.

SFPa SFPb SFPc SFPd SFPe SFPf only already one leh lah hor what ah lor meh ma

3.3.3 The Low/High Divide in SgE SFPs

Earlier, we saw that there was a split between the low (SFP1) and the high

(SFP2 and SFP3) distributional classes in Mandarin and Cantonese based on two main characteristics:

1. Low SFPs are available in both the matrix and embedded clauses, while high SFPs only modify the matrix clause.

2. Low SFPs occur before question tags, while high SFPs occur after question tags.1

We therefore proceed to test for these characteristics in SgE SFPs.

3.3.3.1 Embedded and Matrix Clause Modification

(50) to (53) therefore demonstrate that such a divide does exist:

1There does not appear to be any concrete proposal for the structure of these question tags (see footnote 4 of Erlewine (2016) for more details). However, I think they still serve as a useful diagnostic of the differences between low and high SFPs.

56 (50) He tellF1 me Charles eatF2 only.

sfpa

Reading 1 : ‘He only [told me that Charles ate]F.’ (tell is stressed)

Reading 2 : ‘He told me that Charles only [ate]F.’ (eat is stressed)

(51) He tellF1 me Charles eatF2 already.

sfpb

Reading 1 : ‘He already told me that Charles ate.’ (tell is stressed) Reading 2 : ‘He told me that Charles already ate.’ (eat is stressed)

(52) I know he very tall one.

sfpc

Reading 1 : ‘It’s the case that I know that he is very tall.’ (matrix one) Reading 2 : ‘I know that it’s the case that he is very tall.’ (embedded one)

(53) a. I think [he very tall] leh21.

sfpd

‘You know, I think he’s very tall.’

b. *I think [he very tall leh21].

sfpd c. A and B are having a conversation about a mutual friend. A remarks: I think he’s very tall. B utters the following in response:

I don’t think soi leh21.

sfpd

‘You know, I actually don’t think he’s very tall.’

SFPs from SFPa (50), SFPb (51) and SFPc (52) allow readings in both the embedded and matrix clauses.

57 In contrast, leh21 (SFPd) can only modify the matrix clause and not the

embedded clause. (53a) shows the target of leh21, while (53b) shows that

leh21 cannot modify the embedded clause. (53c) then shows very clearly that

leh21 must modify the matrix clause.

3.3.3.2 SFP Positions with Question Tags

We then turn our attention to the positions of SFPs in SgE with respect to question tags to see if Constant’s (2011) observation from Mandarin holds in SgE as well. Our results from the previous section suggest that low/high

divide is between the SFPc and SFPd distributional classes, hence we begin testing to verify a split there.

However, since leh21 provides an informative reading, it cannot take a

question tag. Adding a question tag to leh21 would have the effect of saying “You know, I like X. Do I?”, where the speaker immediately questions his own

statement. Unfortunately, none of the SFPs in SFPe are compatible with tag questions either. While the reasons for this will be apparent in the next chapter, this leaves us in the undesirable position of having to resort to an

SFP from SFPf to test for properties ascribed to the high side of the low/high divide.

(54) and (55) thus indicate the respective positions of SFPs from both distributional classes in the context of a question tag situation.

(54) Position of ah (sfpf) with respect to tag questions.

a. He got come not ah24?

Tag-q sfpf

‘Did he come or not?’

b. *He got come ah24 not?

sfpf Tag-q

58 As expected, ah from SFPf must come after the question tag and cannot come before it. This confirms that it is therefore a high SFP.

However, one exhibits rather unexpected behaviour, as we see in (55).

(55) Position of one (sfpc) with respect to tag questions.

a. He usually late one not?

sfpc Tag-q

‘Is he usually late or not?’

b. He usually late not one?

Tag-q sfpc

‘Is it the case that he is usually late, or not?’

c. He usually late one not one?

sfpc Tag-q sfpc

‘Is it the case that he is usually late, or not?’

As would be expected of low SFPs, (55a) shows that one can occur before tag question marker not. However, (55b) shows that it can also appear after the tag question marker not, a property we would instead expect to see in high SFPs only. In fact, (55c) shows that it is even possible to have one appear twice: both before and after the question tag marker.

To understand this, we have to return to Constant’s (2011) arguments for ne, which first motivated this distinction in low and high SFPs. He argued that there were two forms of ne: one in SFP1 and the other in SFP2, with the former marking durative aspect and the latter marking contrastive topic. As such, ne appearing before the tag is interpreted as aspectual, while ne appearing after the tag is interpreted as marking contrastive topic. While not explicitly stated, it seems that a situation akin to that in (55c) would have conflicting meanings that make the sentence untenable (which also explains why Constant indicated the sentence’s ungrammaticality with a # instead of

59 a *). In fact, in response to the inevitable counter-argument that two distinct forms of ne should also mean that they should recur (so we should see ne ne, but we don’t), Constant made a convincing case using an explanation based on haplology earlier put forward to explain perfective aspect le and inchoative le. Referring to Chan’s (1980) work that argued for a morpho- phonological haplology constraint against adjacent realisation of two distinct and interpretable morphemes le, he argued that a similar constraint applied to ne, barring two ne-s from appearing sequentially.

We also observe such constraints in the case of one, as shown in (56):

(56) a. He said X one.

‘It is the case that he said X.’

b. A possible proposition that would fit in X:

Charles eat dinner very noisy.

‘Charles eats dinner very noisily.’

c. A possible proposition with one that would also fit in X:

Charles eat dinner very noisy one.

‘It is the case that Charles eats dinner very noisily.’

d. He said that [Charles eat dinner very noisy] one.

‘It is the case that he said that Charles eats dinner very noisily.’

e. ?He said that [Charles eat dinner very noisy one] one.

‘It is the case that he said that it is the case that Charles eats dinner very noisily.’

If we take (56a) to represent a complex sentence with both matrix and embedded clauses, then (56b) and (56c) represent two embedded clauses that can be uttered independently. Replacing X with (56b) to give (56d) is also perfectly acceptable. However, replacing X with (56c) to get (56e) is

60 ungrammatical. It seems there must be some form of morpho-phonological haplological constraint that blocks the pronunciation of both one-s, possibly similar to what has been earlier described for le (Chan, 1980) and ne (Con- stant, 2011).

Such an analysis is backed up by native speaker judgements: while the ten speakers consulted found (56e) acceptable, they rated it as being very marginally so. However, relative to (56e), all speakers rated (55c) very highly (where the two one-s are separated by the tag question marker not and hence no haplology occurs).

To deal with the question of where both one-s must go in a sentence like (55c), let us look at (57):

(57) Derek is known to his friends to enjoy both craft beers and speciality coffee. When his friend is asked what Derek drinks recently, his friend replies:

Nowadays, Derek drink craft beer only one already.

sfpa sfp sfpb

‘These days, it is the case that Derek exclusively drinks craft beer.’

Therefore, it seems one can also appear below already, in addition to above it as in (42).

Interestingly, the haplology analysis proposed for one can also be used to support the underspecification analysis previous proposed for leh in Chapter

2. Since leh21 gives an informative reading, and leh55 gives a contrastive topic reading, we should expect to see grammatical sentences with both types. However, as we see in (58), that is not the case.

61 (58) Identical context as (44), where the speaker utters the following in response to the colleague’s surprise at his massive lunch.

??I never eat breakfast leh55 leh21.

CT sfpd

‘Well, just so you know, I didn’t eat breakfast.’

While the underspecification analysis would predict this to be acceptable, native speakers consulted found this unacceptable, possibly due to haplology. One possible way to get around this is to exploit the fact that there are two variations in how leh is pronounced in SgE: some speakers pronounce it [le], while others pronounce it [læ]. Therefore, if the unacceptability of (58) is due to haplology, then a sentence using different phonetic variants of leh should be more acceptable.

(59) a. I never eat breakfast one [le]55 [læ]21.

sfpc ct sfpd

‘Well, just so you know, I didn’t eat breakfast.’

b. I never eat breakfast one [læ]55 [le]21.

sfpc sfpd ct

‘Well, just so you know, I didn’t eat breakfast.’

c. *I never eat breakfast one [le]21 [læ]55.

sfpc sfpd ct

‘Intended: Well, just so you know, I didn’t eat breakfast.’

As it turns out, this is exactly what native speakers reported: both (59a) and (59b) were considered acceptable to most speakers, indicating that this acceptability was not due to the phonetic variation used. Additionally, re- versing the order of leh55 and leh21 led to a strong unacceptability judgement. Therefore, I conclude that this supports the underspecification analysis pre- viously adopted for leh, and that leh55 (henceforth, contrastive topic leh) must come after one but before leh21 in the linear order and in the syntactic

62 structure.

Having introduced contrastive topic leh into the linear order, we must then reapply the diagnostics for the low/high divide, beginning with the embedded/matrix modification ability.

(60) a. I think [he very tall] leh55. sfp

‘You know, I think he’s very tall.’

b. *I think [he very tall leh55]. sfp c. A and B are having a conversation about a mutual friend. A remarks: I think he’s very tall. B utters the following in response:

I don’t think soi leh55. sfp

‘You know, I actually don’t think he’s very tall.’

Accordingly, contrastive topic leh therefore seems to be a high SFP. Unlike leh21, contrastive topic leh is compatible with tag questions, allowing us to apply that diagnostic.

(61) Position of contrastive topic leh with respect to tag questions.

a. He got come not leh55? Tag-q sfp

‘Did he come or not?’

b. *He got come leh55 not? sfp Tag-q

Since contrastive topic leh appears after the tag, we conclude that it does appear to be a high SFP. This provides additional support for a low/high

63 divide in SgE SFP, and it also allows us to locate it between sfpc and con- trastive topic leh.

We therefore update Table 3.4 to Table 3.5, updating the identities of the distributional classes to reflect their position relative to the low/high divide, and include contrastive topic leh. For reasons that will be clearer at the end

of the chapter, I label contrastive topic leh as SFPH-CT and do not give it a numeric designation.

Table 3.5: SFP Linear Ordering in SgE: Fourth Pass.

SFPL1 SFPL2 SFPL3 SFPL4 SFPH-CT SFPH1 SFPH2 SFPH3

only one already one leh55 leh lah hor what ah lor meh ma

3.3.4 SgE SFP Structural Heights

In contrast to the well-studied structure of SFPs in Chinese varieties, scarcely any research has been done to flesh out the structural heights of SgE SFPs. Based on Erlewine’s (2016) work in Mandarin, Cheong (2016) applied the same approach in SgE to examine the structural heights of sentence-final already and only (albeit referring to them as “sentence-final adverbs”). We will therefore build our understanding of SFP structural heights in SgE off her existing work.

