Wh-Question Constructions in Colloquial Singapore English (Qizhong Chang A0015449E) PDF
Wh-Question Constructions in Colloquial Singapore English (Qizhong Chang A0015449E) PDF
Wh-Question Constructions in Colloquial Singapore English (Qizhong Chang A0015449E) PDF
SINGAPORE ENGLISH
CHANG QIZHONG
M. A., NUS
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
2016
DECLARATION
I hereby declare that the thesis is my original work and it has been written by me in its
entirety. I have duly acknowledged all the sources of information which have been used
in the thesis.
This thesis has also not been submitted for any degree in any university previously.
______________
Chang Qizhong
31 May 2016
i
Acknowledgements
I owe a tremendous debt of gratitude first and foremost to my supervisor, Yosuke Sato, for
his infinite patience and guidance and support. I have benefited greatly from his remarkable
linguistic insight and wealth of knowledge. He is the reason that I am able to finish this
dissertation at all.
I also thank the examiners and members of my dissertation committee, in particular 包老师
and Mitcho, who have provided immensely useful comments and critique on my work.
I would like to thank my wife, Zhi Jun for believing in me and being the best partner one can
ask for. She has kept me going during the rough times and I dedicate this piece of writing to
her and also to my son, Wen Quan, who is the biggest joy in my life. I am grateful also to the
Thank you.
ii
Table of Contents
Summary vi
CHAPTER 1 – Introduction 1
2.1 Methodology 14
iii
2.8 Intervention Effects 40
3.1.1 Malay 71
iv
4.1.2 Bobaljik & Wurmbrand (2015) 143
End 176
References 177
Appendix 191
v
Summary
Colloquial Singapore English (CSE) has traditionally been labelled a language with
‘optional wh-behaviour’. This means that wh-questions freely alternate between having
the wh-expression fronted and being in-situ. This dissertation presents novel data in CSE to
substantiate two important observations: Firstly, wh-behaviour in CSE is not strictly optional,
The obligatory wh-movement in this context is triggered by the need for the clause to be
dubbed Declarative Syntax Questions (DSQ). This is a declarative clause that nonetheless
From this study, we can draw the conclusion that CSE is at heart a wh-movement language;
but the apparent wh-in-situ order of a DSQ gives CSE the appearance of having optional wh-
movement.
vi
List of Tables
vii
List of Symbols and Abbreviations
Ø Null Element
A/A’ Argument
ACC Accusative
ASP Aspect(ual)
BP Brazilian Portuguese
BW Babine Witsuwit’en
C(P) Complementiser (Phrase)
CL Classifier
CSE Colloquial Singapore English
CTH Clause Typing Hypothesis
DAT Dative
DFCF Doubly Filled Comp Filter
DP Determiner Phrase
DS Deep Structure
DSQ Declarative Syntax Question
ECP Empty Category Principle
EPP Extended Projection Principle
F/FocP Focus (Phrase)
ForceP Force Phrase
FSH Featural Subset Hypothesis
IP Inflection Phrase
LF Logical Form
NEG Negation
NIP New Information Phrase
NOM Nominative
NP Noun Phrase
NPI Negative Polarity Item
viii
OCMP Overt over Covert Movement Principle
OP Operator
PF Phonetic Form
PLC Pronounce Lower Copy
PP Preposition Phrase
PRN Pronoun
PrtP Particle Phrase
Q(u) Question
QNP Quantifier Noun Phrase
SCLI Subject Clitic Inversion
Spec Specifier
StdE Standard English
SS Surface Structure
TP Tense Phrase
TopP Topic Phrase
TypP Typing Phrase
VAR Variable
VP Verb Phrase
WALS World Atlas of Language Structures Online
ix
Chapter 1 – Introduction
In this first chapter, I will introduce the general description and analysis of wh-
constructions in Colloquial Singapore English that is the central focus of this thesis. I will
then provide the aims, as well as the layout of the rest of the thesis.
is introduced by, or includes a wh-expression from the following list – what, who, when, why,
where, which, how – and probes for information that is represented by the wh-expression. In a
simple matrix wh-question in English, the wh-expression is typically at the front of the
The simple analysis for the above question is that it is derived from its declarative
counterpart “John bought X/what.”; the wh-expression then moves from the sentence-final
position to the sentence-initial position (among other operations). Working on the assumption
that surface transformational operations do not effect a change in the sentence’s meaning,
Katz & Postal (1964) posited a Q(uestion)-morpheme to account for the meaning contrast
between a statement and its attendant interrogative. This Q-morpheme would have explained,
among other things, Subject-Auxiliary Inversion, and the lack of Q-complementisers such as
whether and if in matrix questions. Baker (1970) later expanded the proposal to include the
Q-morpheme in both direct and indirect questions, basically utilizing Q as the primary
1
motivation for wh-movement. The wh-expression in (1) is therefore assumed to have been
moved to the Specifier position of C(omplementizer) Phrase; and it would have done so as a
them to move (to be interpreted or licensed). In English, the wh-expression cannot be left un-
moved in its original position, or the in-situ position, as shown in (2) vs. (3):
In sentence (2), the wh-expression is left in-situ, and this is routinely judged as
unacceptable to native English speakers as a bona fide (non-echo) question. It is, however,
original position and given special intonational focus. The function of an echo question is to
reiteration of something the speaker might not have caught the first time round. If we do not
take echo-questions into consideration, the consensus seems to be that the wh-in-situ option is
In the same vein, an embedded wh-question in English sees the wh-expression moved
from its base-generated position in the embedded clause to the Spec, CP position. The
landing site of the wh-expression can either be the initial Spec, CP position of the matrix
clause (4), or the intermediate Spec, CP position of the embedded clause (5):
1
There is a growing body of literature that seeks to show that “it is possible for an information-seeking, non-
echo wh-question without fronting the wh-word” (Bobaljik & Wurmbrand, 2015). Examples of such questions
are: “So, your boy’s name is what?” and “Major, you want this stuff where?”. These questions are genuine
requests for information, but are somewhat restricted to a colloquial register.
2
Sentence (4) is a direct question, with the wh-expression moved all the way to the
matrix Spec, CP position. Sentence (5) is an indirect question, with the wh-expression moved
only to the Spec, CP of the embedded clause. If the wh-expression is left in-situ, it can only
(6) John wondered what was going to happen next? (echo interpretation)
(7) *John wondered what was going to happen next. (non-echo interpretation)
A question such as (7), with a non-echo interpretation, is not acceptable. The sharp
where a wh-expression can appear in-situ in a matrix clause (Bobaljik & Wurmbrand, 2015).
has the syntactic structure of a declarative. The DSQ construction actually turns what we
Siobhan: And your next move is what exactly? Because so far, you’ve done a piss-
clauses is unacceptable (2), and that the only possible wh-in-situ StdE matrix clauses are echo
questions (3). However, a DSQ such as (8), which is deemed acceptable in StdE, defies both
those descriptions. (8) is a matrix question, with a non-echo interpretation, and with wh-in-
situ. More details for the DSQ construction will come in Chapter 4.
3
When the pattern of wh-questions in English is considered as a whole, it is no surprise
language, where wh-expressions need to move to the front of the sentence or clause, to form a
of the verb eating, as a thematic licensing requirement. It then needs to be motivated to move
to the front of the sentence. Bresnan (1970) argues for the Q-morpheme to be a wh-
complementiser, with a [+wh] feature in the Comp/C0 head. This is presumably the trigger for
wh-movement. At least two basic operations then have to occur to derive the root
interrogative from its echo counterpart: head movement of the auxiliary was from T to C, and
(10) CP
PRN C’
What
C TP
was
PRN T’
she
T VP
was
V PRN
eating what
4
Unlike head movement, which moves a zero-level X0 category, wh-movement moves
the phrasal projection which is headed by the wh-word. It is possible for the wh-expression to
move even larger chunks of material along with it, such as a prepositional phrase:
local since a wh-expression escapes its own clause and moves to the Spec-CP position of a
higher clause:
(12) What did Mary think that [John would eat ___]?
The movement of what from sentence-final position all the way to sentence-initial
position shows how powerful or arbitrarily long wh-movement operations can be.
wh-expression is left behind in the position out of which it moves. This can also be seen in
The semantic representation, or Logical Form (LF) of a wh-question such as (9) can
be informally paraphrased as ‘Of which x, x, a type of food, is it true that she was eating x?’,
or:
which could then bind one or more question words. By treating Q as an operator, we were
able to determine the scope of in-situ wh-expressions. The postulation of Q as an operator has
been widely accepted since then, though in some analyses Q is no longer an operator and is
5
(13), the quantifier which functions as an interrogative operator which serves to bind the
this form:
a position where it can take scope over the rest of the proposition. This operator is the wh-
expression itself. After the wh-expression moves, a copy of itself remains in the original
position and has the semantic function of a variable which is then bound by the wh-operator.
Interpretation, which states that all elements present at the semantic interface must participate
in a semantic interpretation (Chomsky, 1995). Therefore, we can see that not only is the wh-
movement shown in (10) necessary to achieve the correct word order in English, it is also
question. This analysis of wh-movement is quite standardly assumed for a language such as
English, where wh-expressions typically undergo overt movement to the front of the clause.
In a sense, the analysis very briefly explained above is considered straightforward, as the wh-
movement in syntax (10) mirrors that of the LF-movement in the semantics (14).
by Karttunen (1977). The semantic value of an interrogative cannot be a truth value, unlike
the semantic value of a declarative. That is to say, the utterance “John read a book” denotes a
proposition and has a truth value of either true or false; but the utterance “Which book did
John read?” is a request for a proposition and can only have a value that is a set of
6
propositions which constitute all its possible answers. Therefore, the denotation of the
question “Which book did John read?” is represented as the set of propositions {John read
Animal Farm, John read War and Peace, John read Great Expectations…}, and a true
proposition from this set is the answer to the question. This true proposition provides the
value for the wh-expression which book, which is represented as a variable that proliferates
Given that interrogatives are assumed to be derived from declaratives, there should be
a convergence in where the wh-expression is interpreted in the structure and where the shift
in semantics from propositions to a set of propositions take place. Using the above-stated
example:
(15) gives us the syntactic representation of the interrogative, while (16) gives us its
semantic denotation. The wh-existential operator and the restrictor are outside the scope of
the proposition variable p where λ-abstraction takes place. The trace t of the wh-expression is
interpreted as a logical variable ranging over entities. The λ-abstraction over the propositional
variable p shifts the denotation from a proposition to a set of propositions. The wh-expression
is interpreted in its moved position. Since the wh-expression is also pronounced in its moved
position, the mapping from syntax and semantics is overt and direct.
There are, however, a significant number of languages where the mapping from
syntax to semantics is not this straightforward, i.e. wh-movement is not empirically observed,
or is restricted in some way. This creates a tension in the analysis of wh-in-situ constructions:
on the one hand, we still need to create an Operator-Variable configuration as shown in (14)
for the question to be interpreted as such; yet, the wh-expression is unable to move in overt
7
syntax. To overcome this problem, scholars working on wh-phenomena have come up with
several alternative accounts over the years. These can be grouped into some broad categories
Reinhart (1998)).
than the wh-expression itself (for example, Hagstrom (1998); Watanabe (2001);
Ginsburg (2006); Aoun & Li (1993)). Usually also associated with lexical
- Masked movement, i.e. wh-expression moves, but because the rest of the clause
also moves, it seems like the wh-expression hasn’t moved (for example, Kayne
whether a proper landing site exists in the first plaec, or whether the landing site is
empty and thus able to accommodate the moved wh-expression (for example,
for wh-constructions. We will review some of them in Chapter 3, to see if they could work
for us in explaning the data presented in this thesis. We now move on to look at the pattern of
8
1.2 wh-questions in Colloquial Singapore English
English used in Singapore, mainly in informal settings. A more thorough typological review
of the variety, as well as a comprehensive look at the data on CSE wh-constructions, will
come in the next chapter. For now, let us focus on some pertinent facts about wh-questions in
CSE. In a simple matrix wh-question, the wh-expression can come either in its in-situ position
Both (17) and (18) are perfectly acceptable CSE matrix wh-questions. It has to be
noted that, contrary to StdE, sentence (17) can receive a non-echo question interpretation. Of
course, it can also be given an echo-question interpretation, when the utterance is made with
the appropriate features/conditions – namely, rising intonation and stress on the wh-word,
and/or a suitable discourse context. The free variation between the non-echo question version
of (17) and (18) is commonly referred to in the literature as “optional wh-movement”. That is
to say, the wh-expression has the option of either staying in place or moving to the Spec, CP
position; yet, the propositional content of the question remains the same. This is a marked
difference from StdE matrix wh-questions, which does not seem to have this option.
Given the seemingly free rein over word order in CSE matrix wh-questions, the
pattern that emerges for CSE embedded wh-questions is rather surprising. The wh-expression
in an embedded clause cannot stay in-situ (19) and must be moved to the embedded Spec, CP
position (20):
9
(19) *John wonder Tom buy what? (non-echo interpretation, CSE)
This judgment is fairly robust among native CSE speakers. The pattern that obtains is
actually very similar to that of StdE embedded wh-questions (see (19)’ and (20)’). Therefore,
the deceptively simple paradigm that will be the preoccupation of this thesis is this:
Table A gives us a very neat picture of the basic facts that needs to be accounted for.
Note that in-situ wh-expressions in multiple-wh-questions in StdE and CSE are not included
in this table. The reason for this is that we are restricting our focus in this thesis to single-wh-
questions in both varieties. With this, we can now move on to the next section, which is to
The aims of this paper are manifold, but will remain modest. Firstly, we will present a
wide range of data on CSE wh-questions. Most crucially, by doing so, we can correct the
10
mistaken belief that wh-strategies in CSE are unrestrained, and that there is free variation in
the position of wh-expressions. Many early scholars consider CSE to be a variety that
possesses little to no grammatical rules, exhibiting ‘wild’ and random behavior (Platt, 1975;
Platt & Weber, 1980; Tongue, 1979). These are usually the same people who think CSE is a
broken form of StdE and do not consider CSE a legitimate variety of English.
add the CSE facts to the growing body of literature on wh-phenomena. There has been a vast
amount of work examining the mechanisms of wh-movement, their motivations and their
restrictions. We can refer to Table A once more for the facts on CSE wh-questions that we
want to account for. The most arresting fact that demands explanation in Table A is the
seemingly optional nature of wh-movement in CSE matrix questions. CSE would hardly be
the first language (variety) to exhibit such optional behavior in terms of wh-movement – there
are several other documented languages (e.g. Malay, French, and Babine Witsuwit’en) that
behave in a similar fashion; and linguists have attempted to explain this phenomenon in
various ways. For instance, more scholars (see Kato (2013), Chang (1997), etc.) are now
examining the possibility that there is no true or real ‘optional’ wh-behavior, insofar that each
variation in the word order of wh-expressions can receive a principled explanation. To this
end, CSE can very well lend typological support to studies that pursue such explanations.
CSE’s unique status as a variety of English, coupled with its considerable and varied
substratal influence, makes it a fascinating topic of study and also a useful comparison with
other languages that display the same wh-phenomena. Furthermore, there has, regrettably,
been a dearth of formal studies on CSE wh-questions and hopefully, this paper will go some
Additionally, if you compare wh-strategies within CSE itself, you will notice another
interesting fact: where there is clearly optional wh-behavior in the matrix clause, there is a
11
clear-cut pattern of wh-movement within embedded clauses. This matrix vs. embedded
- Are there languages that exhibit the reverse asymmetry, i.e. only one wh-
- Do all contact languages with a superstrate and a substrate that exhibits different
pattern? Also, does this asymmetry shed any light on the degree of influence each
notwithstanding, we hope to try our best to answer them (some of them, perhaps only
speculatively). It is hoped that through the discussion of the wh-data, we can gain more
insight into the theoretical mechanisms behind wh-movement, as well as into the nature of
CSE itself.
The empirical contributions of this thesis lie primarily in establishing the general
patterns of wh-constructions in CSE, specifically that (a) we can attribute CSE wh-in-situ
12
matrix questions to the observed DSQ phenomena, such as (8); and (b) pointing out the novel
offering a new analysis of the CSE facts, and an attempt to answer the question how best to
The thesis will be arranged in the following manner: this current chapter serves as the
Introduction to what the thesis is about, and its aims. In Chapter 2, we will describe the
linguistic profile of CSE, the detailed patterns of CSE wh-constructions, as well as the
methodology behind corroboration of judgments used in the thesis. Chapter 3 will include an
Chapter 4, we will discuss in detail an analysis of CSE wh-constructions. The proposal that
we will follow suggests that apparent wh-optionality in CSE matrix clauses can be attributed
to a novel construction called Declarative Syntax Questions. Secondly, the proposal accounts
clause as an interrogative. Taken together, these two observations argue for the position that
randomness and optionality as previously believed. Chapter 5 concludes the thesis by giving
a summary of the findings and analysis, pondering over some unresolved questions, as well
13
Chapter 2 – CSE wh-constructions
The majority of this chapter will be devoted to detailing the wh-data in CSE.
as the basis of discussion of this thesis. Also of significance in the data presented is the
argument/adjunct asymmetry that can be found in CSE wh-movement. The full wh-paradigm
in CSE is extremely intricate and complex; and it would be well worth our time to examine it
in more detail. It is hoped that any reader (native and non-native CSE speakers alike) can get
a clear picture of CSE wh-constructions from this chapter, as well as be able to find
interesting patterns, and use the data for other topics of research.
I will start this chapter with a discussion of the methodology adopted to corroborate
the judgments of CSE data presented in this thesis. I will list briefly the steps taken by the
author to ensure that the judgments presented here are as accurate as possible.This will be
followed by a description of the linguistic background and profile of Singapore and the
2.1 Methodology
The data presented in this thesis are obtained from a variety of sources. These include
examples constructed by the author, who is a native CSE speaker, which are then
corroborated with other native speakers; data quoted from other academic sources; naturally
occurring ‘speech’ from online social media and the Web; as well as scripts from TV shows.
14
The novel observation of the unacceptability of wh-in-situ in CSE embedded
that this is not the case, a small scale survey was carried out among native CSE speakers to
corroborate the grammaticality judgments presented here. Basically, a simple set of sentences
was constructed and sent to a small number of undergraduate students who took a class
taught by this author. Subjects were all native CSE speakers. Subjects were instructed to
“Please take a look at the following Colloquial Singapore English (Singlish) sentences
and give your native speaker judgment on them. Instead of using distinctions like
whether (a) you have heard the particular utterance occurring naturally before, or (b)
Please rate these sentences on a scale of 1 (you don’t think this sentence is
different set of sentences, from a separate group of speakers. Judgments were also informally
elicited from the author’s friends and associates on occasion. In total, not more than 50
The specific topic or focus of the survey was not made known to the subjects, but
attention was drawn to the use of wh-expressions, and the syntactic positions they are found
in. This was to avoid a situation where subjects potentially mistakenly attribute their score to
any part of the sentence that is NOT the wh-expression. Additionally, where necessary, the
data was provided in pairs of sentences – one with wh-movement and the other with wh-in-
15
situ – so that a clear sense of which is preferred can be obtained. The socioeconomic details
(gender, age, ethnicity, education attainment, occupation, etc.) of the subjects were not
collected and will not be presented, as this study will not make any comment on social
variables/aspects of CSE speakers. Where necessary (take for example (111)), however,
The study is also not meant to say anything about where the subject falls along the
cline of formality of Singapore English, i.e. whether someone is closer to the acrolect
(Standard Singapore English), or to the basilect (CSE). That is to say, the explanation for
and would be more likely to judge features of CSE (such as wh-in-situ) as unacceptable. That
being said, the variation in question (wh-in-situ vs. wh-movement) was found to be
surprisingly resilient to this assumption, i.e. many highly educated and presumably speakers
interrogative contexts. Perhaps, through personal experience, if not personal usage, this group
of speakers are cognisant of the linguistic reality that such features exist and are widely
produced. This should give us confidence that this is a fairly robust phenomenon in CSE.
A copy of the pilot survey, the follow-up survey, and all the consolidated results will
16
2.2 Linguistic Profile of Singapore
In this section, I will give an introduction to Singapore and to CSE2: its origins, its
defining characteristics, and its status as one of the most rapidly-nativising varieties of
English. Singapore is an island-state with a current total population of 5.5 million, with
roughly 3.3 million citizens. Of the resident population, the ethnic composition is: 74.3%
Chinese, 13.3% Malays, 9.1% Indians, and 3.2% Others (Singapore Department of Statistics,
2015). Singapore was founded by Stamford Raffles in 1819 and remained a British colony till
very early, and had a place together with other languages such as Baba and/or Bazaar Malay,
spoken by the indigenous people; southern Chinese dialects by early immigrants (Hokkien,
Teochew, Cantonese, etc.), Mandarin Chinese, Tamil and various others. Historically, CSE is
borne from the language contact situation between its superstrate, English3, and its substrate
languages, Mandarin Chinese, Chinese dialects, and Malay. Exactly when CSE started its
roots cannot easily be pinpointed to a precise period of time; but by all accounts, an English-
based pidgin gradually replaced Bazaar Malay (a contact variety with Malay superstrate and
Chinese substrate) as the inter-ethnic lingua franca in Singapore, as well as in the region,
around the beginning of the 20th century. This rate of replacement picked up shortly before
Singapore achieved independence, and really took off after Singapore achieved
independence. In 1953, a bilingual education policy was initiated; and by 1956, it was
acknowledged that English, along with a ‘mother tongue’ (an ethnic language) would play a
major role in education. According to Wee (2002), “...this move towards a more pragmatic
2
Some scholars distinguish between Singapore English and Colloquial Singapore English, claiming that CSE is
a more colloquial variety bearing features typical of a creole. I accept this obvious distinction, as variation in
wh-movement is more apparent in CSE than in Singapore English.
3
A historically more accurate account would suggest that the lexifier in CSE is likely not to be Standard
English, but a dialectal variety of English used in the region at the time of formation of early CSE (Mufwene,
1996). However, I will continue to refer to the superstrate in CSE as ‘English’, simply for brevity’s sake. The
same can be said for the use of ‘Malay’ (instead of a variety of Bazaar and Baba Malay), and ‘Chinese’ (with no
intention of omitting the various Chinese dialects), in relevant contexts.
17
view of language can be seen in how the relationship between the endogenous Asian
languages and English is constructed.” English was promoted, for utilitarian purposes – as the
language of economic development – and as an inter-ethnic link language; while the ‘mother
According to the 1957 Census Reports, only about 1 percent of the population self-
reported that English was their mother tongue. This number later increased to 8.9 percent in
1980, which was indicative of a rising trend both in the status and the use of English, which
has persisted till today (Kuo, 1985). If we take the numbers at face-value, it would be hard to
imagine that CSE was prevalent or even just developing during the 1950s, since only 1% of
the population had English as their mother tongue. Of course, this cannot be true, as the
existence of a ‘pidgin-like’ variety of English had been established between merchants, the
colonial government and other settlers even before the war. That variety of English could
very well be the precursor to, or the beginning of, CSE. Therefore, the reasons for a relatively
low percentage of the population reporting English to be the principal household language
even as late as 1980, could be any of the following: people were multilingual and spoke other
languages more than they did English then, and the fact that English had only been recently
installed as the official language of education (English-medium schools only became the
The linguistic ecology in Singapore is highly complex and fluid given the unique
contact and mix of languages in the country (Ansaldo, 2009). More significantly, the trends
of language use have been shifting rapidly in our relatively short history. In the 1980 census,
the population can speak more than one language, and almost all native Singaporeans
understand and speak some form of CSE. Also, in 1980, only 7.9% of Singaporeans spoke
Mandarin Chinese, compared to 62.9% who spoke Chinese dialects. This later became 35.6%
18
Mandarin and 14.3% dialects in 2010 (Singapore Department of Statistics, 2010) – a very
marked change. This was likely the result of a very successful Speak Mandarin campaign,
launched in 1979, that was meant to promote Mandarin so as to unite all dialect groups.
Lastly, a consistently high number of Malay households spoke Malay at home, although there
was a drop from 97.7% in 1980 to 82.7% in 2010. In summary, over Singapore’s formative
years, the following factors: a strong rising trend in English speakers; a ‘pull’ towards
Mandarin Chinese from the Chinese dialects; an increasingly multilingual society; and a
stable base of Malay speakers, put together a clear picture of the prevalence and vitality of
CSE in Singapore.
