Content uploaded by Mark W Post
Author content
All content in this area was uploaded by Mark W Post on Nov 01, 2021
Content may be subject to copyright.
Nitipong Pichetpan* and Mark W. Post
Bare classifier phrases in Thai and other
mainland Asian languages: implications for
classifier theory and typology
https://doi.org/10.1515/lingty-2020-2071
Received October 31, 2019; accepted November 14, 2020;
published online February 4, 2021
Abstract: This article provides a comprehensive analysis of the little-known “bare
classifier phrase”construction in Modern Standard Thai. It describes the syntax,
semantics and discourse functions of Thai bare classifier phrases, and further
proposes a diachronic account of their origin in reduction of post-posed numeral
‘one’. Following this synchronic and diachronic description, this article attempts
to locate Thai within a working typology of bare classifier constructions in main-
land Asian languages, and further argues for the importance of bare classifier
constructions to the theory of classifiers more generally. Following Bisang (1999)
and others, it argues that bare classifier constructions reveal the core function of
classifiers in Asian languages to be INDIVIDUATION –a referential function. It therefore
cautions against some recent proposals to merge classifiers and gender markers
within a single categorical space defined on the semantic basis of nominal clas-
sification, and in favour of continuing to treat classifiers as a discrete linguistic
category –in mainland Asian languages, at least.
Keywords: bare classifier phrases; classifiers; Thai grammar; Thai language
1 Introduction
This article has two main goals. The first is to provide a comprehensive analysis of
the little-known bare classifier phrase in Modern Standard Thai (Glottocode
thai1261, henceforth “Thai”). Description of bare classifier phrases is absent from
*Corresponding author: Nitipong Pichetpan [níʔtipʰoŋpʰíʔtɕʰêːtpʰan], Department of Linguistics,
John Woolley Bldg, A20 Science Road, The University of Sydney, Sydney, NSW 2006, Australia; and
Faculty of Liberal Arts (4th floor), Thammasat University (Rangsit campus), 99 Moo 18
Phahonyothin Road, Khlong Nueng, Khlong Luang, Pathum Thani 12121, Thailand,
E-mail: pnitipon@tu.ac.th
Mark W. Post [mɑɹkpʰəʊst], Department of Linguistics, John Woolley Bldg, A20 Science Road,
The University of Sydney, Sydney, NSW 2006, Australia, E-mail: mark.post@sydney.edu.au
Linguistic Typology 2021; 25(3): 461–506
most major reference works on Tai languages, and at least a few authors have
specifically claimed that such a construction does not exist in Thai. This article
therefore aims to improve the record by providing the first comprehensive syn-
tactic, semantic and discourse-functional description of Thai bare classifier
phrases. It also proposes a diachronic account of their origin.
Following on from our comprehensive synchronic and diachronic account of
Thai bare classifier phrases, this article’ssecondgoalistolocateThaiwithina
typology of bare classifier phrases in Asian languages, at the same time arguing
for their importance to the theory of classifiers more generally. Specifically, we
argue that bare classifier phrases come closest to revealing the “core”gram-
matical function of classifiers in Asian languages, which, in agreement with
Bisang (1999) and others, we argue to be INDIVIDUATION –a referential function.
Given this fundamentally referential functionality of classifiers in Asian lan-
guages, we caution against some recent proposals to merge classifiers together
with gender markers within a single categorical space defined on the semantic
basis of nominal classification, and in favour of continuing to treat classifiers as a
discrete linguistic category –in mainland Asian languages, at least.
The remainder of the article has the following structure: Section 2 frames
our analysis by outlining some aspects of the structures of Thai noun phrases,
focusing in particular on the “classifier phrase”constituent (CLFP). Although
we judge our analysis to be consistent with standard analyses of Tai languages
such as Enfield (2004) on Lao (laoo1244) and Iwasaki and Ingkaphirom (2005)
on Thai, since those sources do not generally identify a CLFP constituent in Thai
NP structure, and since we judge CLFPs to be critical to our analysis of Thai
classifier constructions, we have taken some space to develop a detailed
exposition here; readers who wish to proceed directly to our analysis of Thai
bare CLFPs could skip directly to Section 3, referring back to Section 2 only as
needed. In Section 4, we propose an origin of the Thai bare classifier phrase in
reduction of a co-existing [CLF [ART:one]
MOD
]
CLFP
construction, via gramma-
ticalization and phonological reduction. Section 5 then summarizes the
typological and theoretical implications of our analyses, while Section 6
finally outlines a handful of remaining problems.
2 Preliminaries
2.1 Thai noun phrases
Following Iwasaki and Ingkaphirom (2005: 13; 61), we analyse Thai NPs as either
simplex or complex.Asimplex NP contains a single nominal constituent –a“bare
462 Pichetpan and Post
noun”, as in (1).
1
Acomplex NP includes a head noun and one or more modifiers, all
of which strictly follow the head (2). In these examples and elsewhere in this
article, phrasal heads are underlined.
(1) nám-tòk
water-fall
[N]
NP
‘a/the waterfall(s)’
(2) nákrian leːw
student bad
[N MOD]
NP
‘a bad student’or ‘bad students’
According to Iwasaki and Ingkaphirom (2005: 61), seven types of NP modifier are
found in Thai (3). They include: (a) numerals, (b) demonstratives, (c) interrogative
and indefinite modifiers, (e) genitive phrases, (f) adjectives (g) prepositional phrases,
and (h) relative clauses:
(3) a. tʰoːrásàp sǎːmkʰrɯ
̂aŋ
2
telephone three CLF:machine
[N[NUM CLF]
MOD
]
NP
‘three telephones’
b. rót (kʰan) níː
car CLF:handle PRX
[N[(CLF)DEM]
MOD
]
NP
‘this car’
c. nǎŋsɯ
̌ː(lêm) nǎj
book CLF:sheet WHICH
[N[(CLF)INT]
MOD
]
NP
‘which book?/any book’
d. nǎŋsɯ
̌ːlêm nɯ
̀ŋ
book CLF:volume ART:one
[N[CLF ART:one]
MOD
]
NP
‘a book’
1We do not here address the more difficult matter of whether “bare nouns”are best considered NPs
or simply nominals at clause level, this being beyond the scope of the present article (cf. Has-
pelmath 2017 and Visonyanggoon 2000).
2Here and in other examples in this article, we adopt a convention of bolding text under focus. In
these examples, we have bolded NP modifiers.
Bare classifier phrases in Thai 463
e. bâːn(kʰɔ
̌ːŋ)pʰɔ̂ː
house POSS father
[N[POSS N]
MOD
]
NP
‘the father’s house(s)’
f. mǎː(tua) jàj
dog CLF:body big
[N[(CLF)ADJ]
MOD
]
NP
‘the big dog(s)’
g. nǎŋsɯ
̌ː(lêm) bon tóʔ
book CLF:volume PREP table
[N[(CLF)PREP N]
MOD
]
NP
‘the book on the table’
h. nǎŋsɯ
̌ː(lêm) tʰîːpʰɔ̂ːʔàːn
book CLF:volume REL father read
[N[(CLF)RELC]
MOD
]
NP
‘the book(s) that (the) father read’
Setting aside the somewhat different structure of (3e), we note that the modifying
constituents in (3a–d) and (3f–h) all involve obligatory or optional use of a clas-
sifier. We also note that in all of (3a–h), the head noun may be ellipsed under
conditions of contextual retrievability; only a single such example is provided in
the interest of space (4).
(4) Ø sǎːmkʰrɯ
̂aŋ
(telephone) three CLF:machine
[(N)[NUM CLF]
MOD
]
NP
‘three (telephones)’
In addition to some other facts regarding movement and ordering conventions
(discussed in Section 2.2), head-ellipsability under all conditions in which a
classifier co-occurs argues in favour of identifying the post-nominal bracketed
phrases in (3a–d) and (3f–h) as constituting a single phrasal constituent within a
Thai NP, headed by a classifier; we thus label this constituent a classifier phrase
(CLFP).
3
We might similarly consider the post-nominal bracketed phrase in (3e) to
constitute a possessive phrase, however we will not pursue the idea further in this
article.
While we do not have space here to pursue the matter in detail, we should note
that a Thai CLFP as we have defined it seems qualitatively different from the
3In a similar vein, Gerner (2013: 63) identifies a CL′constituent in his description of Nuosu Yi, as a
“unit greater than a bare classifier but smaller than the whole NP.”
464 Pichetpan and Post
“classifier phrase”constituents identified elsewhere in the literature; for example,
in Cheng and Sybesma (1999) and Jenks (2011). In this literature, a “classifier
phrase”, while it is also headed by a classifier, is posited as standing at a higher
level than NP. It thus either replaces a DP (as in Cheng and Sybesma 1999) or stands
as complement to null D(as in Jenks 2011). Although we appreciate the thrust of
these analyses, which recognize and attempt to explain the same referential
functions of classifiers that we will also identify, the resulting syntactic analysis is
quite different. This topic, while interesting in its own right, falls well outside the
scope of our present enquiry.
2.2 Thai classifier phrases
Examples (3a–h) suggest that Thai CLFPs can have a variety of structures. The head
may be obligatory, as in (3a), or “optional”(= ellipsable), as in (3b). CLFP-internal
modifiers may be simplex, consisting of a single word, or complex, consisting of a
phrase or clause. Finally, while most modifiers follow heads, in only one case is
there a pre-head modifier (3a). Given these different ordering possibilities, we can
then divide classifier phrases into pre-head and post-head types: [MOD CLF] and [CLF
MOD] (cf. Haas 1942; Hundius and Kölver 1983; Jones 1970). Leaving aside (3d)
momentarily, we note that post-head structures are generally associated with
referential qualification, by which we mean the act of somehow qualifying the
referential value of an NP. The single pre-head structure involves a quantifier (nu-
meral or other type of quantifying expression); we can thus associate pre-head
modifiers of a Thai classifier with quantification.
4
An interesting complication surrounds the numeral nɯ
̀ŋ‘one’, which has often
been noted as occurring both before and after classifiers in Thai; no other Thai
numeral has this ability (Bisang 1999; Haas 1942; Hundius and Kölver 1983; Li
1978): see (5)–(6).
(5) (kʰǎw) …kʰwáːkɛ
̂ːwnɯ
̀ŋbaj ʔɔ
̀ːkmaː
(3SG)…grab glass one CLF:leaf DIR:out ASP:come
[N[MOD CLF]
CLFP
]
NP
‘He grabbed (just) one glass out (despite there having been one or more
other glasses there).’(K. Surangkhanang, Mandated Love, 1988 [novel])
4Iwasaki and Ingkhaphirom (2005: 61–62; see also Gebhardt 2009) observe that numerals and
non-numerical quantifiers such as baːŋ‘some’,tʰúk ‘every’,orlǎːj‘many’are syntactically alike.
We assume here that they form a single macro-category in Thai, with numerals referring to precise
quantities and non-numerical quantifiers to imprecise quantities.
Bare classifier phrases in Thai 465
(6) tɕʰaːjnùm …jìp kɛ
̂ːw baj nɯ
̀ŋʔɔ
̀ːkmaː
young.man …pick glass CLF:leaf ART:one DIR:out ASP:come
[N[CLF MOD]
CLFP
]
NP
‘The young man took out a glass (regardless of whether there were or were
not other glasses there.’(Chingchio, Seven Rings, 2016 [novel])
While they derive from the same lexeme, pre-head and post-head nɯ
̀ŋare (in
modern Thai) functionally distinct. Pre-head nɯ
̀ŋfunctions as a numerical
quantifier, specifying a referent’s quantity as ‘one’as opposed to three, four, etc.
However, post-head nɯ
̀ŋmarks a referent as indefinite. In this function, erstwhile
nɯ
̀ŋ‘one’is often toneless, even in orthographic mode (see Section 4.2.1); this is a
likely indication of its grammaticalization. The tone of nɯ
̀ŋcannot be similarly
omitted in pre-head position (7)–(8).