Cheong begins by looking at the structural height of already (SFPb). She shows that it takes scope over subject quantifiers, as in (62).

64 (62) Already takes scope over subject quantifiers.

No one eat pineapple tarts already.

‘As of now, no one eats pineapple tarts.’ (already > no one) *‘No one has eaten the pineapple tarts.’ (no one > already)

Since subject quantifiers are in Spec-TP, if already takes scope over them, then it follows that already (and therefore, SFPb) must be positioned above the TP.

Following that, Cheong also indicates that only has two possible heights, given the ambiguity seen in (63):

(63) Only can be ambiguous (Cheong, 2016, p. 35). a. Spanish teachers in a university complain to colleagues who teach other languages that no one studies languages anymore, but their colleagues then reply that only Spanish has no students; other languages are still studied.

[No student study Spanish]F only.

‘It is only the case that no student studies Spanish; some students study French and Japanese.’ (only > no student) b. A response to the question Is there anyone who only does Spanish, or must they take up at least another language as well?

No student study [Spanish]F only.

‘There is no student who studies only Spanish; every student (also) studies French or Japanese.’ (no student > only)

She concludes therefore that only positioned above the TP yields the reading in (63a), while only positioned below the TP but above the vP would yield (63b). The possibility of only being above the vP but below the TP is unsurprising: Erlewine (2016) has already shown that the class of Mandarin SFPs that includes ´eryˇı (linked with SgE only) also contains the vP but

65 is contained by the TP. Since each distributional class occupies a different position in the structure, we must therefore consider the possibility that there are two forms of only, much like how Mandarin has two forms of ne and two forms of le. The structurally lower only then modifies the vP, and the structurally higher only modifies the TP.

One way to test the possibility that there might be two forms of only is to again see if we can have both forms appear in the same sentence, as in (64):

(64) Charles and Leanne only ate plain rice without accompanying dishes. Derek told B that Charles ate only rice, but did not mention that Leanne did the same, and B told this to A. A later found out that Leanne also only ate plain rice and accuses B of providing insufficient information, but B protests that he was not given that information in the first place.

a: Why you tell me that only Charles eat rice at dinner and not Leanne too! ‘Why did you only tell me that Charles ate rice at dinner, and not tell me that Leanne did so as well!’ b: (?)Derek tell me Charles eat rice only only!

‘Derek only1 [told me that Charles only2 [ate rice]F2]F1.’

Here, it seems impossible to have both proposed forms of only appear in the same sentence. However, we have to consider again that this is impossible due to haplology again. Indeed, B’s utterance in (64) is marked as ? instead of * because some speakers rated it as acceptable, but only marginally. Furthermore, for the speakers who found it clearly acceptable, it had to be produced with a very specific prosody that assigned intonational breaks between the first only and the second only, along with major stress differences: further evidence that the unacceptability might have been morpho- phonological rather than syntactic. We can then try to get around the haplological constraint by substituting the second only for nia, a Hokkien

66 word that has the same meaning and frequently used in situations where SgE is mixed with Hokkien.

(65) Same context as (64), except we substitute nia for the second only 2.

a: Why you tell me that only Charles eat rice at dinner and not Leanne too! ‘Why did you only tell me that Charles ate rice at dinner, and not tell me that Leanne did so as well!’ b: Derek tell me Charles eat rice only nia!

‘Derek only1 [told me that Charles only2 [ate rice]F2]F1.’

The clear grammaticality of (65) among the native speaker informants therefore is strong evidence that the inability to have only appear twice in (64) was due to haplology, and that there are actually two forms of only: a structurally lower one that modifies the vP, and the structurally higher only which modifies the TP. That two forms exist is unsurprising too, since varieties of Chinese have two words that mean ‘only’ and introduce the semantics of exclusivity: one preverbal and one sentence-final. In Mandarin, this would be the preverbal zhˇı and the sentence-final ´eryˇı, while Singaporean Hokkien has the preverbal balu and the sentence-final nia. The uses of these words are shown in (66) and (67) for Mandarin and Singaporean Hokkien respectively.

(66) ‘What does he do on Saturdays?’ (modified from Erlewine, 2016, p. 19)

a. ¯a zhˇı [k`an di`ansh`ı]F. MandarinT he only watch TV

‘He only watches TV.’ ⇒ He doesn’t do anything else.

b. ¯a [k`an di`ansh`ı]F ´eryˇı. MandarinT he watch TV only

‘He only watches TV.’ ⇒ He doesn’t do anything else.

2It is also possible to substitute nia for the first only while keeping the second only, with no effects on interpretation.

67 c. ¯a [zhˇı1 [k`an di`ansh`ı]F1]F2 ´eryˇı2. MandarinT he only watch TV only

‘He only watches TV only.’ ⇒ He absolutely doesn’t do anything else.

(67) ‘What do you all usually eat?’ a. baluIJ chiahIJ p¯ng. Hokkieng´oalˆang

we onlyF [eat rice]F

‘We only eat rice.’ ⇒ We don’t eat anything else.

b. chiahIJ p¯ng ni`a. Hokkieng´oalˆang

we [eat rice]F onlyF

‘We eat rice only.’ ⇒ We don’t eat anything else.

c. baluIJ chiahIJ p¯ng ni`a. Hokkieng´oalˆang

we [only1 [eat rice]F1]F2 only2

‘We only eat rice only.’ ⇒ We absolutely don’t eat anything else.

It is therefore plausible that the calquing process of only in SgE resulted in both preverbal and sentence-final ‘only’-s in Chinese being represented by a single lexical form in SgE that nonetheless maintains distinct syntactic statuses. It might be that the preverbal form also acquired the ability to be sentence-final, giving us the possibility of two sentence-final only-s (with only one being pronounced due to haplology).

Finally, our last task to is to consider where the second only might go. Consider (68):

68 (68) A cafe offers a choice of three set meals: rice, noodles, or a mix of both. A knows B has eaten there and asks B: Have you tried all the set meals at the cafe? B then replies:

I eat rice only already only. Haven’t try the others yet.

sfpL1 sfpL3 sfp

‘I have only tried the set meal that is only rice. I have not tried the others yet.’

We see that the second only comes after already. However, we still need to determine its position relative to the higher SFPs. (69) indicates that the second only must come before one.

(69) A and B are talking about their sons. A tells B that her sons all became tall and athletic when they hit puberty, saying They become tall one. And athletic also. B replies that her son already has hit puberty and grown tall, but did not become athletic.

He become tall already only one.

sfpL3 sfp sfpL4

‘It is the case that he has become tall only, but not athletic.’

We therefore update Table 3.5 to Table 3.6, with the new distributional class for the TP-modifying only taking over the SFPL4 description and the previous SFPL4 now redesignated as SFPL5.

Table 3.6: SFP Linear Ordering in SgE: The Final Picture.

SFPL1 SFPL2 SFPL3 SFPL4 SFPL5 SFPH-CTSFPH1 SFPH2 SFPH3

only one already only one leh55 leh lah hor what ah lor meh ma

Our final task in examining the structural heights is to determine the

69 height of SFPL2 one. Earlier, we saw Cheong’s (2016) approach to doing so for already and only using scope. Following her approach, we examine (70):

(70) [No student [study syntax only]F1 one1 already]F2 one2.

sfpL1 sfpL2 sfpL3 sfpL5

‘It is the case that it is the case that no one studies syntax exclusively.’

Since the lower only clearly only scopes over the vP, we conclude that it must attach above SFPL1 only but below the TP (where already is).

Based on the above, I propose the following tentative first pass structure in Figure 3.3. At this point, I remain agnostic as to the exact identities of each SFPxP, as well as to their precise location relative to the CP. It may well be that these projections are separate projections in the CP layer, or higher layers, or they might also be individually located above/below the traditional Rizzian split-CP heads of Force, Topic and Focus.

70 SFPH3P

SFPH2P SFPH3 hor ah SFPH1P SFPH2 lah what SFPH-CTP SFPH1 leh lor meh SFP P SFP L5 H-CT ma leh55

SFPL4P SFPL5 one

SFPL3P SFPL4 only

TP SFPL3 already subject T ...

SFPL2P

SFPL1P SFPL2 one SFP ... L1 only vP

Figure 3.3: A first pass proposed structure of SgE SFPs.

3.3.5 Rethinking the Low/High Divide

At this point, we start to see clear divisions between groups of projections. We first have two groups of low SFPs with different structural heights: a lower group that attaches just above the vP but below the TP, and a higher group that attaches above the TP. Above that, we have the high group of

71 SFP classes with distinct syntactic behaviour as demonstrated earlier.

Beyond that however, further evidence for the distinctness of the high group of SFPs emerges if we look at the interactions of SFPs with pitch contour (earlier described in Chapter 2). If we rearrange the SFPs shown there by their structural heights, a clear divide emerges.

Table 3.7: Possible Pitch Contours on SgE SFPs: Rearranged. SFP Contour 21 Contour 24 Contour 51 Can Be Q? only X one X already X only X one X leh XX lah XXX what X lor meh X ma X hor XXXX ah XXX

Clearly, pitch contours only target the high group of SFPs: clear evidence that these form a distinct syntactic class. This also supports the earlier deci- sion to classify contrastive topic leh, which is not targeted by pitch contours, as a separate class.

In light of this, I suggest that the low/high dichotomy earlier proposed based on varieties of Chinese is insufficient: we need to distinguish at least four groupings of possible SFP distributional class structural heights. I therefore refer to the lowest vP-modifying group as the low group, the group above the TP excluding contrastive topic leh as the middle group, followed by contrastive topic leh, and the highest group as the high group. These

72 distinctions will prove themselves very useful later when we establish their relative positions within the various layers of syntactic structure.

3.3.6 Situating The Groups

Now that we have a clearer picture of where each group could fit into the structure, we can turn our attention to situating the groups. I first begin by looking at the groups as targets of the two phonological processes that act on SFPs: intonation and pitch contour. Consider (71):

(71) You eat dinner already one lah21, hor24?

sfpL3 sfpL5 sfpH2 sfpH3

‘It’s the case that you’ve eaten dinner, right?’