CSE is a language that can be mutually understood and thus used widely, within the
multi-racial Singaporean community. However, it has never been accorded any official status.
It is therefore difficult to estimate the number of native CSE speakers in Singapore, but by all
accounts, there should be an entire generation of native CSE speakers by now. A recent
survey of 260 students from local public schools produced a self-reported 85% of
respondents who speak CSE (Tan & Tan, 2008). Gupta (1994) defines a native CSE speaker
as ‘those who have acquired Singlish in the home from birth, not subsequent to any other
language’. The informant judgments used in the course of this thesis shall also follow this
with friends and family. The use of CSE is a way of showing familiarity, and reducing
students showed that respondents would feel closer to a CSE speaker, as opposed to a
American English speaker, after listening to audio recordings of the two (Tan & Tan, 2008).
Almost 90% of the respondents also self reported that when speaking to friends and
classmates, family members and relatives, CSE would be the language of choice. A majority
19
of respondents said they ‘enjoyed speaking CSE’, and identified CSE as the most popular
In mainstream media such as television and newspapers, CSE is also used for other
purposes, such as humour and comedic effect (Tan, 2004). The use of CSE in media is
regulated heavily and distinctly marked as different from Standard English. In the Free-to-Air
Television Programme Code set out by the Infocomm Media Development Authority (IMDA,
“Singlish, which is ungrammatical local English, and includes dialect terms and
sentence structures based on dialect, should not be encouraged and can only be
permitted in interviews, where the interviewee speaks only Singlish. The interviewer
Previously, PACE stated that “this should be avoided as it could give the wrong impression,
especially among the young, that Singlish is the standard of spoken English in Singapore”.
This is the result of the Singapore government’s view that CSE is a ‘sub-standard’
variety of English and that its use should not be encouraged in formal situations. This view
was crystallised with the launch of the Speak Good English Movement (SGEM) in 2000,
with the official theme ‘Speak Well. Be Understood.’, accompanied by the explicit objective
“to help Singaporeans move away from the use of Singlish”. The SGEM continues to be
active. Prominent leaders, including the Prime Minister of Singapore, often add their
“It is much easier to speak proper English all the time, than to speak sloppily most of
the time, and then switch to proper English for formal occasions. Because to do the
20
latter is almost like learning two languages. We will learn wrong grammar and
language habits when we speak sloppy English, and when the occasion arises for us to
(Extract from a speech by Prime Minister Lee Hsien Loong at the Launch of the Speak Good
The government’s view of CSE effectively associates some degree of social stigma to
its use; but despite all of the above, the use of CSE permeates through all strata of Singapore
society, and is not limited to only the lower-middle class or the lowly educated, which is a
popular misconception.
There have been differing academic opinions also on how CSE should be treated.
Early work labeled CSE as a ‘creoloid’; a basilect of the Singapore English speech continuum
(Platt, 1975). Later on, two of the main approaches to take centre stage are the Lectal
Continuum Approach (Platt and Weber, 1980), and the Diglossia Approach (Gupta, 1994).
The Lectal Continuum Approach states that CSE is a non-native variety of English. CSE
speakers can be placed along a cline of proficiency, related to their educational level and
variety of English, and that it has an autonomous grammar. In Alsagoff’s Cultural Orientation
Model (Alsagoff, 2010), she argues that a speaker’s choice to orientate himself towards either
features that they want to invoke. Speakers may exhibit some degree of CSE features to
indicate a local perspective in order to stress membership in the community. The infusion of
CSE elements can be subtle – the use of pragmatic particles, grammar and leixis to establish a
local orientation. The choice language variety is thus utilised in identity-building and
21
CSE is also often labelled a New (variety of) English (Kachru, 1985; Pakir, 1991), or
Englishes (Schneider, 2007), CSE is argued to have already reached the status of
endonormative stabilisation, or Phase 4 of the model (Lim & Gisborne, 2009) – this means
that the variety adopts its own identity and no longer looks elsewhere for its norms.
Despite the differences in how CSE should be labelled, it is obvious that CSE cannot
features are considered highly visible and marked to even the most cursory of non-native
observers. The many formal studies on the grammatical features of CSE that have been
produced acknowledge that although there is a high degree of variation in the way CSE is
used and spoken, systematic and identifiable patterns emerge, as well as unique constructions
that are not found in either its superstrate or substrate languages. For instance, when asked to
Copula Deletion
Pro Drop
‘Every year, someone (elided subject) must buy something (elided object) for
22
Lack of tense and agreement features on the Verb (Reduced/Optional
Morphology)
Tag Questions
Passive Constructions
Firstly, the list of features just presented is clearly a non-exhaustive one. Secondly, it
should be noted that the occurrence of any one feature is usually concomitant with other
features in the list. In fact, it is more likely for a cluster of features to be present in a CSE
sentence than for only one feature to be present. For instance, the omission of the copula
23
might seem more ‘natural’ with a lack of verbal inflection and the dropping of determiners:
John is going to the market. vs. John go market. Another way of looking at this would be to
say that the presence of certain features increases the likelihood of occurrence of other
features. It is not clear whether this is simply a process of removing the ‘unnecessary
features’ in the sentence (akin to telegraphic speech, and newspaper headlines), or if this
While the presence of some of the features (21) – (27) are shared among other
English/Chinglish (He & Li, 2009), and copula deletion in Afro-American Vernacular
English (Labov, 1972), other features also serve as “…indexical markers that distinguish one
indigenised variety from another” (Bao, 2010). Thankfully, these features are visible and
relevant data can be obtained simply by corroborating intuitive native speaker judgments.
The unique linguistic matrix of Singapore has created CSE; and its status as a contact
language is undeniable, due to the fact that its grammatical system has developed together
with other indigenous languages, in the endogenous multilingual contact ecologies (Platt,
In the following section, we will immediately start looking at the data on wh-
constructions in CSE.
Let us begin with the basic facts in matrix clauses. The pattern in CSE matrix
in-situ or moving to the matrix Spec, CP position. Wh-adjuncts in CSE, however, face
degraded acceptability judgments when they are left in-situ, and are fine when moved to the
24
Spec, CP position. The list of wh-arguments in CSE remains intact from English, and consists
of: what, who, where; while the list of wh-adjuncts consists of: how, why, when.
In matrix clauses, wh-arguments can either move to the Spec, CP position or stay put
Wh-Arguments
Wh-Adjuncts
Note that the optional dropping of inflections on the matrix verb and/or the optional
omission of do-support – common features of CSE – do not affect the judgments shown in
(28) – (33). (28) – (30) are self-evident and constitute the fairly robust pattern that wh-
arguments are fine either moved clause-initially, or left in-situ in a matrix question. For now,
25
otherwise, of the moved and the in-situ versions of a matrix question such as (28) – (30);
although we will question that assumption later in this thesis. In any case, both the moved and
in-situ versions of a question, for instance, can be given the same answer:
This means that the difference in order of the wh-expressions in the question does not
We now turn our attention to the distribution of wh-adjuncts in matrix questions. (31)
– (33) pose more of an empirical problem since there is no full consensus in speaker
judgments. The speakers who judge the in-situ versions of these questions to be acceptable
are a very slender minority; and among these speakers, there are two further subgroups. The
first subgroup find the in-situ version of a question such as (31), for instance, acceptable
given its intended reading ‘How did John fix the car?’ The second subgroup find (31)
acceptable, but with a different reading along the lines of “How/what is the outcome of John
fixing the car?”. An explanation for the latter reading will be presented in a subsequent
chapter. Similar levels of variation in speaker judgments can be found in examples such as
There are, however, ways to make these sorts of utterances with wh-adjuncts more
acceptable. For instance, embedding the wh-adjunct how in a phrase such as how old
(35) That baby (is) how old? / How old (is) that baby?
26
It is also generally observable that adding an interrogative particle that is common in
CSE, such as ah or leh, or even a non-interrogative particle such as sia, improves the
speakers in accepting that there is a clear distinction between the grammaticality of in-situ
further evidence for this perceived tendency, I present my personal observation of a pattern in
(37) John fix the car for what? / *For what (did) John fix the car?
(38) John watch movie (at) what time? / (At) what time (did) John watch movie?
In my view, this could be a move towards keeping to a consistent in-situ word order
for matrix questions. The phrase for what occurs particularly frequently in the same
probably a shortened take on the phrase for what reason, only that the word reason never
appears. However, for what does not do well when it is moved to the front of the clause. In
contrast, in example (38), the wh-adjunct when is substituted with the phrase what time,
27
(39) A: Do you want to go to the mall?
In this exchange, B’s responses can be understood as “Why are we going to the
mall?” and “When is the movie screening?” respectively. I cannot think of any such phrasal
equivalents for the wh-adjunct how in CSE. These functional innovations are interesting to
me for at least three reasons: Firstly, they show preliminary evidence of contact influence
with a wh-argument (phrase) in order to remain in-situ, this reinforces the clear distinction in
positional variation between wh-arguments and wh-adjuncts in CSE – that only a wh-
argument can remain in-situ, and not a wh-adjunct. Lastly, the examples (35) – (39) betray a
questions: if a wh-adjunct cannot remain in-situ, it is substituted with something that can.
In summary, we see that there is a clear distinction in CSE matrix questions between
the positions of wh-arguments and wh-adjuncts. On the basis of (28) – (30), we conclude that
wh-arguments can either be moved, or remain in-situ in a matrix clause. On the basis of (31)
– (33), we conclude that leaving a wh-adjunct in-situ in a matrix question, with the intended
needs to be moved to the front of the clause. We also note certain interesting variations (35) –
(39) that are in line with this general distinction, and furthermore seek to maintain a wh-in-
4
What time could possibly be analysed as the Chinese ji shi (几时), which has the meaning ‘what time?’, and
for what can possibly be analysed as the Chinese wei(-le) shenme (为了什么) (Tsai, 1994).
.
28
Next, we will look at embedded questions.
what we have just seen in CSE matrix questions, which is the major observation made in this
dissertation. Wh-arguments such as what, who and where are moved to the front of the
In (40), for instance, the embedded question ‘Who (did) Lisa marry?’ appears with the
wh-argument moved to the front of its own clause. The corresponding embedded wh-in-situ
version of the construction is unacceptable. The same pattern obtains for other CSE wh-
arguments.
In the same way, wh-adjuncts such as how, why and when move to the front of the
29
(45) a. John want to know why Lisa cry.
In (43), for instance, the embedded question ‘How (did) Lisa escape?’ appears with
the wh-adjunct moved to the front of its own clause. The corresponding embedded wh-in-situ
version of the construction is unacceptable. The same pattern obtains for other CSE wh-
adjuncts5.
therefore, only verbs that select for an embedded interrogative clause, such as ask, wonder
and want to know are presented here. Verbs that can select for both interrogatives and
statements, or verbs that can only select non-interrogatives, such as believe or think, are not
considered here. In those cases, there are too many confounding factors that prevent us from
5
Earlier, we noted the CSE wh-expressions ‘for what’ (meaning why) and ‘at what time’ (meaning when) being
able to remain in-situ in a matrix clause ((37) – (38)) and acting as a stand-alone answer (39). However, they do
NOT resist wh-in-situ in embedded clauses, as expected in the general pattern shown in (40) – (45):
For this reason, we will not consider these wh-expressions as part of the general pattern.
30
c. How did John think Lisa escape?
In the set of utterances in (46) and (47), we used the verb think, which is compatible
with an interrogative (as in the sentence ‘John thought who left?’), as well as with a non-
interrogative (as in the sentence ‘John thinks Lisa is cute’). All versions of (46) – (47) have a
matrix question interpretation (that corresponds to the (c) reading). That is, utterances (46a) –
(46c) probe the identity of the person Lisa met, while (47a) – (47c) probe the way Lisa
escaped. Noticeably, (47a) is not acceptable, which further solidifies the general ban on wh-
Verbs that do not exclusively select for questions (such as think from (46) – (47))
provide a different paradigm from a question-selecting verb (such as want to know from (40)
– (45)), where a matrix interpretation is not possible: the sentence ‘Who does John want to
know Lisa met?’ has a drastically different meaning from its counterpart in (46c).
Additionally, the pattern obtained in (46) – (47) looks entirely similar to matrix questions.
Therefore, for these reasons, we only consider verbs that select for embedded question
A very important assumption that we have to make here is that the selectional
restrictions and properties of verbs are the same in CSE as they are in StdE; that is to say, a
verb that selects for a question but not a statement in StdE also behaves exactly the same way
in CSE. For instance, a verb such as wonder in CSE may allow for a declarative embedding,
points out, we can test whether this is true, since a wh-expression embedded under wonder
would then be able to take higher scope. In addition, the embedded clause could take a
declarative complementiser that. If that is the case, the prediction is that an embedded in-situ
31
(48) ?You wonder (that) John eat what?
This is, however, only grammatical to a minority of native CSE speakers; and
Now, it might very well be the case that such a verb would have its restrictions
‘relaxed’ and can now also select a declarative statement in CSE. It may even be the case that
the verb’s selectional properties are completely different and now it selects a statement but
NOT a question. In this study, native speaker judgments do not seem to suggest that there are
any radical changes (if at all) in verb selectional properties; or even if such restrictions are
beginning to see change in the variety, they do not affect the core asymmetry that this thesis
seeks to explain.
Construction Example
Note: wh-adjunct cannot remain in-situ (31a) ?John fix the car how?
Interrogative Matrix clause with wh- (28b) What (did) John eat?
movement
Non-interrogative Embedded clause with wh- (46a) John think Lisa meet who?
32
Interrogative Embedded clause with wh- (40a) John want to know who Lisa marry.
movement
Interrogative Embedded clause with wh-in- (40b) *John want to know Lisa marry who.
situ
In the previous sections, we assumed without much argument that the presence of the
auxiliary verb do (or do-support) does not affect the grammaticality of wh-moved CSE matrix
questions (for example, (28) – (33)). This should not be a controversial assumption to make,
since the dropping of do in such a context is very common in CSE. Another phenomenon that
is quite regular in CSE is the dropping of the complementiser that (and also if, whether, etc.)
in embedded contexts. This also occurs in Standard English, but perhaps with a lower
frequency.
However, in some other contexts, this pattern is turned on its head. Consider, in StdE:
33
b. Did John say who (*that) ___ married Lisa?
In (49), the wh-expression forms a long distance dependency with its associated gap.
This construction is possible with certain verbs known as bridging verbs (Erteshik, 1973). In
the sentence pair in (49), the complementiser is prohibited from occurring. The
ungrammaticality of (49a) and (49b) is explained by the Doubly-Filled COMP Filter (DFCF),
COMP position (Chomsky & Lasnik, 1977). It is assumed that a process of complementiser-
deletion then takes place to repair the construction. The DFCF takes the form *[ CP wh that]
and is operational in English, as well as its cognate languages such as German (Bavarian,
phrase, has been argued to be the landing site of wh-movement (Chomsky 1973, Bresnan
1976). Some varieties of English, such as Middle English and Belfast English, have been
reported to display overt evidence of this. The Middle English sentence in (50) (an example
from Han & Romero 2004, obtained from the Penn-Helsinki Parsed Corpus of Middle
English) and the Belfast English sentence in (51) show a wh-expression occurring adjacent to
a complementiser:
(50) First the behoueth to knowe why that suche a solitary lyf was ordeyned.
(cmaelr4-s0.m4,13)
‘First, it behooves thee to know why such a solitary life was ordained.’
(Henry, 1995:107)
In addition to Middle English and Belfast English, CSE is a variety of English that
violates the DFCF, and as a result, provide overt evidence of the landing site of wh-
34
movement. As such, CSE allows a wh-expression to co-occur with a complementiser both in
Therefore, the optionally overt presence of the complementiser as the head of CP does
not affect the grammaticality of CSE wh-questions, as it does in StdE (c.f. (49)). The
possibility of the DFCF showing not only crosslinguistic but also intralinguistic variability in
its application calls into question its status as a general or universal filter.
Additionally, CSE also seems to contrast with StdE in displaying that-trace effects,
which is a ban on subject extraction over an overt complementiser that (Perlmutter, 1971):
(54b) shows that when a subject in the embedded clause is extracted long-distance,
the movement cannot proceed across an overt complementiser that; rather, it can only
35
2.6 Embedded Direct Questions
Another phenomenon that is unique to CSE is that it is possible for direct questions to
be embedded in a context where in StdE only indirect questions are possible. When a
question:
In CSE, inflection is often optional, so a version of (57) with the past-tense inflection
The frequency of tense omission in CSE utterances such as (57) might have in turn
contributed to the following construction, which is an embedded direct question (that is found
To be clear, the embedded question in (58) – “What did John buy?” – is not reported
speech or a quotation. (58) is also not the same as the matrix question “What did I wonder
John bought?”. It is, then, an embedded indirect question with subject-auxiliary inversion.
There are multiple ways to analyse this construction6, which I will not go into here;
but I will simply note that the strong preference for (57) over its wh-in-situ counterpart is
consistent with the word order of a construction such as (58), as opposed to, say, something
like the clearly unacceptable *I wonder did John buy what?. This marked preference for the
6
There is also a link to be made between the phenomenon of embedded direct questions and the more general
one of optional Inversion in CSE sentences, such as in: “Not only the interior was overhauled for a fresh and
brighter look, the outdoor space has also increased substantially”. The lack of inversion of the subject and the
auxiliary is quite frequently observed in CSE.
36
wh-expression to move to the front of the embedded clause mirrors the central wh-movement
asymmetry discussed in this thesis. As far as I can tell, this particular phenomenon (58) is not
found in any of CSE’s superstrate and substrate languages. It might be possible in other
learner’s error (i.e., a failure to distinguish between relative clause formation and question
formation, for instance), and a deliberate construction. However, I would strongly argue
against an analysis which brands (58) as an erroneous utterance made in CSE, since this is
quite a robust observation in CSE. This is certainly not a learner’s error that can be chalked
The existence of a construction such as (58) is also a crucial piece of evidence against
evidence for this claim the fact that do-support and Subject-Auxiliary Inversion exists in
matrix clauses – take for instance sentences such as “What did John buy?” and “What can
John buy?”. The reason that do-support and Subject-Auxiliary Inversion takes place is to
provide an appropriate host for the affixal interrogative complementiser. From this, we can
rule out the possibility of wh-in-situ in the matrix clause, since the requirement that the
derivation then crashes because the phonological requirement on the interrogative head
cannot be interpreted at PF. However, when it comes to embedded questions, Bošković notes
that the interrogative C in English suddenly ceases to be a PF affix, given that neither do-
evident in ungrammatical sentences such as *“I wonder what did John buy?” and *“I wonder
what can John buy?”. Since the embedded interrogative C is now not a PF affix, we cannot
use the previous argument (that the PF affix cannot be interpreted at LF) to rule out wh-in-
37
situ in English embedded clauses. Subsequently, Bošković has to rule out the wh-in-situ
option in English embedded clauses in another way. Using a combination of the Extension
Condition and the principles of LF-insertion, he argues that interrogative C, which has a
strong +wh feature, can only be inserted in overt syntax. The insertion of C in overt syntax
then forces movement of the wh-expression, thus ruling out wh-in-situ. One mysterious
aspect of this analysis, as Bošković himself concedes, is that the interrogative C is sometimes
a PF affix (when it is in the matrix clause) and sometimes NOT a PF affix (when it is in an
embedded clause). If the interrogative C has the constant status of being a PF affix, we would
have the desirable result of using the same explanation to rule out the same phenomenon (wh-
in-situ). Conversely, if we can conclude that interrogative C is simply not a PF affix, then we
can rule out completely using that explanation for the same purposes. This is where the CSE
data comes in to help us make a decision. (58) is a prime example of a rather productive
Utterances such as (58), following Bošković’s reasoning, would prove that the interrogative
C in CSE embedded clauses is in fact, similar to its matrix counterpart, a PF affix. If this is
the case, the rest of Bošković’s analysis is thrown into doubt. More importantly, the status of
the embedded CSE complementiser as optionally being a PF affix would fit nicely with the
central asymmetry this thesis seeks to account for – that wh-movement is preferred over wh-
position in a wh-question to its surface position at the beginning of the clause. A common
diagnostic used to show overt wh-movement is the application of island constraints to the
38
extraction of wh-expressions. If the extraction of wh-expressions from islands is impossible,
(59) *Who you hear the report that Tom slept with ___? (Complex NP Island)
(60) *Who you come home because you don’t want to meet ___? (Adjunct Island)
(61) *Who you wonder whether ___ buy the cake or not? (wh Island)
The above data shows that island constraints apply in CSE as well. The sentences are
ungrammatical because of the illegal movement of the wh-expression across syntactic islands.
Incidentally, in (59) – (61), if the wh-expression had been left in-situ, the sentences would be
grammatical:
(62) You hear the report that Tom slept with who?
(63) You come home because you don’t want to meet who?
For partially moved wh-expressions, as in (46b) and (47b), there is a mixed pattern in
(65) a. *John like [the person [that think who [Lisa meet ___]]]?
(66) a. John like [the person [that think how [Lisa escape ___]]]?
questions:
(67) *Who you hear the report that Tom slept with him?
(68) *Who you come home because you don’t want to meet him?
39
(69) *Who you wonder whether he buy the cake or not?
The pre-requisite assumption here is that if island constraints are operational and the
restriction on resumptive pronouns is observed in StdE, then they should behave in the same
way in CSE as well7. However, this is not always true in CSE, as we can see in their relative
clauses. Therefore, in addition to checking if island effects are found in CSE ((59) – (64)), we
also check to see if resumptive pronouns can be found in questions ((67) – (69)). The
sentences (67) – (69) are indeed ungrammatical because the resumptive pronoun is assumed
to be base-generated in the same position as the moved wh-expression. Taken together, the
diagnostics appear to show that CSE wh-expressions originate in the same base positions and
7
This is not an uncontroversial assumption to make. For instance, CSE relative clauses show a different pattern
from questions – the head noun in CSE relative clauses is not sensitive to island constraints; and it is possible
for resumptive pronouns to appear within CSE relative clauses.
RC[The boy who I hear the report that Tom slept with ___ ] is here.
RC[The boy that I came home because I don’t want to meet ___ ] is here.
RC[The boy that I wonder whether ___ buy the cake or not] is here.
The boy who I hear the report that Tom slept with him is here.
The boy that I came home because I don’t want to meet him is here.
The boy that I wonder whether he buy the cake or not is here.
This means that whatever constraints/principles apply in StdE might not straightforwardly apply in CSE in the
same way. In this case, the diagnostics for wh-movement do not work for CSE relative clauses, even as they are
expected to. In other words, unlike CSE questions, CSE relative clauses do not exhibit signs of wh-movement
having taken place.
40
(70) Intervention
She argues that since both the focusing/quantificational operator and the wh-
(Beck, 2006: 1)
from being interpreted in-situ at LF; or in other words, the presence of an intervention effect
is a diagnostic that indicates a wh-expression does not undergo covert movement. However,
intervention effects are irrelevant for surface movement, since the scrambling of a wh-
Expressions that classically exhibit intervention effects (in languages where such
effects occur) include Negative Polarity Items (NPI), certain quantificational NPs, disjunctive
41
NPs, and focus elements such as only and even. In particular, one of the ‘strongest’ cases of
(Tomioka, 2007). Let us test for intervention effects in CSE with a variety of interveners, in a
variety of contexts:
element, which fulfils the standard intervention structural configuration. In (74), we have an
in-situ wh-argument in a matrix clause that does not show an intervention effect with only. In
(76), we see that a wh-adjunct in an embedded clause does not show an intervention effect
with always. Tentatively, it seems that CSE does not typically display intervention effects.
However, Yeo (2010) did report an intervention effect in CSE with the wh-adjunct why in an
embedded position. He did not find intervention effects otherwise with wh-arguments.
Negation
‘What is the reason x such that you don’t think he is happy for x?’
42
Quantified DP
‘What is the reason x such that everyone thinks he is happy for x?’
variable. It is generally known (Beck, 2006; Hoji, 1985; Kim, 2002) that one way to
overcome intervention effects is through scrambling the wh-expression over the intervener,
observed that intervention effects are much weaker in embedded contexts. Since CSE does
not display any intervention effects in matrix clauses, nor in embedded clauses, this
observation should not interest us very much. Superficially, the lack of intervention effects in
CSE tells us that wh-expressions undergo covert movement in order to avoid intervention.