(7) *(kʰǎw) …kʰwáːkɛ̂ːwnɯŋbaj ʔɔ
̀ːkmaː
(3SG)…grab glass ART:one CLF:leaf DIR:out ASP:come
[N[MOD CLF]
CLFP
]
NP
(8) tɕʰaːjnùm …jìp kɛ̂ːw baj nɯŋʔɔ
̀ːkmaː
young.man …pick glass CLF:leaf ART:one DIR:out ASP:come
[N[CLF MOD]
CLFP
]
NP
‘The young man took out a glass.’
While we might not go so far as to describe post-head nɯ
̀ŋas a fully-fledged
indefinite article, it is playing a very similar functional role; we will henceforth
gloss nɯ
̀ŋas ‘ART:one’when in this function, to signal both its functional
specialization and its continuing lexical association with its source form. Further
discussion of this grammaticalization pathway will be found in Section 4.
5
We thus consider the two patterns [MOD CLF] and [CLF MOD] to be prototypical in
the patterning of Thai classifier phrases, and for this syntactic difference to
fundamentally encode the distinction between QUANTIFICATION (stating a quantity of
referents) and REFERENTIAL QUALIFICATION (qualifying the referential value of an NP,as
given, new, specific, generic, and/or possessing some set of attributes).
5The close continuing formal and semantic resemblance of Thai functional morphemes to their
lexical sources has often been remarked upon (Diller 2001; Morey 2005; Post 2007b, among
others), and presents a persistent representational challenge –in Thai as in most other Mainland
Southeast Asian languages. In this work, we adopt the hybrid approach of glossing such functional
morphemes both in relation to their lexical sources and in relation to the larger functional category
in which they play functional roles. For another example, ma: ‘ASP:come’is glossed as an aspectual
operator whose core semantic value derives from its lexical source, a serialized verb meaning
‘come’.
466 Pichetpan and Post
It is important to note that both pre-head and post-head CLFPs can be “stacked”
in Thai within a single NP –and even within the same NP. In such cases, modifi-
cation has leftward scope (9).
6
(9) tapʰâːptua jàj lǎːj tua
softshell.turtle CLF big INDF.NUM:many CLF
[N[CLF MOD]
CLFP
[MOD CLF]
CLFP
]
NP
‘the many big softshell turtles’(https://tinyurl.com/y9ax336y, [web content])
Thai CLFPs can typically occur in three types of relationship with the head noun of
the NP in which they are notionally a constituent: (a) continuous (b) discontinuous
and (c) ellipsed-head.Inacontinuous relationship, a CLFP directly follows the
modified head noun (cf. (5) and (6) above). In a discontinuous relationship, the CLFP
does not follow a noun head immediately; instead, we find insertion of an inter-
vening element such as an aspectual marker. This type might be schematized as
[N]
NP
i
…[CLFP]
NP
i
(in which
i
symbolizes the distributed structure of a single NP);
(10)–(11) illustrate the occurrence of discontinuous NPs with both post-head and
pre-head CLFP structures.
(10) dâj plaː-krabeːnmaːtua nɯ
̀ŋ
get fish-ray ASP:come CLF:body ART:one
[N]
NPi
…[[CLF MOD]
CLFP
]
NPi
‘He got a ray (fish).’(Google.com, June 20, 2019 [web content])
(11) lɯ
̂ak plaːmaːnɯ
̀ŋtua
choose fish ASP:come one CLF:body
[N]
NPi
…[[MOD CLF]
CLFP
]
NPi
‘Choose one fish.’(Google.com, June 20, 2019 [web content])
Finally, a “head-ellipsed”CLFP may “modify”an ellipsed nominal head, in the
constructions [∅[CLF MOD]
CLFP
]
NP
or [∅[MOD CLF]
CLFP
]
NP
(in Thai, an NP head is generally
ellipsible when the designated referent is understood from context; cf. Yates and
Tryon 1970) (12)–(13):
6An anonymous reviewer suggests that modification here could be “intersective”. This does not
seem to be the case, inasmuch as a change in order yields a different meaning (in tapʰâːptuːa jàj lǎːj
tuːa, all the ‘many’turtles must be ‘big’, whereas tapʰâːplǎːjtuːatuːa jàj selects a subset of ‘many’
turtles which are ‘big’, implying a non-selected subset of turtles which are ‘small’). This implies
gradual leftward modification.
Bare classifier phrases in Thai 467
(12) miːpòːtkàːdmáj …kʰɔ
̌ː∅baj nɯŋ
have postcard INT …ask.for (N)CLF:leaf ART:one
[N]
i
[[∅
N
]
i
[CLF MOD]]
NP
‘Do you have postcards? Can I have one?’(adapted from Google.com, June
20, 2019 [web post])
(13) kʰuːpɔːŋmiːlɯ
̌amǎj…kʰɔ
̌ː∅nɯ
̀ŋbaj
coupon have surplus INT …ask.for ∅one CLF:leaf
[N][[∅
N
][MOD CLF]]
NP
‘Are there coupons available? Can I have (just) one?’(adapted from
Google.com, June 20, 2019 [Facebook comment])
Since they occur in both pre-head and post-head CLFP configurations, and since
they occur with all numerals in these functions, we consider the basic capacity for
Thai CLFPs to occur in continuous, discontinuous and head-ellipsed relationships to
an NP head to be prototypical. However, not all CLFP types appear to have equal
access to each type of construction. For example, [CLF DEM], [CLF RELC:tʰîː-clause] and
perhaps also [CLF ADJ] types do not generally occur in discontinuous NP construc-
tions: (14)–(16).
(14) *kʰáːw dâj plaːmaːtua nán
3SG get fish ASP:come CLF:body DEM:that
[[N]…[CLF MOD]]
NP
(intended: ‘He got that fish.’)
(15) *kʰǎwsɯ
́ːplaːmaːtua tʰîːnàk sìp kiloː
3SG buy fish ASP:come CLF:body REL weigh ten kilogram
[[N…[CLF RELC]]
NP
(intended: ‘He bought the fish that weighed 10 kilograms.’)
(16) ?kʰáːw dâj plaːmaːtua sǐː-dam
3SG get fish ASP:come CLF:body ADJ:colour-black
[[N]…[CLF MOD]]
NP
(intended: ‘He got the black fish.’)
It is therefore possible that CLFPs could be further subclassified on this basis; we will
return to this question in Section 3.2.
Returning now to the functions of Thai CLFPs, we follow several authors (cited
below) in assigning the functions shown in Table 1 with the corresponding set of
Thai CLFP types.
With regard to the pre-head macro-type, scholars of Thai generally agree that
it expresses QUANTIFICATION (e.g. Bisang 1999; Haas 1942; Hundius and Kölver 1983;
468 Pichetpan and Post
Placzek 1978; Singnoi 2008; Uppakitsinlapasan 2002).
8
Haas (1942: 203) appears
to have been the firsttopointoutthatThaiclassifiers in the (N)[NUM CLF] pattern
are similar to English numeratives –both are employed in the context of
counting. Correspondingly, Uppakitsinlapasan (2002: 72–73) insists that lák-
saʔnàʔnaːm(classifiers) are compulsory after pràʔmaːnawíʔs`
eːt(adjectives for
counting, or number words) such as sɔ
̌ːŋ‘two’.Pre-
CLF numerals and classifiers
constitute pakatìʔsǎŋkʰajǎːwáliː(cardinal numeral phrases), and are used to
modify nominal heads in the context of quantification.
Yet certain researchers have proposed a revision to this framework, suggesting
that a classifier used in the context of numerical quantification actually involves
distinct functions of CLASSIFICATION and INDIVIDUALIZATION (or INDIVIDUATION, in Hundius
and Kölver’s terms). The CLASSIFICATION function serves to define countables as
members of a type-set, whereas INDIVIDUATION licenses construal of classified refer-
ents as individuals and, therefore, as enumerable (Bisang 1999; Hundius and
Kölver 1983; Placzek 1978). More recently, Singnoi (2008) has called into question
the “classificatory”function as such, given that this “function”, if it is one, is often
opaque. For example, the classifier lêm can be applied to a diverse set of objects,
Table :Functions of CLFP types in Thai.
Macro-type Sub-type Sub-functions Macro-functions
PRE-HEAD [NUM CLF]INDIVIDUATIVE QUANTIFICATION QUANTIFICATION
POST-HEAD [CLF ATTR]
CONTRASTIVE FOCUS
REFERENTIAL CONNOTATION
ATTRIBUTIVE MODIFICATION
REFERENTIAL QUALIFICATION
[CLF RELC]REFERENT CHARACTERIZATION
REFERENTIAL SPECIFICATION
[CLF DEM]DEFINITE REFERENTIAL SPECIFICATION
[CLF ART:one] INDEFINITE REFERENTIAL SPECIFICATION
7Post-head predicative adnominal modifiers which lack an intervening linker behave similarly in
Thai; they include ADJ,V, and gapped relative clause-like structures, and are described as
“attributive clauses”by Post (2008: 356–358). We use ATTR as a convenient shorthand for such
modifiers in this study.
8Classifiers are sometimes divided into sortal and mensural subtypes, while other scholars have
associated the category of “classifiers”with the sortal subtype only, terming mensural classifiers
as “measure words”(Hundius and Kölver 1983; Bisang 1999). We follow Haas (1942; cf. also
Uppakitsinlapasan 2002; Singnoi 2008) in using the term “classifier”to encompass both sortal and
mensural classifiers in Thai. We do so because both sub-classes commonly occur in the same
position in a CLFP, albeit with slight syntactic differences in some expansive constructions (cf. Jones
1970).
Bare classifier phrases in Thai 469
including ‘books’,‘carts’, and ‘knives’, each of which corresponds to a different
semantic class (Singnoi 2008: 83). Clearly, there was an earlier meaning of lêm that
lent itself to use as a classifier in each of these cases, through semantic extension;
however, this is best understood as the outcome of exactly such a process: se-
mantic extension and historical change; not as a case of synchronic “classifica-
tion”per se.
As for INDIVIDUATION, this function seems to be more clearly justified in the case
of Thai classifiers. As the above-cited authors have already noted, Thai nominals
inherently lack any specification of number. Accordingly, and since construal of an
enumerated quantity presupposes the existence of a set of enumerable in-
dividuals, the obligatory inclusion of a classifier in this construction would argue
in favour of the classifier’s fulfilling exactly this functional need. This being the
case, we would argue that in the case of Thai pre-head CLFPs, it is the numeral
component that expresses the function of QUANTIFICATION per se, whereas the role of
the classifier is to express an entity’sINDIVIDUATION. In other words, pre-head CLFPsdo
not simply express QUANTIFICATION in the sense of enumeration or measurement of
referents per se, but instead entail both INDIVIDUATION and QUANTIFICATION.
Regarding post-head CLFP patterns, the association of classifiers with referen-
tial functions is here even more apparent. For example, when an adjective alone is
used to modify a head noun in Thai, it serves a purely attributive function and is
referentially neutral (Bisang 1999). The addition of a classifier to this construction
was accordingly taken by some authors (e.g. Hundius and Kölver 1983) as
amounting to REFERENTIAL CONNOTATION. However, the pattern [N[CLF ADJ]] is in fact
distinguished from [N[ADJ]] in Thai by virtue of the former’scontrastive value: what
is being profiled when a classifier is used is an entity or entities with a specific value
as opposed to entities of the same kind with another range of values (Bisang 1999;
Hundius and Kölver 1983). For example, contrast rôm sǐː-kʰǐaw ‘umbrella green’
with rôm kʰan sǐː-kʰǐaw ‘umbrella CLF:handle green’. In the first case, what is
denoted is an abstract class of umbrellas with a property of greenness. In the
second case, the phrase contrasts some set of individually green umbrellas with
some other set whose members may have one or more different colours.