Structurally, we can divide (71) into two zones:

Low/Middle High z }| { z }| { Y ou eat dinner already one lah21, hor24 | {z } | {z } intonation pitch contour

Clearly, intonation and pitch contour target two separate stretches of the syntactic structure. We now consider (72) to see where contrastive topic leh fits in this:

(72) You eat dinner already one leh55.

sfpL3 sfpL5 ct

‘But it’s the case that you’ve already eaten dinner.’

Low/Middle CT z }| { z }| { Y ou eat dinner already one leh55 | {z } intonation

Therefore, I conclude that contrastive topic leh must therefore be syn- tactically grouped with the low and middle SFPs, while the high SFPs are

73 syntactically distinct in the structure.

At this point, it would be reasonable to propose that the low group mod- ifies the vP and is possibly part of the extended projection of the vP. The middle group, by virtue of being above the TP, must therefore be in the CP-domain. Since contrastive topic leh is part of the same syntactic group, I suggest that it is located within the CP-domain as well.

What is interesting however, is the location of the high group. Since it is not targeted by intonation, I suggest that the high group is actually not in the upper reaches of the CP-domain, as might be traditionally assumed, but in a separate domain that deals with discourse and speech acts (Wiltschko, 2016, 2017). I refer to this as the discourse domain, and explore this analysis in the next chapter.

74 Chapter 4

SFPs in the Discourse Domain

4.1 Motivating the Discourse Domain

Within Minimalism, the three-domain structure of the clause is well accepted: the CP is the pragmatic domain, the TP domain handles tense and as- pect, and the vP houses argument and event structure (Van Gelderen, 2012, pp. 65-67). Following the work done in the cartographic enterprise, each domain can be split into various functional projections. It is in the spirit of such an approach that a discourse domain above the CP has been proposed.

The first proposal for a discourse domain came from Ross (1970), where he suggests that there is a layer of structure that dominates the root clause of a sentence and encodes the illocutionary force of the utterance. Ross analyses this layer of structure as a type of clause, with a subject (the speaker), ob- ject (addressee) and predicate (a verb of saying), with performative deletion happening to reduce the output into the overt clause we hear. Since even declarative clauses can now be considered performative (saying I am hungry is effectively saying I say that I am hungry), his analysis is known as the performative hypothesis. However, there were also many problems that chal- lenged the hypothesis, leading to its rejection by the field (Wiltschko, 2017).

Wiltschko (2017) argues though that the general insight that speech acts

75 should be syntactically represented is still valuable and has recently been revived in part by the discovery of functional projections, what she refers to as the neo-performative hypothesis. Much like how event structure has been decomposed through generative semantics to allow the postulation of func- tional categories within the vP-layer (Ramchand, 2008), speech act structure has been decomposed through postulation of several functional categories above the CP-domain (Wiltschko, 2017). Accordingly, various disparate types of linguistic units in a variety of languages have been analysed as encoding speech act structure: evidential markers (Speas and Tenny, 2003), sentence- peripheral particles (Haegeman, 2014), response particles (Krifka, 2013) and even please in English (Woods, 2015).

The evidence is clearly in favour of discourse-oriented syntactic structure above the CP-domain, and accordingly, I adopt Wiltschko’s (2017) proposal of a complex discourse domain with two distinct speech act layers to account for the presence of the high group of SFPs in SgE.

4.2 The Complex Speech Act Structure

The neo-performative hypothesis aims to capture the empirical generalisations that Ross (1970) made, while avoiding the same criticism as the original pro- posal. Accordingly, where Ross (1970) proposed a regular matrix clause of the type S[NP VP] as the superordinate structure, neo-performative approaches such as Speas and Tenny (2003) propose a complex speech act phrase above the CP with two layers. The higher head (sa) has the speaker in Spec-saP, taking a sa*P as its complement. In turn, utterance content is in Spec-sa*P, with the lower sa* head taking the addressee as its complement. Speech act structure is therefore argued to behave like a double object construction: the speaker is the agent, the utterance is the theme, and the hearer is the goal.

However, Wiltschko postulates a different model of speech act structure, led in part by the fact that empirical evidence exists that the addressee role is structurally higher than the speaker role (Heim, Keupdjio, Lam, Osa-

76 Gomez, and Wiltschko, 2014; Lam, 2014). For example, Cantonese has two SFPs: me1 and ho2. Me1 combines with a declarative to produce a request for confirmation with a negative bias on behalf of the speaker. Ho2 also derives a request for confirmation, but suggests that speaker assumes the addressee believes the proposition p introduced by ho2. Accordingly, Lam (2014) classifies me1 as a speaker-oriented particle, as it expresses the speaker’s belief (or lack thereof). Ho2 is taken to be a addressee-oriented particle, as it expresses the speaker’s belief about the addressee’s belief. Both particles can co-occur, but their ordering is restricted: me1 must come before ho2, as we see in (73) from Lam (2014, p. 64):

(73) Jimmy is the first in a taxi queue. A taxi arrives, but someone cuts the queue, saying loudly that he is in a hurry. Everyone in the queue is angry. Jimmy says this to the second person in the queue.

seng1 zau6 dak1 gaa3 laa3 me1 ho2 Cantonesedaai6 big voice then okay sfp sfp sfp sfp

‘What, can one get by by being loud? I assume you’d agree it’s a valid question, right?’

Given what we have earlier seen about SFP linear order and structural height in varieties of Chinese, the addressee-oriented ho2 must therefore be structurally higher than the speaker-oriented me1. Accordingly, Lam (2014) proposes the following structure shown in Figure 4.1 for what she describes as a “complex ForceP”.

ForceAP

ForceSP ForceA ho2

XP ForceS me1

Figure 4.1: Lam’s (2014) proposal for a complex ForceP in Cantonese.

77 In addition, Wiltschko notes that there is an additional dimension involved when an SFP is added to a declarative in Cantonese. Quoting Bach and Harnish (1979), she describes the typical conditions of use for declaratives as follows:

(i) speaker believes proposition(p) conveyed by own utterance.

(ii) speaker wants addressee to adopt p into addressee’s set of beliefs.

(Bach and Harnish, 1979)

However, the addition of an SFP in Cantonese modifies the declarative into a request for confirmation, or in other words, it changes what the speaker expects the addressee to do in response to the utterance. Where a simple declarative would not necessarily require a response, the addition of the SFP now requires a response from the addressee. Beyssade and Marandin (2006) refer to this aspect of the speech act as the Call on Addressee (CoA), and Wiltschko (2017) notes that this is not just limited to Cantonese, but can be observed in English, as in (74).

(74) a.You are leaving. English b.You are leaving, eh? English

In (b), eh modifies what we see in (a) to do two things: it modifies the CoA, and it also modifies the speaker’s commitment. Unlike Bach and Harnish’s (1979) condition (i) that applies in (a), the speaker in (b) is clearly not fully committed to the proposition.

Accordingly, Wiltschko proposes a different model of a complex speech act structure, given in Figure 4.2. In this model, there are two distinct layers that comprise the speech act structure: a layer that encodes speech act participants’ commitments towards p, and a higher layer where the CoA is represented. These layers are referred to as GroundP and ResponseP respectively: the label GroundP evoking Clark and Brennan’s (1991) notion of grounding, as

78 well as the notion of a common ground in discourse between participants (Wiltschko, 2017).

ResponseP

GroundP

p

Figure 4.2: Wiltschko’s (2017) proposal for a complex speech act structure.

Since we earlier saw that it is possible for speech acts modifiers to express the speaker’s assessment of the addressee’s commitment towards p, it is proposed that GroundP is complex: there is one position for the speaker’s

commitment (GroundS), and one for the addressee’s commitment (GroundA). Naturally, the latter is structurally higher than the former. In this model therefore, the Cantonese example in (73) can be analysed as follows:

1. The proposition p (that one can get by just by being loud) forms the CP, in the lowest structure of the model.

2. Me1 (in GroundS) then modifies the speaker’s commitment to p, since it indicates that he is not entirely committed to it.

3. Ho2 (in GroundA) then seeks the addressee’s commitment to p, in the form of a confirmation response to the speaker.

Finally, following the work of Pierrehumbert and Hirschberg (1990), Heim et al. (2016; 2014) suggest that intonational contours are best analysed as intonational morphemes in the case of Canadian English, Cantonese and Medumba, and can modify the CoA in this model. Such an analysis is unsur- prising: it has been suggested for Cantonese elsewhere (Sybesma and Li, 2007; Wakefield, 2016), and English (Gussenhoven, 2004; Trinh and Crnic, 2011).

79 Therefore, we recognise that it is indeed possible to consider phonological contours as a form of suprasegmental morphemes.

The articulated grounding layer proposed is shown in Figure 4.3, and we now turn our attention to the analysis of high SFPs in SgE as being part of the discourse domain.

ResponseP

Phonological groundP Contours

GroundA GroundP

p GroundS

Figure 4.3: Wiltschko’s (2017) proposal for an articulated complex speech act structure.

4.3 SgE SFPs and the Discourse Domain

In the previous chapter, I proposed that the high group of SFPs are actually located in the discourse domain, instead of the complementiser domain as might be traditionally assumed. Figure 4.4 presents a simplified structure for reference.

80 SFPH3P

SFPH2P SFPH3 hor ah SFPH1P SFPH2 lah what CP SFPH1 leh lor meh ma

Figure 4.4: Position of the high SFPs in the discourse domain.

We already know that pitch contours in SgE are able to target the high group of SFPs. Recasting them slightly from how they were earlier defined, I suggest that it is possible to view them as modifying the CoA, with the old meanings (earlier seen in Chapter 2) and the recast meanings given in Table 4.1.

Table 4.1: Pitch contours available in SgE recast. Pitch Contour Meaning 21 (small fall) old: The speaker does not expect dispute about p. recast: Addressee is called upon to accept p. 24 (small rise) old: The speaker wants to soften the force of p. recast: Addressee is called upon to recognise that the speaker is trying to offer p gently. 51 (sharp fall) old: The speaker will accept no dispute about p. recast: Addressee is being forcefully encouraged to accept p without question.

Incorporating them into the structure, we get Figure 4.5.

81 ResponseP

SFPH3P Pitch Contour Contour 21 Contour 24 SFP P SFP H2 H3 Contour 51 hor ah SFPH1P SFPH2 lah what CP SFPH1 lor leh meh ma

Figure 4.5: Structure of the discourse domain in SgE.