This observation can be potentially problematic for an analysis that assumes otherwise,
although it is also entirely possible that there is something slightly special or different about
Another interesting and important property of CSE wh-expressions is that they can be
semantic variables, i.e. a wh-variable that does not have an operator associated with it, and
43
(79) He what also don’t want to eat.
(80) Who comes first, then who gets the prize loh!
(81) John want to buy what, his father (also) will pay (for it) one.
“Whatever John wants to buy, his father will pay for it.”
*“What item, if John wants to buy that item, will his father pay for?”
expressions can occur in CSE. In (79), the free choice item what is displaced from its object
The same meaning is not obtained when what is left in its original position, however. The wh-
expression who, which seems to be a wh-correlative of the Chinese type, found in the subject
positions of both the antecedent and the consequence clauses of the bare conditional in (80),
also does not have an interrogative meaning, but is interpreted as a universal quantifier, i.e.
x [x is a person & x comes first] (x gets the prize). Lastly, in (81), what is not displaced
from its object argument position (c.f. (79)) and is also interpreted as an indefinite; i.e. the
wh-expression does not probe for the item that John wants to buy. This seems to be a
There are several languages that exhibit indefinite wh-construals. The fact that CSE
complex forms such as whatever in English) is often attributed to substratal influence from
44
(82) Ni yao shenme, xuexiao hui gei ni shenme.
However, it would be an overly simplistic reduction to think that the Chinese wh-indefinite
paradigm has been transferred wholesale to CSE. For a start, in certain constructions where a
The use of the in-situ wh-expression shenme as an indefinite (to mean ‘anything’) is
possible in Chinese, as can be seen from (83); however, there is just no good way of
expressing the same meaning using a wh-expression as an indefinite in CSE. In fact, if the
question tag or not is removed from the CSE question, the in-situ wh-expression what will
take on an interrogative meaning, rather than a non-interrogative one. Conversely, there are
CSE constructions with wh-expressions as free relatives that do not have an equivalent in
Chinese as well:
45
Also, although not technically an indefinite, the ubiquitous rhetorical question ‘You
think what?’ (the apparent in-situ version of the StdE version ‘What do you think?’), as
(87) *Ni xiang shenme? Qian zhang zai shu shang ma?
(89) We decided that who come first, then who get the prize loh!
(90) ?Tom wondered if John want to buy what, his father will pay (for it) or not?
We took the previous examples (79) – (81) and embedded them within another clause.
(88) and (89) remain declarative sentences and there is no change in the grammaticality or the
structure of the constructions. I changed (81) into a question – “If John want to buy what, his
father will pay for it or not?” – and then embedded it as a complement to a matrix verb that is
only compatible with interrogatives (wonder). Interestingly, the result is that the indirect
Although it is not clear at this stage why this is so, the crucial observation in (90) is that the
morphology, such as wh-word + -ever (e.g. whatever, wherever, whoever) and any- + wh-
word (e.g. anyhow, anywhere). These expressions are, expectedly, also found in CSE. The
distribution and use of some of these expressions, however, have changed and I will be
46
It is clear that wh-expressions can be used in non-interrogative ways in CSE, just as is
the case in Chinese, although their distributions are not an exact match. In this regard, CSE
resembles Chinese more than it goes StdE, where bare wh-expressions cannot be used as
indefinites at all8.
Where behaves in much the same way in CSE as the other wh-arguments, and has a
fairly free distribution, being able to be left in-situ or to undergo wh-movement to the
beginning of a matrix clause. This is the case when where is used as an argument (91), or
In his account of CSE wh-data, Yeo (2010) previously suggested that there was an
animate/inanimate subject distinction when it came to whether where would be left in-situ or
undergo movement. According to him, CSE speakers strongly prefer to front the wh-
expression when the subject is human and animate, while there is no such preference for
inanimate subjects. Therefore, a question like (92a) would be dispreferred according to him.
It is noted here that Yeo does not offer any explanation for why this might be the case; and
8
Although Berman (1991) argues when wh-expressions appear in embedded clauses as in the examples below,
they behave on a par with indefinites in the sense of Heim (1982):
a. The principal usually finds out which students cheat on the final exam. (Berman, 1991: 62)
b. With few exceptions, Mary knows which students submitted which abstracts to which conferences.
(Berman, 1991: 65)
Note – The CSE sentences corresponding to (a) and (b) above use bare wh-expressions instead:
a’. I can find out who cheat in the final exam!
b’. MaryI know who submit what to where!
47
also, he only claims this asymmetry in regards to where and not the other wh-expressions,
which is mysterious. For the record, my CSE informants and I do not possess any systematic
Yeo.
There are several other observations to be made about the use of where in CSE. It is
sometimes accompanied with copula deletion (when in a copula construction such as (93)), or
The omission of copula and auxiliaries such as ‘do’ is quite commonplace in CSE and
examples (93) – (94) simply show that the omission applies without exception to where-
questions. Another interesting observation is the alternation between the use of where and at
B: (At) where?
The preposition at usually selects a location Noun Phrase as a complement and forms
a Prepositional Phrase that also denotes a location. In a typical question such as (95), where is
48
in-situ and can be optionally preceded by the preposition at. At can also optionally precede
where in a reply to a statement such as in (96). However, when the wh-expression is moved
to the front of the copular clause such as in (97), it cannot be optionally preceded by the
The distribution of the at where expression in CSE bears a passing similarity to what is found
in Chinese, although they are not identical: crucially, the preposition (zai in Chinese) is not
Bag at where
There is one other ubiquitous expression involving where in CSE that is commonly
marker got (see Lee et al., 2009 for more details on the functions of CSE got), the expression
where got is frequently used in CSE to rhetorically (and occasionally, literally) question the
existence of something:
(101) Where got people don’t know who is Donald Trump one?
“How is it possible that someone who doesn’t know who Donald Trump is
exists?”
49
The phrase where got is admittedly a very strange construction to StdE speakers who
are non-native CSE speakers, but should be quite familiar to Chinese speakers even if they
are non-native CSE speakers as well. Even though the wh-expression where is used, there is
nothing in the phrase that denotes physical location. Instead, the expression has multiple
indirect speech act (102). The phrase where got bears more than a passing similarity to the
Chinese expressions nali you9 (nali = where, you = have). Lee et al. (2009) label got as an
involving where, we see that the wh-expression has moved into a pre-verbal position.
Lastly, the adverbial anywhere can also be used in a preverbal position in CSE, with a
In the first place, we note that moving anywhere to the preverbal position, as in (104),
is possible in CSE but not in StdE. Secondly, there is a slight meaning difference between
(103) and (104) – the former has the meaning ‘do not throw your rubbish at all’ while the
latter is closer to ‘do not throw your rubbish in the wrong places (throw them in the bins)’.
9
There are other possible expressions in Chinese with nali (where), such as nali hui (hui = know) and nali
keneng (keneng = possible). They respectively mean, roughly, “how would I know” and “how is it possible”.
However, there are no such variants in CSE (*where know, *where possible).
50
Therefore, in addition to being left in-situ and moved to the front of the clause, where
in CSE shows an interesting construction with the preposition at, and forms the unique where
got construction, which is likely to be derived from Chinese. In the subsequent sections, we
will look at a few more constructions similar to where got, where we might be able to make a
the wh-adjunct why in CSE. It is already established that why does not appear in an in-situ
position in both matrix (105) and embedded clauses (106). It can also appear at the front of
*‘What is the reason x such that you think he came to class for x?’
*‘What is the reason x such that you think for x he came to class?’
‘What is the reason x such that you think for x he came to class?’
‘What is the reason x such that you think he came to class for x?’
unacceptable question no matter whether we take the embedded reading or the matrix reading
51
of the sentence. (107), with the wh-expression fully moved to the front of the matrix clause,
has a matrix interpretation; while (108), with the wh-expression moved to the front of the
It was claimed in Yeo (2010) that a question similar to (107), but with negation in the
embedded clause – my examples do not contain negation to avoid any potential intervention
‘What is the reason x such that you think he didn’t come for x?’
‘What is the reason x such that you think for x that he didn’t come?’
Yeo claims that (109) is ambiguous between the matrix and the embedded
interpretations. However, I do not find this consensus among CSE native speakers that I
consulted with. A vast majority of them preferred the matrix reading, while some remarked
that they could possibly get the embedded reading only if the question was presented with
Another contrast in the data presented by Yeo (2010) and this author is the availability
mentions one of his informants noting that this construction sounds “very Chinese”, and
52
speculates that speakers who are more proficient in Mandarin Chinese and less in CSE would
find this ok. Indeed, this pre-verbal distribution of why in CSE is on par with that of its
ethnicity or the langauges they speak – are not known to avoid constructions that are surface-
similar to languages that are not spoken by their ethnic group, or languages that they simply
do not speak. The non-Chinese CSE native speakers I checked (111) with agree that it is a
Like the previous wh-adjunct described above, how cannot be left in-situ in a matrix
clause (113), it must be moved to the front of the clause (114). How cannot be left in-situ in
an embedded clause (115), but it can appear in the Spec, CP (clause-initial) position in
53
(115) *You know John win the contest how?
*‘How did you know (the fact) that John won the contest?’
‘How did you know (the fact10) that John won the contest?’
(113) and (114), being matrix clauses, have fairly straightforward readings, although
(115) is ungrammatical due to the in-situ position of how. (115) is ungrammatical for any
interpretation of how. A typical answer to (116) would be something like ‘Yes I do’; although
it is certainly possible to interpret (116) as an indirect speech act (Austin, 1962), and answer
it like you would answer (114). Lastly, when how is in the matrix clause-initial position of a
clause that embeds another clause, it maintains a matrix interpretation and not an embedded
one, as in (117).
There are several different interpretations that can be invoked by how when it is used
in a question; and this can be seen in the type of responses/answers given to the how-
question. Taking (114) as an example question, a Manner response to that question could be
something like “He won with aplomb”; an Instrumental response could be “He won by
bribing the judges”; and a Resultative response could be “He won it quite successfully”.
(113) is unacceptable in CSE on all 3 of these readings, while (114) is acceptable on all of the
interpretations.
10
Interestingly, this question can also be asked felicitously even if it is not actually true that John won the
contest, i.e. John did not win the contest. If that is the case, the question probes for the source of information, as
in the paraphrase ‘Where did you learn that John won the contest (because that source is inaccurate)?’.
54
There is another construction in CSE involving how in an in-situ position, which is
surprising because we have consistently observed that wh-adjuncts do not appear in those
positions:
*‘The car that you fixed yesterday – how did you do it?’ (Ambiguous between
In (118), there are two distinct clauses – the relative clause Noun Phrase and a
truncated wh-question appended after that. There is a short break (indicated by the # sign)
between the two phonological units as evidence of this. It might even be more accurate if the
break is represented by the | sign, which represents separate intonational phrases, since
speakers can produce a slight falling intonation at the end of the Noun Phrase (specifically,
on yesterday), while how is produced with rising intonation. This is even more obvious if a
discourse particle is added to the end of the first intonation phrase. The problem with an
analysis based on phonological pauses is that natural CSE speech is often rapid and these
breaks are not always observed. Therefore, it is not unreasonable to simply assume that how
can appear in-situ in CSE. However, it is noted that for questions like (118) with an in-situ
how, the interpretation is exclusively Resultative, with native speakers unable to get the
Manner or the Instrumental interpretation. This somewhat weakens the in-situ-how argument;
interpretations, whereas its in-situ-how counterpart has access only to the Resultative
interpretation, and the two versions should be on par with each other, but are not. The natural
explanation is thus: the so-called in-situ-how question is structurally different from the
55
regular fronted-how question, as illustrated in (118). In other words, there is no in-situ-how
There are various other truncated how-constructions in CSE that indirectly support the
acceptable) is more likely part of a separate structural unit. They appear with varying
(119) How to V
I buy a ticket then they don’t let me in # how can (like that)?
‘I bought a ticket but they are not letting me in, how can they do this?’
(121) If…how
56
(124) Exasperation/Distress
The how-to-Verb construction in (119) can also be found in StdE, especially when
found at the end of the question, and can be truncated to a single how. However, the phrase
cannot be truncated if it is fronted. It can also take on a rhetorical meaning, when someone is
describing an impossible task, as in “This question so difficult, how to do?”. In (120), the
phrase how can (like that) generally means ‘how can they do that?’. In (121), the then how
informally even in StdE; in CSE, it can additionally be found at the end of the sentence. See
how (123) is probably a truncation of the common phrase see how it goes, and has roughly
the same meaning. Finally, in (124), how is used to express exasperation and distress, usually
after an unpleasant or undesirable situation. There is a clear break between the main clause
and how. The data presented here show a wide variety of possible truncations that involve the
wh-expression how in CSE. In particular, (119), (120), (121) & (124) constitute strong
consistent with the generalization made in the previous section that CSE wh-adjuncts do not
appear in-situ.
Lastly, we take a look at an expression derived from how – anyhow. This adverb can
also be found in StdE; however, its meaning and distribution in CSE seems to have deviated
significantly from how it is used in StdE. In StdE, the adverb frequently occurs in the clause-
57
initial position, and sometimes in a clause-final position. In CSE, in addition to appearing in
This construction bears similarity to the preverbal anywhere in (98), although the use
of anyhow is much more productive in CSE, and has also led to other innovations such as
coming before generic action verbs such as whack, reduplication (anyhow anyhow) and
In this section, we will look at the makeup and structure of tag questions in CSE. Tag
questions are not strictly confined to the realm of wh-questions, since they do not typically
contain wh-expressions; and indeed, are rather thought of as an alternative strategy to forming
auxiliary or copula verb, and a pronoun; these will agree in terms of grammatical features
with the matrix subject and verb. CSE tag questions deviate from StdE mainly in the tags
themselves:
(127) You want this apple or not? The apple sweet or not?
58
(128) You want to watch a movie or what?
The first and most distinct difference in CSE question tags is that grammatical feature
agreement is no longer obligatory: in (126), the copula is in the tag does not agree in number
with the copula are in the matrix clause, and the pronoun it in the tag does not agree in
number with the matrix subject the boys. In fact, not only is agreement optional in question
tags, it seems preferred – is it is a frequently used tag in CSE and will be found far more
frequently than a tag such as aren’t they. Another commonly used tag in CSE – or not (127) –
which also sees informal use in StdE, is used to form yes-no questions by using a general X-
Neg-X structure (found in many other languages including Chinese). Another frequently used
tag in CSE or what (128) seems similar to or not, but it is functionally different as it does not
simply result in a yes-no question, since the question tag itself opens up the possibility of a
different answer other than a simple refusal. Or what is also used as a slang term in StdE
which adds emphasis or suggests an option. It is noted that what in the tag or what does not
actually refer to or probe for an entity, which it typically does. It is also not clear if it is part
(129) Do you want to watch a movie, or what (else do you want to do)?
Do you want to watch a movie, or (do you) not (want to watch a movie)?
The ellipsis analysis of a question tag such as or what and or not is an interesting one
and cannot be completely ruled out, but we will not delve any deeper here. This section
shows us that the strategy of combining a declarative sentence with a question tag is very
productive in CSE, especially with tags such as is it, or not, and or what.
59
2.14 Question Particles
repertoire of discourse particles in the variety and they perform an equally wide range of
linguistic functions, from shifting the type of a sentence, determining the pragmatic force of
that shares the form of a wh-expression – what – although it has no interrogative force and is
not compatible with questions at all. Discourse particles share certain common properties
such as: being ‘syntactically optional’, coming at the clause/sentence-final position, and
carrying distinct intonation patterns that affect their interpretation (Kwan-Terry, 1991).
Discourse particles have largely been acknowledged to be the result of the influence from
substrate languages such as Chinese, Malay, and Chinese dialects. In Lim (2007), it was
suggested that “…in an early era Bazaar Malay and Hokkien contributed the particles lah, ah
and what, and Cantonese in a later age provided loh, hor, leh, meh and ma, inclusive of tone.”
In this section, we will take a look specifically at the question particles ah, meh, leh and hor.
Ma is a question particle that forms yes-no questions, just like its counterpart in
Chinese. For this reason, ma sees some use in CSE, but almost exclusively among Chinese
speakers. It is extremely rare for Malay and Indian Singaporeans, or even Chinese
Singaporeans who are not fluent in Mandarin, to use ma productively in everyday speech.
The most prominent and frequently used question particle in CSE is ah. There are two
distinct ah particles, one with rising intonation (áh) and one with a falling intonation (àh).
The former is compatible with syntactically interrogative utterances, while the latter is
60
(131) You think what he eat for dinner áh?
The two examples above are structurally interrogative and therefore áh is compatible
with both of them. In these contexts, áh expresses an interrogative meaning. Rising áh can
also occasionally be used with a sentence with declarative syntax, but it has little to no real
interrogative force as the utterance “…becomes rhetorical and is usually just to reiterate or
check a fact with no response required from the interlocutor…” (Lim, 2007):
Example (132) simply functions as a reminder to the recipient of the message to ‘not
work too late tonight’. Outside of this function, áh should not be compatible with a
Example (133) is a question that requires a response and is not merely rhetorical nor a
reminder. Although the utterance is syntactically declarative and has a mid fall or low pitch
softening a command, etc. (Gupta, 1992) – but since these have nothing to do with question-
61
formation, I will not touch on them. CSE discourse particles are notoriously multi-functional
(Ler, 2006) and it should not be a surprise that particles that can be used in questions are not
Another CSE question particle – hor – is only compatible with utterances with
declarative syntax. Hor has rising intonation and it is used to assert a proposition and to
garner support for that proposition from an interlocutor (Kim & Wee, 2009):
The utterance in (134) seeks to assert the fact that John is very cute, and additionally
also seeks agreement/support on this fact from whomever the proposition is conveyed to. In
this context, hor is quite similar to right in StdE. It can also be used in a rhetorical question,
such as:
In (135), B’s question is rhetorical because B has already come to the conclusion that
A is very free based on what A said, and B is not really genuinely asking A if A is free. Like
falling àh, hor is not compatible with an utterance with interrogative syntax. Its other non-
Leh is another CSE question particle, which is used with rising intonation and high
pitch. It is generally used with declaratives, but more often than not, simply with Noun
Phrases. It can be informally described as having a ‘what about…?’ meaning when used:
62
B: “What about your daughter?” (Does she need to go to school? / When is
In (136), B’s question is well-formed in CSE but is technically ambiguous because the
particle leh can question whether A’s daughter is going to school at all, or it could question
when A’s daughter is going to school. We can also see this ambiguity from the different
possible replies by A. From (136), we can also see that in addition to leh’s ‘what about’
meaning, there is an element of comparison in its function (Lim, 2007), when A’s son is
compared with A’s daughter. Leh is generally not compatible with interrogatives and with
other wh-expressions, although there are a number of exceptions. The wh-expression why, for
B’s question is well-formed and can in fact be shortened to simply ‘why leh?’ most of
the time for a question of this sort. It seems that leh is used when the question probes for a
reason (137) or when there is a comparison (136). However, it does not work too well with
the other wh-expressions. In fact, leh seldom appears even in what about-type questions,
There is a certain redundancy with having both ‘what about’ and leh in the same
sentence (138), and the construction is given a mixed judgment by native CSE speakers. Leh
63
also has a non-interrogative function, which is uttered with low tone. It is used in this way as
a pragmatic softener.
The last CSE question particle is meh, which has a mid to high pitch. Meh is only
seems related to the degree of sureness/suspicion the speaker asking the question possesses
(Lim, 2007) – the closer the question is to being a rhetorical one, the lower the pitch the
(139) This pen is yours meh? (I don’t know if it is yours) High Pitch
In (139), meh is used with a declarative that questions whether the pen belongs to the
speaker’s interlocutor. There are two types of meh that can be used in such a question: if the
speaker is legitimately questioning, and has no presuppositions on the ownership of the pen,
the particle will be used with high/rising pitch; if the speaker has a strong presupposition that
the pen does not belong to his/her interlocutor, the particle will be used with mid pitch.
Although meh is not compatible with interrogatives (sentence type) and wh-expressions, any
utterance with meh can only be interpreted as a question, even if it is a rhetorical one.
Let us also examine the distribution of particles in CSE. Particles generally occur at
the sentence-final position, as they have semantic scope over the entire utterance. That is the
usual position particles occupy, as can be seen in the many examples above. We should also
take a look at one other position particles can be found in. In some cases, a particle can be
found at the end of a phrase or a clause, such as when it follows a topic or contrastive focus,
64
or anything that has been left-dislocated for some kind of discourse or information-
configurational purpose:
In (140), the NP John is followed by the particle lah, which generally indicates its
status as the Topic of the sentence. In (141), the NP John is in a clefted position and is
followed by hor, which sets up a question that verifies the identity of the person who kicked
the ball. The particle also indicates that John is in contrastive focus, since he is only one of a
number of candidates that the speaker knows/suspects to have kicked the ball. In (142), we
see an example very relevant to our discussion in this section: the wh-expression has been
moved to the front of the question and the question particle ah is pied-piped with the wh-
expression (or vice-versa). Besides ah’s function as a question particle, we could speculate
that by virtue of the wh-expression being fronted, in conjunction with the particle ah forming
a distinct intonational unit with it, therefore producing a pause after the phrase, ah doubles up
Moreover, the fact that there are now two instances of rising intonation – one at the
end of ah and another at the end of bought – tells us that this utterance could be analysed as
two distinct but obviously related questions. The what ah phrase at the beginning of (142)
could very well be an elided question of the sort frequently found in CSE. In other words,
there are principled reasons for us to treat a construction such as (142) differently from the
same question but with the question particle ah left at the end of the sentence. It should also
be noted that empirically, the fronted question particle ah does not occur very frequently in
CSE, and in fact (142) is regarded as questionable by some native CSE speakers.
65
To sum up this short discussion, generally, question particles do not differ greatly
from other CSE discourse particles in terms of distribution. They are usually found at the
sentence-final position, and can sometimes be located at the end of a displaced phrase such as
a Topic or Focus.
The question particles that we have described here include ah, hor, leh and meh. To
be clear, as long as the particle is part of a grammatical construction that can be interpreted as
we do not only consider particles that are compatible with interrogatives, since it turns out
that many CSE questions are in fact utterances with declarative structure but with ‘question
intonation’ and question particles. Table C summarises the properties of the discussed
discourse particles.
The two generalisations that can be formed here are: Firstly, many question particles
in CSE turn out to be compatible with declarative syntax rather than interrogative syntax;
secondly, many question particles in CSE are incompatible with wh-expressions. These two
generalisations can be captured by Cheng’s (1997) Clause Typing Hypothesis (CTH), which
66
was proposed to account for the difference between wh-movement languages such as English
(144) a. No language has yes/no particles (and thus wh-particles) and syntactic wh-
movement.
particles; while English types a sentence via syntactic wh-movement. The CTH was
controversial, due to the strong predictions made by it which were ultimately not borne out.
An Economy-driven corollary of the CTH predicts that all languages with wh-movement will
lack particles, while all languages with wh-in-situ will have particles. It also predicts that a
language can only either have wh-movement or particles as a typing mechanism, but it will
not have both, nor will it have neither. The case of Malagasy, a language with both wh-in-situ
and wh-fronting (Potsdam, 2006) falsifies both these predictions. CSE is also one of the
67
questions with interrogative syntax AND particles. Cheng (1997) herself notes the existence
of such languages that violates the predictions of the CTH. To prevent the violation, she
claims that wh-fronted questions in these languages are actually cleft sentences:
(145) [CP [DP miin] [CP OPi illi [IP Mona shaafit-uhi]]]. (Egyptian Arabic)
clefted constructions ‘Who is it that Mona see?’ or ‘It is who that Mona see?’, followed by
deletion of the complementiser and the cleft, leaving the final ‘Who Mona see?’. I am not
sure that this analysis would be on the right track, since properties associated with a cleft,
such as contrastivity and presuppositional force, are not always present in CSE wh-fronted
questions:
B: Nobody.