Unlike the [CLF ADJ] pattern, the [CLF DEM] construction is very close to counting as a
definiteness marking construction. However, it is notable that while a bare demon-
strative entails both DEICTIC and DEFINITENESS functions, it remains unspecified for NUMBER in
Thai. To force a singular reading, it is necessary to use a classifier. Consider (17)–(18).
(17) rôm níː
umbrella PRX
[NDEM]
NP
‘this/these umbrella(s)’(Hundius and Kölver 1983: 172)
470 Pichetpan and Post
(18) rôm kʰan níː
umbrella CLF:handle PRX
[N[CLF DEM]]
NP
‘this umbrella’(Hundius and Kölver 1983: 172)
Once again, we see here a close association of classifiers with INDIVIDUATION. Yet we
must also clearly state that INDIVIDUATION, while it may be prerequisite to a singular
reference in Thai, is not equivalent to singularity per se. We can clearly see this
from cases in which additional context biases or forces a plural interpretation of a
classifier phrase (19).
(19) rót kʰan sǐː-khǎːw tem paj mòt
vehicle CLF:handle colour-white full ASP:go complete
‘(The place) was full of white cars.’
Finally, as we briefly mentioned above, the [CLF ART:one] pattern is generally
considered to mark INDEFINITE REFERENCE in Thai (Bisang 1999; Haas 1942; Hundius
and Kölver 1983). In this referential function –in contrast to the pre-head QUANTI-
FICATION function –the erstwhile numeral may occur either stressed and (accord-
ingly) carrying a tone, or unstressed and toneless (20)–(21).
(20) rôm kʰan nɯ
̀ŋ
umbrella CLF:handle ART:one:stressed
[N[CLF ART:one]]
NP
‘an umbrella’
(Hundius and Kölver 1983: 180)
(21) rôm kʰan nɯŋ
umbrella CLF:handle ART:one:unstressed
[N[CLF ART:one]]
NP
‘an umbrella’
Note that it is not usually possible for a Thai noun to be directly modified by nɯ
̀ŋ.
9
Thus, while it is difficult to assess the independent functional contributions of CLF
and ‘one’in this construction (as was possible with adjectives and de-
monstratives), we suppose that these functions are likely to be INDIVIDUATION and
SINGULARITY, respectively. We will return to this matter in Section 4, following our
discussion of bare CLFPs in Section 3.
9Some exceptions exist in poetic language, listing, and certain colloquial usages.
Bare classifier phrases in Thai 471
3 Bare CLFPS in Thai: structure, functions, and
distribution
This section aims to provide a comprehensive description of Thai bare CLFPs. We
begin with a discussion of related literature in Section 3.1, where data sources are
also discussed. Basic structures are presented in Section 3.2, while Section 3.3
focuses on discourse distribution and semantico-pragmatic functions.
3.1 Previous descriptions and data for this study
Thai bare CLFPs have received an inadequate characterization in the literature to
date. At least a few authors go so far as to deny their existence; for example,
Hundius and Kölver (1983) claim that bare CLFPs constitute ill-formed utterances in
Thai, and provide the following “unacceptable”example (22).
(22) *rôm kʰan
umbrella CLF:handle
[N[CLF]]
NP
(Hundius and Kölver 1983: 172)
Solnit (1994) also doubts the existence of bare CLFPs in at least some Thai varieties.
According to him, if bare CLFPs are used at all in “Standard Thai”, they are infre-
quent.
10
More recently, the putative absence of bare CLFP in Thai played a key role in
Jenks’(2011, 2018) analysis of Thai referentiality and NP structure.
These analyses all appear to suffer either from their reliance on context-free
elicitation, in which NPs are analysed independently of their larger clause struc-
tures and discourse contexts, and/or in which the variety of Thai under exami-
nation is an idealized construct. Yet both spoken and written data collected for this
study provide ample support for our quite different intuition that bare CLFPs are in
fact well-established in modern vernacular Standard Thai, defined broadly as “the
language spoken and written on a daily basis by the overwhelming majority of
10 Since Solnit does not clearly define his “Standard Thai”, we are not certain of the extent to
which any of the various spoken and written varieties of Thai language should be understood as
encompassed by his conception. If we were to speculate, it would seem that Solnit’s“Standard
Thai”might refer to the “officialized”language of administration, mainstream media, and
schooling, which is generally considered “historically or pedagogically or socially correct”
(Bickner and Hudak 1990: 163). It could perhaps be considered that rigid exclusion of less “offi-
cialized”registers of Standard Thai could result in a corpus that might support Solnit’s claim
regarding the unavailability/infrequency of bare CLFPs in Thai.
472 Pichetpan and Post
native Thai speakers in Thailand, as well as in the global Thai-speaking diaspora
today”. While Standard Thai (thus defined) is traditionally identified with the
Central Thai dialects, in the modern era it has spread far beyond this context, and
in fact the overwhelming majority of Thai language used on the internet and in
social media –including within the increasingly large global Thai-speaking
diaspora, whose population is by now well into the millions –can be identified as
“modern vernacular Standard Thai”. It is in fact quite easy to find examples of the
bare CLF construction in such contexts, all of which strike both the first named
author (who is a native speaker of Standard Thai) and our native Thai-speaking
consultants as straightforwardly well-formed utterances. We will first provide a
few such examples, drawn from a variety of spoken and written sources (23)–(25):
(23) ʔaːtɕaːn…kʰɔ
̌ːmamûaŋlûːknáʔ
teacher …ask.for mango CLF:fruit EMP
[N[[CLF]
CLFP
]
NP
‘Teacher! May I have a mango?’(Luangpho Noi, 2019, April 27 [audio
file])
11
(24) pen pʰɔ̂ːmɛ̂ː…miːkʰâːwtɕaːn
be father mother …have rice CLF:plate
[N[[CLF]
CLFP
]
NP
kɔ̂ːtɔ̂ːŋhâj lûːk kin ʔìm kɔ
̀ːn
SFOC AUX give child eat full ADV:first
‘As parents, (if) there is a plate of rice, they must let the children have it
first.’
(Th.answer.yahoo.com, 2019 [reply to blog post])
(25) sùːpburǐːjîːhɔ̂ːʔaraj kʰráp kʰɔ
̌ː∅muan sìʔkʰráp
smoke cigarette brand what POL ask.for ∅CLF:roll EMP POL
[N]
NPi
[[∅]
N
[[CLF]
CLFP
]
NPi
‘What cigarette are you smoking? Would you give me one?’(Google.com,
2019 [reply to blog post])
The abovementioned authors’scepticism regarding the well-formedness of bare
CLFPs notwithstanding, at least some other authors –all of them, seemingly, native
speakers of Thai –have actually recognized the use of bare CLFPs in Thai. Two of
the –albeit few –examples of bare CLF constructions which have been published in
the linguistics literature are given in (26)–(27) (cf. also Singnoi 2008).
11 This attested example may be directly compared with the putatively “ungrammatical”example
in Jenks (2011: 232).
Bare classifier phrases in Thai 473
(26) pàːkkaːdâːm
pen CLF:handle
[[N][CLF]
CLFP
]
NP
‘A pen!’(Aroonmanakun 1994: 20)
(27) kʰɔ
̌ːmîːt lêm síʔ
ask.for knife CLF:sheet EMP
[[N][CLF]
CLFP
]
NP
‘Give me a knife.’(Panupong 1995: 68)
Readers with knowledge of Thai language are likely to have an immediate reaction
that, of the above examples (26)–(27), (27) strikes one as “better sounding”than (26).
We believe that this is because of the additional context (i.e., the more elaborate clause
structure) provided in (27); as we will show in Section 3.3 below, Thai bare CLF con-
structions are biased toward certain clause structures and discourse contexts, and can
sound “ungrammatical”in the absence of such a contextual bias.
For the purpose of the present study, we assembled data from a diverse range of
sources, including (a) searches of digital corpora comprising multiple anonymized
authors, speech genres and texts, (b) searches of written texts such as Thai popular
novels and periodicals, and (c) a survey of consultant acceptability judgements.
Concerning (a), our digital corpus searches were “opportunistic”,andweredesigned
to elicit both (i) actual occurrences of the bare CLF construction and (ii) information
regarding syntactic context and discourse function. For this purpose, we primarily
employed Google.com, through which both web content and social media postings
were accessed. We also queried the SEAlang Library Thai Text Corpus (n.d.)
12
as well
as the Thai National Corpus (TNC) (n.d.).
13
However, data illustrating bare CLFP usage in
this article are never derived from those queries; since texts in these corpora are drawn
from academic books, newspapers, and “highbrow”literature, bare CLFPstendnotto
appear in these corpora. However, we nevertheless employed both corpora to obtain
examples of related structures, as well as to construct example sentences for the
questionnaire to be discussed in Section 3.3.3.
We should emphasize here that although our study is based primarily on
naturally-occurring Thai language data, it is not intended as a corpus study per se.
For example, corpus-linguistic methods such as abstractions and statistically
probing analyses were not applied, and we did not establish a sampling method or
employ extensive controls for text type, genre, relative frequency, etc. Instead, the
databases we searched were exploited flexibly to derive basic information
12 Retrieved from http://sealang.net/thai/corpus.htm.
13 Retrieved from http://www.arts.chula.ac.th/∼ling/tnc3/.
474 Pichetpan and Post
regarding the existence, structure, co-occurrence possibilities and functional do-
mains of bare CLFPs in Thai.
3.2 Basic structure of Thai bare CLFPS within a Thai NP:[(N)[CLF]
CLFP
]
NP
We can now situate Thai bare CLFPs within the overall structure of a Thai NP as
follows: [(N)[CLF]
CLFP
]
NP
. As shown, a Thai bare CLFP must contain a single overt
classifier head, and must modify a single logical NP head; however, the NP head may
or may not be syntactically overt.
Like other CLFP types, Thai bare CLFPs can occur in either continuous,discon-
tinuous or head-ellipsed relationships to an NP head (28)–(30).
(28) kʰɔ
̌ːmîːt lêm síʔ
ask.for knife CLF:sheet EMP
[[N][CLF]
CLFP
]
NP
‘Give me a knife.’(Panupong 1995; repeated from (27) above for convenience)
(29) kʰáːw dâj sɯ
̂ajɯ
̂ːtmaːtua
3SG get T-shirt ASP:come CLF:body
[[N]…[CLF]
CLFP
]
NP
‘He got a T-shirt.’(adapted from Instagram [post content])
(30) miːnaːmbàt mǎj|kʰɔ
̌ː∅baj siː
have business.card INT |ask.for (N)CLF:leaf EMP
[N][[∅
N
][CLF]
CLFP
]
NP
‘Do you have a business card? Can I have one?’(adapted from Sakunthai
[novel excerpt])
The capacity of bare CLFPs to occur in all three CLFP constructional subtypes is important
to their analysis. First, it indicates that bare CLFPs exhibit the same basic structural
potential of CLFPs in Thai more generally. Second, it aligns bare CLFPswithpre-andpost-
head numerical modifiers of a CLFP head –which also occur in continuous, discontinuous
and head-ellipsed structural types –and not with other CLFP types, such as those whose
MOD constituent is an demonstrative or adjective; we return to this point in Section 4.
3.3 Distribution and functions of Thai bare CLFPS
This section discusses the distribution and functions of Thai bare CLFPs. We will
argue that discourse distribution and frequency of bare CLFPs are more restricted
Bare classifier phrases in Thai 475
than other CLFP types, in line with a relatively marked set of functions. Specifically,
we will find that bare CLFPs typically entail reference to new individuals (non-
specific, and/or not-previously-introduced-in-discourse), and tend to occur in
constructions dedicated to such functions.