Going back to the Cantonese example in (73), it happens that me1 and ho2, which have been described as GroundS and GroundA respectively, have equivalents in SgE. Lim (2007, p. 461) links meh55 in SgE with me55 in Cantonese, noting that the latter “indicates a highly marked ‘surprise’ question, checks truth of unexpected state of affairs (common in rhetorical questions)”. She also links hor24 in SgE with meh35 in Cantonese, with the latter “expect[ing] confirmation of a statement or suggestion.” Indeed, it is even possible to have a close SgE equivalent of (73), given in (75).

(75) Identical context to (73), where Jimmy says the following to the second person in the queue:

Loud voice can anyhow meh, hor24? sfp sfp

‘What, can one get by by being loud? I assume you’d agree it’s a valid question, right?’

As it turns out, the other SFPs in SFPH2 do indeed turn out to be speaker- oriented: lah expresses a speaker’s opinion, what indicates the speaker’s disagreement with the proposition being responded to, lor indicates the

82 speaker’s resignation/lack of strong feelings about the proposition it marks, and ma indicates the speaker’s belief that p was reasonable/natural.

Similarly, hor either seeks to give the addressee an order or seek confirma- tion and while ah seeks information about p from the addressee. Therefore,

I propose to identify SFPH2 with Wiltschko’s GroundS and SFPH3 with

Wiltschko’s GroundA in SgE, giving us the updated structure in Figure 4.6.

ResponseP

groundP Pitch Contour Contour 21 Contour 24 GroundP GroundA Contour 51 hor ah SFPH1P GroundS lah what CP SFPH1 lor leh meh ma

Figure 4.6: Updated structure of the discourse domain in SgE.

We are now left with the task of identifying SFPH1, consisting only of leh. Unfortunately, it does not seem to correspond to any of the projections proposed by Wiltschko (2017). However, recall that leh has the semantics of comparison and contrast: leh21 is used to indicate either new information, or information the addressee has not considered, while leh24 is used when a speaker wants to soften the impact of a statement that might not go down well on the addressee. We see these in (76).

83 (76) a. When A sees B eating a massive lunch, A asks, surprised, Why are you eating so much?!

B: I never eat breakfast leh21.

sfpH1

‘As you might know, I didn’t have breakfast, hence the massive lunch.’ b. A cooks lunch of roast beef for B and serves it to B, who suddenly becomes rather sheepish.

B: Forgot to tell you, I’m vegetarian leh24.

sfpH1

‘Actually, I forgot to tell you I’m vegetarian.’

In both examples here, there is a common element of the speaker providing information about himself that the addressee does not know. Additionally, the speaker in both examples demonstrates awareness of the addressee’s reaction and beliefs. Therefore, I suggest that both forms of leh here have both speaker orientation and addressee orientation. I summarise these in Table 4.2.

Table 4.2: Speaker and Addressee Orientation in leh. Form Orientation

leh21 speaker Speaker’s belief (knowledge) of p: that speaker has not had breakfast. addressee Speaker’s belief that addressee’s beliefs do not include p. addressee Speaker’s belief that addressee’s beliefs find it nat- ural for people who haven’t had breakfast to eat a big lunch.

leh24 speaker Speaker’s belief (knowledge) of p: that he is vege- tarian. addressee Speaker’s belief that addressee’s beliefs do not include p. addressee Speaker’s belief that addressee’s beliefs will make addressee upset that speaker cannot eat the roast beef cooked.

84 Leh is therefore unlike the other projections, which have either speaker orientation only, or addresse orientation only. Therefore, I propose that leh instantiates a type of grounding that has so far not been proposed to be part of

the structure: GroundC (C for common, evoking Stalnaker’s (2002) proposal).

In GroundC, the proposition marked is neither solely speaker-oriented, nor solely addressee-oriented, but shared between both (producing the effect of don’t you know...p? in English).

Lexicalisation of common grounding however, is not limited to SgE: it seems that the German discourse particle denn might also instantiate such mutual speaker-addressee shared grounding. German is already known to

have GroundS and GroundA discourse particles: an example of each would be eh and doch respectively (Thoma, 2014), hence the presence of GroundC in German should not come as a surprise either. An example of denn’s use is given in (77), taken from Bayer and Obenauer (2011, p. 450).

(77) a. wohnst du? GermanWo where live you

‘Where do you live?’

b. wohnst du denn? GermanWo where live you DENN

‘Where do you live? (I am wondering)’

While (77a) is a simple question that seeks information, (77b) conveys that the speaker is “concerned” about the proposition that would be yielded by the answer. In this case, denn serves to tone down the question and make it sound less blunt. If denn is used in a question with a negative element, it also signals reproach, with the speaker expecting a justification from the addressee.

85 (78) ist denn hier los? GermanWas what is DENN here happening

‘What on earth is going on here?’

Bayer and Obenauer (2011) describe denn in (77b) as signalling refer- ence to a common ground between the speaker and addressee beyond the presupposition in (77a): no such additional common ground would be as- sumed in (77a). Here, denn expresses speaker belief that the addressee has information/knowledge on what has happened. This is speaker belief on the addressee’s belief: addressee orientation. However, denn also expresses speaker belief on what the speaker wants: speaker orientation (P. Milton, pers. comm.). This explains why using denn in a quiz show context is infelicitous (since it assumes the addressee has the information desired, which may not be true). However, in the same context, if the host asks a question that he knows the contestant clearly knows the answer to, and does so jokingly (because everyone present knows too that the contestant definitely knows the answer), denn is felicitous (S. Pfitzner, pers. comm.). Therefore, denn does seem to

be another example of a GroundC particle.

I further speculate that other possible semantics that might be found in such a position would be reminders and fragments, such as as you know. What is interesting about this however, is that by marking a proposition with leh, the speaker also indicates to the addressee that the proposition was possibly earlier given, and the addressee has to search existing shared knowledge to find this reference (since this knowledge is evidently known by both speaker and addressee). This search might or might not be successful, as we see in (79) and (80).

(79) Successful search: A and B are discussing sports their friend C might be good at and should therefore try.

a: I think since he likes indoor sports, maybe basketball?

86 b: But he’s not tall leh21. ‘But he’s not tall though?’ a: Oh yeah, he’s quite short.

In (79), we see an example of a successful search: the speaker indicates to the addressee that the proposition p (p: C is not tall) should be known to the addressee, and the addressee realises that and acknowledges his/her knowledge of p.

(80) Unsuccessful search: A and B are discussing what sports their friend C might be good at and should therefore try.

a: I think since he likes indoor sports, maybe basketball?

b: But he’s not tall leh21. ‘But he’s not tall though?’ a: Is he? I didn’t realise!

In contrast, in (80), we see an example of an unsuccessful search: the speaker again indicates to the addressee that the proposition (p: C is not very tall) should be known to the addressee, but the addressee indicates that p is not known to him/her and therefore could not already be in the common ground.

This possibility of success/failure in the act of reminding has been ob- served elsewhere: Bar-Hillel (1971) notes that remind is not a success verb in English, since it is possible to keep reminding for some time. What is crucial is the observation that leh can trigger a search for information that could be in the common ground and can serve as an antecedent. We will return to this observation in the next section.

87 Therefore, I update Figure 4.6 to Figure 4.7, using notation similar to established notation for projections in the verbal domain.

ResponseP

groundP Pitch Contour Contour 21 Contour 24 ground*P GroundA Contour 51 hor ah GroundP GroundS lah what CP GroundC lor leh meh ma

Figure 4.7: The final proposed structure of the discourse domain in SgE.

4.4 The Case of hor24

At this point, we return to the case of hor24, which we have generally treated as an SFP, but with additional tag question characteristics, discussed in Chapter 2. Its use is reproduced in (81).

(81) A is trying to convince B to let her host a group dinner at her place instead of a restaurant.

a: (to B) Everyone says my cooking is the best, hor24? ‘Everyone says that my cooking is the best, right?’

a: (turns to C) Hor24? My cooking is the best, right? ‘Right? My cooking is the best, eh?’

In (81), we see that hor24 can appear sentence-initially like the tag com- ponent of a tag question. It clearly references the earlier proposition (p:

88 everyone says that A’s cooking is the best). It would be infelicitous for A to turn to C and utter the same however if what A said to B was not earlier said. In fact, if C happened to have been inattentive and missed the earlier proposition, then he would have been incapable of a felicitous

response. This clearly suggests that hor24 is therefore not a standalone ut- terance when used sentence-initially: it must reference an antecedent, and hence such sentence-initial use looks more like a case of ellipsis in tag questions.

Tag questions in English have been proposed to be regular yes/no questions that have undergone e-given ellipsis (Merchant, 2001) from their antecedent clauses (Sailor, 2009). Sailor argues that a proposition must be first made explicit in the discourse for it to be confirmed in a tag question. Licensing el- lipsis of this proposition merely requires its interpretation be recoverable from the immediate discourse. There is thus no specific mechanism that “generates” candidates for ellipsis: ellipsised clauses do so because their interpretation can be recovered. Hence, no special “tag question construction” exists.

Sailor’s insights are illuminating and allow us to explain other sentence- initial uses of hor that superficially appear to be interjections, such as in (82):

(82) A has just discovered that B has dared to do something that A specifi- cally warned him not to do.

a: Hor51/hor21! You dare to do that! ‘IS THAT RIGHT? HOW DARE YOU!’ / ‘Is that right? How dare you!’

Again, the use of the other two variants in (82) looks like ellipsis: I specu- late that the full form of the utterance might be something of form Like that hor!, with the like that undergoing ellipsis (because it is recoverable from the immediate discourse context). The semantics of hor might thus be described as being very similar to the expression is that right? in English.

89 Therefore, we see that hor behaves very much like a tag question: superfi- cially sentence-initial uses of hor involve ellipsis. In such contexts, hor then searches for its antecedent in the discourse context. This is strikingly similar to our earlier analysis of leh, with the main difference being that hor has a stronger commitment and bias: used felicitously, hor is more likely to be responded to with a confirmation. This difference might result from hor being addressee-oriented and hence elicits information the addressee is expected to have, while leh is in the common ground where information is being negotiated.

I suggest that a possible way of accounting for this syntactically is that hor and leh contain a feature that can be valued by meaning, a mechanism Wiltschko (2016) has referred to as m-valuation. However, unlike syntactic feature valuation by agree, failure of m-valuation does not create an un- grammatical sentence. Since m-valuation operates on pragmatic features, failure (such as in the unsuccessful search we saw with leh) leads to a prag- matic failure of speaker expectations being incorrect. Due to scope and space constraints, I cannot explore this further.