The fact that the presupposition and contrast can be cancelled shows that the cleft
construction analysis might not be accurate. I will end the discussion here by noting the fact
that many of the questions in CSE possess declarative syntax is interesting and is something
68
Chapter 3 – A Review of Analyses of wh-constructions
Research on questions has had a long and illustrious history in the development of
language acquisition, among others. Of these fields, the impact on syntactic research has
probably been the most profound. Questions can be generally grouped into wh-questions and
yes-no questions, with the former being more widely discussed than the latter (yes-no
questions tend to be structurally less complex and rather easier to analyze, cross-
linguistically). Research on wh-questions in the Generative and Minimalist tradition has been
centered on two major strands: Firstly, what drives wh-movement? What factors cause,
attract, or force wh-expressions to be displaced from their original positions? Secondly, what
restricts wh-movement? What factors prevent or constrain wh-expressions from moving out
Also, what factors restrict the range of wh-movement to shorter distances, when long-range
movements are not observed? Given the wide typological variation of wh-phenomena and
language-specific word orders, we can thus appreciate how challenging (if at all possible) it
must be to construct a coherent and non-contradictory theory about wh-questions which has
both the descriptive scope and empirical accuracy. We will review some of the literature on
wh-questions, as well as studies pertinent to CSE wh-phenomena. However, the way this
chapter will be organized will be thematic: Firstly, we will look at existing analyses of
languages in the linguistic ecology of Singapore – StdE, Malay and Mandarin Chinese. The
rationale behind doing so is this: one of these analyses could be a ready-made explanation for
the facts at hand. Secondly, we will look at analyses that specifically deal with discourse-
seemingly free variation in the CSE data, particularly in matrix clauses. Then, we will look at
69
analyses of languages that others have claimed also to exhibit ‘optional’ wh-movement. Since
that is one of the defining characteristics of CSE wh-constructions, there might be something
useful for us in there as well. Lastly, I will examine analyses that were proposed specifically
to account for CSE wh-constructions. In particular, we will focus on a detailed and technical
We have previously looked at the historical and socio-political conditions that were
necessary, and ideal, for a contact language such as CSE to develop and thrive. The lexifier
language, or the superstrate, is English; while the substrate languages include Malay,
Mandarin Chinese, and possibly Tamil. There is some debate as to which substrate language
– Chinese or Malay – is the more significant source or influence for structural and functional
innovation in CSE, among scholars who work on CSE (see for instance Bao (2005), Bao &
Lye (2005) and Bao (2009), who argue for systemic Chinese influence; and Sato (2013), who
argues for influence from Malay). Interestingly, there is also significant typological
congruence between Chinese and Malay. The main arguments for Chinese as the primary
from Chinese, and the majority of Singapore’s population being Chinese. However, the
arguments for Malay as the primary substrate language are mainly based on socio-historical
relevance: Malay was the lingua franca of the region and of Singapore for a good long time
up until English eventually took over. It is obvious that both languages have significant
import on CSE.
I am not sure if there is any theoretical or explanatory upside to taking only one
position (to the exclusion of the other) in this argument, although it definitely seems unwise
70
to discount the influences exerted by any of the substrate languages in the contact ecology of
CSE. The only exception that I can think of, that warrants such an exclusive stance, is an
analysis that is based on systemic substratal transfer of grammatical features, such as in Bao
(2005), Bao & Lye (2005) and Bao (2009). In such an analysis, the so-called ‘Cafeteria
Principle’ (Bickerton, 1981), which states that a contact language can ‘pick and choose’ the
linguistic features from its superstrate and substrate(s) to adopt, is rejected. Instead,
“substrate transfer targets the entire cluster of properties…”, and “…rules out the ad hoc
mixing of competing features from typologicall distinct languages” (Bao & Lye, 2005). I will
3.1.1 Malay
There are many varieties of Malay but for our purposes, we will focus on the obvious choice
terms of wh-phenomena, Malay is also very important to the linguistic ecology of CSE.
nominal wh-expressions (both arguments such as apa ‘what’ and siapa ‘who’, and adjuncts
such as di mana ‘at where’) can freely move (147) or stay in-situ (148), while adverbial wh-
expressions such as kenapa ‘why’ and bagaimana ‘how’ must undergo movement (149) and
71
‘Where did Ali buy a condominium?’
The pattern shown here immediately reminds us of a similar pattern in CSE, where
wh-arguments are free to move or remain in-situ in matrix clauses, while wh-adjuncts cannot
remain in-situ and have to move ((28) – (33) from the previous chapter).
The pattern for embedded clauses is also very interesting. The full range of optionality
is in display here, with the 3 options – Full Movement (151), Partial Movement (152), and
wh-in-situ (153):
(151) Siapai (yang) [CP Bill harap [CP yang ti akan membeli
baju untuknya]]?
(152) Ali memberitahu kamu tadi [CP apai (yang)[IP Fatimah baca ti]]?
72
The examples here correspond to the CSE data presented in (46a-c). A popular
analysis of Malay wh-patterns is provided by Cole & Hermon (1998, 2000). For Cole &
Hermon (henceforth, C&H), Q is universally strong and the range of variation in wh-
movement across languages need not appeal to variation in the strength of Q. Therefore, there
lexical items/morphemes. C&H argue that the crucial explanation for the variety of
movement and non-movement options in Malay is that a wh-expression can either have the
option of its question operator (Op) and the variable (Var) joined together into a single word;
or generated separately, in the lexicon. Question operators are universally generated as null
operators and the wh-forms as variables. A question displaying Full Movement (such as
(151)) simply follows from the feature checking requirements of matrix Q, which is assumed
to be universally strong by C&H. Full movement obeys both weak and strong islands in
Malay:
(154) a. *Di manai [kamu fikir [Ali suka [perempuan yang tinggal ti]]]
(155) *[CP Siapai [t’i yang [Ali mengahwini ti]] mengecewakan ibunya]
73
Ungrammatical Overt Movement from Adjunct Island
(157) *Apai (yang) [awak agak [di manaj [Mary beli ti tj]]]
movement of the wh-expression to the beginning of the subordinate clause (or the Specifier of
the subordinate CP), followed by LF movement of the operator to root scopal position. The
two relevant pieces of evidence are: Firstly, partially moved wh cannot move overtly out of
an island:
74
Secondly, Subjacency also applies when an island boundary intervenes between the
surface position of the partially moved wh-expression and the Specifier of the CP
already eat
This suggests that partially moved wh involves not only overt syntactic movement of
the wh-expression but also covert movement of the operator to scopal position.
result of the needs of the moved element itself, a version of Greed. C&H formulates this as
the wh-Operator Condition, which states that a question operator must be located in the
the language has a focus construction, and a strong Focus feature in embedded Spec, CP that
triggers overt movement of the wh-expression (which must be a merged [Op+Var] lexical
item).
75
For a wh-in-situ question (such as (153)), the operator is merged at root Spec, CP and
unselectively binds a wh-variable in its scope. Therefore, the operator and the variable are
generated separately, and there is no movement at any level. The evidence for this lies in two
facts: Firstly, Malay wh-in-situ does not obey any island constraints:
Secondly, the meng-prefix, which is standardly omitted when the object of a verb
undergoes wh-movement over a verb that would otherwise permit the prefix (163a), appears
76
Both of these movement diagnostics suggest that no covert movement has taken place
in the case of wh-in-situ. Further independent evidence that wh-expressions can be used as
pure variables that are bound by non-wh-operators, as in the case of wh-in-situ Malay
constructions, is found in Reduplication (164) and with the word pun ‘also’ (165):
that do not range over entities, but over propositions. Therefore, this accounts for why they
can never be found in-situ. Also, if wh-adverbials in Malay are never simply variables, it is
predicted that they can never receive a non-interrogative interpretation (i.e. they can never be
bound by an operator other than the interrogative operator). This prediction is borne out by
the fact that reduplication and pun ‘also’ (as shown above) do not work with wh-adverbials
In summary, C&H’s analysis of Malay wh-movement (Full, Partial and in-situ) is that
configuration.
languages that display wh-in-situ. Aside from Cole & Hermon’s (1998, 2000) analyses, it has
77
been used in multiple accounts, such as Tsai (1994) and Chomsky (1995), among others. This
interpretation mechanism is also favored in the Minimalist Program. The proposal that both
the wh-feature in C0 and the wh-feature of the wh-expression are interpretable meant that
there was no more need for the wh-feature of the wh-expression, or the wh-expression itself to
move to C0. It is then assumed that in-situ wh-expressions are interpreted via unselective
binding. However, there was one prominent opponent to unselective binding. Reinhart
(1998), working with Minimalist assumptions, argues that unselective binding is inadequate.
Essentially, Reinhart agrees that the heart of such analyses – that there can be no further LF
movement of wh-in-situ and they must be interpreted and assigned scope without moving –
Superiority and apparent Empty Category Principle (ECP) effects with adjuncts. Reinhart
interpretation is that it can be answered only with the identity of the person corresponding to
who, i.e. the scope of the wh-in-situ is the embedded clause. The second interpretation is a
paired-list reading, where what has scope over the matrix clause, roughly corresponding to
the reading in (167). Considerations of Economy come into play as adjoining the in-situ what
to who should be dispreferred to the adjunction of what to where, which constitutes a shorter
move. Therefore, in order to permit the interpretation in (167), we would either have to say
that such (LF-) movement does not exist; or that this movement operation violates economy.
78
Huang (1982) claims that LF movement, unlike syntactic movement, does not obey
Subjacency. Reinhart disagrees with this fact on two levels. The first objection is a theoretical
one. There are no levels in the Minimalist Program and there is only one derivation (LF)
which can be Spelled-Out to the PF interface. Therefore, it is impossible to state that up to the
branching to PF a certain constraint has to be obeyed, and beyond that it does not have to be.
concludes that Subjacency is a general constraint on Move Alpha, and that there is no
difference in this respect between PF and LF movement. Since this is the case, it is obvious
that LF movement of wh-in-situ is simply not possible. She also points out further problems
(a) the ‘Donald Duck’ problem, and (b) how to handle wh-adverbial/adjuncts.
Firstly, Reinhart observes that a serious problem exists in the interpretation of wh-
Wrong
(170) For which <x, y>, if we invite y and y is a philosopher, then x will be
offended.
79
(172) Lucie will be offended if we invite Donald Duck.
Right
(173) For which <x, y>, y is a philosopher, and if we invite y, x will be offended.
In the question (169), the N-restriction which philosopher is found within the if-
clause. Thus, if we leave the restriction in-situ, we will obtain a semantic representation such
as (170), and equivalently, (171). This N-restriction is found in the string “…and y is a
philosopher” within the representations in both (170) and (171). Given these semantic
representations, an answer such as (172) – “Lucie will be offended if we invite Donald Duck”
– will be interpreted as correct, but it is not. Since Donald Duck is not a philosopher, it must
be true of him that if he was a philosopher and he was invited, Lucie would be offended. In
fact, any entity that is not a philosopher could be used as the value for the variable y, since its
restriction occurs in the antecedent clause of an implication, and the statement will still return
a correct truth value. Thus, if we do not want (172) to be in the set of possible true answers to
the question (169), we will have to ‘pull out’ the N-restriction from the implication. What this
means is that the value for y has to satisfy the condition of being a philosopher before the rest
of the proposition can be true. The result of ‘pulling out’ this restriction can be seen in the
semantic representations in (173) and (174), where the string “y is a philosopher” is no longer
conjoined to the implication “if we invite y”. This correctly allows the values for y to be all
and only those individuals who are philosophers and for whom the implication is true.
Donald Duck would no longer be a correct answer to the question given the revised
representations.
80
In summary, the interpretative problem here is how to assign wide scope to a wh-
phrase which shows properties of staying in-situ, without essentially ‘pulling out’ its
restriction (i.e., without moving it at LF). Reinhart proposes a solution to this interpretation
problem, which is to use choice functions. Choice functions are functions that apply to a non-
empty set and yield an individual member of the set. They are notated here as ‘CH(f)’. If we
use a choice function that applies to the set of philosophers, in a context similar to the
(176) {P|(∃<x, f>) (CH (f) & P = ^((we invite f(philosopher)) (x will be
operator that is arbitrarily far away. In other words, the N-restriction ‘philosopher’ can
remain in-situ within the implication – the if-clause – and the representation above can still
capture the correct truth conditions of the question in (169). The representation in (176) states
that a choice function exists such that if we invite the philosopher that is selected by the
function, someone will be offended. This is equivalent to saying that the philosopher-phrase
has a wide scope interpretation, if its set is non-empty. The wide scope reading of existentials
can be accounted for by quantification over choice functions. By extension, since wh-phrases
are also existential quantifiers, the same mechanism can be applied to them. The use of
choice functions as a means to interpret a question differs from the semantics of questions
proposed by Karttunen (1977), who views wh-NPs essentially as existential NPs, and the
question as denoting the set of propositions that are true answers to it.
Reinhart’s theory also derives another observation both in Malay (as noted by C&H in
(149) – (150)), and CSE ((31) – (33)) – how to capture apparent ECP effects in the case of
81
adjunct wh-phrases, or the argument-adjunct asymmetry in in-situ wh-phrases described
How and what way from (177) and (178) respectively are both adjuncts; however, the
former being in-situ causes the question to be unacceptable while the latter does not. This is
because a wh-adverbial adjunct such as how is fundamentally different from a wh-NP adjunct
such as what way, and this difference is crucially picked up by the choice function. Wh-
adverbials do not have an N-set, therefore, they have no N-role or variable. They denote
functions that range over higher-order entities. Therefore, this means that they cannot be
interpreted via a choice function, which applies to a non-empty set and yields an individual
member of the set, as there is simply no set of individuals that the function can apply to, nor
is there an appropriate variable for the function to bind. It is for this reason that Sato’s (2008)
Indonesia – claims that a Choice Function analysis of wh-in-situ is better than an Unselective
In a more recent study, Sato (2013) takes the basis of C&H’s analysis and applies it to
can consist of both the interrogative operator and the wh-word conjoined (as in English), or
where the wh-word is base-generated separately from its interrogative operator (as in
Chinese). I will summarise the analysis and the corroboration of CSE facts below:
82
- The interrogative operator is combined with the wh-variable as a single lexical
entry. This conjoined element then moves overtly to matrix Spec, CP. This derives
- The interrogative operator is base generated in matrix Spec, CP; wihle the wh-
variable remains in-situ. The former then unselectively binds the latter. This
Then, only the operator moves to the matrix Spec, CP position. This derives CSE
whereas CSE wh-in-situ should not obey island constraints. This is predicted in
(59) – (64).
- Partial wh-movement should obey island constraints. This is predicted for wh-
arguments (65).
predicted (see (31a), (32a) and (33a)) but exceptions such as (35) and (36) exist.
The predictions that Sato’s proposal makes for CSE are largely accurate so far.
However, there is a particular construction that does not seem to receive an explanation in
this analysis. Examples (40b), (41b) and (42b) show that a wh-argument is unable to remain
83
verb. It is unclear why this is so, since there is nothing in the analysis that prohibits an in-situ
wh-argument in that context. A potential workaround for this problem could come in the form
of an alternative account of Partial Movement according to C&H – that the language has a
focus construction and a strong Focus feature in the embedded Spec, CP triggers obligatory
movement of the wh-expression (which must be a merged [Op+Var] lexical item). However,
this particular explanation is not terribly convincing since in-situ positions can conceivably
be Focus positions as well. There is also no independent evidence that there is a Focus-based
motivation behind the movement to embedded Spec, CP position; nor is there a principled
reason why this is also not the case for sentences such as (46a).
To sum up this section, Sato (2013) proposes that there is a restructuring of a micro-
parameter within CSE that changes the composition of wh-phrases in the lexical component,
which is based on Cole and Hermon’s (1998) analysis for Malay. The basis for this proposal
is that CSE wh-questions bear much surface similarity to Bazaar Malay wh-constructions: in
terms of disallowing wh-adverbials in-situ, allowing partial wh-movement, and the non-
and predict the majority of the pattern of CSE wh-constructions, save for a few exceptions. It
is also noted by Sato that the phenomenon of partial movement is not found in Mandarin
Chinese, and other dialects like Cantonese and Hokkien. His analysis also claims to present
“…the first strictly grammatical evidence” in favour of substratal influence from Malay, on
CSE.
84
3.1.2 Mandarin Chinese
There are many Chinese languages, dialects and varieties, but we will focus on the variety of
Mandarin Chinese (henceforth, Chinese) spoken in Singapore. Unlike CSE or Malay, Chinese
Cheung (2014) on wh-fronting in Chinese), but is included in the discussion here because of
its importance to CSE’s linguistic ecology. Chinese is widely assumed to be the substrate of
CSE in the context of language contact, given its historical and geographical significance on
the island-state of Singapore. Past studies of CSE have even (simplistically) branded CSE
grammar as a ‘mixed system of English and Chinese grammatical rules’. The general wh-
pattern in Chinese is that wh-arguments remain in-situ in both matrix and embedded contexts,
whereas wh-adverbials have to move to sentence-initial position in matrix clauses, and are
85
(Note that it is equally fine in (182) for the wh-expression weishenme ‘why’ to precede the
Despite the existence of sentences such as (181) and (182), Chinese is standardly
known as a wh-in-situ language. Huang’s (1982) seminal work on Chinese offered a LF-
covert raising operation at LF to the Spec, CP position. The main advantage of this proposal
is that it unifies languages such as English, which has wh-expressions raising overtly to the
same position, with wh-in-situ languages such as Chinese. The evidence for this proposal is
manifold: For instance, the in-situ shei ‘who’ in (180) gives an embedded question
question interpretation if the matrix verb was yiwei ‘think’. In English, this is simply
attributed to the fact that a verb such as ‘wonder’ selects for an interrogative [+Q]
complement, whereas a verb such as ‘think’ selects for a non-interrogative [-Q] complement.
In Chinese, these same requirements can be fulfilled by proposing that shei ‘who’ raises to
the embedded Spec, CP position in (180), making the embedded clause an interrogative
complement suitable for the verb xiang-zhidao ‘wonder’; and it will raise to the matrix Spec,
CP position in exactly the same sentence but with the verb yiwei ‘think’, making the
issue of verb selectional requirements, especially for wh-in-situ languages. Grimshaw (1979)
proposed that predicates must bear features which select for the ‘semantic type’ of their
not need to correspond to syntactic categories; for example, a CP can be semantically typed
wonder or ask would select for a Q-complement, verbs such as believe select for a declarative
complement, while verbs such as know can take both. Therefore, it is necessary for the wh-
86
expressions in wh-in-situ languages to covertly move at LF to semantically type the clause
while the interpretation of adjunct wh-in-situ does not. This asymmetry is explained if we
assume a LF-raising analysis – the ECP allows arguments to be raised out of islands but not
adjuncts (183):
For a sentence such as (183), the scope of shei ‘who’ is in the matrix clause while the
scope of weishenme ‘why’ is in the embedded clause, and this is accounted for by the LF-
However, there are also several empirical discrepancies which lead us to doubt that
the covert wh-movement analysis is completely foolproof. The most damaging argument
Chinese, Japanese and even English. Huang (1982) observed that covert wh-movement of a
11
This is not as clear cut as it used to be, given that there have been other proposals on what can be used to
satisfy selectional requirements. See for example Cheng (1997) for her famous Clausal Typing Hypothesis: that
question particles fulfil the typing function.
87
(184) Ni xiang-zhidao shei mai-le shenme?
Specifically, (184) shows that even though the wh-expression shei (who) ostensibly
moves across a wh-Island, the question is nonetheless acceptable. Data from English
For Huang, covert movement does not need to obey the island constraints that overt
movement obeys, i.e. the bounding theory applies only to overt movement and not covert
in Japanese, as pointed out by Watanabe (1992). Firstly, we see both the presence of
know-want Q
88
Both (186) and (187) contain the in-situ wh-expression nani-o; however, (186) is
acceptable while (187) is not. Japanese is a wh-in-situ language and it is surprising that the
latter is considered ungrammatical because it has violated the wh-Island Constraint. This
suggests that there might not be a uniform treatment of covert wh-movement in these two
tazuneta no?
asked Q
tazuneta no?
asked Q
dare-ni outside the island, and is judged to be grammatical. Sentence (189) has both its wh-
expressions dare-ga and nani-o in-situ within the island, and is curiously judged to be
structure (overt syntax) to Spec, CP. Subjacency effects are thus expected of this S-structure
movement (in (187) and (189)). Watanabe’s proposal runs contrary to the idea that wh-
89
Additionally, we can see an asymmetry in Binding phenomena with overt/covert wh-
In (190), the overtly moved wh-expression is co-referential with either the matrix
subject John or the embedded subject Bill. However, in (191), the same wh-phrase, which is
now in-situ, can no longer take the matrix subject John as its referent. This suggests that the
wh-phrase does not undergo covert wh-movement out of the embedded clause, such that
Therefore, moving on from Huang, Aoun and Li (1993) propose that in-situ-wh do not
need to covertly raise to Spec, CP at LF, but are instead co-indexed and interpreted with
respect to a null Question Operator that is overtly raised to the appropriate Spec, CP in
syntax. They contend that it is possible for in-situ wh-elements to be accounted for entirely in
overt Syntax, rather than the covert LF component, both in Chinese as well as in English (c.f.
proposals that postulate a covert wh-raising process at LF, such as Huang (1982)). Aoun and
Li propose that there is a Question projection within the clause, and its Spec position is filled
by the Q-operator. The head of this projection can have any of the four combinations of the
features [+/-Q] and [+/-wh], in which the feature pair [+Qu, +wh] will result in a [+wh]
question, a Q-operator is generated in the Spec position, and subsequently moves to the
appropriate Spec of Comp (the remaining feature pairs are [+Qu, -wh], which results in a
yes/no question; [-Qu, -wh] results in statements; and [-Qu, +wh] results in exclamatory
90
interrogative element). An example of a non-interrogative indefinite wh-element in Chinese
In (192), the wh-expression shenme ren ‘what person’ does not actually probe for the
identity of a particular entity, but has the meaning of the indefinite ‘anyone’. For Aoun and
What this also means is that a wh-element in Chinese functions as a polarity item (and not as
Subsequently, in a series of papers, Tsai (1994, 1999, 2003) proposed that the best
way to assign wide scope to Chinese in-situ wh-expressions is through Unselective Binding
of an in-situ wh-variable by an operator directly merged into CP/IP. He develops the general
determined by their lexical composition and where they are merged in the structure. Firstly,
Tsai (1994) makes the distinction between wh-arguments and wh-adverbials in Chinese: the
former are always in-situ and are not sensitive to island effects, therefore are not assumed to
have undergone movement; while the latter also appear in-situ, but their inability to appear in
islands is taken as evidence of covert movement having taken place. He then argues that
91
(193) Ruguo A mai-le shenme, ta yiding hui lai gaosu wo.
come tell me
“If A couldn’t submit his homework for some reason, he will surely tell me.”
‘something’; but in (194), weishenme ‘why’ cannot be bound in the same way and mean ‘for
some reason’. The same is true for dou ‘also’ constructions in Chinese. We have already
noted that something like weishenme ‘why’ can be bound by a non-interrogative operator.
This suggests that the difference between wh-nominals and wh-adverbs is a semantic one.
While both are variables, each is restricted to being bound by a different type of operator:
those that bind entities, or those that bind propositions. It is further argued that operator
features were not universally strong (such that they do not trigger any movement); and that
only nominals, not adverbs/adjuncts, can serve as sole providers of non-pronominal variables.
Tsai (1999) stated that only (wh-) nominals “may introduce pure (i.e. [- pronominal])
variables in-situ” and enter into unselective binding; and (wh-) adverbs, in contrast, are
“intrinsic operators and must undergo movement to create variables (i.e. traces left behind by
movement)”. This asymmetry was explained by Tsai’s Lexical Courtesy Hypothesis, which
makes use of the general Economy consideration that Merging a Q-operator into an empty
position and then Unselectively Binding an in-situ wh-expression is less costly than getting
the in-situ wh-expression to move into an empty position and creating an operator-variable
92
(195) Lexical Courtesy Hypothesis
Formation.
to the position (height) in the structure they are merged. In English, the Q-operator is merged
at D0, or at the word level. An operator-variable pair is formed prior to the wh-movement
required to check its feature on Spec, CP. In Chinese, the existence of bare conditionals gives
level. Since they are inserted in Spec, CP, no further movement is involved. In Japanese, the
fact that wh-in-situ can be licensed within a Complex NP but not in a wh-island, and that
there are no bare conditionals in the language, suggest to us that the Q-operator is merged
into the Spec of either PP or DP, or at the phrasal level. Schematically, the parameterization
standard analysis. The fact that Chinese wh-expressions are taken to be pure variables also
directly explains their well-known status as indefinites and polarity items in the language:
93
(197) Botong shenme dou chi
‘Isn’t it the case that someone came here, so the door is open?’