3.3.1 Discourse distribution
In Thai, variability in the form and distribution of classifiers can be correlated with
differences in linguistic registers. For example, Dixon (1977) notes that some Thai
classifiers take different forms in formal and informal registers. Similarly, specific
classifiers are often associated with higher speech registers; more general classi-
fiers seem more suitable for less formal speech.
Simi lar observations hold in the case of bare CLFPs. Notably, unlike other CLFP types,
bare CLFPs tend to be restricted to less formal registers in usage: namely, those of
vernacular speech or speech-like writing. Consider again the examples of bare CLFPs
presented in (23)–(25) above. These examples were drawn from casual conversations as
represented in popular novels, or else from speech-like digital texts from online social
networks like Facebook or Twitter (as accessed via a web search). We must therefore
consider a possibility that bare CLFPs could simply be a sociolinguistic variant associated
with informal Thai speech or younger speakers.
However, we have found that bare CLFPs are moreover likely to be associated with
particular discourse functions, such as requests or commands, and that these functions
are simply more likely to occur in vernacular speech and speech-like writing than in
more formal genres. For example, in the Thai popular novels we studied, bare CLFPsare
usually found when a character makes a request or issues a command, such as in (30)
above. Use of bare CLFPs in declarative statements is also possible; for example, the
constructed declarative statement in (29) above was accepted by our consultants.
Nevertheless, such examples seem to be much less frequent in natural discourse (or at
least, the majority of examples that we discovered were not of this type).
Another difference between bare CLFPs and other CLFP typesisthat,unlikeother
types, we have not been able to attest a case in which a bare CLFP occurs in pre-verbal
position; constructed sentences such as (31) are rejected by our consultants.
(31) *tɕʰáːŋtua miːsɔ
̌ːŋŋaː
elephant CLF:body have two tusk
*[[N][CLF]
CLFP
]
NP
(intended: ‘An elephant has two tusks.’)
These facts suggest that the distributional biases we have observed regarding bare
CLFPs do not reflect the sociolinguistics of usage but are actually conditioned by the
functions of the construction itself. We will now elucidate those functions.
476 Pichetpan and Post
3.3.2 Functions of bare CLFPS
In this sub-section, we argue that Thai bare CLFPs are most often found in con-
structions associated with INDEFINITE REFERENCE. Namely, they tend to be found in
existential-presentative constructions, as well as in (other) references to specificand
non-specificmentions of new (not previously-introduced) referents in a discourse.
3.3.2.1 Presentative reference
Bare CLFPs can be used in existential presentation, as when introducing a new
referent into a discourse. Accordingly, NPs which contain a bare CLFP frequently
occupy a position in the clause that is associated with information that is new and
under focus (Givón 1993: 63; Lambrecht 1994: 262). In Thai, focused NPs may follow
miː‘have; exist’in two existential sub-constructions: [miː[NP]] (32) and [miː[NP [VP]]]
(33). Note that in both of these examples, the bare CLFP kɛ̂ːw‘CLF:glass’supports
introduction of a new referent náːm‘water’.
(32) miːnáːmkɛ
̂ːwbon tóʔ
have/exist water CLF:glass PREP:on table
[[N][CLF]
CLFP
]
NP
‘There is a glass of water on the table.’
(33) miːnáːmkɛ
̂ːwjùːnaj pʰâːp
have/exist water CLF:glass be PREP:in painting
[[N][CLF]
CLFP
]
NP
‘There is a glass of water in the painting.’(Google.com, online [web
content])
As these examples suggest, NPs modified via a bare CLFP in an existential presen-
tative construction tend to be associated with specific indefinite reference (see
Section 3.3.2.2). Once established, the item can then be tracked in later mentions,
as in (34): here, the NP náːmkɛ̂ːw‘a glass of water’has a specific reading as a newly
introduced referent, and can be reintroduced in the following clause with the
anaphoric pronoun man ‘3SG’.
(34) miːnáːmkɛ
̂ːwbon tóʔ|ʔaw man
have/exist water CLF:glass PREP:on table |take 3SG
[[N][CLF]
CLFP
]
NPi
[PRO]
NPi
maːhâj nɔ
̀ːj
ASP:come give EMP
‘There is a glass of water on the table. Take it for me please.’
Bare classifier phrases in Thai 477
3.3.2.2 Specific indefinite reference
As was briefly mentioned in Section 3.3.2.1, Thai bare CLFPs can entail specific
indefinite reference, indicating an item which is known to the speaker but assumed
to be unknown to the hearer. NPs containing a specifically-referential bare CLFPs are
always singular. This finding is consistent with Piriyawiboon’s (2010: 68) account
of Thai classifiers more generally, in which she claims that only classifiers can
enable denotation of both specificity and singularity in Thai. Consider examples
(35)–(36): in (35), the bolded object NP contains only a bare noun, and is unspecified
for both referentiality and number; based on the context of interpretation, it could
either be the case that the speaker has one or more specific houses in mind, or not.
However, in (36) the addition of a bare CLFP renders the expression specifically and
singularly referential. Here, the speaker knows which individual entity is being
referred to (though the addressee is assumed not to).
(35) piːtʰîːlɛ
́ːwkʰǎwsɯ
́ːbâːn
year last 3SG buy house
[N]
NP
‘He bought a/some house/houses last year.’
(36) piːtʰîːlɛ
́ːwkʰǎwsɯ
́ːbâːnlǎŋ
14
year last 3SG buy house CLF:built.structure
[[N][CLF]
CLFP
]
NP
‘He bought a (specific) house last year.’
3.3.2.3 Non-specific reference
Bare CLFPs may also be used in non-specificreference to individuals, wherein a
referent’s identity is unknown or irrelevant to both speaker and addressee. Such
cases are common in requests/commands, which is to be expected in light of the
14 An anonymous reviewer suggests that the NP:bâːnlǎŋ ‘a certain house’in (36) could be scopally
ambiguous as either specific or non-specific, with a continuation like ‘I went there once’in prin-
ciple able to disambiguate the expression (similarly in the following example (37)). While we agree
that a continuation would indeed condition such a reading, Thai native speakers consulted for this
study nonetheless expressed a strong epistemic preference for a specific reading. The reverse was
true in the case of (37); our consultants were strongly biased toward a non-specific reading, and
considered a continuation sentence to force a specific reading to be odd. Therefore, while we
concede that it may be possible to sustain an analysis in which bare CLFPs are considered to be
scopally ambiguous, we find that the two functions are well enough distinguished epistemically by
Thai speakers in particular constructional contexts that it is worth the effort to distinguish these
functions analytically.
478 Pichetpan and Post
Gricean Cooperative Principle.
15
For example, consider again (25) above. In this
example, the speaker does not specify a particular cigarette, but leaves selection of
a specific individual up to the addressee.
Although such uses in request/command-type constructions are common,
bare CLFPs can also be found to make non-specific references in declarative con-
texts –especially, perhaps, in the irrealis mood. In (37), the speaker does not have a
particular glass of water in mind, and is in fact not committed to the actual exis-
tence of such a glass of water. Accordingly, the bare CLFP can be considered to here
indicate non-specific reference.
(37) kʰǎwjàːkdɯ
̀ːmnáːmkɛ
̂ːw
3SG want drink water CLF:glass
[[N][CLF]]
NP
‘He wants to drink a glass of water.’
(constructed example based on questionnaire described in Section 3.3.3)
It is worth noting here that in competition with bare CLFP constructions, two other
nominal/NP structures are also used to introduce new referents in presentational
sentences, to refer to specific referents, and to refer to undefined individuals in
Thai. They are NPs with [CLF ART:one] (Bisang 1999; see also Chen 2003, for Man-
darin; Heine 1997, for indefinites) and bare nouns (Hundius and Kölver 1983;
Visonyanggoon 2000). Though alike with respect to such identification of (non)
specificity, bare CLFPs, similar to [CLF ART:one] constructions (cf. Borer 2005; Chang
2014), always involve specification of singularity (see (32)–(34) and (36)–(37)). By
contrast, bare nouns are number-neutral, and interpretable either as singular or
plural (as in (35)) according to context (Jenks 2011).
To summarize, Thai bare CLFPs are associated with three types of indefinite
reference: (existential) presentative reference,specific indefinite reference, and non-
specific (indefinite) reference. The question thus arises as to whether bare CLFPsin
Thai can be more directly associated with specific or non-specific indefinite
reference. To answer that question, we designed a questionnaire, the results of
which are reported on in the following sub-section.
3.3.3 Association of bare CLFPS with non-specific indefinite reference
To address the question of whether bare CLFPs can be more directly associated with
specific or non-specific indefinite reference, we designed a questionnaire to
15 Namely: in a request situation, speakers should not provide more information than is neces-
sary –for example, concerning the specificity of a referent –when asking an addressee to perform a
task (Haspelmath 2017).
Bare classifier phrases in Thai 479
evaluate Thai native speakers’acceptability judgements concerning the use of bare
CLFPs in relation to other indefinitely-referring constructions in Thai. The subtypes
of indefinite reference we selected were those discussed in the preceding section
Section 3.3.2, i.e., existential-presentative,specific reference, and non-specific
reference. Two constructions [N[CLF ART:one]] and bare [N] were selected for com-
parison with bare CLFP, since those constructions are commonly associated with
indefiniteness in Thai (Bisang 1999; Haas 1942; Hundius and Kölver 1983).
Development of the questionnaire involved several phases. First, we collected
exploratory data from web searches, from online corpora, and from perusing
several Thai novels and magazines. We then constructed a set of sentences based
on the information we obtained. A set of fifteen sentences was then made
contingent on the extent to which each sentence was sufficiently indicative of the
targeted indefinite reference sub-categories, following the taxonomy developed in
Figure 1 (see the Appendix for the survey questionnaire). Finally, sentences were
Questionnaire
[CLF ART:one]
phrases
Presentative use
Item #1
Item #2
Specific use Item #3
Non-specific use
Item #4
Item #5
Bare Ns
Presentative use
Item #6
Item #7
Specific use Item #8
Non-specific use
Item #9
Item #10
Bare CLFs
Presentative use
Item #11
Item #12
Specific use Item #13
Non-specific use
Item #14
Item #15
Figure 1: Taxonomy of
indefinitely-referring
constructions as a framework
for developing a questionnaire
on the preferred functional
contexts of Thai bare CLFPs.
480 Pichetpan and Post
randomly sorted so that different constructions and sub-categories of indefinite-
ness were distributed throughout the questionnaire.
We next elicited acceptability judgements from a group of native Thai
speakers. We asked them to rate sentences on an acceptability scale of 0–5,
ranging from complete unacceptability (0) to full acceptability (5). Having accu-
mulated the resulting figures, we next calculated average values pertaining to all
items of the same referential sub-categories for an individual construction type.
We employ the symbol “+”to refer to instances when standard deviation is higher
than one (1.00). In our interpretation of the data, an average response that is a little
more or less than 3.00 is regarded as indicating broad “acceptability”among our
consultants; the higher the average value, the greater the acceptability. Figure 2
presents the outcome.
Our results indicate that while [N[CLF ART:one]] and bare [N] are judged as about
equally (and highly) acceptable in all three indefinite contexts, and while bare
CLFPs are also broadly acceptable, there is nevertheless a clear preference for the
use of bare CLFPsinnon-specific reference; use of bare CLFPs in existential-
presentative and specific reference is markedly less acceptable.
The results of this study suggest that we might posit a prototypical function of
bare CLFPs in establishing reference to indefinite non-specific individuals –singular
individuals to whose specificity a speaker is uncommitted, and which are assumed
to constitute new information to an addressee in the context of a discourse. That
being the case, we can make use of this establishment to explain why the distri-
bution of bare CLFPs appears to be associated with postverbal positions, in which
Thai nominals or NPs are by default inferred as non-specific (Takahashi 2008; cf.
Enfield 2004, for Lao; Li 2011, for Mandarin; see also Lambrecht 1994: 262). We
come back to this association again in Section 4.1.3.