4.5 Conclusion

In this chapter, I proposed a speech act structure-containing discourse domain above the CP-domain in SgE, following similar proposals in German and Cantonese. This speech act structure is complex and contains at least two layers: one dealing with addressee-oriented material and the other dealing with speaker-oriented material. With the exception of leh, high SgE SFPs can be classified as either speaker or addressee-oriented. However, there is another projection for common ground, which hosts leh, and possibly denn in German. Finally, I examined leh and hor, finding that both seem capable of ellipsis that requires antecedents in the preceding discourse, possibly through m-valuation.

90 Chapter 5

SgE and The Final-over-Final Condition

5.1 The Final-over-Final Condition

Within the recent generative literature on word order typology, there has been much discussion about the Final-over-Final Condition (FOFC), a proposed universal on structure-building and linearization. Proposed by Holmberg (2000), it was revised by Biberauer et al. (2008, 2014; 2009) to its current form presented in (83).

(83) The Final-over-Final Condition (Biberauer, Holmberg, and Roberts, 2014, p. 171):

A head-final phrase αP cannot dominate a head-initial phrase βP, where α and β are heads in the same extended projection.

FOFC predicts that of the possible combinations among head-initial and head-final categories within an extended projection, it is not possible to have a head-final category dominate a head-initial category, as in Figure 5.1.

91 αP αP αP αP

α βP βP α α βP βP α

β γ γ β γ β β γ (a) HI over HI (b) HF over HF (c) HI over HF (d) *HF over HI

Figure 5.1: Predictions of FOFC.

The configurations of strictly head-initial or strictly head-final are very common, possibly for learnability reasons (harmony is preferred), but dishar- mony is also allowed. However, this disharmony has to be of a specific type: that is, head-initial projections dominating head-final projections. While seemingly innocuous, this generalisation holds across categories, and typologi- cally diverse languages, in the both verbal and nominal extended projections, and even constrains linguistic change (Biberauer, Holmberg, and Roberts, 2014; Biberauer, Newton, and Sheehan, 2009).

Biberauer, Holmberg and Roberts (2014) then go on to make the empirical claim that structure of the kind disallowed by FOFC are indeed not found in the world’s languages based on a survey of such structure across a large sample of languages around the world, thereby establishing a strong case for FOFC as a possible syntactic universal. Nevertheless, claims have been made that there are exceptions to FOFC. Many of these claims of FOFC exceptions have focused on particles and their syntax, a famous counterexample being Chinese, where head-final SFPs dominate head-initial clause (Paul, 2014, 2015). Since SgE SFPs are also head-final and dominate a head-initial clause, they also appear to violate SgE. I therefore discuss such apparent violations in the next section.

92 5.2 Particles and FOFC

5.2.1 FOFC and Chinese SFPs

To support the claim that Chinese SFPs violate FOFC, Paul (2014, 2015) provides a Rizzian split CP analysis of these SFPs, which we have earlier seen in Chapter 3. However, we have also seen Erlewine’s (2016) work that shows not all Chinese SFPs are uniformly positioned high in the clause. He argues the class previously identified by Paul as Clow is actually positioned far lower in the structure, dominating the vP. I present his proposal again in Figure 5.2:

AttitudeP

CP Attitude SFP3 TP C SFP2 subject T ...

SFP1P

SFP ... 1

vP

Figure 5.2: Erlewine’s (2016) proposal for Mandarin SFPs.

However, the SFPs in this structure would still appear to violate FOFC: the head-final SFP1 dominates a head-initial vP, while the higher SFPs dominate the head-initial TP. To explain the apparent FOFC violation of the SFP1 class, Erlewine suggests an approach based on Richards’ (2016) Contiguity-based analysis of FOFC. This analysis proposes that FOFC only holds in individual Spell-Out domains (complements of phase heads) but

93 not across extended projections. Therefore, Erlewine proposes that if the low SFPs are taken to be heads of the lower phase (and hence are Spell-Out domains), then the low Chinese SFPs do not actually violate FOFC.

However, the Contiguity analysis does not explain the fact that super- ficially FOFC-violating elements consistently exhibit defectivity, such as in East Asian languages with system-defining homophony. Neither does it ex- plain how these final elements frequently do not meet Wiltschko’s (2014) non-projection tests. Such an analysis would therefore have to ascribe these cross-linguistically very consistent properties to coincidence, which is a very unsatisfactory outcome. Therefore, I follow Biberauer (2017) in concluding that a Contiguity-based analysis is unsatisfactory (see note 82 of Biberauer (2017) for details).

Any proposal seeking to explain these apparent FOFC violations must therefore include an explanation for such particle defectivity. In her review of potentially FOFC-violating particle-containing structures, Biberauer (2017) proposes several formal configurations for such structures which would not lead to FOFC violations. Among them, she proposes that some of the claimed FOFC-violating particle structures in the literature actually contain a parti- cle categorially distinct from the head initial structure, either by bearing a distinct categorial feature, or by lacking a categorial specification altogether (and are hence defective elements). Therefore, they do not actually violate FOFC.

Along these lines, it has been proposed that Chinese SFPs are acategorial and neither associated with [+V] nor [+N] specifications (Biberauer, Newton, and Sheehan, 2009). Accordingly, if they are acategorial, they are therefore always categorially distinct from the head-initial structure and not subject to FOFC. However, since acategoriality results from lacking a categorial specification, it is possible that there might be a diversity of acategorial particles differing in other aspects beyond lacking a categorial specification. We therefore take a closer look at such possible diversity in the next section.

94 5.2.2 Types of Particle Acategoriality

Biberauer (2017) distinguishes two kinds of acategorial particles: those with formal features ([F]s), and those that completely lack [F]s and are hence formally inert. Listing the Marshallese Q-particle le, Bulgarian and Macedo- nian Q-particle -li and German wh-final modal particles as good examples of the former, she notes that it is clear from their distributional constraints that they bear [F]s. For instance, Marshallese ke is obligatorily clause-final in negative structures, but has positional freedom in affirmative structures. German modal particles also seem to need [F]s to interact with the speech act layer in the discourse domain via some form of agree, since they appear to be positioned at the vP (Bayer and Obenauer, 2011; Cardinaletti, 2011).

For those completely lacking [F]s, she suggests they often appear in systems with multiple instances of apparent homophony, citing Vietnamese as an example. However, proposing acategorial particles without [F]s leads us to another problem. If merge operates on the basis of [F]s, how would such items be merged into the derivation? Biberauer suggests that if we assume the existence of lexical arrays (LAs), then one possible solution is to merge such elements after elements that can be merged on the basis of their [F]s have been merged. Her proposal predicts that such elements will always be spelled out at the edges of phasal domains. This dovetails with what we observe in SgE: the low SFPs appear at vP-edge, while the middle SFPs and contrastive topic leh appear somewhere plausibly CP-edge. Additionally, we know that already and only seem to have homophonous counterparts that can appear preverbally, similar to their uses in British English.

5.2.3 Acategoriality and Borrowing

However, distinguishing acategorial particle types also lets us make predictions about how particles are borrowed in contact situations. Frequently, such particles tend to be acategorial, and are therefore merged phase-peripherally and sentence-finally.

95 Biberauer (2017) argues that such borrowing involves speakers taking elements from other languages without an [F]-specification and merging these elements into an existing [F]-based structure, citing SgE and Hong Kong English (example from Gibbons (1987) below). As [F]-less elements, these borrowed particles must then be merged at phase-edge as predicted, which turns out to be the case.

(84) matter when the first time I do philosophy HK EnglishItdoesn’t

le1, I met the same problem with you gaa3. sfp sfp

‘It doesn’t matter when I first did philosophy; when I did, I encountered the same problems as you did.’

(85) leh55, I also had problems the first time SgEPhilosophy ct

I do leh21.

sfpH1

‘When it comes to philosophy, you know, I also had problems the first time I did it.’

In fact, (85) shows that leh in SgE can be merged at two types of phase edges: at the edge of the nP(leh55), and the edge of the CP (leh21).

Since the origin of many SgE SFPs have been shown to be Sinitic, this then lends credence to the idea that Chinese SFPs, or at least those borrowed into SgE, are acategorial particles completely lacking an [F]-specification. Accordingly, speakers of SgE and Hong Kong English are merely borrowing [F]-less elements into their language.

96 5.2.4 Acategoriality, FOFC and SFPs

We now return to the proposal that Chinese SFPs are acategorial. Already, the evidence that there are borrowed particles merged at the phase-edge in languages influenced by Chinese supports this proposal for some Chinese SFPs (at least, for those that were borrowed by Hong Kong English and SgE).

Correspondingly, Biberauer (2017) suggests that grammaticalised Q- particles (such as ma in Mandarin) are acategorial but still possess formal features. While lacking a categorial specification, they still possess a pro- jecting [Q] feature that marks the sentence as a question and takes the the clause as a complement. Such a proposal also accounts for the fact that ma is prosodically integrated with the clause. More importantly, it is compatible with Paul’s (2015) work suggesting that ma encodes interrogative force. I therefore conclude that Chinese also does have acategorial SFPs possessing formal features.

Taking into account the evidence presented, I therefore adopt the proposal that acategoriality is a possible way to explain the apparent FOFC violations in Chinese. Returning to SgE, I suggest that such an approach might be useful in our study of SgE SFPs. If we can show that an SgE SFP is acategorial, then the proposed violation by the SFP in question disappears.

5.3 Resolving FOFC and SgE SFPs

At this point, we are ready to propose a resolution for the apparent FOFC violations by SgE SFPs. Motivated by the observation that in SgE, phonology targets different stretches of structure for intonation and pitch contour re- spectively, I refer to the two segments as intonational SFPs and pitch contour SFPs. I then deal with the apparent violations in each targeted segment separately.

97 5.3.1 Intonational SFPs

Recall that the intonational SFPs comprise three groups: the low group above the vP, the middle group located at the CP-edge, and the contrastive topic leh, presented in order of ascending structural height. There is already clear evidence that contrastive topic leh is acategorial: it does not c-select and can merge with both nominals and clauses (and possibly others). The latter fact is also significant, in that these are phase edges: behaviour predicted by Biberauer’s proposal on acategorial [F]-less elements. Therefore, I propose that contrastive topic leh does not violate FOFC by virtue of being acategorial.

The next step therefore, is to examine the low and middle SFPs for acategoriality as well.

5.3.1.1 The Low and Middle SFPs

To make any claim of acategoriality for the low and middle SFPs in SgE, we first have to examine their positions in the syntactic structure. Figure 5.3 shows their relative heights.