94
Therefore, the proposal that languages are: firstly, parameterized in terms of their
lexical makeup – whether the wh-operator and the variable are generated together, or
separately – and secondly, parametrized in terms of the syntactic position where the operator
is Merged; provides an elegant solution for whether the language allows wh-in-situ and
To conclude this section, just like in Malay, we can see a lot of congruence between
the Chinese and the CSE data. However, again like in the previous section, the same
construction does not seem to receive an explanation in this analysis – examples (40b), (41b)
and (42b) showing that a wh-argument is unable to remain in-situ in an embedded clause
why this is so, since there is nothing in the analysis that prohibits an in-situ wh-argument in
that context.
driving the syntax behind wh-constructions. At the heart of this discussion is a simple
b. Where John?
Are we then able to explain the difference between (203a) and (203b) as the former
placing more emphasis on the identity of the person whose location the speaker wants to
ascertain, and the latter placing more emphasis on what exactly about John the speaker wants
95
to know – his location as opposed to his identity, for instance? Is this difference in meaning
also predictable?
There are already some accounts of CSE wh-question formation that are grounded in
Ho’s (2000) main argument is that “…discoursal factors can influence the choice of one wh-
question construction over another…”. Let us examine her assumptions. First, she assumes
that in a question like (203b), the wh-expression is “…in the TOP(ic) position”, and is “…the
focused constituent” (Ho, 2000: 21). I am not sure if what she means by ‘TOP position’ is
similar to a high functional projection such as TopP in the left periphery (à la Cinque, 1999);
and when she says ‘focused constituent’, whether she means that in a general non-technical
‘unmarked’ and ‘canonical’ in CSE, but offers a ‘discoursally motivated’ explanation of the
wh-expression being in the clause-initial position anyway, with considerations such as:
introduction of a new discourse topic, ensuring cohesive discourse flow, and politeness.
Although I’m not sure what the unmarked wh-construction in CSE should be (as well as the
basis for that decision), it does strike me as odd that an unmarked construction still requires
that “…it is perceived that focus is no longer placed on the wh-element. Rather, focus is
assigned to some other element in the discourse for differing motivating factors” (Ho, 2000:
30). The factors that she offers in this case are: specifying the frame of reference, shifts in
discourse structure, contrastive effect, and maintaining the temporal order of information.
How I understand Ho’s explanation is that a wh-in-situ in CSE is a marked construction, and
it is marked because something else in the sentence is more important and takes precedence
96
in the clause-initial position (for the reasons she specified above). However, the explanation
does seem like a ad-hoc analysis of wh-constructions to me; and on top of that, we can
plausibly fulfil the suggested factors (for instance, contrastive focus) without having to force
wh-in-situ. In a study that primarily seeks to answer the question of how the in-situ vs.
movement choice is made, Ho’s claim that “…it may suffice to conclude that a particular
constituent is placed in the initial position simply because it is the focus of attention of the
speaker or conversation, and is the single factor that influences wh-question formation.”
(Ho, 2000: 49, emphases mine) probably comes across as a little strong. Discourse salience is
definitely a crucial factor in ultimately deciding what form an interrogative takes, but it does
Another largely similar analysis is given by Tan (2011), who argues that wh-fronting
and wh-in-situ can be accounted for by a combination of discourse factors and contact
language influence. Simply put, wh-fronting is due to influence from the superstrate; or
“…motivated by the need to place the focus on the entity of concern.” While the former
reason is already considered to be received wisdom, it is the nature of this influence that is
interesting to us, though this was never explicated in any detail. The latter reason is a
reiteration of Ho’s argument above. Tan then provides his arguments for what causes wh-in-
situ: Firstly, wh-in-situ occurs most frequently after a declarative sentence. The wh-question
then “…tends to mimic the preceding structure of the declarative form, resulting in wh-in-
situ.” This mimicking behavior is claimed to be the result of substratal influence from
Mandarin and Hokkien – although it is also unclear how this comes to be. The second reason
offered by Tan is that wh-in-situ occurs “…when the focus of the question is on the non-wh
words, thus fronting the entity of concern and leaving the wh-word in-situ.”, which is again
very similar to what Ho has argued above. Pragmatic accounts of wh-phenomena tend to be
quite post-hoc and unfalsifiable, in the sense that any explanation offered for a particular
97
utterance can never be wrong. For instance, any constituent that is fronted in a sentence is
always the Topic or Focus of the sentence by virtue of appearing in that position. Therefore,
when a wh-expression is moved, one simply has to say it was discoursally the most salient;
and when a wh-expression remains in-situ, it was because something else was more salient
than the wh-expression. It is for this reason that I think that a purely pragmatic approach to
These two CSE studies bear some similarity to studies which invoke pragmatic
Brazil. It is often characterized as an ‘optional wh-in-situ’ language, as both fronting and in-
98
However, wh-movement in matrix clauses is optional only with a phonetically null
‘Which is the book such that you talked with the author that wrote it?’
to islands (210-213). Hornstein et al. (2005) says that the BP facts cannot be explained, but
can be described using a system of strong and weak features, where all obligatory movement
is linked to a strong feature, and all cases of ‘optionality’ are illusory as each option is simply
99
associated with a wh-feature of varying strength (emphasis from source). This is, admittedly,
a very weak claim. The featural composition of the complementiser analysis is shared by
Pires & Taylor (2007). However, Pires & Taylor’s main argument is that single wh-questions
in BP can have the wh-expression in-situ, subject to semantic and pragmatic requirements
It is common ground that φ in a group if all members accept (for the purpose
of the conversation) that φ, and all believe that all accept that φ, and all
These constructions, according to Pires and Taylor, are distinct from echo-questions,
as they: are interrogative in nature; do NOT require an immediately prior antecedent; and do
not receive rising intonation and focal stress. These contexts are:
A: I made desserts.
expected)
100
(217) Reference-Question (asks for a paraphrase or repetition of an immediately
prior antecedent)
Pires and Taylor claims that the 4 contexts listed above (215-218) are licensed as the
set of possible answers to them is part of the Common Ground, which is defined as
“information that was previously given in discourse or in the extralinguistic context, and
which is shared (or assumed by the speaker to be shared) by speaker and hearer”. One
important prediction of this analysis is that wh-in-situ, in these contexts, is not freely optional
in Portuguese, contrary to what has been standardly assumed. This prediction is corroborated
questions.
These four contexts suggested by Pires and Taylor are interesting to us as we can
101
(219) [+Specific] Question (requests more specific information about something
A: I made desserts.
expected)
prior antecedent)
These exchanges (219) – (222) seem perfectly natural in CSE. However, there are two
problems here. Firstly, as far as restrictions and prohibitions go, these contexts do not
actually prevent someone, for instance, from asking ‘What you reading?’ instead in (222). So,
we can only accept these as general tendencies. Secondly, if it is the case that we can only
admit these contexts as tendencies, then we would expect the above list to be non-exhaustive.
That is to say, any number of plausible contexts can be constructed to ‘predict’ wh-in-situ.
And if that is so, then I’m afraid the analysis becomes more descriptive than explanatory.
She proposes that in BP, there is the long wh-movement option to Spec of CP, and a short
wh-movement option to Spec of F(ocus)P – Kato assumes that wh-expressions can move to
102
the sentence-internal F (in Belletti’s (2001) perspective) to have their features checked – and
these two options are decided by the type of head chosen in the numeration. Basically, a true
wh-in-situ question (not an echo question) is only apparently so as the wh-expression has
vacuously moved to Spec, FP position. The remnant IP then moves to Spec, CP. Kato offers
two pieces of evidence for the proposed new position of the in-situ wh-expression. Firstly, it
Telephoned who
‘Who called?’
Secondly, focalized and wh-constituents can vary in order clause internally, since in
Belletti’s architecture of the left periphery, there are Top(ic)P projections both below and
above FP:
This analysis explains why ‘true’ wh-in-situ questions are interpreted as questions and
display the same scope properties as ‘moved’ wh-questions. There is, however, a problem for
CSE in the first piece of evidence for Kato’s remnant movement analysis. A question such as
‘Who called?’ (from (223)) would never be in the same word order in CSE as it is in BP.
Lastly, Figueiredo Silva & Grolla (2010) put forth an interesting and novel proposal
to account for the above data, where wh-in-situ is thought to be obligatory. They argue that
103
the absence of higher functional projections to accommodate the movement of wh-expression
(indicated in angled brackets in (226)). Although constructions analogous to (226) like ‘do
what?’ i.e. with an elided subject and modal verb, are quite regularly found in CSE, it seems
rather drastic to assume that wh-in-situ is obligatory and that CP and IP are regularly
truncated.
The final two accounts of BP fall under the category of masked movement. As we
have seen earlier, linguists have been using covert wh-movement to account for apparent
locality and scope effects with wh-in-situ. It was in the 90s that some of them started to re-
analyse covert movement as overt movement. A particularly strong argument was made by
Kayne, who argued that “…it is possible and advantageous to dispense with covert movement
and replace it with a combination of overt movements” (Kayne, 1998: 183). The rationale
behind this proposal is that Scope reflects the interaction of Merger and overt movement and
it must be expressed hierarchically; also, Universal Grammar does not permit covert phrasal
movements, and the effect of covert phrasal movement cannot be achieved by feature raising.
Many have since tried to apply his ideas to wh-questions, particularly wh-in-situ. The general
idea is that an apparent in-situ wh-expression has actually undergone overt movement to the
left periphery. This is then followed by a series of overt movement operations of the
remainder of the clause to an appropriate position such that it masks the previous movement
of the wh-expression. For this reason, this group of analyses has been variously termed
104
‘masked movement’, ‘remnant movement’, ‘disguised movement’ and even ‘super
movement’. For ease of reference, I will simply use the first description.
Bellunese, bare wh-expressions in matrix clauses display obligatory Subject Clitic Inversion
(SCLI), and appear in-situ at the end of questions. It is well-known that French speakers
typically do not invert subject clitics in normal speech, with the exception of que, which
requires SCLI. However, in apparent in-situ structures, SCLI is banned and que is not
available in French. On that basis of comparison, the authors suggest that the wh-expression
does move to the left periphery in Bellunese, just as they do in French. In turn, the remnant IP
also moves to the left periphery. In addition, Munaro et al claim that nonassertive clitics,
which appear in Bellunese wh-questions and yes-no questions (as well as optative,
counterfactual and disjunctive constructions) have the function of expressing the ‘force’ or
‘type’ of the proposition that contains them; and that their absence in French has triggered
highly articulated left periphery, which is based on discourse configurational principles, that
can account for the differences between the ultimate positions of wh-expressions in wh-in-situ
and overt wh-movement (Here, I ignore other projections such as OperatorP, TopicP, etc. that
also appears in the left periphery for Munaro et al., as they are not relevant to the discussion):
105
(227) CP
(Interrogative)ForceP
GroundP
NewInformationP (~ FocusP)
IP
Since wh-expressions have to be in focus, they are moved to the New Information
projection (NIP). In French wh-in-situ clauses, there is subsequent movement of the remnant
IP to GroundP, which hosts presupposed elements such as IP. However, ForceP is claimed to
be missing, which is why the wh-expression does not have the option of moving even higher
up, thus yielding what looks like a wh-in-situ structure. In French overt wh-movement
clauses, ForceP is present, and the wh-expression further moves to ForceP, producing the
correct word order. The problem with this proposal is that this truncated structure for wh-in-
situ that Munaro et al. have come up with is not sufficiently motivated; and also, it would be
odd to think of other languages with wh-in-situ to entirely consist of such truncated
structures. The underlying idea for a masked movement proposal such as Munaro et al. is that
projections which serve as the landing sites of moved wh-expressions and remnant IPs.
However, the nature and type (and existence) of these projections seems to differ greatly
from language to language, which ultimately leads to variations in word order. As such,
proposals that depend on micro-parametric variations in projections in the left periphery have
106
an arbitrary flavor to them12. Lastly, masked movement proposals also do not work well in
optionality of overt wh-movement vs. wh-in-situ in Spanish. She is also in support of Kayne’s
(1998) position that there are no covert movement operations. For Uribe-Etxebarria, Spanish
wh-in-situ involves firstly the overt A-bar movement of the wh-expression to Spec, CP and
Step 1 qué starts in its neutral position: [IP tú le diste qué a María]
Step 2 qué moves to Spec, CP: [CP quéi [IP tú le diste ti a María]]
Step 3 IP remnant moves to XP: [XP [IP tú le diste ti a María]j [CP quéi tj]]
She uses the following facts about Spanish wh-in-situ questions to support her claim:
Firstly, they do not have a natural, neutral word order. Also, there is a requirement that the
wh-expression must be sentence-final. Secondly, only right dislocated elements can follow
the in-situ wh-expression in this sentence-final position. Lastly, wh-in-situ questions allow
pro-subjects, but crucially require a Clitic Left Dislocated constituent in preverbal position.
These properties are all different from those exhibited by regular wh-questions (formed by
overt fronting); and furthermore, they are unexpected under an analysis that assumes that the
wh-expression stays in its base-generated position and does not undergo overt movement
12
Although Cinque (1999) defends the Universal Base Hypothesis, which is that languages have a largely
invariant underlying clausal structure and that there is a fixed universal ordering of functional projections.
107
(Jiménez, 1997). Like Munaro et al’s proposal, Uribe-Etxebarria’s analysis of wh-in-situ is
does not resort to using a highly articulated left periphery. Uribe-Etxebarria claims that her
remnant movement analysis explains why Spanish wh-in-situ is not limited to only wh-
arguments, but also includes adjuncts; why wh-in-situ questions do not display intervention
effects, such as from Negation and modals; and lastly why in-situ wh-expressions that are
embedded within interrogative islands are able to have a matrix scope interpretation. Her
proposal, however, have come under criticism in Reglero (2005), and Reglero & Ticio
(2008), for running into problems such as: Uribe-Etxebarria’s prediction that wh-in-situ
within an island should not be acceptable is not borne out; and the analysis has difficulties
would not work for CSE, since all the steps listed in the derivation in (228) are perfectly fine
wh-constructions:
Additionally, her claim that this predicts in-situ wh-adjuncts is not borne out in CSE.
First, we will need to assume the Copy theory of movement (Chomsky, 1995), which states
that a moved element leaves behind a copy of itself, rather than a trace. Movement in overt
syntax therefore creates a chain with at least two copies (the original and the moved), these
copies are then interpreted at both the LF and PF interfaces. The two interfaces do not have to
act in sync; that is, they can both interpret the highest copy – this creates the typical wh-
108
movement scenario. If PF chooses to pronounce the lower copy, while LF interprets the
higher copy, a wh-in-situ scenario is created (Bobaljik, 2002). The main idea here is that the
grammar can choose to pronounce either the highest copy of the moved element, or the
lowest copy (the original), or indeed any of the copies in between, depending on the kinds of
copies will then be deleted at PF. Usually, the head of a nontrivial chain ends up being the
sole survivor of this deletion process; but in certain circumstances, a lower copy of the chain
is pronounced instead. This is called the ‘Pronounce Lower Copy’ (PLC) analysis, and
Bošković (2002) shows this with Romanian, which is a multiple-wh-fronting language where
We see in sentences (230) and (231) two different wh-expressions in a question: the
sentence is grammatical only when both of them are fronted (230), but it is ungrammatical
when the second wh-expression is left in-situ, or not fronted (231). However, when the
second wh-expression is identical to, or homophonous with the first fronted wh-expression, it
109
(233) Ce precede ce?
homophonous wh-expressions that rules out a sentence like (232), which would normally be
expected to be fine. Instead, the only acceptable way to ask the question “What precedes
forms, the higher copy of the second ce-chain is deleted, while the lower copy survives, and
is pronounced. Villa-García (2014) also adopts the PLC to explain the observation that the
subject may not intervene between a non-D-linked wh-expression and the verb in Spanish
When the subject comes between the wh-expression and the verb (236), the sentence
is ungrammatical; as opposed to (235), where the subject is found at the end of the sentence.
The PF requirement apparently in play here is that a New Information focused element –
Silvia – bears sentential stress in sentence final position, while higher copies of the element
110
are deleted. It has also been argued that morphological restrictions on identical elements, the
second position requirement providing support for a prosodically weak element, intonational
requirements, and even clitic weakening and stress assignment processes can all trigger lower
copy pronunciation.
The PLC and Copy theory of movement is also used in Reglero’s (2005) and Reglero
& Ticio’s (2008) analyses of Spanish wh-in-situ, in conjunction with the stress assignment
Basically, a version of the Nuclear Stress Rule is used in Spanish to assign stress to a
constituent and identify it as New Information Focus. Stress assignment interacts with copy
deletion, to decide which copy to pronounce. The highest copy will always be pronounced
unless it leads to a PF violation (such as a case where main stress would not be assigned); in
which case, the lower copy is pronounced instead to ensure that main stress can be assigned.
Additionally, in Spanish, all material is assumed to be metrically visible, and the domain of
stress assignment is the intonational phrase. Other PLC accounts include Reintges’ (2007)
morphological agreement proposal for Coptic Egyptian, and Manetta’s (2013) sluicing
The advantage of the analyses described here is that it would be motivated by several
with the Inclusiveness Condition (Chomsky, 2001), a grammatical principle which bars the
introduction of new elements (features) – in this case, traces – in the course of a derivation.
This confines the power of syntax to the (re-)arrangements of lexical items. The Copy theory
also greatly simplifies Reconstruction analyses, in the sense that they can now be treated as
with copies also reduces the number of theoretical primitives in the system and reduces the
complexity of the grammar. However, a big downside of the PLC analysis is that it might not
111
readily work in a derivational Minimalist framework, for the reason that it is usually syntactic
operations feeding phonology (Spell-Out); and not PF constraints overriding syntax. The
constraints that can trigger the PLC are too diverse in nature; although most of them are
It is questionable whether a PLC analysis would work in CSE. Firstly, I’m not sure if
The wh-expression what and the discourse particle what, for instance, are homophones but
rhythm, would have any stress assignment rules such as in Spanish. It also remains to be seen
this would mean that whatever constraint(s) in operation within the language in order for a
In the next section, I will look at the analyses offered for certain languages that have
In this section, we focus specifically on languages which display wh-patterns that are
similar to that in CSE. Although they perceivably have no discernable or direct links to CSE,
we might be able to apply some of the proposals and analyses offered to account for these
112
languages, to CSE. In doing so, we can make certain generalisations of CSE that has not been
hitherto discovered; and to bring CSE under the auspices of an existing framework would be
Columbia. The language has a basic SOV and fairly rigid word order, but wh-expressions
have the ability to either stay in-situ or be fronted at the beginning of the clause, with no
optional wh-movement language. The BW data presented below were all extracted from
Denham (2000). In simple matrix questions, the object wh-expression may remain in-situ
113
In an embedded question, it is possible for the wh-expression to be fronted at 3
positions – the in-situ position (242), the beginning of the embedded clause (243), and the
beginning of the matrix clause (244). This is true of both direct and indirect questions.
Denham (2000) showed quite convincingly that the sentences in (238-244) involve
wh-movement – through Island sensitivity diagnostics, for example – and are not merely
Focus or Clefting constructions. Her explanation for optional wh-movement in BW is that the
selection of C from the numeration itself is optional; and since C is the functional category
directly responsible for wh-movement, when it is not selected from the numeration, no
not affect interpretation. Therefore, the difference between an interrogative with a moved wh-
expression and its wh-in-situ counterpart in BW that has exactly the same meaning is simply
that the former derivation selected C in its lexical array, while the latter derivation did not. If
C is selected and then projected, the wh-expression is motivated to move to that position. In
which houses features relevant for clausal typing and scope marking. Although this is a
straightforward explanation of the facts, we are left wondering what the reasons may be for C
not to be selected and projected in the first place. Additionally, is a Typ(ing) projection –
114
which performs exactly the same role as C would if it has been present – independently
motivated otherwise?
Another language that has been widely discussed in the literature of wh-constructions
due to its intricate and interesting wh-patterns is French. Some scholars believe that French
has a ‘mixed’ system in forming wh-interrogatives – allowing both wh-movement and wh-in-
situ options. In matrix clauses, it is indeed possible for the wh-expression to be fronted13
13
There is some opposition to the idea that such optionality exists in French matrix clauses in the first place.
Denham (2000) claims that when a wh-expression is fronted, ‘est-ce-que’ is required:
Since the literal translations of both sentences are different, and different forms of the wh-expression are used in
both sentences, Denham argues that the fronted version is not the result of optional wh-movement from the in-
situ version. Rather, it is some kind of wh-clefting that presumably satisfies a strong Focus feature.
115
However, only the movement option (247) is possible when there is an overt C in the
sentence, while the in-situ option (248) is ruled out. Similarly, in embedded clauses, it is not
possible to leave the wh-expression in-situ (250), therefore it must be moved to the front of
Bošković’s (2000) explanation14 for the above data uses Chomsky’s (1995) idea of
Merge: proposing that phonologically null elements could in principle enter the structure at
LF. In (245), null C with a strong wh feature enters the structure at LF, triggering LF wh-
movement that checks the strong feature but not overt wh-movement, thus explaining the wh-
in-situ. In (246), null C is presumably inserted at the same time as the wh-expression, thus
triggering overt wh-movement. The same is true of (247), but now with an overt C.
Therefore, we can see that whenever C is overtly inserted, overt wh-movement has to occur
(248). To account for the ungrammaticality of (250), we need to appeal to the Extension
Condition, which states that Merge can only take place at the root. Therefore, a derivation
such as (250) that involves a Merge operation that does not expand the tree, i.e. Merge
applies in an embedded position, has to be ruled out. (249), in contrast, is grammatical since
the null C is presumably Merged in overt syntax before the higher clause is built, therefore
inducing overt wh-movement. In Bošković’s account, then, it follows that the LF movement
14
Bošković’s analysis is also a workaround to the proposal that if a language has no CP layer at all, there would
not be a viable landing site for wh-expressions to move to, considered by Fukui (1986) and Kuroda (1988) for
Japanese. The ramifications of not having the CP layer, however, extend beyond the realm of wh-constructions,
since there are many other operations that target or involve the CP domain. That is to say, there is a huge theory-
internal downside in abolishing the CP projection altogether.
116
of wh-expressions (which is essentially head movement, or adjunction to X0 positions) to
check off the strong +wh feature would be subject to locality effects and conditions. When a
wh-expression moves to C, which is an A’-head position, all other A’-heads are therefore
potential blockers or interveners. This also explains why the following patterns emerge:
We can see from the above data that C (251) and Negation (252), which are
intervening A’-heads, makes LF wh-movement impossible. Therefore, the only way to check
addition to A’-heads being interveners for wh-movement, certain XPs and also elements in
A’-specifier positions such as focus markers, universal quantifiers and some adverbs are
found to be able to block wh-in-situ as well. Secondly, the presence of a complementiser (C)
does not always result in overt wh-movement. Mathieu’s last objection is a theoretical one:
Bošković’s notion of acyclic Merger (the insertion of a null C at LF) goes against the
French wh-constructions involve the overt movement of the question (wh-)operator to Spec,
117
CP. There, the operator indicates the scope of the wh-expression and binds it, as well as to
check the strong feature of C. The movement of the wh-operator leaves a non-referential
Let us examine the above arguments made by Mathieu against Bošković, in the
context of CSE. Firstly, CSE generally does not seem to exhibit Intervention effects in wh-
questions ((74) – (76)). Also, there are no overt complementisers in CSE matrix wh-clauses
like there are in French. However, the argument against acyclic Merge is relevant to CSE,
which exhibits the full range of options – in-situ, partial movement and full movement (46a-
Let us end this section by giving a quick summary of alternative accounts of French
wh-constructions. Zubizarreta (2003) argues that the French wh-in-situ construction is a case
Evidence for this comes from the fact that contrastive focus, but not informational focus, is
associated with the property of exhaustivity. Therefore, it is predicted that a French wh-in-
situ question would not be compatible with a marker of non-exhaustivity such as par exemple
(253) a. Avec qui par exemple (est-ce que) la jeune artiste a dansé?