4.69 4.63
3.25
4.38 4.50+
2.75
4.56
4.00+ 4.13+
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
[ᴄʟꜰ ᴀʀᴛ:one] bare [ɴ] bare [ᴄʟꜰᴘ]
ytilibatpeccAfoeergeD
Existential-presentative Specific Non-specific
Figure 2: Averaged acceptability of three referential marking patterns for different indefinite
referential subcategories; (higher average value = greater acceptability).
Bare classifier phrases in Thai 481
4 A proposed origin of the (N)[CLF] pattern in Thai
In this section, we will adopt suggestions previously made by Aroonmanakun
(1994) and Singnoi (2008) and develop an argument that Thai bare CLFPs are derived
relatively recently from the co-existing [CLF [ART:one]]
CLFP
construction. We first
review similarities and differences among bare CLFPs and the structurally
and functionally related constructions [one CLF]
CLFP
,[CLF [ART:one]] and bare [N]
(Section 4.1). On the basis of this description, we will propose an origin of bare CLFP
through grammaticalization, functional reanalysis, and phonological reduction
(Section 4.2).
4.1 Comparative analysis of [CLF]
CLFP
, [one CLF]
CLFP
,[CLF
ART:one]
CLFP
, and bare [N]
As we briefly mentioned in Section 3.3, bare CLFPs appear to count as one of the
several indefinitely-referring constructions available in Thai. In one of the most
comprehensive studies of classifiers in Thai, Hundius and Kölver (1983) align the
function of indefiniteness marking most closely with the post-head CLFP con-
struction [(N)[CLF ART:one]
CLFP
]
NP
. However, they also discuss the fact that refer-
entiality is not a simple matter of NP marking in Thai, claiming it to be “largely a
matter of contextual interpretation”(Hundius and Kölver 1983: 180). By this they
appear to mean that bare nouns also have the capacity to be referential, depending
on the context. We accordingly regard the bare [N] construction as a second con-
struction capable of yielding indefinite reference in Thai.
In the following sub-sections, we will therefore compare the bare CLF pattern to
bare [N]s, both the pre-head and post-head classifier phrases in which the classifier
is modified by numeral nɯ
̀ŋ‘one’(Section 2.2). In addition to the post-head variant,
which we will argue to be the diachronic source of Thai bare CLFPs, we are also
including the pre-head variant because of its close structural and functional
relationship to the “post-head”construction.
4.1.1 Bare CLFP and [CLF ART:one]
CLFP
compared with bare [N] in referential use
We first compare the bare CLFP and [CLF ART:one]
CLFP
constructions with bare [N]in
referential use. All three of these patterns may motivate identification of the same
three categories of indefinite reference as were discussed in Section 3.3.2.2 (pre-
sentative, specific and non-specific). However, bare [N]s differ from the other two in
that their use can entail a range of possible referential interpretations relativized to
482 Pichetpan and Post
contexts and a wide variety of construals relating to the characteristics of a kind/
class. For example, (38) shows that the bare [N]mamûaŋ‘mango’can refer not only
to an individual mango fruit, but also to multiple such fruits, or to one or more
pieces of such a fruit, or to one or more plants from which the fruit is derived.
(38) ʔaːtɕaːn…kʰɔ
̌ːmamûaŋnáʔ
teacher …ask.for mango EMP
[N]
NP
‘Teacher …may I have (a) mango(es)/a piece of mango flesh/(a) mango
tree(s)?’
By contrast, both bare CLFPs and [CLF ART:one]
CLFP
are constrained in their reference
to singular indefinite individuals. Examples (39) and (40) demonstrate that both
máʔmûaŋlûːkand máʔmûaŋlûːknɯŋcan only refer to individual mango fruits.
(39) ʔaːtɕaːn…kʰɔ
̌ːmamûaŋlûːknáʔ
teacher …ask.for mango CLF:fruit EMP
[[N][CLF]
CLFP
]
NP
‘Teacher …may I have a [single] mango [fruit]?’
(40) ʔaːtɕaːn…kʰɔ
̌ːmamûaŋlûːknɯŋnáʔ
teacher …ask.for mango CLF:fruit ART:one:unstressed EMP
[[N][CLF ART:one]
CLFP
]
NP
‘Teacher …may I have a [single] mango [fruit]?’
4.1.2 Bare CLFP and [CLF ART:one] compared with [one CLF] in referential use
In modern Thai, nɯ
̀ŋ‘one’is the only numeral which can follow a classifier.
According to Li (1978), this was the original position of the numeral ‘one’as such in
Tai languages, while all other numerals preceded classifiers. If this is the case, then
pre-head numeral nɯ
̀ŋ‘one’as we find it in modern Thai would appear to be a
secondary development via analogy with other quantifiers (including lǎːj‘many’,
etc.). Whatever the diachronic facts may ultimately turn out to be, given the
evident lexical relatedness and functional similarity between pre-head [one CLF]
CLFP
and post-head [CLF ART:one]
CLFP
, it will therefore be important to determine which of
these two constructions –both of which could, in principle, serve as a source
construction for bare CLFPs in Thai –is functionally closer to bare CLFPs.
In functional terms, bare CLFPs and [CLF ART:one]
CLFP
are used primarily for
reference. However, as we also discussed in Section 2.2 with regard to pre-head CLFP
in Thai more generally, the [one CLF]
CLFP
construction does not seem to have the
Bare classifier phrases in Thai 483
main function of reference, but rather profiles enumeration. Consider (41) as
compared to (39), (40) and indeed also (38) above.
(41) ʔaːtɕaːn…kʰɔ
̌ːmamûaŋnɯ
̀ŋlûːknáʔ
teacher …ask.for mango one CLF:fruit EMP
[[N] [one CLF]
CLFP
]
NP
‘Teacher …may I have one mango fruit [not two or three]?’
As shown in (41), the [one CLF]
CLFP
construction has the pragmatic function of
isolating an individual for contrastive quantification (‘one, and not some other
quantity’) (Singnoi 2001). A sentence such as (41) could thus be felicitously used in
response to a question “How many mangoes do you want?”By contrast, (38)–(40)
focus on class or kind, and could be felicitously used in response to a question
“What [kind of thing] do you want?”Bare CLFP and [CLF ART:one]
CLFP
constructions
thus appear more closely aligned in the function of reference than either is to [one
CLF]
CLFP
, namely in isolating an single individual of a particular type rather than in
focusing on ‘one’as a numerical quantity.
4.1.3 Bare CLFP vs. [CLF ART:one], bare [N] and [one CLF] in pre-verbal subject
position
We briefly stated in Section 3.3.3 that bare CLFPs seem confined to post-verbal
position. Here we present a more explicit analysis of this phenomenon in relation
to our functional analysis of the bare CLFP construction presented in Section 3.3.2.
Sentences (42)–(45) test the ability of bare CLFP,[CLF ART:one], [one CLF] and bare
[N] to occur in preverbal position. As shown, only the bare CLFP is disallowed.
(42) *tɕʰáːŋtua miːŋaːdiaw
elephant CLF:body have tusk single
*[[N][CLF]
CLFP
]
NP
‘An elephant has a single tusk.’(intended for non-specific reference)
(43) tɕʰáːŋtua nɯŋmiːŋaːdiaw
elephant CLF:body one have tusk single
[[N][CLF ART:one]
CLFP
]
NP
‘An elephant has a single tusk.’(specific reference)
(44) tɕʰáːŋnɯ
̀ŋtua miːŋaːdiaw
elephant one CLF:body have tusk single
[[N] [one CLF]
CLFP
]
NP
‘One elephant has a single tusk.’(specific quantification)
484 Pichetpan and Post
(45) tɕʰáːŋmiːŋaːdiaw
elephant have tusk single
[N]
NP
‘(An/the) elephant(s) has/have a single tusk.’(specific/definite reference)
Thai is thus like Mandarin Chinese in restricting bare CLFPs to post-verbal position
(Cheng and Sybesma 1999; Jenks 2011; Li 2011; among others) –however, Thai bare
CLFPs appear to be somewhat less restricted than are Mandarin bare CLFPs in this
post-verbal position. For example, Jiang (2012; see also Yang 2001) discusses the
fact that a Mandarin bare CLFP must be immediately verb-adjacent, only tolerating
the intervention of a suffixed aspectual particle le ‘ASP:finish’(46); no other word
may intervene (47). Similarly, bare CLFPs may occur only within the first of two post-
verbal coordinated structures, not the second, unless the preceding coordinator is
stressed (48)–(49).
(46) wǒjiào-le (yī)-g`
e nóngmín
1SG teach-ASP (one)-CLF farmer
[[one-CLF]
CLFP
N]
NP
‘I taught a farmer.’(adapted from Jiang 2012: 189)
(47) wǒjiào-guò *(yī)-g`
e nóngmín
1SG teach-ASP (one)-CLF farmer
[[one-CLF]
CLFP
N]
NP
‘I taught a farmer.’(intended translation) (adapted from Jiang 2012: 189)
(48) Yuēhàn mǎi-le (yī)-běnshūh´
e*(yī)-zhībǐ
John buy-ASP (one)-CLF book and (one)-CLF pen
[[one-CLF]
CLFP
N]
NP
[[one-CLF]
CLFP
N]
NP
‘John bought a book and a pen.’(adapted from Yang 2001: 69)
(49) Yuēhàn mǎi-le (yī)-běnshūhai-you ?(yī)-zhībǐ
John buy-ASP (one)-CLF book and (one)-CLF pen
[[one-CLF]
CLFP
N]
NP
[[one-CLF]
CLFP
N]
NP
‘John bought a book and a pen.’(adapted from Jiang 2012: 212)
Note here that yǒuin hái-yǒu(underlined) can be thought of as the immediately
preceding verb that licenses a second-conjunct bare CLFP. Still, the use of bare CLFPs
in non-first conjunct is described as marginally acceptable (Jiang 2012: 212).
Thai bare CLFPs occur with fewer restrictions. Naturally, there is no verb-
adjacency requirement, as the differing noun phrase syntax of Thai places head
nouns between a classifier phrase and a verb (23). In addition, the possibility of
Bare classifier phrases in Thai 485
discontinuous classification in Thai means that classifiers may occur at even
greater distance from a verb (29). As far as we can see, Thai does not have any
incompatible aspectual particles, like guò in Mandarin, that inhibit the use of bare
CLFPs in postverbal position. And, in postverbal coordination, Thai bare CLFPs can
occur in both first and second conjuncts, as per syntactic parallelism, despite the
use of a preceding unstressed coordinator (50).
(50) pʰǒmkʰɔ
̌ːnám-ʔùn kɛ
̂ːwkàʔmanaːwlûːk
1SG.Mask.for water-warm CLF:glass and lime CLF:fruit
[N[CLF]
CLFP
]
NP
[N[CLF]
CLFP
]
NP
‘May I have a glass of warm water and a lime (fruit)?’
While some authors have explained the preferentially postverbal occurrence of
bare CLFPs in Sinitic languages in terms of “scope of existential closure”and
“lexical government”(Cheng and Sybesma 1999; Jenks 2011; Jiang 2012, inter alia),
we will suggest an alternative account for the strictly postverbal occurrence of bare
CLFPs in Thai in terms of their association with non-specificity.
We argue that the unacceptability of bare CLFPs in pre-verbal position relates to Thai
speakers’preferential association of bare CLFPswithindefinite non-specificreferential
mentions (Section 3.3.3). As has been shown by Iwasaki and Ingkaphirom (2005) and
Diller (1997) among others, Thai exhibits topic-comment sentence organization, and
tends to place new (unknown, not previously established) information post-verbally,
while given (known, already-established) information is placed at the beginning of a
clause (as in Sinitic languages; cf. Li 2011; Xu and Liu 2007). As in (51), the sentence topic
pʰîː-sǎːw‘elder sister’is placed in the sentence-initial position, while the previously
unknown referent mɔ
̌ː‘physician’is introduced post-verbally.