98 DiscourseP ...

SFPL5P

SFPL4P SFPL5 one

SFPL3P SFPL4 only

TP SFPL3 already subject T ...

SFPL2P

SFPL1P SFPL2 one SFP ... L1 only vP

Figure 5.3: Structural heights of the low and middle SFPs in SgE.

In the structure, one seems to follow only, while already somehow must come above the TP (on the basis of its scope). It is also possible to have an additional one and only after already. The question then arises: could we have multiple already-s? The answer is yes, as we see in (86):

99 (86) Charles, Leanne and Derek are meeting for a karaoke session. Usually, they will have dinner together first before proceeding to sing. However, Leanne doesn’t join Charles and Derek for dinner. Puzzled, Derek asks why this is the case and Charles replies:

She nowadays eat dinner at home already one already.

sfpL3 sfpL5 sfp

‘In a recent change, it is now the case that she has started eating dinner at home.’

It seems that we might have to split the projections even further, but this is not limited to already: it turns out that there can be at least another one in the structure, as we see in (87).

(87) He got grow taller one already one or not one?

sfpL2 sfpL3 sfpL5 Tag-q sfp

‘Is it the case that it is the case that it is the case that he has already grown taller?’

Clearly, we cannot possibly continue positing additional distinct projec- tions each time we find it is possible to add another SFP into the structure. While further extension is difficult (native speakers reported that there were too many embedded propositions to process, since each additional one cre- ated yet another layer of embedded proposition), I speculate that it would probably be the case that we can repeat this order freely should processing considerations not be a limiting factor. However, if we adopt an [F]-less acategoriality explanation for already, only and one, then it follows that there should be no limit to the number of times each item can occur, bar processing constraints. We can therefore account for the possibility of sentences such as those in (88).

(88) a. He eat dinner one only already one.

sfpL2 sfp sfpL3 sfpL4

‘It is already the case that it is the case that he ate dinner only.’

100 b. He eat dinner already one only already.

sfpL3 sfpL5 sfp sfp

‘Already, it is only the case that he already ate dinner.’

However, with a fixed ordering approach, the only way to explain the additional SFPs in both (88a) and (88b) is to propose even more layers of repeating structure, which is unsatisfactory.

Such an approach also greatly simplifies our explanation for why these SFPs appear where they do: these positions are phase edges and such SFPs are merged last in the lexical array after all other elements that can be merged on the basis of their [F]s are merged. Theoretically, this is more desirable. It therefore simplifies our treatment of strings like only...one. Under the previous approach, we would have been unsure whether the underlying structure would

have been sfpL1...sfpL2 or sfpL1...sfpL5. We were thus forced to posit two

distinct one-s in sfpL2 and sfpL5 respectively. Treating them as acategorial items then allows us to instead suggest there is one featurally-deficient item that can be merged freely (at the phase edges).

However, we still face a potential problem: if already can merge at any phase edge, then why must it have scope over the TP? I suggest this could be because already grammaticalised from a temporal adverb (preverbal already). Unlike regular adverbs, temporal adverbs have been suggested to attach in the Spec-TP instead of at the vP to allow them appropriate scope (Alexiadou, 2000; Ernst, 2002). While grammaticalised, SFP already still possesses similar semantics that require it to attach relatively high up in the structure, above the TP. This is in line with what Aldridge’s (2011) and Cardinaletti’s (2011) analyses of Mandarin and German/Italian suggest.

At this point, I briefly return to the other uses of one described in Chapter 2, both of which were nominal. If we pursue an acategorial analysis for one (and hence, one is underspecified), then we can see the nominal properties of both pronoun one and nominaliser one as the result of the same underspecified

101 one being merged with a covert nominaliser. What is interesting however, is that this also relates to the particle de in Mandarin. We earlier noted that Mandarin de has been linked in the scholarship to SgE one. De has also been proposed to be the overt realisation of nominaliser n in Mandarin (Zhang, 1999), which is also a phase head. Therefore, we see yet again that the acategorial one is merged at a phase edge as Biberauer predicts, lending support to this proposal to one.

If we accept that already, only and one are acategorial, then several unresolved issues begin to come together. We first remove the need to stip- ulate that there are two (or more) positions in the structure for the same item to be merged: one above the vP and the other above the CP. Instead, these are merely acceptable positions for acategorial particles to be inserted, being phase edges. We also remove the need to posit additional projections to explain the iterability of already, only and one. Therefore, I adopt this analysis for already, only and one. As acategorial particles, already, only and one are therefore not considered to be FOFC-violating.

I present the attachment possibilities for already, only and one in Table 5.1 and the revised proposal for the structure of the intonational segment of the larger syntactic structure in Figure 5.4.

Table 5.1: Attachment possibilities for already, only and one. SFP vP phase edge CP phase edge already X only XX one XX

102 DiscourseP ...

SFPH-CT

Contrastive Topic ... leh55 CP

C TP ∅ subject T ...

vP

Figure 5.4: The structure of the intonational segment in SgE. Dashed lines indicate phase boundaries.

5.3.2 Pitch Contour SFPs

Finally, we turn to the SFPs targeted by pitch contour. While we have already seen evidence that some of these SFPs, such as leh and hor might be underspecified, it is difficult to extend such an analysis to all of the SFPs in this category. Therefore, I turn to the alternative approach that Biberauer (2017) suggests in resolving apparent FOFC violations: that these particles are categorially distinct, hence they do not violate FOFC either.

While a discussion of phonological targeting of syntactic constituents would be outside the scope of this study, it seems plausible to suggest on the basis of what we have seen so far, in the phonology, the clause (bounded by the CP) and the elements marking the grounding of this clause are treated distinctly. Accordingly, it seems that intonation targets the clause, and pitch contours target the grounding elements.

103 If we accept that already, only, one and leh55 are acategorial, then it follows that the [+V] extended projection would be able to project right through them to the end of the CP phase periphery. Therefore, I propose that the failure of intonation to target the grounding SFPs is because these grounding SFPs are categorially distinct from the [+V] extended projection that intonation targets. They are therefore free to realise pitch contours overlay additional semantic content on their basic underspecified meanings.

Therefore, I propose that this is evidence that these grounding SFPs do not violate FOFC either.

5.4 Summing Up

In this chapter, we looked at FOFC and its apparent violations by SFPs in SgE. By adopting an analysis that regards already, only, one and leh55 as acategorial, we were able to greatly simplify the proposed structure and explain the distribution of these SFPs as being at phase edges. We could also then explain other apparently unconnected uses of these SFPs, such as the distribution of nominal one. More importantly, if this analysis is correct, then these SFPs no longer violate FOFC.

We also explored the apparent violations of FOFC by the grounding SFPs situated in the discourse domain. While it was not possible to apply the same acategorial analysis uniformly across all such SFPs, the fact that the phonology separately targets discrete stretches of the syntactic structure gave us evidence of that these grounding SFPs were structurally distinct. Closer examination of the stretch targeted by intonation then suggested it was the clause (which formed the [+V] extended projection), suggesting that the grounding SFPs were therefore not in this extended projection, since they remained untargeted. Therefore, despite being head-final, these grounding SFPs would also not violate FOFC.

104 Chapter 6

Conclusion

This thesis began with the goal of developing a syntax of SFPs in Singaporean English. Since existing scholarship on SFPs in SgE has been somewhat scat- tered, I began Chapter 2 with an empirical description of the entire SFP system, describing each SFP as part of a larger system. In that process, I had to deal with the question of tone in SgE SFPs, since various scholars have differing treatments for SFP variants that have the same surface form but with different pitch contours. Pace Wong (2004), I adopt a unitary approach to such SFP variation, following Lee (2016b) in proposing that such variants are compositionally formed from an underspecified SFP that carries some semantics and a pitch contour that overlays additional pragmatic meanings onto it. This approach aligns with Lim’s (2007, 2011) diachronic work on SFP origins, untangling why some SFPs seem to carry lexical tone while others have pitch contours. Finally, I show these pitch contours are not merely restricted to the SFPs with variation: it is also possible to reanalyse some of the other SFPs without reported variants as possessing base meaning modified by one of the pitch contours, with semantic clashes ruling out other pitch contour combinations.

Having thus established the parameters of variation for SFPs, I then examined their linear ordering in Chapter 3 using a cartographic approach. By systematically testing ordering possibilities, I established their relative

105 linear positions, finding that they group into distributional classes. I then applied diagnostics developed for Chinese SFPs to see if there were structural similarities in the distribution of these classes. While the broad strokes were similar, it turned out that SgE SFPs exhibited a considerably more complex structure than Mandarin SFPs, though less complex than that of Cantonese as proposed by Sybesma and Li (2007). The earlier work in Chapter 2 on pitch contours then allowed me to demonstrate that pitch contours only targeted a certain group of distributional classes, which I referred to as the “high”’ group. I also found that intonation targets the rest of the sentence excluding the SFPs targeted by pitch contour. Since other evidence suggested this group targeted by pitch contour appeared to have a very distinct syntactic location, I suggested they might actually be above the CP-domain, in a distinct domain dedicated to discourse and speech acts, following Wiltschko (2017).

This suggestion was explored in Chapter 4, allowing us to identify a class of SFPs as being speaker-oriented, and another structurally higher class as being addressee-oriented. However, it turned out there was also a lower position occupied by another class, of which leh was the sole member. I then suggested this lower position was for common ground particles that both contain speaker orientation and addressee orientation, drawing support from similar particles found in German (denn). Finally, I returned to take a closer look at hor24, drawing comparisons with leh. I then proposed hor24 behaves like a tag question, and unlike other SFPs, allows its antecedent to undergo ellipsis. I then concluded leh and hor were capable of obtaining their antecedents from the preceding discourse.

In Chapter 5, I examined apparent violation(s) of FOFC by SgE SFPs. I then showed the SFPs previously classified as low and middle SFPs were acategorial and lacked [F]s. This then enabled me to propose that they only surface at phase edges, provided the semantics were compatible. Therefore, despite being head-final, these SFPs do not violate FOFC. I then looked at the grounding SFPs in the discourse domain, arguing that the fact that they do not receive intonation is evidence they are not considered as part of the

106 clausal extended projection. Therefore, their head-finality would also not violate FOFC. I then concluded that SgE does not violate FOFC.