Similarly, Chang (1997) argues that wh-in-situ questions in French are felicitous only
with a strongly presuppositional context (see also Boeckx et al., 2000), as they “seek details
118
on an already established (or presupposed) situation. Therefore, it is infelicitous to answer a
appears that CSE wh-in-situ is not licensed by exhaustivity nor presuppositional contexts:
(254) Scenario: John ate a burger, fries and a salad for dinner. A is asking B what
John ate.
The exchanges above show us that both wh-movement (254a) and wh-in-situ (254b)
forms of the question are acceptable to ask when speaker A interprets B’s answers
when A interprets B’s answers non-exhaustively (257), and even if the presupposition is
cancelled (258).
intensionality and selectional requirements. Boeckx et al. (2000) argue that for a dialect of
intensionality of the matrix verb – embedded wh-in-situ is possible with intensional verbs
such as ‘think’ and ‘believe’, while it is not possible with non-intensional verbs such as ‘say’
119
and ‘regret’. Denham (2000) shows that wh-fronting takes place with matrix verbs that select
for embedded interrogatives, such as demander ‘ask’; while wh-in-situ takes place with
matrix verbs that cannot select for an embedded interrogative, such as penser ‘think’. The
latter suggestion is especially interesting for CSE because it exactly predicts the facts in (40)
– (42).
Cheng & Rooryck (2000), while in agreement with Bošković that French wh-in-situ
involves LF wh-movement, propose another analysis for the apparent wh-optionality. French,
which does not have a Q-particle, relies instead on an intonation morpheme to license both
wh-in-situ and yes-no questions. Cheng and Rooryck observes that French yes-no questions
and wh-in-situ questions share the same rising intonation; while questions with wh-movement
does not have this rising intonation (the presence of this rising intonation also correlates with
whether the question has a strong presuppositional or a neutral context). This rising
intonation is the PF Spell Out of an abstract intonation morpheme that is merged into C 0 in
the derivation of a question. This intonation morpheme is underspecified and thus needs to be
valued by the Q-feature of an in-situ wh-expression before it can license wh-in-situ; if not, it
licenses only yes-no questions by default. They further claim that it is optional for the
analysis might not work perfectly for CSE since wh-movement and wh-in-situ do not seem to
Lastly, we have Poletto & Pollock (2005)’s proposal that French-wh is simply a clitic,
based on evidence such as: similarities with pronominal clitics, wh-doubling phenomena and
obligatory Subject Clitic Inversion (SCLI). They propose that the wh-clitic targets the IP
clitic field, which then undergoes SCLI, which is basically Remnant IP movement to the left
periphery. This analysis draws parallels between the impossibility of SCLI in embedded
clauses, and the ungrammaticality of embedded que, in French. This also would not work
120
because there is just no evidence that CSE wh-expressions behave like clitics, nor are there
In summary, there are some useful aspects of analyses for the French data that we
may be able to use for CSE, in particular the general idea (à la Bošković) of inserting
something into the derivation that triggers obligatory wh-movement, as well as Denham’s
Chinese, and Standard English, the literature on CSE wh-questions is rather limited, given the
relative youth of the variety and general lack of attention it receives from linguists who are
not themselves native CSE speakers. The wh-patterns found in CSE, however, are as varied
as they are fascinating; and the variety has proved itself to be an excellent testing ground for
CSE wh-constructions have moved from the earlier, descriptive/comparative and pragmatic
accounts we have already seen, to the more recent syntactic analyses by Kim et al (2008,
2009), Yeo (2010), and Sato (2013). Early discussions of wh-question formation in CSE were
in the context of language contact, given the status of CSE as an emerging variety of English,
and involved comparisons with the languages in its ecology. The influential role of Chinese
languages (such as Mandarin, Hokkien and Cantonese) on CSE syntax, for instance, has been
extensively covered in the literature. Wh-question formation has also been discussed through
121
the lens of language acquisition (Kwan-Terry 1986, 1991; Gupta 1990) – English-learning
There are other, more recent, studies in CSE wh-constructions that adopt structural
explanations. Kim et al. (2008) explain the possible wh-strategies in CSE as a combination of
both English and Chinese wh-strategies, which is quite a common position to take. Their
analysis is based on two ideas: firstly, languages can choose to either generate an operator
within the wh-expression, or at the sentence-initial position; secondly, languages can also
choose whether operator movement can pied-pipe the wh-expression along with it, or not.
The analysis attempts to cover a range of CSE data, from the full paradigm of wh-movement,
partial movement and wh-in-situ ((259a-c), shown earlier in (46a-c)), Island effects (260) and
Island Effects
(260) a. *Whati John like [NP the man [CP that think [CP [IP Mary eat ti]]]?
b. John like [NP the man [CP that think [CP [IP Mary eat what]]]]?
c. *John like the girl because [IP Tom think [CP whati [IP she eat ti]]?
Negation as Intervener
c. *John does not think [CP whati [IP Mary like ti]]?
122
For full wh-movement (259a), an operator is generated within the wh-expression, and
when the operator moves to sentence-initial position, the wh-expression is pied-piped along
with it:
(262) (OP + What)i John think [CP ti [IP Mary like ti]]?
This structure explains why there are island violations in CSE, such as in (260a) and
why there are no intervention effects, such as in (261a) – the reason being that this pied-piped
wh-movement is a PF phenomenon.
For wh-in-situ (259b), a null Q-marker licenses the operator in the sentence-initial
This structure predicts that there are no island violations with wh-in-situ, such as in
(260b), and also no intervention effects, such as in (261b) – since there is no movement in the
first place.
application of pied-piping. Kim et al argues that since pied-piping is allowed in StdE, but not
allowed in Chinese, the end result is that pied-piping becomes optional in CSE. Thus, what
happens is that first the operator is generated with the wh-expression; then, this moves to the
intermediate Spec, CP position while pied-piping the wh-expression along with it; finally, the
(264) [OPi [John think [CP (ti + what)j [IP Mary like tj]]]?
Since pied-piping is now optional, the operator can choose to pied-pipe the wh-
expression all the way to the front of the sentence, thus deriving (259a)/(262) once again.
123
Kim et al. eschew a Government-Binding motivated analysis in favour of a Minimalism
motivated one. Their account covers broadly the wh-in-situ, wh-movement and partial-
movement options present in CSE. What this study has effectively done is to shift the burden
of wh-formation to the domain of the lexicon – that is, optionality still exists at the level
where the wh-operator is licensed, insofar that it has not been explained or motivated.
However, one major drawback of the analysis is that it is still unclear what triggers the
optional pied-piping operation in cases of partial movement, since Kim et al. simply explain
Another account is Kim et al. (2009), which presents the findings that CSE wh-
adjuncts do not have the option of staying in-situ, as opposed to the optional wh-pattern
with what we have seen earlier (wh-arguments in matrix clauses (28) – (30) vs. wh-adjuncts
in matrix clauses (31) – (33); wh-arguments in embedded clauses (40) – (42) vs. wh-adjuncts
in embedded clauses (43) – (45)). The CSE wh-paradigm was analysed as an adoption of a
combination of StdE and Chinese wh-strategies, but with one crucial gap in the paradigm.
While CSE wh-arguments could move, just like in StdE, and also remain in-situ, just like in
Chinese; wh-adjuncts need to move, like in StdE, but cannot remain in-situ, although their
Chinese counterparts can do so. They note that a combination of island and intervention
Intended Meaning: For which x, Mother beat the boy who ate x?
124
CNPC and Chinese wh-adjunct in-situ
(266) *Mama da-le [DP [IP weishenme chi pingguo de] nanhai]
Intended Meaning: For which reason x, Mother beat the boy that ate apples x?
intervention effects (267), but Chinese wh-adjuncts show both subjacency violations (266)
and intervention effects (268), Kim et al. conclude that in Chinese, wh-adjuncts only move
covertly. It was then suggested that while wh-adjuncts move covertly in Chinese, they prefer
to move overtly in English. Kim et al. explains the ungrammaticality of in-situ wh-adjuncts
with a general principle that the possibility of overt wh-movement blocks covert wh-
movement from taking place. They termed this the Overt-over-Covert Movement Principle:
125
According to Kim et al., the Overt-over-Covert Movement Principle can be subsumed
The Earliness Principle effectively says that all else being equal, a derivation that can
be derived the earliest in terms of the hierarchy of levels is always preferred over another that
can be derived at a later level, and also implies that overt (SS-level) operations are preferred
over covert (LF) operations. Kim et al. argues that the Overt-over-Covert Movement
Principle is thus a substantiation of the Earliness Principle, and that it accounts for CSE wh-
adjuncts preferring to overtly move rather than take a covert movement approach. The basis
of their analysis is that “…a language created in a contact situation is shaped not only by the
influences of its parent languages but also by universal principles”. The universal principle in
However, this principle can either be criticised as being too strong a statement, or
being simply inaccurate. It predicts that in a language where both overt and covert movement
is possible, the former would always block the latter; and this is shown not to be true in a
language such as German, or in just about any language which has both overt and covert
movement.
One final piece of work we should consider is Yeo’s (2010) innovative syntactic
analysis of CSE wh-constructions. Yeo’s work is probably the longest and most detailed
technically rigorous piece of work on CSE-wh thus far, and deserves a thorough discussion.
functions). Qu then takes wh-expressions as its complement to form QuP. A question particle
such as ah serves as the head of QuP. Secondly, he proposes the Featural Subset Hypothesis
126
(FSH), which says that EPP features can enter subset relations with other features (such as
Q), to motivate optional movement. The FSH was proposed by Yeo as a means of separating
movement from agreement. It was a handy tool for the purposes of his analysis, but one
wonders if the mechanism is too powerful. Yeo’s own words – “the FSH allowed us a way of
controlling the strictness of EPP satisfaction, with maximum strictness on one hand,
movement being parasitic on Agree and maximum looseness on the other, where any
category will do” – serves as an implicit admission to the unbridled power of the FSH. Lastly,
Yeo reworked the idea of Q-migration, while borrowing from two other analyses, namely m-
merger (Matushanky, 2006) and reprojective movement (Donati, 2006), to form a new
operation called reprojective m-merger. This new operation was borne of the desire to move
away from head adjunction (for the purposes of Qu movement), and to replace it with a head-
to-specifier movement strategy. Eventually, this process is supposed to let Qu ‘escape’ from
within an island into a scope-taking position. But in order for this to happen, some structural
(271) CP
whati C’
C TP
I buy ti ah
We can run through the general architecture and operations proposed by Yeo very
quickly. I will use for illustration part of the derivation process for the sentence “What you
particle are largely similar. A typical QuP consists of a question particle as the Qu head,
which takes a wh-expression as its complement. In (271), this is ah and what respectively.
What then moves to the specifier of the QuP to obtain the correct word order. Subsequently,
what moves to the Spec, CP position, presumably to check off an EPP feature. At this stage
127
of the derivation, the particle ah is essentially left in an island and is unable to escape from
the CP.
(272) HP
C H’
[EPP, WH] H CP
whati C’
tC TP
I buy ti ah
Yeo’s solution is to basically create a new specifier position for the particle ah to
move to. For that to happen, reprojection of the (C head in the) CP has to take place first. In
(272), reprojection means that the C head is moved out of CP and then takes CP as its
complement by merging with it, projecting anew in the derived position (HP). This was
Adjunction, Extension Condition (Chomsky, 1995) and the Head Movement Constraint
(Travis, 1984), but is ‘repurposed’ by Yeo in this analysis. Crucially, the new H head bears
an EPP feature as well; since the EPP feature on C has already been ‘used up’ by the
movement of the wh-phrase, we will need something else to motivate movement of the Qu
head (the particle ah in this case). M-merger then works to combine the bundle of features of
both the heads. A process of relabelling (or label absorption) then replaces HP with CP.
128
(273) CP
ah
whati C’
C TP
C H I buy ti tah
At this intermediate stage of the derivation (273), the picture is already quite clear to
us. By creating two specifier positions in the CP (or in his own words, “…a specifier was
able to essentially turn into an adjunct”), Yeo kills many birds with one stone – firstly, he is
able to account for CSE partial wh-movement; secondly, he is also able to allow both the
question particle and the wh-phrase to continue moving on up. The reason for this is that both
specifier positions are escape hatch positions from which movement can take place, and both
positions are considered equidistant to any target landing site (for the purposes of Attract
Closest), although Yeo himself concedes that “One may of course argue that this is merely a
technical roundabout…”.
Finally, Yeo proposes a Particle Phrase projection (PrtP) at the top of the derivation
structure. This is a projection to house the particle, to ensure correct word order, and for
scope-taking purposes. He claims at first that PrtP is “projected whenever possible”, going as
far as coming up with the generalisation that “a language that has sentence-final or sentence-
initial particles always projects PrtP if possible”. In the face of derivations where the PrtP is
not projected (such as when QuP moves to Spec, CP and that suffices to assign scope), Yeo
retreats to the position that “…there is no formal requirement for PrtP to project, only a
preference”. The PrtP is a strange creature – it is one of the few (if not only) right headed
129
projections in CSE; also, the dual purpose of housing the question particle/ensuring correct
word order and being the site for scope assignment sometimes pulls it apart in different
directions.
QuP moves to
What ah you think I
embedded Spec, CP, Qu in Matrix QuP No
buy?
then to matrix Spec, CP
From Table D, we can see certain derivations that would put Yeo’s analysis in an
(such as “What ah you think I buy?”), the matrix QuP is said to be the site of scope
assignment, instead of PrtP, as previously proposed. The reason is that the particle now
130
cannot be separated from the wh-phrase, for the correct word order to obtain, and thus cannot
move into a PrtP (which is not projected as a consequence). Therefore, we have an alternation
between QuP and PrtP as the site of scope assignment. Furthermore, in a sentence where the
particle is left in embedded Spec, CP position (such as “What you think ah I buy?”), it is
The above debate on the actual function of the PrtP in the derivation should be
prefaced with a discussion of the role of the question particle in an analysis of wh-
constructions. For Yeo, since the QuP consists of the Qu head – which can either be the
particle, or null – and its wh-phrase complement, a phrase such as what ah is always taken to
be a constituent unless it is assumed they move separately to unique specifier positions. This
creates rather unnatural data such as “What ah you think I buy?”, which I have previously
suggested should be better treated as some kind of Topic or Focus construction due to its
distinct intonation break from the rest of the clause. Yeo himself also rightly concedes that a
similar construction – “You think what ah I buy?” – is marginal. Both examples, where [wh-
expression+ah] are displaced together, as a unit, are actually seldom found in naturally
occurring CSE.
Secondly, the importance of QuP in the analysis leads to the need to posit a null
particle when one is not actually uttered. In other words, even in a language that does not
have particles, a null particle has to be assumed. The presence of a null particle is based on
the congruence between the rising intonation of a question and the rising intonation of a
question particle. This argument is a problematic one: Firstly, rising intonation is found in
many different particles but not all of them are interrogative or compatible with questions.
utterance? If a question has the same sort of rising intonation with or without a question
particle, perhaps it is the case that the question particle simply adopts the intonation pattern of
131
the question it appears in. In other words, the particle itself does not inherently have any fixed
the rising áh and falling àh is what leads to their compatibility with different question
types/intonation patterns, and not the other way round. Incidentally, in his arguments in
favour of a null particle as head of QuP, Yeo cites Chinese as support – particles in Chinese
are not optional because they are used to type the clause, therefore if the particle is not
overtly present, there is a null particle posited in LF – and yet rejects Cheng’s Clause Typing
Hypothesis at the same time. At the heart of the issue is the question of how to type an
utterance. Is a sentence with interrogative syntax but no rising intonation still a question?
The close association between the question particle and the wh-expression in Yeo’s
analysis creates another problem in that it cannot satisfactorily deal with multiple particles.
(274) A: But it’s beautiful in that…how…I mean, Finn got got a chance to realise
himself, right?
In any utterance with multiple particles, the first consideration is that of scope. Yeo
speculates that scope differences in multiple wh-questions rest solely on whether PrtP is
projected in the first place. This is a significant retreat from his original proposal that PrtP is
always projected whenever possible. The second consideration here is whether the grammar
can make a choice or decision on which particle to make phonologically overt. In other
words, which particle ends up being the Qu head, and how do the Qu(s) decide which wh-
phrase to take as its complement? Yeo agrees that this is problematic for his analysis.
132
Overall, the technical apparatus of the QuP, the FSH, and the innovative combination
of reprojective m-merger and Q-migration allows Yeo to account for a wide distribution of
CSE wh-constructions. The fact that we have been able to dissect it so extensively is a
testament to the level of detail in the analysis. However, as I have pointed out above, the
analysis suffers from several problems, most of them caused by the importance of the
question particle’s role in the proposal. The most serious (empirical) problem is that some
parts of the data that Yeo’s analysis is based on – specifically wh-constructions such as
“What ah you think I buy?” and “What you think ah I buy?”, where the particle does not
appear at the end of the clause – are not corroborated by my informant group of CSE native
speakers. For these reasons, and others that were mentioned in this section, I will outline a
133
Chapter 4 – Proposal
This dissertation introduces and tries to account for the wh-asymmetry between
matrix and embedded clauses in CSE – simply put, wh-movement is obligatory in embedded
clauses while it is non-obligatory in matrix clauses. As far as I know, this asymmetry has not
been pointed out or discussed in any of the CSE literature so far. Establishing this asymmetry
as a valid pattern in CSE will lead to the conclusion that CSE can no longer carry the
simplistic label of being a language with ‘optional wh-movement’. Although the term
‘random’ and ‘haphazard’ (see Platt, 1975; Platt & Weber, 1980), the question remains if it is
the best way to characterize a language such as CSE. If it is in fact the case that CSE’s
misleading, then how else should CSE be labelled? Will we end up having to choose between
the two conventional labels for wh-behaviour in languages: and say that CSE is at heart a wh-
movement language, or a wh-in-situ language? The answer that we will eventually reach by
the end of this chapter is that CSE is indeed a wh-movement language as we understand it,
but also possesses a special construction termed Declarative Syntax Question, from the
Let us take a quick look at the typological breakdown of the world’s languages into
wh-movement, wh-in-situ and optional wh-movement categories, using one of the largest
databases of the world’s languages – the World Atlas of Language Structures Online
(WALS). Under the section ‘Position of Interrogative Phrases in Content Questions’ (Dryer,
2013), out of a total of 902 languages analysed for wh-behaviour, there are 264 languages
where interrogative phrases are obligatorily initial, 615 languages where interrogative phrases
are not obligatorily initial, and only 23 languages classified as having a ‘mixed system’.
134
These numbers generally show that languages with non-initial-wh outnumber those with
initial-wh. It also shows, rather surprisingly, that there are only a small number of languages
in the world that are properly classified as having a ‘mixed wh-system’, which means both
wh-movement and wh-in-situ are available in those languages – what we have been calling
‘optional wh-movement’ languages so far. This is surprising only because there seems to be a
significant part of the wh-literature devoted to discussing these languages. More crucially, the
languages that are usually taken to display ‘optional wh-movement’, such as French and
Brazilian Portuguese, are not included in the ‘mixed wh-system’ classification in the WALS.
While the reason for this discrepancy in classification could have something to do with the
criteria or structural properties that the WALS adopt, that is a separate discussion for a
different time. What these numbers suggest to us, at least, is that ‘mixed system’ languages
aren’t as prevalent as we think they really are. In other words, we need to be more stringent
about what it means to have undergone wh-movement, or to remain in-situ, and ultimately
If that is the case, there are two theoretical options open to CSE: Firstly, CSE is a wh-
movement language on a par with Standard English, and all apparent wh-in-situ behaviour
can be ‘explained away’. This means that we will need a special explanation for CSE matrix
Japanese, and all apparent wh-movement phenomena can be otherwise accounted for. This
means that we will need to account for why there is wh-movement in both matrix and
embedded clauses in CSE. On the face of things, it appears that the first option is
theoretically more feasible than the second one, since there are fewer contexts we now have
to account for (matrix wh-in-situ, vs. both matrix wh-movement and embedded wh-
movement). This is the option that we will take in the analysis. Also, we will show that
135
apparent wh-in-situ in CSE is simply a case of a construction named Declarative Syntax
Questions (Bobaljik & Wurmbrand, 2015). Following Bobaljik & Wurmbrand, we will argue
clausal typing. Taken together, the two proposals will account for the wh-asymmetry in CSE.
It will also solve the puzzle of whether CSE should be labelled a wh-movement language or a
Chinese. Looking beyond CSE, the analysis also makes the cross-linguistic generalisation
that all (non-wh-in-situ) languages have access to the Declarative Syntax Question
construction, which may give a language the appearance of being a mixed system language
We started the previous section discussing the typological utility of having wh-
parameters and the importance of why we needed to clearly distinguish between what it
with a number of claims and their accompanying implications that we will now try to account
for:
- CSE wh-in-situ is only apparent wh-in-situ and not ‘true’ wh-in-situ as we know it. As
such, CSE will be regarded simply as a wh-movement language and not a language
with both wh-movement and wh-in-situ options. Also, this means that the
136
- The apparent wh-in-situ in CSE can be attributed to a construction named Declarative
languages.
OR
The above generalisation was made by Bobaljik & Wurmbrand (henceforth, B&W),
who were also the first to make the observation that DSQs are systematically utilized in overt
wh-movement languages.
languages. We can start with Standard English. In StdE, contrary to most standard
expression that has not undergone any sort of (wh-)movement. This construction is most
137
(276) Ellison: If you…if you really wanna torture yourself this holiday, go ahead, be
(277) Fitz: There’s a lot of overlap between the two companies, but only Transia has
(278) Ward: For thousands of years, they’ve remained in stasis, orbiting our solar
system.
These examples show us that DSQs are quite easily found in English. I have collected
many more examples of DSQs during the course of writing, and have compiled them in the
Appendix for reference. Although (276) and (277) are truncated/elided utterances, (278) is a
complete sentence and has a clearly declarative syntax with the wh-expression in-situ.
Crucially, (278) is not a rhetorical question, since Daisy’s answer following her own question
is only speculative and her interrogative intent was genuine in the first place. In addition to
what, DSQs are also possible with the full range of wh-expressions in English:
138
English is one of the most typically quoted examples of a wh-movement language, or
a language lacking wh-in-situ. The above examples (279 – 283) turn that characterisation on
its head, and show that both wh-arguments and wh-adjuncts can apparently be found in in-situ
positions (see also: Postal, 1972; Ginsburg & Sag, 2000; Pires & Taylor, 2007; Zocca
DeRoma, 2011). Crucially, none of the examples above are instances of echo questions,
which constitute the usual context in which wh-in-situ occurs. Echo questions are questions
with specific properties such as: they are usually a request for a speaker to repeat a specific
part of a preceding utterance, either as a point of clarification, or if that part of the utterance
was just not heard properly or difficult to understand. Since that is the case, an echo question
usually mimics almost entirely the structure of the preceding utterance, with only the wh-
expression replacing what needs to be clarified. Echo questions sometimes also imply
incredulity, when it is difficult to believe what was just said. As a result, there is usually an
exaggerated intonation or stress on the echo-wh-expression. Echo questions also have the
The question uttered by B in (284) has all the properties of an echo question: it is a
request for repetition/clarification of only a specific part of the preceding utterance – the
name of a location – and therefore has a strict linguistic antecedent; the structure of the
question is identical to that of the preceding utterance; disbelief and incredulity is expressed,
along with a raised pitch and tone; and the wh-expression is infixed within a single lexical
word. The DSQ examples presented in (276 – 283), on the other hand, do not share any of
139
- Do not require a specific linguistic antecedent (and therefore, trivially, do not need to
- Are a genuine request for information rather than simply a request for repetition or
clarification;
Since this is the case, DSQs are clearly distinguishable from echo questions.
While it may seem that DSQs are somewhat colloquial in register, we cannot simply
dismiss them as a quirk of non-Standard English. They can be found in a wide variety of
contexts, including legalistic questions (285), game-show questions (286), test questions
(287) as well as instructor’s comments (288), and most commonly, out-of-the-blue questions
(289):
(285) You met who for what purpose in the hotel room, Mr Smith?
(286) Brian is the name of the pet of the main character in which famous TV series?
situations?
The different contexts in which DSQs can be found (285 – 289) show how productive
they are in English. Ginsburg & Sag (2000) note an especially frequently used context (289)
– out-of-the-blue questions – where the speaker requests further information to flesh out a
140
salient and possibly non-linguistic context, in the absence of any linguistic antecedent.