(51) pʰîː-sǎːwkʰǎwtɛ
̀ŋŋaːn kàp mɔ
̌ː
elder.sister 3SG be.married with physician
GIVEN NEW
‘His elder sister got married to a physician.’
(Iwasaki and Ingkaphirom 2005: 301)
Since the preverbal position in Thai is thus by definition unavailable to referents
who are not known, or to whose existence or reality the speaker is uncommitted,
the prohibition against bare CLFPs in pre-verbal position is thus likely to be an
outcome of its close association with indefinite non-specific reference.
4.1.4 Interim summary
Table 2 summarizes the commonalities and differences of [CLF ART:one]
CLFP
phrases,
[one CLF]
CLFP
phrases, bare [N], and bare CLFPs with respect to constructional
486 Pichetpan and Post
structures (Section 2.2 and Section 3.2), referential functionality and specification
of singularity (Sections 3.3.2–3.3.3 and Sections 4.1.1–4.1.2), and restriction to
postverbal constituents (Section 4.1.3).
4.2 Origin of bare CLF in [CLF ART:one]
As Table 2 shows, among all candidate source constructions, bare CLFPs are
structurally and functionally closest to [CLF ART:one]
CLFP
: (i) both occur in the three
basic types of CLFP constructions: continuous,discontinuous, and head-ellipsed; this
is unlike at least some other CLFP types, which exhibit restrictions in some domains.
Additionally, (ii) like [CLF ART:one]
CLFP
, bare CLFPs are used for a full range of indef-
inite constructions, including existential-presentative reference,specific reference,
and non-specific reference. Finally, Thai bare CLFP and [CLF ART:one]
CLFP
are alike in
forcing an interpretation of singular number, unlike bare [N].
However, we have also noted differences between these two CLFP patterns;
namely, bare CLFPs cannot occur pre-verbally, whereas [CLF ART:one]
CLFP
phrases
may. We have argued that this restriction relates to the prototypical function of
bare CLFPs as introducing non-specific referents, as demonstrated by our consul-
tants’acceptability judgements (see Section 3.3.3). We argue, in line with pre-
vious suggestions by Aroonmanakun (1994) and Singnoi (2001), that these facts
are best explained by supposing that bare CLFPs may have developed out of the
[CLF ART:one] construction. We will add to these earlier accounts by showing
how this development is indeed a plausible outcome of existing patterns of
Table :Structural, functional, and distributional features of the patterns [CLF]
CLFP
,[CLF ART:one]
CLFP
,
[one CLF]
CLFP
, and bare [N].
Characteristics of basic constructions, functions,
and syntactic distribution
[CLF
ART:one]
CLFP
[one
CLF]
CLFP
Bare
[N]
[CLF]
CLFP
Continuous NP construction Yes Yes –Yes
Discontinuous NP construction Yes Yes –Yes
Covert-head NP construction Yes Yes –
Yes
Presentative reference Yes No Yes Dispreferred
Specific reference Yes No Yes Dispreferred
Non-specific reference Yes No Yes Yes
Singular number identification Yes Yes No Yes
Preverbal position Yes Yes Yes No
16 These three construction types are irrelevant in the case of bare [N] due to logical incompatibility.
Bare classifier phrases in Thai 487
functional reanalysis and phonologicalreduction,giventhesetofstructural
and functional similarities between bare CLFP and [CLF ART:one]
CLFP
just discussed.
4.2.1 Grammaticalization of [CLF ART:one] →[CLF Ø]
The development of (singular) indefiniteness-marking functions from the numeral
‘one’is of course a well-attested phenomenon cross-linguistically (Givón 1981;
Heine 1997, among others). As we discussed in Section 2.2, Thai nɯ
̀ŋ‘one’has
developed indefinite singular referential functionality when occurring with a
classifier in a post-head CLFP. As we also discussed, this is in line with the general
ordering conventions of Thai CLFP in which pre-head operators quantify while post-
head operators focus on referential qualification.
The question thus arises as to the relative functional contributions of each of
the two components CLF and ART:one to the resulting indefinite singular expression.
We will return to that question further below. First, however, it is important to note
that nɯ
̀ŋ‘one’exhibits gradient phonological erosion in its post-head context
only; pre-head ‘one’does not exhibit the same erosion. For example, post-head
‘one’frequently exhibits tone loss –a commonly-recognized concomitant of
grammaticalization in Thai as observed by Diller (2001) and Post (2007b), among
others (52)–(53). Tone loss is not generally observed among pre-head numeral
modifiers in Thai (54)–(55).
(52) náːmkɛ̂ːwnɯ
̀ŋ
water CLF:glass ART:one
[N[CLF ART:one]
CLFP
]
NP
‘a glass of water’(transliterated from Anantaprayoon (2018: 200))
(53) náːmkɛ̂ːwnɯŋ
water CLF:glass ART:one
[N[CLF ART:one]
CLFP
]
NP
‘a glass of water’
(Google.com, [Twitter comment]; note: transliteration from attested
example in Thai orthography; see below for discussion of this point.)
(54) náːmnɯ
̀ŋkɛ̂ːw
water one CLF:glass
[N[one CLF]
CLFP
]
NP
‘(just) one glass of water [not two or three]’
488 Pichetpan and Post
(55) #náːmnɯŋkɛ̂ːw
water one CLF:glass
[N[one CLF]
CLFP
]
NP
(grammatically understandable, but not native-like pronunciation)
Tone loss is in this case associated with a more general de-stressing of the erstwhile
numeral ‘one’in its post-head referential function. However, note that de-stressing
is not a standard outcome of post-head position in Thai, which is a fundamentally
iambic language (Bennett 1994); other post-head referential operators are not
necessarily destressed. In (56), note that post-head demonstrative nán ‘DST’does
not undergo tone loss.
(56) kâwʔîːtua nán
chair CLF:body DST
[N[CLF DEM]
CLFP
]
NP
‘that chair’
By contrast, de-stressing and tone loss in post-head ‘one’is so well-entrenched
in modern Thai that it is often reflected in the orthographic practices of
native Thai writers. Although the traditional form หนึงnɯ
̀ŋ, is usually found in
formal or academic writing,
17
avariantนึง nɯŋis often found in speech-like
writing in vernacular Thai, for example in social media (see again (52)
compared to (53)).
What is perhaps even less well-known is that post-head ‘one’can also exhibit
segmental erosion, namely from /nɯŋ/to[ŋ
̩]. Although such reductions are not
commonly reflected in the orthographic practices of native Thai writers, examples
such as (53) are commonplace in vernacular Thai speech.
(57) náːmkɛ̂ːwŋ
̩
water CLF:glass ART:one
[N[CLF ART:one]
CLFP
]
NP
‘a glass of water’
In fact, one notable outcome of the survey reported in Section 3.3.3 was that our
consultants on a number of occasions commented that while they would accept a
bare CLFP in a given construction as presented, the sentence would be improved by
addition of post-head ŋ
̩‘ART:one’. These facts suggest that both the grammatical/
17 For readers who are unfamiliar with Thai orthography: here, the initial, synchronically un-
pronounced consonant หworks together with the above-line tone marker _̍,máːjʔ`
eːk, to indicate
low tone. If they are both removed, as they are in the following example, the result will be a
toneless (or mid/neutral toned) syllable.
Bare classifier phrases in Thai 489
discourse-functional and phonological conditions for the identification of a
grammaticalization cline motivated by specialization of the function indefinite
singular reference are present in Thai. We may conceptualize this as beginning with
reanalysis and specialization of post-head ‘one’as a referential modifier, followed
by progressive phonological erosion. We argue that this has ultimately led to full
deletion of the erstwhile post-head numeral –which, due to high predictability,
would be pragmatically retrievable in this context –and emergence of the bare CLFP
construction (Figure 3).
It is important to note that although we are conceptualizing this development
as a temporal progression, we have no evidence that such a progression has in fact
taken place over historical time in Thai. Furthermore, we must note that all four of
the realizations schematized in Figure 3 remain available to modern Thai speakers,
even if the bare CLFP construction is dispreferred in certain contexts, regiolects, or
written registers. We might therefore ultimately prefer the more elaborate stage-
wise construal outlined in Figure 4.
4.2.2 Relative functions of CLF and ART:one in [CLF ART:one] and the development of
the bare CLF construction
To fully comprehend the emergence of the bare CLF construction as a case of
progressive erosion and eventual deletion of ART:one in a [CLF ART:one] con-
struction –leaving, in essence, the CLF inplaceasthesoleconstituentofthenew
construction –it will be necessary to understand the individual functional
Earlier stage Later stage
(N)[CLF nɯ
̀ŋ] → (N)[CLF nɯŋ] → (N)[CLF ŋ
̩]→(N)[CLF ]
Figure 3: Grammaticalization of post-head ‘one’, phonological erosion and emergence of the
bare CLF construction.
Stage I Stage II Stage III Stage IV
[CLF nɯ
̀ŋ] → [CLF nɯ
̀ŋ] →[CLF nɯ
̀ŋ] →[CLF nɯ
̀ŋ]
↘ [CLF nɯŋ] →[CLF nɯŋ] →[CLF nɯŋ]
[CLF ŋ
̩]→[CLF ŋ
̩]
[CLF ]
Figure 4: Grammaticalization of bare CLFP from [CLF ART:one] construed as an
expanding constructional vocabulary.
490 Pichetpan and Post
contributions of each constituent of the source construction. We return now to
this question, which was briefly posed at the outset of Section 4.2.1.
As we mentioned above, ‘one’is a frequent source of (singular) indefinite
reference morphology cross-linguistically. It seems similarly clear that the prin-
cipal contribution of post-head ‘one’to the [CLF ART:one]
CLFP
construction in Thai is
the concept of singular indefiniteness. [CLF ART:one]
CLFP
contrasts with all CLFP
structures other than bare CLFPs in focusing pragmatically on singular indefinite-
ness to the exclusion of all other types of construal (Section 2.2).
However, we have also discussed the idea –developed in detail elsewhere by
Hundius and Kölver (1983) and Bisang (1999) –that Thai classifiers fundamentally
serve to (overtly) individuate referents, whereas bare [N] are essentially unspecified
for referentiality in Thai. The question before us thus relates to the functional
similarity between the INDIVIDUATION function of classifiers and the SINGULAR INDEFINITE
meaning derived from erstwhile numeral ‘one’.
Our suggestion here will be that the function of INDIVIDUATION lends itself to
recruitment, through constructional reanalysis and grammaticalization, as a marker of
SINGULAR INDEFINITENESS even in absence of the numeral ‘one’.Evidently,SINGULAR INDEFI-
NITENESS is the functional domain which the Thai bare CLFP construction targeted, and into
whose distributional range it has partially spread. However, it is important to note that
this is what happened in Thai; it does not necessarily follow that INDIVIDUATION,evenasa
property of classifiers as a form-class per se, will motivate the same discourse-functional
outcomes for grammaticalization in any other language. Some of the other possibilities
will be discussed in the following section.
5 Towards a typology of bare classifier
constructions in mainland Asian languages:
implications for the theory of classifiers
5.1 Towards a typology of bare CLFP constructions in mainland
Asian languages
The preceding sections have provided the first comprehensive structural and
functional description of Thai bare CLFP constructions, and they should serve to add
Thai to the list of mainland Asian languages which exhibit bare CLFP constructions.
We argue that Thai is an especially important addition to this list, since not all
mainland Asian classifier languages allow a bare CLFP construction, and Thai has
previously been included among those Asian languages which “lack”such a
Bare classifier phrases in Thai 491
construction (Cheng and Sybesma 1999, 2012 [2008]; Jenks 2011). Indeed, it would
seem that closely related Southwestern Tai languages such as Lao in fact do lack
such a construction; no description of bare CLFPs is found in Enfield (2007), and
native speakers of Lao whom we consulted for this study uniformly rejected those
bare CLFP constructions which our native Thai-speaking consultants considered to
be acceptable. Such facts would argue in favour of a relatively recent origin of this
construction in Thai, as we suggested was the case in Section 4.