While this thesis has developed a proposal of SFP syntax in SgE, it is difficult to do justice to such a complex topic within this limited space. However, I hope this work lays the ground for future study of SgE SFPs and contributes to the larger discussion of particles cross-linguistically. It is sometimes held that contact languages are less complex than other types of languages, due to the conditions under which they developed. Having but scratched the surface of what is possible in SgE and yet discovered immense structural complexity, I have hopefully shown here that such a view is not always accurate. Instead of simplifying, the development process of SgE appears to have taken features from its substrate languages and combined them together to give a structurally even more complex language that can lexicalise pragmatic information. Therefore, I suggest that SgE offers an excellent opportunity for us to probe the nature of the syntax-pragmatics interface.

107 Bibliography

Aldridge, Edith (2011). “Neg-to-Q: The historical origin and development of question particles in Chinese”. In: The Linguistic Review 28.4, pp. 411– 447. Alexiadou, Artemis (2000). “On the syntax of temporal adverbs and the nature of Spec,TP”. In: Rivista di Linguistica 12.1, pp. 55–75. Alsagoff, Lubna and Chee Lick Ho (1998). “The relative clause in colloquial ”. In: World Englishes 17.2, pp. 127–138. issn: 1467- 971X. doi: 10.1111/1467-971X.00087. url: http://dx.doi.org/10. 1111/1467-971X.00087. Bach, Kent and Robert Harnish (1979). Linguistic communication and speech acts. Cambridge, Massachusetts: MIT Press. Bao, Zhiming (1995). “Already in Singapore English”. In: World Englishes 14.2, pp. 181–188. issn: 1467971X. doi: 10.1111/j.1467-971X.1995. tb00348.x. — (2005). “The aspectual system of Singapore English and the systemic substratist explanation”. In: Journal of Linguistics 41.2, pp. 237–267. issn: 0022-2267. doi: 10.1017/S0022226705003269. — (2015). The Making of Vernacular Singapore English. Cambridge: Cam- bridge University Press. isbn: 9781139135375. doi: 10.1017/CBO9781139135375. Bao, Zhiming and Huaqing Hong (2006). “Diglossia and register variation in Singapore English”. In: World Englishes 25.1, pp. 105–114. issn: 08832919. doi: 10.1111/j.0083-2919.2006.00449.x. Bar-Hillel, Yehoshua (1971). “Out of the Pragmatic Wastebasket”. In: Lin- guistic Inquiry 2.3, pp. 401–407.

108 Bayer, Josef and Hans Georg Obenauer (2011). “Discourse particles, clause structure, and question types”. In: Linguistic Review 28.4, pp. 449–491. issn: 01676318. doi: 10.1515/tlir.2011.013. Beyssade, Claire and Jean-Marie Marandin (2006). “The Speech Act As- signment Problem Revisited : Disentangling Speaker’s Commitment from Speaker’s Call on Addressee”. In: Empirical Studies in Syntax and Seman- tics. Ed. by Olivier Bonami and Patricia Cabredo Hofherr, pp. 37–68. url: http://llf.linguist.jussieu.fr/cssp/eiss6/beyssade-marandin- eiss6.pdf. Biberauer, Theresa (2017). “The Final-over-Final Condition and Particles”. In: The Final-over-Final Condition: A Syntactic Universal. Ed. by Michelle Sheehan et al. Cambridge, Massachusetts: MIT Press. Biberauer, Theresa, Anders Holmberg, and Ian Roberts (2008). “Structure and linearization in disharmonic word orders”. In: Proceedings of WCCFL 26, pp. 96–104. — (2014). “A syntactic universal and its consequences”. In: Linguistic Inquiry 45, pp. 169–225. Biberauer, Theresa, Glenda Newton, and Michelle Sheehan (2009). “Limiting synchronic and diachronic variation and change: The final-over-final con- straint”. In: Language and Linguistics 10.4, pp. 701–743. issn: 1606822X. B¨uring,Daniel (2003). “On D-Trees, Beans and B-Accents”. In: Linguistics and Philosophy 26.5, pp. 511–545. Cardinaletti, Anna (2011). “German and Italian modal particles and clause structure”. In: Linguistic Review 28.4, pp. 493–531. issn: 01676318. doi: 10.1515/tlir.2011.014. Chan, Marjorie K. M. (1980). “Temporal Reference in Mandarin Chinese: An Analytical- Semantic Approach to the Study of the Morphemes le, zai, zhe, and ne”. In: Journal of the Chinese Language Teachers Association 15.3, pp. 33–79. Chao, Yuen Ren (1968). A grammar of spoken Chinese. Berkeley: University of California Press. Cheng, Lisa Lai-Shen (1991). “On the typology of wh-questions”. PhD thesis. Massachusetts Institute of Technology.

109 Cheong, Si En Phoebe (2016). “Sentence-final Already and Only in Singapore English”. Honours Thesis. National University of Singapore. Clark, Herbert H. and Susan E. Brennan (1991). “Grounding in communi- cation”. In: Perspectives on socially shared cognition, pp. 127–149. isbn: 9781557981219. doi: 10.1037/10096-006. url: http://www.cs.cmu. edu/%7B~%7Dillah/CLASSDOCS/Clark91.pdf. Constant, Noah (2011). “On the Independence of Mandarin Aspectual and Contrastive Sentence-Final ne”. In: Proceedings of the 23rd North Ameri- can Conference on Chinese Linguistics (NACCL-23). Ed. by Zhuo Jing- Schmidt. Vol. 2. Eugene, pp. 15–29. — (2014). “Contrastive topic: Meanings and realizations”. PhD thesis. Uni- versity of Massachusetts Amherst. Erlewine, Michael Yoshitaka (2010). “Sentence-final only and the interpreta- tion of focus in Mandarin Chinese”. In: The Proceedings of the 22nd North American Conference of Chinese Linguistics (NACCL 22) and the 18th Annual Meeting of the International Association of Chinese Linguistics (IACL 18) 2, pp. 18–35. — (2012). “Structurally distant haplology”. In: Snippets 26, pp. 6–8. doi: 10.7358/snip-2012-026-erle. — (2016). “Low sentence-final particles in Mandarin Chinese and the Final- over-Final Constraint”. In: Journal of East Asian Linguistics, pp. 1–39. issn: 15728560. doi: 10.1007/s10831-016-9150-9. Ernst, Thomas (2002). The Syntax of Adjuncts. Cambridge: Cambridge Uni- versity Press. Gan, Lee Meng (2000). “A study of the ’hah’ and ’hor’ particles in Colloquial Singapore English”. Honours thesis. National University of Singapore. Gibbons, John (1987). Code-mixing and code choice: A Hong Kong case study. Clevedon, UK: Multilingual Matters. Gupta, Anthea Fraser (1992a). “Contact features of Singapore colloquial English”. In: Sociolinguistics today: International perspectives. Ed. by Kingsley Bolton and Helen Kwok. London: Routledge, pp. 323–345.

110 Gupta, Anthea Fraser (1992b). “The pragmatic particles of Singapore collo- quial English”. In: Journal of Pragmatics 18.1, pp. 31–57. issn: 03782166. doi: 10.1016/0378-2166(92)90106-L. — (1994). The Step-tongue: Children’s English in Singapore. Clevedon, UK: Multilingual Matters. Gussenhoven, Carlos (2004). The Phonology of Tone and Intonation. Cam- bridge: Cambridge University Press. isbn: 9780521812658. Haegeman, Liliane (2006). “Conditionals, factives and the left periphery”. In: Lingua 116.10, pp. 1651–1669. issn: 00243841. doi: 10.1016/j.lingua. 2005.03.014. — (2014). “West Flemish Verb-based Discourse Markers and the Articulation of the Speech Act Layer”. In: Studia Linguistica 68.1, pp. 116–139. issn: 14679582. doi: 10.1111/stul.12023. Heim, Johannes, Hermann Keupdjio, Zoe Wai Man Lam, Adriana Osa-Gomez, Sonja Thoma, et al. (2016). “Intonation and particles as speech act modi- fiers: A syntactic analysis”. In: Studies in Chinese Linguistics 37.2, pp. 109– 129. issn: 10171274. doi: 10.1515/scl-2016-0005. Heim, Johannes, Hermann Keupdjio, Zoe Wai-Man Lam, Adriana Osa-Gomez, and Martina Wiltschko (2014). “How To Do Things With Particles”. In: Proceedings of the 2014 Annual Conference of the Canadian Linguistic Association. isbn: 9780816676460. doi: 10.1057/ajcs.2015.8. Hiramoto, Mie (2015). “Sentence-final adverbs in Singapore English and Hong Kong English”. In: World Englishes 34.4, pp. 636–653. issn: 1467971X. doi: 10.1111/weng.12157. Holmberg, Anders (2000). “Deriving OV order in Finnish”. In: The derivation of VO and OV, Number 31 in Linguistic Aktuell. Ed. by Peter Svenonius. Amsterdam: John Benjamins, pp. 123–152. Kachru, Braj B. (1982). The Other Tongue: English Across Cultures. Urbana, Illnois: University of Illinois Press. — (1985). “Standards, codification, and sociolinguistic realism: The in the outer circle”. In: English in the World: Teaching and Learning the Language and the Literature. Ed. by R. Quirk and H. G. Widdowson. Cambridge: Cambridge University Press, pp. 11–30.

111 Kim, Chonghyuck and Lionel Wee (2009). “Resolving the paradox of Singapore English hor”. In: English World-Wide 30.3, pp. 241–261. issn: 01728865. doi: 10.1075/eww.30.3.01kim. Koh, Ann Sweesun (1990). “Topics in colloquial Malay”. PhD thesis. Univer- sity of Melbourne. Krifka, Manfred (2013). “Response particles as propositional anaphors”. In: Proceedings of SALT 23. Vol. 23. April, pp. 1–18. Kwan-Terry, Anna (1991). “Through the looking glass: A child’s use of particles in Chinese and English and its implications on language transfer”. In: Child language development in Singapore and Malaysia. Ed. by Anna Kwan-Terry. Singapore: University Press, pp. 161–183. Lam, Zoe Wai-man (2014). “A Complex ForceP for Speaker- and Addressee- oriented Discourse Particles in Cantonese”. In: Studies in Chinese Lin- guistics 35.2, pp. 61–80. issn: 10171274. Lan, Yingjie (2016). “Can move or not, ah: Exploring Optional Wh-movement in Wh-in situ Languages”. Senior Thesis. Waseda University. Law, Ann (2004). “Sentence-final focus particles in Cantonese”. PhD disser- tation. University College London. Lee, Hun-Tak Thomas (1986). “Studies on quantification in Chinese”. PhD thesis. University of California, Los Angeles. Lee, Junwen (2016a). “How Colloquial Singapore English Lah Interacts with Speech Acts: A Modal Analysis”. — (2016b). “Lah Revisited: A Modal Analysis”. Ler, Soon Lay Vivien (2006). “A relevance-theoretic approach to discourse particles in Singapore English”. In: Approaches to Discourse Particles. Ed. by Kerstin Fischer. Oxford: Elsevier. Chap. 9, pp. 149–166. Li, Charles N. and Sandra A. Thompson (1981). Mandarin Chinese: A func- tional reference grammar. Berkeley: University of California Press. Lim, Lisa (2007). “Mergers and acquisitions: On the ages and origins of Singapore English particles”. In: World Englishes 26.4, pp. 446–473. issn: 08832919. doi: 10.1111/j.1467-971X.2007.00522.x.