Therefore, it should be quite evident from the data that these DSQs all have a genuine
interrogative intent and are legitimate interrogative speech acts. Now, let us move on to
B&W (2015) present data from languages such as German, Brazilian Portuguese,
Spanish, French, Dutch and Icelandic that show the existence of DSQ in those languages as
well. For the sake of space, I will reproduce only some of the examples that were presented:
(290) German
(291) Icelandic
‘…and you were where when Jóna drove off the road?’
(292) French
Since this construction can be found across a wide variety of languages that,
typologically speaking, have no relation with one another other than the fact that they are wh-
141
movement languages, we can tentatively hypothesise that the DSQ construction is in theory
However, there is one environment where a DSQ construction cannot be found, and
English
(294) *John asked Susan the beautiful lady standing in front of him is who?
‘John asked Susan who the beautiful lady standing in front of him is.’
German
Icelandic
142
b. þingmaðurinn spurði hvar Obama væri fæddur
Brazilian Portuguese
Examples (293) – (295), (296a), (297a) & (298b) clearly show that DSQs, in whatever
language they appear in, are ungrammatical when they are embedded as an interrogative
clause, or simply when they occur as/in an indirect question. In contrast, (296b), (297b) and
(298a) show that when the wh-expression is moved to the front of the embedded clause, the
construction is grammatical. Now, let us examine B&W’s proposal that accounts for DSQs.
The central fact that drives this thesis is that wh-in-situ is possible in CSE matrix
When one attempts to broach an explanation for this asymmetry, it is natural to turn to C, or
the CP layer, to look for a potential solution. This is based on the assumption that the target
or final landing site for wh-movement is usually the CP. One possible way to account for this
143
asymmetry – inspired partly by Fukui (1986), Kuroda (1988) and Bošković’s (1998) separate
works – was based on the decidedly counter-intuitive scenario that there was no CP
architecture in CSE matrix clauses, which would then preclude the necessity of wh-
movement in matrix clauses; while the CP is intact in embedded clauses, which would then
force movement of wh-expressions in embedded clauses. However, not only would this lead
Another common (and less drastic) analysis is to assume a different featural makeup
of matrix C and embedded C, since the presence/nature of features in C could also work to
induce/hinder wh-movement. The challenge with this line of reasoning is that this difference
in feature makeup has to be motivated in some way; if not, then the difference(s) will end up
being as mysterious as the outcomes it was meant to account for in the first place. B&W’s
analysis adapts the two ideas articulated above – by claiming that an interrogative C is
present in a regular question but absent in a DSQ; and that the feature makeup of that
around the two challenges raised above, the analysis links feature differences to wh-
- The difference between wh-movement and true wh-in-situ languages involves the
features of CWH: in a wh-movement language, CWH has features that require an overtly
filled specifier whereas in a true wh-in-situ language, the dependency between CWH
144
and a wh-expression may be satisfied without overt movement, either by covert
- DSQs are (syntactically) declarative TPs or CPs which lack CWH and in which the wh-
The first claim states that there is a dependency between an interrogative CWH and wh-
interrogative clause that has been typed as such via syntactic means. A DSQ is syntactically a
and correspondingly there is no CWH in a DSQ. We see the importance of CWH in motivating
wh-movement, and in the typing mechanism. The lack of such a dependency will clearly
account for the fact that DSQs are not sensitive to island effects and intervention effects, as
previously observed. It is noted that this dependency is present in true wh-in-situ languages,
and island/intervention effects are not observed there either. The general architecture is
presented below:
CPWH
CWH TP
wh-XP
145
(300) Declarative Syntax Question (DSQ)
(C) TP
[declar]
[wh-XP]FOC
The arrows linking CWH and the wh-XP in (299) show the dependency between the
two elements. There are no arrows in (300) since a DSQ does not have any syntactic
dependency at all. Since there is no wh-operator in C in (300), this means that the wh-
its own. B&W suggest that the best way to do this is to propose that wh-expressions have an
interpretable and valued wh-feature: iQ: wh. A wh-expression having an interpretable and
valued wh-feature has many advantages in this analysis: Firstly, the wh-expression is
syntactically independent of the C head and does not need to be licensed or bound by it.
Secondly, because of this, the wh-expression is free to contribute its own interrogative
interpretation. This feature proposal has to apply to all wh-expressions (both in DSQs and in
true syntactic interrogatives) since there is no overt morphological evidence to suggest that
moved and in-situ wh-expressions are different in any way, other than the way they are
interpreted. There is an additional stipulation, however, to account for languages such as CSE
and German, which have indefinite wh-expressions. These wh-expressions cannot receive an
interrogative interpretation, and thus must be distinguished from those that can. Therefore,
indefinite wh-expressions will have an uninterpretable wh-feature instead (see also Pesetsky
B&W assume that the interpretability of features is independent of its valuation, since
they take licensing to be established via Agree, which is valuation driven. This means that a
feature, whether interpretable or uninterpretable, can be either valued or unvalued. They also
146
use a slightly modified definition of Agree, called ‘Reverse Agree’, which is taken from
Wurmbrand (2014):
a. β c-commands α AND
would probe downward till it finds a suitable valued Q-feature (in a wh-expression), and that
feature then Agrees with that C by valuing it upwards. This method of (upward-)valuation
has the drawback that we will still need to depend on an EPP feature to trigger actual wh-
now recaptured as the optionality of having an EPP feature – whether an EPP feature is
mechanism gets around this problem since if valuation has to take place from a c-
Given these assumptions, we can now systematically derive the following: a matrix
wh-movement. For comparison, we can also look at how similar structures are derived in wh-
in-situ languages. In the first scenario, the derivation is carried out as such: An unvalued but
interpretable Q-feature (iQ: ___) is merged with an interrogative CWH. Since this feature is
unvalued, it will require valuation by something which both has a valued Q-feature and c-
147
commands it, such as a wh-expression in the sentence. This requirement for valuation triggers
movement of a wh-expression with a valued Q-feature to Spec, CP. From this position, the
wh-expression values the unvalued Q-feature (iQ: ___) in CWH. Additionally, the valuation of
CWH (now, iQ: wh) syntactically types the CP as an interrogative. The derivation structure is
language)
CP [iQ: wh]
XP 3. Type C’
iQ: wh
C TP
iQ: wh
2. Value …tXP…
1. Move
Next, we can follow (302) with the derivation of the second structure: an embedded
syntactic interrogative, with wh-movement having taken place. Following the typing of the
CP as an interrogative clause, the resultant typed CP can subsequently be merged with a verb,
such as wonder or ask, that selects for an interrogative clause. Only a CP typed with the
feature [iQ: wh] can be selected in this way and all other clauses not typed as interrogative
148
(303) Derivation of Embedded Syntactic Interrogative (in wh-movement language)
VP
Select
XP C’
iQ: wh
C TP
iQ: wh
…tXP…
VWH in (303) denotes a predicate which is only able to select a clause that is overtly
and syntactically typed as an interrogative. It is important to note that syntactic typing of the
commanding wh-expression. Therefore, this means that the wh-expression must overtly move
to Spec, CP before any typing takes place. wh-movement within the embedded interrogative
cannot take place after the CP is merged with VWH, since at that point of the derivation, the
CP would not be typed and thus would not be selected by the predicate. The valuation
requirements, selectional requirements and the timing of the operations can be taken as a
whole to force overt movement of the wh-expression to the front of the embedded
interrogative clause. The syntactic domain (CP) in which these movement operations have to
be resolved before Selection and Merge can take place also corresponds to the phase
boundary postulated in Chomsky’s (2001) Phase Theory. The requirement for wh-expressions
to overtly move in embedded syntactic interrogatives is observed for CSE, English and other
wh-movement languages.
As for true wh-in-situ languages such as Chinese and Japanese, there are (at least) two
ways of deriving matrix and embedded interrogatives. The first method is very similar to
what was laid out in (302) and (303), but instead of overt movement of the wh-expression,
149
covert movement takes place instead ((304) & (306)). What is moved also differs – in
addition to the wh-expression itself, the covert movement of entire phrases, feature
movement, movement of morphemes rather than words, and overt movement of phrases
followed by pronunciation of the lower copy have all been variously suggested before. The
second method is simply to postulate that a valued Q-feature is merged into C, which then
unselectively binds another Q-feature to provide its interpretation as a wh-question ((305) &
(307)). In any case, both options will lead to a structure where it is possible to have an in-situ
wh-expression in an embedded interrogative clause ((306) & (307)), which is the one
derivation that distinguishes between true wh-in-situ languages and wh-movement languages,
since it is simply not possible for a wh-movement language to have an in-situ wh-expression
CP [iQ: wh]
3. Type C’
iQ: wh
C TP
iQ: wh
2. Value …XP…
1. Covert Move
150
(305) Derivation Structure of Matrix Syntactic Interrogative (in wh-in-situ language,
CP [iQ: wh]
3. Type C’
C TP
iQ: wh
…XP…
1. Merge
2. Binds
covert movement)
VP
Select
XP C’
iQ: wh
C TP
iQ: wh
…XP…
Unselective Binding)
VP
Select
C’
C TP
iQ: wh
…XP…
151
In this section, we have shown a formal description of how syntactic interrogatives –
clauses that are overtly/syntactically typed as interrogatives – are derived in both wh-
movement and wh-in-situ languages. Next, we will show that the DSQ construction can also
be found in CSE, and can shed some light on the central matrix/embedded asymmetry we
Focus that can be interpreted as a question. B&W’s analysis is crucially based on the same
slightly different from this thesis. B&W started off with the aims of reaffirming “…the role
that syntactic selection plays in the domain of clausal embedding…”, and arguing against
analyses that “…propose to assimilate the optional wh-in-situ facts of English, French or
other languages to the wh-in-situ constructions of, say, Chinese, Japanese or Turkish”.
B&W’s analysis, although still quite new and yet to receive significant attention in the
literature, is very promising indeed; and seems especially suited to account for a ‘optional
wh-movement’ language such as CSE. For its part, CSE could serve as an important data
point, and possibly make some new contributions to developing B&W’s analysis.
Crucially, for the purposes of this thesis, we want to see if the DSQ construction can
be found in CSE, according to our tentative hypothesis (275). It should come as no surprise
that it does. We have previously presented data (from (28) – (33)) showing that although it is
unproblematic for wh-arguments to be in-situ in CSE matrix questions, there is only very
152
questions. Data such as (28) – (30) (repeated below as (308) – (310)) show that DSQs are
found in CSE, while (31) – (33) (repeated below as (311) – (313) seem to suggest otherwise:
Wh-Arguments
Wh-Adjuncts
However, we must note that the theoretical possibility of having the DSQ construction
in a language does not mean that it is automatically or always acceptable when such a
construction occurs. As a case in point, we turn to the apparent asymmetry (in acceptability)
between a question with an in-situ wh-adjunct how (repeated from (31)) and a question with
153
Both these sentences qualify as DSQs because they are syntactically declarative, with
the wh-expression in-situ. However, the former is generally considered less acceptable than
the latter by speakers. Does the existence of something like (314) prevent us from saying that
the DSQ construction exists in CSE? The answer is that it should not, since other factors
existence of (315) shows that we can add CSE to the list of wh-movement languages where
Additionally, we will have to satisfy ourselves that a DSQ in CSE is indeed not
that lead to apparent wh-in-situ outcomes, such as Pronounce Lower Copy, or Remnant
Movement. Let us look at how DSQs in CSE behave with regards to movement diagnostic
Adjunct Island
Wh Island
Subject Island
(318) The capital of which country has the highest crime rate?
*Which country the capital of ___ has the highest crime rate15?
15
Interestingly, the same question without the preposition of – “Which country the capital has the highest crime
rate?” is perfectly fine in CSE. This kind of sentence is usually accounted for as a topic-comment construction.
154
Coordinate Structure Island
(320) You need a driver who can drive what (kind of) car?
*What (kind of) car you need a driver who can drive ___?
Firstly, the data in (316) – (320) show that the syntactic distribution of DSQs is
relatively free in CSE. Across multiple different syntactic islands, the in-situ wh-expressions
did not render the above CSE DSQs ungrammatical, as all corresponding wh-movement out
of the island is impossible. What these island movement diagnostic results – the ability to
appear inside an island and the inability to move out of it – tell us is that there is no
movement involved in deriving a DSQ; and more importantly, that a DSQ receives its
interpretation as a question via a mechanism other than movement. To further confirm these
properties, we will turn to a blocking effect generally weaker than island effects –
Intervention effects.
(Hagstrom, 1998; Beck, 1996), semantic (Beck 2006; Cable 2007) and pragmatic in nature
respectively. The diagnostic itself is quite clear – if the presence of an intervener in a position
above the wh-question leads to ungrammaticality, an intervention effect is said to have taken
place. The list of potential interveners differs slightly across languages, but traditionally
155
‘strong’ ones that induce effects include Negation, Negative Polarity Items, Nominal
The examples in (321) – (323) are all legitimate DSQs because they have declarative
syntax and the wh-expressions are in an apparent in-situ position. In all the examples, a
seems that there is no evidence of any intervention effects present in CSE. The lack of
intervention effects in CSE tells us that wh-expressions undergo covert movement in order to
avoid intervention. A quick test of the interaction between Island and Intervention effects
Up until this point, the data presented might not seem at all surprising to anyone who
is familiar with wh-in-situ phenomena. In fact, one might be tempted to treat the DSQ
construction in CSE the same way as a wh-in-situ construction from typical wh-in-situ
languages such as Chinese, Japanese and Turkish. We have already seen a number of
analyses in the previous chapter that can deal with this. To get around the island-insensitivity
and lack of intervention effects, one could propose some kind of null Complementiser which
binds wh-expressions without having to resort to movement (either overt or covert), or use
156
some version of Unselective Binding or Agree to get the job done. However, that would not
work for CSE because of the crucial observation that this thesis is based on – that there is an
In summary, from multiple examples of DSQs in CSE (316) – (324), we have also
noted that they are fine not only as matrix questions but also in embedded clauses
examples of this restriction in CSE throughout the paper; although we had not given the DSQ
We also see how the proposal aids us in distinguishing between apparent wh-in-situ
(DSQ) and true wh-in-situ. The crucial property of the DSQ construction – that wh-movement
that lets us firstly define, and then distinguish between what is true wh-in-situ and what is
only apparent wh-in-situ – which goes some way in answering the questions posed in the
discussion at the beginning of this chapter. Traditionally, the yardstick for determining
to island and intervention effects. We see now that these tests do not truly distinguish
between the different language types; but merely tell us whether there is movement involved.
The discussion of the DSQ facts so far leads us to conclude that it is the ability to be
embedded in an interrogative clause that is the crucial factor. We can make a tentative
generalisation here:
157
According to (325), even when CSE displays wh-in-situ in matrix clauses, it will still
not be considered a true wh-in-situ language because it is not possible for a wh-expression to
remain in-situ in an embedded interrogative clause. The type of wh-in-situ patterns displayed
in CSE can only be described as ‘apparent’ wh-in-situ. Since that is the case, CSE is simply
another wh-movement language that has more in common with English and German in terms
Therefore, in this section, we have presented convincing data that show the following:
- The DSQ construction is observed in CSE, where it is both productive and it appears
frequently;
- The distribution of DSQ in CSE is relatively free, other than the restriction that it
Using these facts, we are able to generalise what it means to be a true wh-in-situ
language, and rule CSE out of that category. In the next section, we will apply the DSQ
analysis to the CSE facts and attempt to derive as much of the wh-paradigm as we can, while
noting what the analysis fails to explain. Before that, however, we leave a short remark on
4.1.4 wh-in-Focus?
One of the main claims made in B&W’s analysis is that a DSQ is syntactically a
declarative and not typed as an interrogative even as it is interpreted as one. This means that a
DSQ lacks a CP marked with iQ: wh, explaining why it cannot be selected by an
158
interrogative-selecting predicate, and therefore why it cannot be embedded by such a
predicate. Under their proposal, a DSQ does not involve a C-XP-wh dependency. We have
taken these properties of a DSQ construction, as well as the construction itself, to motivate a
distinction between wh-movement languages that display apparent wh-in-situ phenomena and
true wh-in-situ languages. However, one of the questions that falls out from this analysis is:
typed as one? B&W argues that languages where the DSQ construction can be found are also
“…languages in which elements in focus may remain in-situ” (B&W, 2015: 23; emphases
my own). This explains the FOC subscript on the wh-expression in the structural
representation shown in (300). What they are essentially proposing is that DSQ constructions
are Focus constructions; furthermore, that the wh-expression(s) in a DSQ construction are
actually Focused elements. Therefore, they use the term ‘wh-in-focus’ to describe these wh-
it is the case that DSQ constructions are Focus constructions, then deriving an interrogative
interpretation from a DSQ might be similar to, or aligned with however Focus is derived.
It is actually widely known and well discussed in the literature that there are many
expressions and Focus phrases behave similarly – both being insensitive to island constraints.
In some languages, wh-expressions and Focus elements occupy the same structural positions
(as shown above for CSE; the same is also true for Malayalam (Jayaseelan, 2003), Hungarian
(Brody, 1990), and Chadic (Tuller, 1992)). It has been proposed in some of these languages
Focus feature present in wh-expressions, and this triggers movement of the wh-expressions to
the same positions that Focus(ed) elements are found. Phonologically, wh-expressions that
carry an interrogative interpretation share the same intonation characteristics – such as lexical
159
stress or pitch accent – as a Focus(ed) element. An interrogative wh-phrase is marked with
the presence of such a phonological property (Liu & Yi, 2006) while an indefinite wh-
expression lacks it. This is true in almost all languages that utilizes wh-indefinites, such as
We are not given very much information on how the interpretation of meaning part of
B&W’s analysis works other than that “…an interrogative interpretation is derived via
pragmatics” (B&W, 2015: 21). This is also alluded to in the representation of a DSQ in (300)
shown earlier. To be fair to B&W, they did not have the space to go into the details of how
this would pan out in their paper, and also simply because they did not intend to do so in the
first place. As for this current thesis, our central focus is to account for the matrix/embedded
will simply assume, following B&W, that an interrogative interpretation for DSQ follows
In this section, we will take stock of the technical apparatus that we will need to
describe the wh-paradigm in CSE. It is a pleasant surprise that we have been able to use
existing analyses, and have not needed to resort to extra stipulations to derive the patterns in
CSE. This should increase our confidence that CSE has more in common with other well-
studied languages than we previously thought, and has an underlying systematicity that
makes it an excellent testbed for emerging theories on natural languages. First, let us lay out
the entire range of wh-constructions in CSE that we have looked at in this paper:
160
Construction Example
(DSQ)
(31a) ?John fix the car how?
Interrogative Matrix clause with wh- (28b) What (did) John eat?
movement
Non-interrogative Embedded clause with wh- (46a) John think Lisa meet who?
Interrogative Embedded clause with wh- (40a) John want to know who Lisa marry.
movement
Interrogative Embedded clause with wh-in- (40b) *John want to know Lisa marry who.
situ
Now, let us give a proposed derivation of each of the presented examples. A matrix
This is the most typical word-order of wh-movement languages such as StdE and CSE, with
161
(326) Derivation of Matrix Syntactic Interrogative
CP [iQ: wh]
XP Type C’
iQ: wh
C TP
iQ: wh
Value …XP…
Move
In the proposal that we have borrowed from B&W, the wh-expression would overtly
feature (iQ: ___) merged into C, which requires valuation from a c-commanding, Q-valued
wh-expression:
CP
C TP
[Decl]
DP VP
John
eat [what]FOC
162
There is no movement in (327). There is no Q-feature in C because it is syntactically a
declarative clause. The interpretation of the wh-expression is derived in-situ via pragmatics.
CP
C TP
[Decl]
DP VP
John
think CP
C TP
[Decl]
DP VP
Lisa
meet [who]FOC
VP
Select
XP C’
iQ: wh
C TP
iQ: wh
…tXP…
163
In (329), the wh-expression in the embedded clause has to undergo wh-movement,
first to value (via Reverse Agree) the unvalued Q-feature in C, and secondly to type the
embedded clause as an interrogative, before the clause itself can be selected by the
interrogative-selecting verb wonder. If the wh-expression is left in-situ and does not move,
then the clause does not get typed as an interrogative and the derivation breaks down since it
will then be selectionally incompatible with the verb wonder, therefore ruling out an
The derivations in (326) – (329) basically capture the full range of wh-phenomena
that we are prepared to discuss in this dissertation. We will discuss some unresolved
questions, the way forward for future research, and conclude the dissertation in the next and
final chapter.
If we take a closer look at the structure proposed in (328) to explain a sentence such
as (46a), and assuming that the derivation proceeds by phases, when the embedded
declarative CP is being assembled, we have no way of knowing whether the matrix CP will
problem. However, if the matrix CP is interrogative, then the wh-expression must somehow
move up to the intermediate Spec, CP position before moving up again when the matrix
interrogative CP is merged in. This means that when the embedded C is merged in, the wh-
164
However, our analysis manages to circumvent this problem. We can go through the
merged in.
(331) Derivation of the syntactic interrogative John think Lisa meet who
merged in.
Assuming that the system is unable to look ahead to check what would be merged in
at step (d), a derivation such as (331) will naturally crash at step (331e). However, if there
isn’t a CWH present in the numeration, the derivation will proceed without a hitch, as in (330),
165
4.3.2 Partial Movement
The fact that partial movement of a wh-expression occurs in CSE, as in (46b), despite
this movement not being motivated by our current analysis, is a problem for the DSQ
analysis. There are two suggestions on how to resolve this problem. First, we can postulate,
like B&W does, that only matrix clauses are DSQs. That is to say, embedded clauses
involving partial movement and full movement are not DSQs and involve a dependency
(332) Derivation of the syntactic interrogative John think who Lisa meet
merged in.
derived.
merged in.
166
c. This triggers movement of the wh-expression to embedded Spec, CP, a
position from which it may value the unvalued iQ: ___ of C, the matrix Spec,
CP position.
Alternatively, we could retain the assumption that they are DSQs, and think of partial
Independently, we already know that CSE is a Topic-prominent language (Platt & Ho, 1993;
Bao, 2001; Bao & Lye, 2005), and that a wide range of structures such as conditionals, Verb
Phrases, Noun Phrases and Adjective Phrases are available as topics in CSE. This means that
Topicalisation is a productive process in CSE. That being said, since in-situ wh-expressions
in DSQs are actually in-situ Focus, it is a mystery how they can move to Topic positions –
The prohibition against wh-adjuncts staying in-situ is problematic for us given the
availability of wh-adjuncts to participate in DSQs, such as in (281) – (283), (290) and (292).
This leads us to wonder if wh-adjuncts in CSE are really base-generated higher in the clause
167
than they are assumed to be (Erlewine, personal communication, 2016; Tsai, 2008). We have
assumed without question that wh-adjuncts are generated at sentence-final positions, but
because they ultimately cannot remain in-situ, are forced to move (Kim et al, 2009). If we
reject Kim et al’s claim, and assume that it is possible wh-adjuncts are generated nearer to the
subject, or even just near to the verb, then we no longer need to stipulate their movement. As
it turns out, we have already seen multiple instances of wh-adjuncts occurring either after the
subject, or before the verb, such as with why (111), special how phrases ((119), (120) &
(122)), anyhow (125). In addition, we see that other wh-arguments regularly occur at these
positions as well –what (79), where (100), and anywhere (104) etc. If this assumption is true,
a sentence such as (43) – (45) would be impossible to generate in the first place. There would
then no longer be any contrast between wh-adjuncts and wh-arguments left to explain.
168
Chapter 5 – Conclusion
At the beginning of the dissertation, I posed some general questions about wh-
typology. Although we still don’t have all the answers, we might have gotten a bit closer to
A1: The asymmetry in CSE is thus: matrix wh-in-situ is possible but embedded wh-in-
possible because of the Declarative Syntax Question construction. The latter is not
CP in order to type the clause as an interrogative, before the clause can be selected by
an appropriate verb.
A2: No, it is not. This asymmetry is noted in a number of languages as pointed out
above: including English, German, Dutch, Icelandic, etc. The Declarative Syntax
languages.
Q3. How would we still classify CSE typologically in terms of wh-behavior? Would it
169
A3: As a result of the proposal, CSE should clearly be classified as a wh-movement
language. We will not need to create new categories in the wh-paradigm: languages
that exhibit ‘mixed’ wh-behavior similar to CSE should also be considered wh-
movement languages. True ‘mixed’ wh-systems are quite rare in the world.