Looking outward to the occurrence of bare CLFP constructions in other Asian
languages, the outlines of a typology of bare CLFPs can be found in Jenks (2011),
Jiang (2012) and Wang (2015). Typologically significant parameters of variation
among languages with bare CLFP constructions which they discuss might be stip-
ulated in a preliminary fashion as (i) position of the CLF in relation to its head N(ii)
referential value of bare CLFP (definite/indefinite/both) (iii) position of bare CLFP in
the clause, i.e. relative to V.
18
We have not been able to determine values for all
three of these parameters for most of the languages for which we have data; for
example, Wang (2015) focuses on the position of CLFP relative to Vfor 120 Sinitic
languages, but does not indicate the position of CLF relative to N, while some other
sources lack clear information regarding position relative to V. A preliminary
summary is presented in Table 3.
As Table 3 shows, bare CLFP constructions are found in most major language
families of mainland Asia. Although most languages in which bare CLFPs have
been clearly attested are concentrated in southern China and northern Southeast
Asia –in the environs of the Yue Sprachbund –languages such as Mandarin and
Thai expand the geographical scope considerably.
Table 3 also shows that where bare CLFPs occur, they always function to indi-
viduate singular referents. However, bare CLFPs can yield definite or indefinite
singular reference in different languages. Cantonese is a well-known language in
which bare CLFPs most often yield definite reference (58) (a few Northern Tai lan-
guages such as Saek or Wuming Zhuang appear to follow the same pattern), while
in Nuosu Yi –as in Thai –bare CLFPs yield indefinite reference except when fol-
lowed by a definite article su (59). We find a statistical correlation between pre-
nominal position of the CLFP and definite reference (and vice versa), however
18 More speculatively, and working toward an explanatory account of the role of bare CLFP con-
structions in the overall referential systems of Asian classifier languages, one might further
include values associated with the occurrence or non-occurrence of articles, and/or the referential
potential of related structures such as bare [N], [CLF DEM], and/or [CLF ONE] (Jenks 2011, 2018 in
particular notes the frequent inability of bare [N] to yield definite reference in languages in which
bare CLFP code the same function). Inclusion of such values at the present stage, however, would
expand this paper well beyond its intended scope.
492 Pichetpan and Post
Mandarin provides a clear counterexample to this tendency –which may, there-
fore, be an artifact of areal contiguity or some other factor (60).
(58) ngo5 wan2 dou2 zek3 maau1
1SG find successfully [CLF cat]
[[CLF]
CLFP
N]
NP
‘I found the cat.’(Cantonese; Matthews 2006: 230)) (not ‘I found a
cat/cats.’)
Table :Working typology of bare CLFP constructions in Asian languages.
Phylum Language pre/
post-N
DEF/
INDEF
number References
Trans-
Himalayan
Nuosu Yi (sich)POST-INDEF SG Gerner ()
Mandarin (beij)PRE-INDEF SG Jiang (); Wang ()
Cantonese
(cant)
PRE-BOTH SG Cheng and Sybesma ();
Matthews and Pacioni ();
Matthews ()
Wu (wuch)PRE-BOTH SG Li ()
Austroasiatic Vietnamese
(viet)
PRE-DEF SG Nguyen (); Simpson, Soh,
and Nomoto ()
Hmong-Mien White Hmong
(hmon)
PRE-DEF SG Simpson et al. ()
Green Hmong
(hmon)
PRE-DEF SG Bisang ()
Weining Ahmao
(larg)
PRE-DEF SG Bisang ()
Tai-Kadai
Yay/Bouyei
(bouy)
PRE-DEF SG Hudak (); Central Institute
for Nationalities ()
Wuming Zhuang
(yong)
PRE-DEF SG You ()
Nùng (nung)PRE-DEF SG Saul and Wilson ()
Lungming
Tai/Zuojiang Zhuang
(zuoj)
PRE-DEF SG Hudak ()
Thai POST-INDEF SG This article
19 Though the Northern Tai languages in Table 3 are seen for their bare CLFPs producing DEFINITE-
NESS, like the more well-known Cantonese, we still have an impoverished understanding of such
bare CLFPs in whether they can occur post-verbally in definite use, due to lack of full analyses in the
existing literature. Also, unlike Cantonese, some of them, such as Saek, seem to necessitate special
and different classifier forms for bare CLFP constructions, as opposed to those for other CLFP-type
constructions (Meng Weijian, pers. comm.). In the interests of space and to maintain the focus of
this article, we will leave these matters to future investigation.
Bare classifier phrases in Thai 493
(59) i33ti34 gu33 ʑo33 si44 la33
coat CLF catch take come
[N[CLF]
CLFP
]
NP
‘Bring a coat (to me).’(Nuosu Yi; Liu and Gu 2011: 320)
(60) wǒshì ge hěn wangu de r´
en
1.SG be CLF:GEN very stubborn ATTR person
[[CLF]
CLFP
…N]
NP
‘I am a very stubborn person.’(Mandarin; adapted from Lü 1944: 168, also
cited in Jiang 2012: 182)
In some languages, bare CLFPs yield both definite and indefinite interpretations,
depending on clause type. The best-described such language seems to be
Cantonese, in which pre-verbal bare CLFP must be interpreted as definite (61). Post-
verbally, indefinite non-specific readings are available in irrealis contexts (62). In
realis contexts, definite readings seem preferred (58); in neither case is a generic
reading available.
(61) zek gau zungji sek juk
CLF dog like eat meat
‘The dog likes to eat meat.’(not ‘A dog/dogs like to eat meat.’) (Cheng and
Sybesma 1999: 511)
(62) ngo soeng maai bun syu (lei taai)
I want buy CLF book come read
‘I want to buy a book (to read).’(Cheng and Sybesma 1999: 511)
Turning to the origin of bare CLFP constructions cross-linguistically, we are
unfortunately not yet in a position to generalize across most bare CLFP con-
structions. However, it is important to note that the Mandarin Chinese bare CLFP
construction is argued to have the samediachronicsourceaswehaveposited
for Thai –namely, a pre-head [[ONE CLF]
CLFP
N]
NP
construction (Lü 1944; Huang
2009; Wang 2015, among others). In Mandarin, as in Thai, deletion of the
erstwhile numeral ‘one’correlates with the emergence of INDEFINITE SINGULAR
referential values in their newly emergent bare CLFP constructions. Also similar
to what we find in Thai, while Mandarin bare CLFPscanbeusedtocodeSPECIFIC
INDEFINITES, authors nonetheless report a more typical association with NON-SPE-
CIFIC reference –an association which we have demonstrated quantitatively in
the case of Thai (cf. Jiang 2012: 184). There are important differences between
the Thai and Mandarin cases; in particular, while as we discussed in Section 2.2
above Thai quantitative and reference-qualifying operators occur on different
sides of the classifierhead,inMandarintheybothoccurinpre-headposition
(meaning that it is much more difficult, and perhaps impossible, to clearly
distinguish these functions in the case of Mandarin). Nevertheless, the
494 Pichetpan and Post
developmental pathways of each language’sbareCLFP constructions are
remarkably similar –as, indeed, are their outcomes.
However, the more frequent association is between bare CLFP constructions and
definite reference, particularly in preverbal positions.
20
Presumably, bare CLFP
constructions in such languages would have a different diachronic source than
that found in Thai and Mandarin Chinese. Unfortunately, we have not been able to
identify such an account in the sources we have consulted to date.
21
The eventual
development of such an account will thus constitute an important contribution to
the study of bare CLFP systems.
5.2 Implications for the theory of classifiers
Although they are perhaps not as rare as may have once been supposed, bare CLFP
constructions nonetheless appear somewhat marked among the classifier lan-
guages of Asia. Several classifier languages seem to lack such a construction,
including Lao, Japanese, and Southern Min (cf. Cheng and Sybesma 2012 [2008]:
134, ex. 20b). Furthermore, in at least some languages with a bare CLF construction
(such as Thai and Mandarin Chinese), bare CLFPs are considered marginal, infre-
quent, or associated with informal speech. It is perhaps because of these facts that
the potential importance of bare CLFP to the theory and typology of classifiers seems
to have been overlooked. In many of the most well-known works treating classifiers
in Asian languages, authors have focused primarily on the use of “numeral clas-
sifiers”in quantifying constructions, sometimes describing classifiers differently
as “noun classifiers”when functioning non-quantitatively (Aikhenvald 2019,
2000; Haas 1942; Vittrant and Tang in press).
Yet we will argue that bare CLFP constructions are critical to determining the
core functionality of classifiers per se, since they are the only classifier construc-
tions in which classifier functionality is represented discretely –in the absence,
that is, of any additional modifying material. Despite cross-linguistic variation in
the capacity for bare CLFP constructions to code definite or indefinite reference –a
parameter of variation whose precise motivation is not yet obvious –we find that
20 See Wang (2015) for this correlation in Sinitic languages.
21 Li and Bisang (2012) argue that Mandarin, Wu and Cantonese bare CLFP constructions exist in a
grammaticalization cline, with the definiteness-marking function emerging as a symptom of more
advanced grammaticalization, in Sinitic languages at least. Their account would suggest that bare
CLFP might always emerge (in the initial instance) from reduction of [ONE CLF] constructions, with
definiteness features always emerging later. We do not adopt a position on such a hypothesis,
other than to say that its plausibility should be investigated with respect to other bare CLFP lan-
guages, including those of Hmong-Mien and Tai-Kadai stock.
Bare classifier phrases in Thai 495
bare CLFP constructions in Asian languages seem to always correlate with (singular)
individual reference.Wefind no Asian classifier language in which bare CLFP can
denote plural or generic referents, regardless of syntactic function, clause type, or
the genealogical status or areal context of the language in question. This fact
would argue strongly in favour of acknowledging the core function of classifiers in
Asian languages to be that of INDIVIDUATION –of delineating discrete instance(s) of a
kind denoted by [N]. This is the function that is coded by all bare CLFP constructions
in all known Asian languages which have them, and it is perhaps the only invariant
property of classifiers within that same set of languages.
The idea that classifiers function primarily to individuate referents is hardly new:
it may have been introduced by Greenberg (1972), and is underscored by Adams (1989:
9), Croft (1994), Bisang (1999) and Gerner (2013: 66), among others. Adams has noted
that classifiers are often used with lower numerals only, a fact which may relate to the
difficulty of construing members of very large sets as individuals. Similarly, classifiers
are often used in discourse anaphora, to track referents with which an addressee is
presumed to be familiar (Post 2007b). When classifiers are “optional”, as they often are
with non-numerical quantifiers in Cantonese, the presence of a classifier functions to
highlight awareness of the existence of discrete individuals within a set (63).
(63) ngóh yáuh hóu dō(jī)bāt
I have very many CLF pen
‘I have very many (individual) pens.’(adapted from Matthews and Yip
1994: 94)
In other words –and (perhaps somewhat counterintuitively) despite what the
semantics of classification may seem to imply –when an Asian language uses
classifiers, it does so not to classify, but rather to individuate. Grammatically
and discourse-functionally, classifiers thus fall within the domain of other
referentializers, such as articles. It is thus hardly coincidental either that the ma-
jority of classifier languages lack articles –whose prototypical functions they
themselves may serve to code –or that the grammaticalization pathway that we
have identified in the development of bare CLFP constructions in Standard Thai
so closely resembles the cross-linguistically most frequent pathway for the
grammaticalization of indefinite articles, namely via the numeral ‘one’(Givón
1981, 1984: 432–435; Heine 1997: 66–82; Hopper and Martin 1987). It is likely that
increased prevalence of classifiers with individuating functionality in Asian lan-
guages is closely associated with a typological preference for analytical as opposed
to inflectional expression of functional operators in these languages (e.g. number,
and gender), however we do not intend to pursue this claim in detail here.