112 Lim, Lisa (2011). “Tone in Singlish: Substrate features from Sinitic and Malay”. In: Creoles, Their Substrates, and Language Typology. Ed. by Claire Lefebvre. Amsterdam: John Benjamins, pp. 271–288. Marie, Wilma Vimala (1988). “A study of sentence-final particles in Singapore English”. Honours thesis. National University of Singapore. Matthews, Stephen and Virginia Yip (1994). Cantonese: A Comprehensive Grammar. London: Routledge. Merchant, Jason (2001). The syntax of silence: Sluicing, islands, and the theory of ellipsis. Oxford: Oxford University Press. Mufwene, Salikoko (1996). “The Founder Principle in Creole Genesis”. In: Diachronica 13.1, spring, pp. 83–134. issn: 0364-3417. doi: 10.1075/dia. l3.1.05muf. — (2004). The ecology of language evolution, p. 255. isbn: 9781402079122. doi: 10.1353/lan.2005.0033. arXiv: arXiv:1011.1669v3. url: e: %7B%5C%%7D5CBook%7B%5C%%7D5CMufwene(2004-The%20ecology%20of% 20language%20evolution).pdf. Paul, Waltraud (2014). “Why particles are not particular: Sentence-final particles in Chinese as heads of a split CP”. In: Studia Linguistica 68.1, pp. 77–115. issn: 14679582. doi: 10.1111/stul.12020. — (2015). New Perspectives on Chinese Syntax. Ed. by Walter Bisang and Niina Ning Zhang. Vol. 271. Berlin: Walter de Gruyter GmbH. isbn: 0024-3841. doi: 10.1016/j.lingua.2005.03.011. arXiv: arXiv:1011. 1669v3. Pierrehumbert, Janet and Julia Hirschberg (1990). “The meaning of intonation contours in the interpretation of discourse”. In: Intentions in Communi- cation. Ed. by Philip R. Cohen, Jerry Morgan, and Martha E. Pollack. Cambridge, Massachusetts: MIT Press, pp. 271–311. Platt, John (1975). “The Singapore English Speech Continuum and Its Basilect ’Singlish’ as a ’Creoloid’”. In: Anthropological Linguistics 17.7, pp. 363– 374. — (1987). “Communicative functions of particles in Singapore English”. In: Language topics: essays in honour of Michael Halliday, Vol 1. Ed. by Ross Steele and Terry Threadgold. Amsterdam: John Benjamins, pp. 391–401.

113 Platt, John and Mian Lian Ho (1989). “Discourse particles in Singaporean English: substratum influences and universals”. In: World Englishes 8.2, pp. 215–221. issn: 1467-971X. url: http://dx.doi.org/10.1111/j. 1467-971X.1989.tb00656.x. Ramchand, Gillian Catriona (2008). Verb Meaning and the Lexicon: A first- phase syntax. Cambridge: Cambridge University Press. isbn: 9780521842402. doi: 10.1017/CBO9780511486319. arXiv: arXiv:1011.1669v3. Richards, Norvin (2016). Contiguity Theory. Cambridge, Massachusetts: MIT Press, p. 400. Rizzi, Luigi (1997). “The fine structure of the left periphery”. In: Elements of grammar. Ed. by Liliane Haegeman. Dordrecht: Kluwer, pp. 281–337. Ross, John Robert (1970). “On Declarative Sentences”. In: Readings in English Transformational Grammar. Ed. by Roderick A. Jacobs and Peter S. Rosenbaum. Waltham, MA: Ginn and Company. Chap. 13, pp. 222–272. Sag, Ivan and Mark Liberman (1975). “The intonation disambiguation of indirect speech acts”. In: Papers from the Eleventh Regional Meeting. Chicago: Chicago Linguistic Society. Sailor, Craig (2009). “Tagged for Deletion: A Typological Approach to VP Ellipsis in Tag Questions”. Masters thesis. University of California, Los Angeles, pp. 1–83. url: http://cwsailor.bol.ucla.edu/papers/ sailor%7B%5C_%7Dthesis%7B%5C_%7D10-3-11.pdf. Sato, Yosuke and Chonghyuck Kim (2012). “Radical pro drop and the role of syntactic agreement in Colloquial Singapore English”. In: Lingua 122.8, pp. 858–873. issn: 00243841. doi: 10.1016/j.lingua.2012.02.006. url: http://dx.doi.org/10.1016/j.lingua.2012.02.006. Simpson, Andrew (2014). “Sentence-Final Particles”. In: The Handbook of Chinese Linguistics. Ed. by James C.-T. Huang, Audrey Y.-H. Li, and Andrew Simpson. Chichester, UK: John Wiley & Sons. Chap. 7, pp. 156– 179. Soh, Hooi Ling (2012). The aspectual marker dah in Colloquial Malay. Taipei, Taiwan. Soh, Hooi Ling and Meijia Gao (2006). “Perfective aspect and transition in Mandarin Chinese: An analysis of double–le sentences”. In: Proceedings of

114 2004 Texas Linguistics Society Conference. Ed. by Pascal Denis. Somerville, MA: Cascadilla Proceedings Project, pp. 107–122. Speas, Peggy and Carol L. Tenny (2003). “Configurational properties of point of view roles”. In: Asymmetry in grammar. Ed. by Anna-Maria Di Sciullo. Vol. Vol. I: Sy. Amsterdam: John Benjamins, pp. 315–344. Stalnaker, Robert (2002). “Common Ground”. In: Linguistics and Philoso- phy 25, pp. 701–721. issn: 1573-0549. doi: 10.1023/A:1020867916902. url: papers2 : / / publication / uuid / 78B5C0C8 - 3284 - 4CB5 - BE59 - 94EBD770DE16. Sybesma, Rint and Boya Li (2007). “The dissection and structural mapping of Cantonese sentence final particles”. In: Lingua 117.10, pp. 1739–1783. issn: 00243841. doi: 10.1016/j.lingua.2006.10.003. Tan, Ludwig A.-K. (2007). “Null Arguments in Singapore Colloquial English”. PhD dissertation. University of Cambridge. Tang, Sze-Wing (1998). “Parametrization of features in syntax”. PhD thesis. University of California, Irvine. Thoma, Sonja (2014). “Bavarian discourse particles- at the syntax pragmatics interface”. In: Proceedings of NWLC 2013. Pp. 41–58. Tomioka, Satoshi (2010). “A Scope Theory of Contrastive Topics”. In: Iberia 2.1, pp. 113–130. Traugott, Elizabeth Closs (1982). “From propositional to textual and expres- sive meanings: some semantics-pragmatic aspects of grammaticalization”. In: Perspectives on historical linguistics. Ed. by Winfred P. Lehmann and Yakov Malkiel. Amsterdam: John Benjamins, pp. 245–271. — (1988). “Pragmatic strengthening and grammaticalization”. In: Proceedings of the Fourteenth Annual Meeting. Berkeley Linguistics Society, pp. 406– 416. — (1989). “On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change”. In: Language 65.1, pp. 31–55. Traugott, Elizabeth Closs and Ekkehard K¨onig(1991). “The semantics- pragmatics of grammaticalization revisited”. In: Approaches to grammati- calization. Ed. by Elizabeth Closs Traugott and Bernd Heine. Amsterdam: John Benjamins, pp. 189–218.

115 Trinh, Tue and Luka Crnic (2011). “On the Rise and Fall of Declaratives”. In: Proceedings of Sinn & Bedeutung, pp. 645–660. issn: 00280712. doi: 10.1126/science.1059746. Van Gelderen, Elly (2012). Clause structure. isbn: 9780521883702. doi: 10. 1017/CBO9780511712203.012. Wakefield, John C. (2016). “Sentence-final particles and intonation: Two forms of the same thing”. In: Speech Prosody 2016. Boston University. doi: 10.21437/SpeechProsody.2016-179. url: http://www.isca-speech. org/archive/SpeechProsody%7B%5C_%7D2016/abstracts/176.html. Wee, Lionel (2002). “Lor in colloquial Singapore English”. In: Journal of Pragmatics 34, pp. 711–725. — (2004). “Reduplication and discourse particles”. In: Singapore English: a grammatical description. Ed. by Lisa Lim. Amsterdam/Philadelphia: John Benjamins, pp. 105–126. Wiltschko, Martina (2014). The Universal Structure of Categories: Towards a Formal Typology. Cambridge: Cambridge University Press. — (2016). Response markers as a window into linguistic modularity. — (2017). “Ergative constellations in the structure of speech acts”. In: The Oxford Handbook of Ergativity. Ed. by Jessica Coon, Diane Massam, and Lisa deMena Travis. Oxford: Oxford University Press. Wong, Jock (2004). “The particles of Singapore English: A semantic and cultural interpretation”. In: Journal of Pragmatics 36.4, pp. 739–793. issn: 03782166. doi: 10.1016/S0378-2166(03)00070-5. — (2005). “”Why You So Singlish One?”: A Semantic and Cultural Interpre- tation of the Singapore English Particle One”. In: Language in Society 34.2, pp. 239–275. Wong, Peck San (1994). “The use of ’one’ in Singapore Colloquial English”. Honours thesis. National University of Singapore. Woods, Rebecca (2015). “Modelling the syntax-discourse interface: A syntactic analysis of please”. In: Proceedings of ConSOLE XXIII, pp. 360–382. Zhang, Niina Ning (1999). “Chinese DE and the DE-construction”. In: Syn- taxis 2, pp. 27–49. Zhu, Dexi (1982). Yufa jiangyi [On grammar]. Beijing: Shangwu yinshuguan.

116