Q4. Are there languages that exhibit the reverse asymmetry, i.e. fixed wh-behavior in
in embedded clauses? If so, or if not, what does that mean for us theoretically and
typologically?
A4: No natural human language exhibits this reverse asymmetry, as far as I know.
motivation for wh-movement (or the lack thereof) in a matrix clause, would then fail
Q5. Do all contact languages with a superstrate and a substrate that exhibits different
wh-behavior (such as CSE) end up with a ‘mixed’ wh-behavior pattern? Also, does
this asymmetry shed any light on the degree of influence each substrate language
A5: No, but only in the sense that CSE is not really a ‘mixed-system’ language. It also
depends on how wh-movement and clause typing works in the particular contact
language. As for the degree of substrate influence, it does not provide a definitive
answer, other than that CSE should be classified as a wh-movement language, just
like English. The proposal also argues against assimilating an analysis of Chinese-
170
5.2 Optionality
In general, there are many questions, and a good amount of confusion surrounding
Optionality as a concept in grammar. Scholars have been trying to formalise the idea of
optionality for a long time, and in doing so, locate the source of optionality in (the) language.
in the system using the tools we have. In the context of wh-movement, this has resulted in a
myriad of ways of formally adopting optionality in syntax, including but not limited to: the
optional selection of C (Denham, 1997); the optional assignment of the EPP feature
(Chomsky, 2000 and Pires & Taylor, 2007); the optional occurrence of an intonation
morpheme in the Numeration (Cheng & Rooryck, 2000) and the optional occurrence of other
lexical items in the Numeration in general; optionally strong or weak features that do or do
not trigger movement (Sabel, 2003); and the optional (choice of) lexical makeup of wh-
optionality must exclude some of the above explanations, for the simple reason that they
involve different derivations – in that case, there would not be a ‘one LF to multiple PFs’
correspondence that we have been using to characterise optionality so far. A cursory glance at
this list tells us that there are indeed many different areas within syntax where we can
optionality in wh-movement as there are ‘triggers’ for wh-movement in the first place.
However, this should not be seen as a desirable state of affairs, as this suggests that we are
yet unclear which part of syntax formal optionality is a part of. A general criticism of any one
analysis from the above list is that we are merely kicking the optionality can down the road,
by pushing the burden of explanation to another part of the system. For instance, we can
assume that an EPP feature in non-interrogative phase heads is the reason for triggering wh-
171
movement, and that the feature can be optionally present. The question then changes from
‘what makes wh-movement optional?’ to ‘what makes the EPP feature optional?’. We will
still have to answer that question eventually, even as we have successfully fulfilled our
narrow aim of operationalising optionality within the syntax 16. In fact, to stipulate optional
(2006) suggests using Re-Insertion of the XP that wh-phrases are found in, into the Sub-
Numeration, to be Re-Merged at a higher point in the derivation, to obviate the need to do so.
This is simply an illustration of the sort of response to the stipulative nature of some of the
several observations become apparent: Firstly, that there are many more languages with both
wh-in-situ and wh-movement options than we had realized; secondly, the phenomena known
language would acquire scope overtly, while a wh-phrase in a wh-in-situ language would
acquire scope covertly. The first observation is borne from the fact that most languages have
access to at least both these wh-question formation strategies. Take for instance English,
questions (i.e. non-echo-questions) have been offered as evidence that English has access to
the wh-in-situ strategy as well as the wh-movement strategy. Many contact and nativised
varieties of English, such as CSE, also share this property. This brings into question the rigid
been conveniently classified typologically as only being one or the other – but the truth is
probably more in the region of: Language X adopts one strategy most of the time, but the
16
When taken to its logical conclusion, uncovering the reason(s) for optionality paradoxically reduces that
particular paradigm to one of non-optionality; since there is now a principled reason why there are multiple PF
candidates for the same LF.
172
other strategy is also available to it. In other words, the predominant or preferred wh-strategy
is only a subset of all the available wh-strategies in the language. If that is the case, then what
language? For a wh-movement language, does it mean: (a) a wh-phrase in the language can
possibly undergo wh-movement; or (b) a wh-phrase in the language can only undergo wh-
movement? Statement (a) is a bit too general and it seems to be readily applicable to almost
any language, since it is difficult to imagine a language which imposes a strict ban on any
wh-movement; while Statement (b) sounds too restrictive and will probably therefore be
does it mean: (a) a wh-phrase in the language can possibly remain in-situ; or (b) a wh-phrase
in the language can only remain in-situ? Again, the same comments apply: (a) is too strong
while (b) is unlikely to be an accurate representation of the facts. This suggests that taking a
a macroparametric one.
interest has been generated in finding ‘truly’ optional wh-constructions in languages over the
past 2 decades (see Cheng & Rooryck, 2000; Denham, 2000; Mathieu, 1999; Poletto &
Pollock, 2005; Sabel, 2003; Uribe-Etxebarria, 2002 and many others). The phrase ‘true
optionality’ was coined to convince an audience who were yet unsure whether alternative
operations that generated different syntactic objects could equally and optimally satisfy
displaying ‘true optionality’ was actually only a first step in proving that optionality exists in
the language. After this was established, the next logical step of the endeavor was to show
whether various options in the wh-paradigm were equally or differentially favoured. For
17
This has the same meaning as it is used in a phrase such as ‘true wh-in-situ language’ (B&W, 2015).
173
example, if two wh-strategies are equally favoured (i.e. a 50-50 split) in the grammar, then
that could constitute a much narrower meaning of the phrase ‘true optionality’. However, we
have seen, such as in the case of CSE embedded questions, that this is not always the case –
since one strategy seems to be favoured over the other. In fact, I would venture the claim that
a system exhibiting ‘true optionality’ (in this narrow sense) is far more rare than a system
where multiple options exist, but with some being more prevalent than others.
One major debate that the proposal outlined here lends itself to is how languages are
typologically categorized in terms of their wh-behaviour. Primarily, there are three categories
– wh-movement languages, wh-in-situ languages, and mixed systems (where both movement
and in-situ are possible). The majority of the world’s languages fall in the first two
categories, with wh-in-situ languages numbering more than wh-movement languages. The
wh-typological status of a language is usually not called into question unless a special
construction (that shows a different pattern) in that language is brought to light through
linguistic research. Even so, that special construction might simply be treated as an exception
or anomaly in the language that is subject to extraordinary conditions. Examples include echo
emerging language variety such as CSE, however, started with an unclear status. Even though
its superstrate language was English, it also had significant substratal influence from Chinese
and Malay, which presented a contradictory picture in wh-behaviour. Early formal accounts
of CSE simply resorted to describing it as having random, haphazard and optional wh-
movement.
174
Later on, it was suggested that instead of being described as having optional wh-
movement, CSE could be described as a language with access to BOTH18 wh-movement and
wh-in-situ strategies. Having access to more than one wh-strategy was a plausible suggestion
that was considered, for instance, by Cole and Hermon (1998) and others. Coming up with
the technical apparatus to handle multiple wh-strategies would require innovation but it was
shown to be workable. However, it was not until B&W’s DSQ proposal that we were able to
find a simpler, more generalisable explanation for CSE’s wh-behaviour. The DSQ
many (if not all) wh-movement languages, including CSE. The existence of the DSQ
under certain verbs solidified our confidence that CSE be rightfully classified as a wh-
Apart from finally finding CSE a typological home in terms of wh-status, this
classification has other implications. For one, we no longer needed to assimilate apparent wh-
Chinese. This is an important implication because for a long time, once a construction has
been noted to be insensitive to islands (or any other movement diagnostic), it would be
treated as analogous to wh-in-situ licensing. In doing so, we might have glossed over subtle
Another implication, which is relevant only to other related phenomena in CSE, is that we
can now take wh-movement as the default strategy in the language. The other two more
general typological conclusions that we can draw are: Firstly, we can now distinguish more
clearly between apparent wh-in-situ phenomena and ‘true’ wh-in-situ phenomena. For a
18
Whether this description is akin to the typological status of being a mixed system (some wh-expressions
obligatorily initial, some not) is unclear, although it should be noted that mixed system languages amount to less
than 3% of the world’s languages, and none of them are typologically similar to CSE in any other way.
175
language to be considered truly wh-in-situ, it must minimally allow in-situ wh-expressions in
not able to do so, then that language cannot be classified as a wh-in-situ language. We have
become ‘stricter’ on what it means for a language to be wh-in-situ. Secondly, we reaffirm the
typological fact that mixed system languages are indeed quite rare (see footnote 18). Lastly,
with the characteristics of the DSQ construction so clearly laid out now, I think the next step
can tell, there are no wh-movement languages that do not have a version of the DSQ
construction. If this is the case, then the DSQ construction can be regarded as a typological
176
References
Alsagoff, L. 2010. English in Singapore: Culture, capital and identity in linguistic variation.
Aoun, J. & Li, A. 1993. Wh-elements in-situ: Syntax or LF? Linguistic Inquiry 24: 199-238.
Austin, J. 1962. How to do things with words. London: Oxford University Press.
Bach, E. 1968. Nouns and Noun Phrases, in E. Bach and R. T. Harms (eds.) Universals in
Baker, C. 1970. Notes on the Description of English Questions: the Role of an Abstract Q
Bao, Z. 2001. The origins of empty categories in Singapore English. Journal of Pidgin and
Bao, Z. 2005. The aspectual system of Singapore English and the systemic substratist
Bao, Z. & Lye, H. M. 2005. Systemic Transfer, Topic Prominence, and the Bare Conditional
Bayer, J. 2006. Wh-in-situ. In M. Everaert & H. van Riemsdijk (eds.) The Blackwell
177
Beck, S. 1996. Quantified Structures as Barriers for LF Movement. Natural Language
Beck, S. 2006. Intervention Effects Follow from Focus Interpretation. Natural Language
Belletti, A. 2001. Inversion as Focalization. Subject inversion in Romance and the theory of
Berman, S. 1991. On the Semantics and Logical Form of wh-clauses. PhD dissertation.
Bobaljik, J. 2002. A-chains at the PF-interface: Copies and ‘Covert’ Movement. Natural
Bobaljik, J. & Wurmbrand, S. 2015. Questions with Declarative Syntax tell us what about
Boeckx, C., Stateva, P. & Stepanov, A. 2000. Optionality, Presupposition, and wh-in-situ in
French. In Romance Syntax, Semantics and L2 Acquisition: Selected Papers from the 30th
Bošković, Ž. 1998. LF movement and the Minimalist Program. In Proceedings of the North
Bošković, Ž. 2000. Sometimes in SpecCP, sometimes in-situ, in R. Martin et al. (eds.) Step
by step: Essays on Minimalism in honor of Howard Lasnik: 53-87. Cambridge, MA: MIT
Press.
178
Bošković, Ž. 2002. On Multiple wh-Fronting. Linguistic Inquiry 33: 351-383.
Bošković, Ž. & Lasnik, H. 1999. How Strict is the Cycle? Linguistic Inquiry 30(4): 691-703.
2: 354-393.
Brody, M. 1990. Some Remarks on the Focus Field in Hungarian. UCL Working Papers in
Linguistics 2: 201-225.
Brody, M. 1995. Lexico-Logical Form, a Radically Minimalist Theory. Cambridge, MA: MIT
Press.
Cable, S. 2007. The Grammar of Q: Q-particles and the Nature of wh-Fronting, as Revealed
Columbia.
Cheng, L. 1994. Wh-words as Polarity Items, in Paul Li, Chu-Ren Huang and Jane Tang
Cheung, C. 2014. Wh-fronting and the left periphery in Mandarin. Journal of East Asian
179
Chomsky, N. 1973. Conditions on Transformation, in S. R. Anderson and P. Kiparsky (eds.)
A Festschrift for Morris Halle: 232-286. Holt, Rinehart and Winston, New York.
Chomsky, N. 1993. A minimalist program for linguistic theory, in K. Hale and S. J. Keyser
Uriagereka (eds.) Step by Step: Minimalist Essays in Honor of Howard Lasnik: 89-155. MIT
Press.
Chomsky, N. & Lasnik, H. 1977. Filters and Control. Linguistic Inquiry 8: 425-504.
Syntax 1: 221-258.
Cole, P. & Hermon, G. 2000. Partial wh-movement: Evidence from Malay, in U. Lutz, G.
Müller and A. von Stechow (eds.) wh-Scope Marking: 85-114. Amsterdam and Philadelphia:
John Benjamins.
Donati, C. 2006. On Head Movement, in L. Lai and N. Corver (eds.) wh-movement: Moving
180
Dryer, M. 2013. Position of Interrogative Phrases in Content Questions, in M. Dryer, M.
Fukui, N. 1986. A Theory of Category Projection and its Applications. PhD Dissertation,
MIT.
Ginsburg, J. & Sag, I. 2000. Interrogative Investigations: the Form, Meaning and Use of
Grimshaw, J. 1979. Complement Selection and the Lexicon. Linguistic Inquiry 10: 279-426.
Gupta, A. 1990. A Study of the Acquisition and Use of Interrogatives and Questions in the
Multilingual Matters.
http://themiddleground.sg/2015/08/05/singweesh-wednesday-anyhowly/
181
Hagstrom, P. 1998. Decomposing Questions. PhD Dissertation, MIT.
Han, C-H. & Romero, M. 2004. The Syntax of Whether/Q…Or Questions: Ellipsis
Combined with Movement. Natural Language and Linguistic Theory 22(3): 527-564.
He, D. & Li, D. C. S. 2009. Language attitudes and linguistic features in the ‘China English’
Heim, I. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD Dissertation,
Henry, A. 1995. Belfast English and Standard English: Dialect Variation and Parameter
Ho, A. 2000. Discourse Factors in Singapore English wh-questions. Ms., National University
of Singapore.
Hoji, H. 1985. Logical form constraints and configurational structures in Japanese. PhD
University Press.
Huang, J. 1982. Logical Relations in Chinese and the Theory of Grammar. PhD Dissertation,
MIT.
https://www.imda.gov.sg/~/media/imda/files/regulation%20licensing%20and%20consultatio
ns/committee%20reports/pac%20combined%20report%20fy2015.pdf?la=en
182
Infocomm Media Development Authority. 2017. Free-to-Air Television Programme Code.
Retrieved from:
https://www.imda.gov.sg/~/media/imda/files/regulation%20licensing%20and%20consultatio
ns/codes%20of%20practice%20and%20guidelines/acts%20codes/01%20industrytvcontentgu
idelinesftatvprogcode.pdf?la=en
69-99.
Jiménez, M. L. 1997. Semantic and Pragmatic Conditions on Word Order in Spanish. PhD
Kachru, B. 1985. Standards, Codification and Sociolinguistic Realism: The English Language
in the Outer Circle. In R. Quirk and H.G. Widdowson (eds.), English in the World: Teaching
and Learning the Language and Literatures: 11-30. Cambridge: Cambridge University Press.
Karttunen, L. 1977. The Syntax and Semantics of Questions. Linguistics and Philosophy 1: 1-
44.
Katz, J. & Postal, P. 1964. An Integrated Theory of Linguistic Descriptions. Cambridge, MA:
MIT Press.
183
Kim, C., Lau, R. & Selvanathan, N. 2008. Optionality in the Lexicon and Singapore English
Seoul, Korea.
Kim, C., Chang, Q., Lau, R., & Selvanathan, N. 2009. Singapore English wh-questions: A
Gap in the Paradigm. Journal of the Southeast Asian Linguistics Society 1: 127-140.
Kim, C. & Wee, L. 2009. Resolving the paradox of Singapore English hor. English World-
Kim, S. 2002. Intervention effects are focus effects. Japanese/Korean Linguistics 10: 615-
628.
Kuo, E. 1985. Language in the Family Domain in Singapore: An Analysis of the 1980 Census
Kuroda, S-Y. 1988. Whether we agree or not: a comparative syntax of English and Japanese.
Kwan-Terry, A. 1991. Through the Looking Glass: A Child’s Use of Particles in Chinese and
English and its Implications on Language Transfer. In Anna Kwan-Terry (ed.), Child
University Press.
Labov, W. 1972. Language in the Inner City. Philadelphia: University of Pennsylvania Press.
Lasnik, H. & Saito, M. 1984. On the Nature of Proper Government. Linguistic Inquiry 15:
235-289.
184
Lasnik, H. & Saito, M. 1992. Move α, Conditions on its application and output. Cambridge,
Lee, N., Ling, A.P., and Nomoto, H. 2009. Colloquial Singapore English got: Functions and
Lim, L. 2007. Mergers and Acquisitions: On the Ages and Origins of Singapore English
Lim, L. & Gisborne, N. 2009. The Typology of Asian Englishes: Setting the Agenda. English
Liu, F. & Yi, X. 2006. Parallel encoding of focus and interrogative meaning in Mandarin
Mathieu, E. 1999. Wh-in-situ and the Intervention Effect. UCL Working Papers in Linguistics
11: 441-472.
Matushansky, O. 2006. Head Movement in Linguistic Theory. Linguistic Inquiry 37: 69-109.
Mufwene, S. 1996. The Founder Principle in Creole Genesis. Diachronica 13 (1): 83-134.
Munaro, N., Poletto, C. & Pollock, J-Y. 2001. Eppur si muove! On comparing French and
185
Obenauer, H. 1994. Aspects de la syntaxe A-barre: Effets d’intervention et mouvements des
Pakir, A. 1991. The Range and Depth of English-Knowing Bilinguals in Singapore. World
Perlmutter, D. 1971. Deep and Surface Structure Constraints in Syntax. Holt, Rinehart and
Pesetsky, D. 1987. Wh-in-situ: Movement and Unselective Binding, in E. Reuland and A. Ter
Pesetsky, D. 1989. Language-particular Processes and the Earliness Principle. Ms., MIT.
Pesetsky, D. & Torrego, E. 2007. The Syntax of Valuation and the Interpretability of
Features, in S. Karimi, V. Samiian and W. Wilkins (eds.) Phrasal and Clausal Architecture:
Pires, A. & Taylor, H. 2007. The Syntax of wh-in-situ and Common Ground. Romance
Platt, J. 1975. The Singapore English Speech Continuum and its Basilect ‘Singlish’ as a
Platt, J. & Ho, M. L. 1993. Dynamics of a contact continuum: Singaporean English. New
186
Platt, J. & Weber, H. 1980. English in Singapore and Malaysia: Status, Features, Functions.
French and some North Eastern Italian dialects. Recherches linguistiques de Vincennes 33:
135-156.
Potsdam, E. 2006. More concealed pseudoclefts in Malagasy and the Clausal Typing
(eds.) Proceedings of the 24th West Coast Conference on Formal Linguistics: 334-342.
Reglero, L. & Ticio, E. 2008. Wh-in-situ and the Spanish DP: Movement or No Movement?
Proceedings of the 31st Annual Penn Linguistics Colloquium Vol. 14, Issue 1: 311-324.
Reintges, C. 2007. Variable pronunciation sites and types of wh-in-situ, in N. Corver and J.
187
Sato, Y. 2008. Wh-in-situ in Bahasa Indonesia and Choice Function. Proceedings of the 2007
Sato, Y. 2013. Wh-questions in Colloquial Singapore English: Adaptive Traits from Malay
and Typological Congruence. Journal of Pidgin and Creole Languages 28: 299-322.
http://www.singstat.gov.sg/docs/default-source/default-document-
library/publications/publications_and_papers/population_and_population_structure/populatio
n2015.pdf
http://www.singstat.gov.sg/docs/default-source/default-document-
library/publications/publications_and_papers/cop2010/census_2010_release1/cop2010sr1.pdf
Stjepanović, S. 2003. A Word Order Paradox Resolved by Copy Deletion at PF, in P. Pica
and J. Rooryck (eds.) Linguistic Variation Yearbook: 139-177. John Benjamins Publishing.
Sybesma, R. 1999. Overt wh-movement in Chinese and the Structure of CP, in H. S. Wang,
F. F. Tsai and C. F. Lien (eds.) Selected Papers from the 5th International Conference of
188
Tan, G-Y. 2011. wh-Movement in Singapore English. Ms., National University of Singapore.
Tan, K. P. 2004. Ethnic representation on Singapore film and television, in A. E. Lai (ed.)
Beyond rituals and riots: Ethnic pluralism and social cohesion in Singapore: 289-315.
Tan, P. & Tan, D. 2008. Attitudes towards non-standard English in Singapore. World
Tongue, R. K. 1979. The English of Singapore and Malaysia 2nd edition. Eastern Universities
Press: Singapore.
Travis, L. 1984. Parameters and Effects of Word Order Variation. PhD Dissertation, MIT.
Tsai, D. 1994. On Economizing the Theory of A-Bar Dependencies. PhD Dissertation, MIT.
Tsai, D. 2003. Lexical Courtesy Revisited: Evidence from Tsou and Seediq wh-
Tsai, D. 2008. Left Periphery and how-why Alternations. Journal of East Asian Linguistics
17: 83-115.
Tuller, L. 1992. The Syntax of Postverbal Focus Constructions in Chadic. Natural Language
189
Villa-García, J. 2014. Constituent Questions in Romance and the Subject-Gap Restriction: A
New Argument for Lower Copy Pronunciation. 44th Linguistic Symposium on Romance
Languages.
Watanabe, A. 2001. Wh-in-situ Languages, in M. Baltin and C. Collins (eds.) The Handbook
725.
L. Schürcks, S. Franks and T. Radev-Bork (eds.) Minimalism and Beyond: Radicalising the
Zocca DeRoma, C. 2011. Divide et Impera – Separating Operators from their Variables.
Zubizarreta, M. L. 1998. Prosody, Focus and Word Order. Cambridge, MA: MIT Press.
papers from the 31st Linguistic Symposium on Romance Languages: 359-380. Amsterdam:
John Benjamins.
190
Appendix
Instructions
Please take a look at the following Colloquial Singapore English (Singlish) sentences and
give your native speaker judgment on them. Instead of using distinctions like
sentence on a 5-point scale. Judging sentences is not an exact science – think of whether (a)
you have heard the particular utterance occurring naturally before, or (b) if you would
Please rate these sentences on a scale of 1 (you don’t think this sentence is acceptable at
Note – the first preliminary round of judgments was elicited from 18 local, undergraduate
students in NUS.
191
Results
Sentences/Correspondents A B C D E F G H I J K L M N O P Q R
Sentences/Correspondents A B C D E F G H I J K L M N O P Q R
Sentences/Correspondents A B C D E F G H I J K L M N O P Q R
192
Judgment Elicitation (Second Round)
Instructions
Please take a look at the following Colloquial Singapore English (Singlish) sentences and
give your native speaker judgment on them. Simply label each sentence as either
“Acceptable” or “Unacceptable” in your own view. Judging sentences is not an exact science
– consider the combination of whether (a) you have heard the particular utterance being
spoken naturally before, or (b) if you would produce the utterance yourself.
Note – the second round of judgments were elicited from 10 local, undergraduate students in
UniSIM.
The sentences are presented here with their corresponding example numbers from the text:
193
b. John want to know Lisa leave when. 0√ 10 X
What is the question asking (you can choose more than one):
194
Supplementary DSQ Examples
Standard English
1. Brent: Unbelieveable. You break into my quarters, hold me captive, wreck the place, and
for what?
2. Greer: You are a smart girl. The simulation is not convincing you. Maybe a little field trip
Sameen: To what?
Greer: To showing you the people you refuse to betray are not the heroes.
Preacher, S01E04
195
5. Daniel Radcliffe: Dylan is where, exactly?
Archer, S07E10
7. Kyle's Dad: Look officers, Why don't you just go ask him why he quit Twitter?
8. Sarah: If I walk in there now, after 10 months, think about how hard she can make custody.
Sarah: Kira.
196
10. Prof Stein: Go talk to him.
11. Mon-El: So we're stuck in the most dangerous planet in the galaxy, we don't have our
Supergirl, S02E09
197
14. Jemma: Pick it up.
Jemma: Cut your wrist. If you're the LMD, then I'll see your substructure and I'll know.
15. CNN News Anchor: Thank you very much, Tom Foreman. A lot of sound and fury, Mark
16.
17.
198
CSE
1.
2.
199
3.
4. Matrix wh-fronting
5. Embedded wh-fronting
200