We would, however, like to address recent claims by Corbett and Fedden
(2016; Fedden and Corbett 2017, 2018) that classifiers and gender markers –
contra earlier claims by Dixon (1977) and Aikhenvald (2000), among others –are
not distinct types of systems, but should instead be considered to constitute
496 Pichetpan and Post
different positions within a single categorical space; a space defined, in their
terms, by the semantic principle of categorization.
22
We do not dispute that
classifiers and gender markers can be read as semantically categorizing referents
in more or less comparable ways; as Aikhenvald (2000) has also discussed, a
great range of linguistic systems –for example, existential verbs –semantically
categorize referents in more or less comparable ways; that does not necessarily
mean that they belong in the same categorical space, such space being tradi-
tionally defined by structural, behavioural and functional properties. On all of
these grounds, in addition to the semantic grounds of referentialization, there is
an at least equally strong argument for considering classifiers to fall within
the same “categorical space”as articles; indeed, in the generative literature,
classifiers are sometimes argued to instantiate Dof DP (Cheng and Sybesma 1999).
We acknowledge that there may be languages in which both gender systems and
classifier systems may be found –just as, indeed, there are languages in which
both articles and classifiers may be found;
23
in the end, every language is a
productofitshistory,andisexplainedbythespecific histories of opportunities
and constraints that characterize its development of coincidental semantic
properties and functionality. Yet we find that a stronger case can be made for
continuing to treat classifiers –in Asian languages, at least –as constituting a
discrete grammatical category based (following Bisang 1999) on both of their core
semantic and grammatical/discourse-functional properties of classification and
individuation, respectively.
6 Some remaining problems
Despite our investigation of structural, distributional, and functional properties of
bare CLFPs in Thai and their probable origin, there continue to be at least two
problems remaining: the first relates to a certain type of Thai classifier, namely
repeaters, while the second relates to the unavailability of historical spoken/
spoken-like language corpora.
As far as we can see, most types of classifiers in Thai, i.e. sortal classifiers,
mensural classifiers, and partial repeaters, can involve the pattern of bare CLFPs;
22 In a 2017 post on LINGTYP, a leading listserv devoted to discussion of linguistic typology,
Martin Haspelmath took Corbett and Fedden’s observations a step further by proposing a new
superordinate category label “genifier”, or later “nomifier”–being a cross-linguistically-operative
“comparative concept”which would in principle encompass both gender markers and classifiers.
23 Nuosu Yi is one such language (Jiang 2012), and so is the entire Tani branch of Trans-Himalayan
languages; one relatively well-described case is that of Galo (galo1242) (Post 2007a).
Bare classifier phrases in Thai 497
however, repeaters seem to form an exception. “Repeaters”are defined as (pseudo-)
classifiers which exactly resemble nominal head forms. They can occupy the same
syntactic position as other classifier types. Repeater constructions are often used for
nouns, which designate new or previously unclassified items (cf. Senft and Levinson
2000). Based on the repeater phenomenon, all Thai nouns can thus be found
in commonly known classifier constructions, such as [CLF ART:one]
CLFP
. Nonetheless,
in bare CLFP constructions, no repeater classifier can occur.
The absence of repeaters in bare CLFP constructions remains one of the un-
resolved issues in our analysis. One explanation could relate to avoidance of
ambiguity; specifically, if a repeater occurred in bare CLFP construction, the
outcome might bear too close a resemblance to reduplication of the head noun,
which has a distributive function in Thai: (64) exemplifies the distributive
reduplication construction, while (65) shows that this is not interpretable as a
bare CLF construction.
(64) kʰon kʰon
man man
[NN]
NP
.
RDUP
‘each person’
(65) *kʰon kʰon
man CLF:man
[N[CLF]
CLFP
]
NP
‘a person’(intended translation of a putative bare CLFP construction)
Theoretically, reduplicated forms are likely to be historically earlier in the devel-
opment of Thai grammar. Repeaters in bare CLFP constructions, which are
considered historically later, cannot compete formally with reduplication as such.
As a result, the repeater phenomenon is disallowed for the bare CLFP pattern.
However, our assumption demands a further justification by historical data, which
in turn leads to another remaining problem: a serious lack of historical data rep-
resenting authentic Thai speech.
Although diachronic changes of the [CLF ART:one] phrase pattern to the bare CLF
pattern can be assumed by the theory of grammaticalization, the grammaticalization
clines we have presented are theoretical, inasmuch as we lack historical documents
that would bear on this question. As discussed in Section 3.3.1, bare CLPFsare
associated with speech or speech-like writing. A diachronic description of bare CLFPs
is accordingly regarded as perhaps impossible, seeing that we lack adequate his-
torical spoken and spoken-like written data. In fact, researchers need to examine
historical(spoken) corpora to prove with greater certainty that the bare CLFP pattern,
as historically later, could be regarded as derived from the [CLF ART:one] phrase
pattern. We could then be sure that the pattern of bare CLFPs was not established as
early as that of [CLF ART:one] phrases in the first place, but actually developed out of it.
498 Pichetpan and Post
Abbreviations
1 First person
3 Third person
ADJ Adjective
ADV Adverb
ART Article
ASP Aspect marker
ATTR Attributive
AUX Auxiliary verb
CLF Classifier
CLFP Classifier phrase
DDeterminer
DEM Demonstrative
DIR Directional
DP Determiner phrase
DST Distal demonstrative
EMP Emphatic
GIVEN Given information
INDF Indefinite
INT Interrogative
MMasculine
MOD Modifier
NNoun
NEW New information
NP Noun phrase
NUM Numeral
POL Polite register
POSS Possessive marker
PREP Preposition
PRX Proximate demonstrative
RDUP Reduplicant
REL Relativizer
RELC Relative clause
SFOC Sequential focus
SG Singular
VVerb
Author contribution: The alphabetically-listed authors have contributed equally to
this article. An earlier version of this work was presented at the 29th meeting of the
Southeast Asian Linguistics Society, held at Tokyo Foreign Studies University, and
we thank the audience for a helpful discussion. We also thank our primary Thai
language consultants for this article: Paul Vaivasa, Jurairat Sriamorn, Surachet
Phichitphongphao, and Yamonpat Pattarakup, together with two anonymous LT
reviewers, but retain responsibility for any remaining errors. Thai data have been
romanized via an IPA-based model.
Bare classifier phrases in Thai 499
Construction Functional context Question
[CLF ART:one] Presentative use
miːkʰa
̂ːwtɕaːnnɯ
̀
s/nɯskɔ̂
ːtɔ̂
ːsha
̂j
have rice CLF:plate ART:one then must give
lu
̂ːk kin kɔ̀
ːn
child eat first
‘If there was a plate of rice, we would let the children eat first.’
naj hɔ̂
ːsmiːna
́ːmkε̂ːwnɯ
̀
s/nɯs
in room have water CLF:glass ART:one
‘In the room was a glass of water.’
Specific use
mɯ
̂ːaki
̂
ːkʰa
̌wdɯ
̀ːmna
́ːmkε̂ːwnɯ
̀
s/nɯs
just.now SG drink water CLF:glass ART:one
‘Just a moment ago, he drank a glass of water.’
Non-specific use
kʰa
̌wja
̀ːkdɯ
̀ːmna
́ːmkε̂ːwnɯ
̀
s/nɯs
SG want drink water CLF:glass ART:one
‘He wanted to drink a glass of water.’
pʰo
̌mkʰɔ̌
ːna
́ːmkε̂ːwnɯ
̀
s/nɯssi
́
ʔ
SG ask.for water CLF:glass ART:one FP.IMP
‘Please give me a glass of water.’
Appendix Survey questionnaire of functional contexts of Thai bare CLF con-
structions (vs. [CLF ART:one] constructions vs. bare noun constructions)
500 Pichetpan and Post
(continued)
Construction Functional context Question
Bare Ns Presentative use
miːkʰa
̂ːwkɔ̂
ːtɔ̂
ːsha
̂jlu
̂ːk kin kɔ̀
ːn
have rice then must give child eat first
‘If there was rice, we would let the children eat first.’
naj hɔ̂
ːsmiːna
́ːm
in room have water
‘There was water in the room.’
Specific use
mɯ
̂ːaki
̂
ːkʰa
̌wdɯ
̀ːmna
́ːm
just.now SG drink water
‘Just a moment ago, he drank water.’
Non-specific use
kʰa
̌wja
̀ːkdɯ
̀ːmna
́ːm
SG want drink water
‘He wanted to drink water.’
pʰo
̌mkʰɔ̌
ːna
́ːmsi
́
ʔ
SG ask.for water FP.IMP
‘Please give me some water.’
Bare CLFs Presentative use
miːkʰa
̂ːwtɕaːnkɔ̂ːtɔ̂
ːsha
̂j
have rice CLF:plate then must give
lu
̂ːk kin kɔ̀
ːn
child eat first
‘If there was a plate of rice, we would let the children eat first.’
Bare classifier phrases in Thai 501
(continued)
Construction Functional context Question
naj hɔ̂
ːsmiːna
́ːmkε̂ːw
in room have water CLF:glass
‘In the room was a glass of water.’
Specific use
mɯ
̂ːaki
̂
ːkʰa
̌wdɯ
̀ːmna
́ːmkε̂ːw
just.now SG drink water CLF:glass
‘Just a moment ago, he drank a glass of water.’
Non-specific use
kʰa
̌wja
̀ːkdɯ
̀ːmna
́ːmkε̂ːw
SG want drink water CLF:glass
‘He wanted to drink a glass of water.’
pʰo
̌mkʰɔ̌
ːna
́ːmkε̂ːwsi
́
ʔ
SG ask.for water CLF:glass FP.IMP
‘Please give me a glass of water.’
Note that the translations should not be read as translated transcripts, but rather as free translations of the Thai original.
502 Pichetpan and Post
References
Adams, Karen Lee. 1989. Systems of numeral classification in the Mon-Khmer, Nicobarese and
Asian subfamilies of Austroasiatic. Canberra: Dept. of Linguistics, Research School of Pacific
Studies, Australian National University.
Aikhenvald, Alexandra Y. 2000. Classifiers: A typology of noun categorization devices. Oxford:
Oxford University Press.
Aikhenvald, Alexandra Y. 2019. Noun categorization devices: A cross-linguistic perspective. In
Alexandra Y. Aikhenvald & Elena Mihas (eds.), Genders and classifiers,1–29. Oxford: Oxford
University Press.
Anantaprayoon, Apichit. 2018. A model of developing the mediators by Buddhist peaceful means:
A case study of Chonburi Court. PhD dissertation. Bangkok, Thailand:
Mahachulalongkornrajavidyalaya University.
Aroonmanakun, Wirote. 1994. Thai. In Frans Plank (ed.), Conversation on noun phrases, EUROTYP
Working Papers, Series 7, Number 20, p. 20. Berlin: The European Science Foundation,
EUROTYP Programme.
Bennett, J. Fraser. 1994. Iambicity in Thai. Studies in the Linguistic Sciences 24(1). 39–57.
Bickner, Robert J. & Thomas J. Hudak. 1990. The nature of “Standard”Thai. Journal of South Asian
Literature 25(1). 163–175.
Bisang, Walter. 1993. Classifiers, quantifiers and class nouns in Hmong. Studies in Language
17(1). 1–51.
Bisang, Walter. 1999. Classifiers in East and Southeast Asian languages: Counting and beyond. In
Jadranka Gvozdanović(ed.), Numeral types and changes worldwide, 113–186. Berlin: de
Gruyter.
Bisang, Walter. 2012. Numeral classifiers with plural marking: A challenge to Greenberg. In Da