From LIKE/LOVE to habitual: the case of Mainland East and Southeast Asian languages

2.1 Introduction

The LIKE/LOVE construction in the languages of MESEA is the concatenation of a LIKE/LOVE verb with another verb. The LIKE/LOVE verbs in question are 爱 ài ‘love’ in Chinese, thích ‘like’ in Vietnamese, ชอบ ˈcʰɔ̂ɔp ‘like’ in Thai, ចូលចិត្ត coolcət ‘like’ in Khmer, ມັກ māk ‘like’ in Lao, and ʧaiʔ ‘like’ in Burmese.^[7] In Chinese, Vietnamese, Thai, Khmer, and Lao, which have the basic VO order, the LIKE/LOVE verb is followed by another bare verb while in Burmese, which has the basic OV order, the LIKE verb is preceded by another verb which is suffixed by a nominalizer (see examples in Section 2.2).

The reason why these LIKE/LOVE verbs are dealt with as constructions is that they are not compositional with regard to indicating habitual aspect. Neither the LIKE/LOVE verb nor the other juxtaposed verb alone can bring about the habitual meaning. It is the entire construction that gives rise to it. This can be confirmed by the fact that the LIKE/LOVE verbs in these six languages can all function as a full verb if co-occurring with a nominal/pronominal object, as exemplified by sentences from Khmer, Lao, and Burmese in (4). In such cases, only the lexical meaning ‘like’ obtains. The habitual reading is not possible.

(4)

As a full verb

a.

Khmer
ញុំ	ចូលចិត្ត	អ្នក
kɲom	coolcət	neak
1sg	like	2sg
‘I like you.’

b.

Lao
ຂ້ອຍ	ມັກ	ແມວ
kʰɔ̏ːy	māk	mɛ́ːw
1sg	like	cat
‘I like cats.’

c.

Burmese
θù	təyoʊʔ	hí̃	əyá̃	ʧaiʔtè
3sg.m	China	dish	very	like.nfut
‘He really likes Chinese dishes.’

The LIKE/LOVE constructions in the languages of MESEA have followed a common grammaticalization path. They all developed into expressing habitual aspect, yet with varying degrees of development. Apart from their variation in degrees, different stages of development co-exist within one and the same language. For instance, in Chinese, synchronically, the verb 爱 ài ‘love’ has three uses: (i) a full verb, (ii) an element that indicates both the meaning of love and the meaning of habitual aspect, and (iii) a habitual marker.

In the following sections, I will document the grammaticalization pathway from LIKE/LOVE constructions to habitual aspect in the languages of MESEA. Firstly, I will provide a detailed account of the varying stages of the grammaticalization based on the model proposed by Heine (2002) (Section 2.2); secondly, I will show that there is an intermediate three-step development during the initial stage of grammaticalization (Section 2.3); and thirdly, I will demonstrate how these LIKE/LOVE constructions diverge in linguistic behavior even when they arrive at a more advanced stage of grammaticalization (Section 2.4).

2.2 From LIKE/LOVE to habitual: stages of grammaticalization

Heine (2002) proposes a context-induced four-stage scenario to explain the intermediate stages in the process from a lexical item/construction to a grammatical item/construction. As indicated in Table 1, at Stage I, only the source meaning occurs; at Stage II, there is a bridging context that gives rise to an inference in favor of a new meaning; at Stage III, there is a switch context that is incompatible with the source meaning; at Stage IV, the grammatical meaning is conventionalized and does not need the support of context.

According to Heine’s model, the grammaticalization process of the LIKE/LOVE constructions in the languages of MESEA shares the same properties at Stage I and Stage IV yet differs at Stage II and Stage III. All LIKE/LOVE verbs still retain their original lexical meanings as illustrated in (4) above, which is characteristic of Stage I; no language has reached the conventionalization stage (Stage IV), which means that the constructions at issue are not conventionalized, and the habitual interpretation still needs the support of context. As regards differences, one of them lies in whether they use the LIKE/LOVE verbs in switch contexts, which is characteristic of Stage III. Among the six national languages investigated, three languages (Chinese, Thai, and Lao) have researched this stage. Their LIKE/LOVE constructions are compatible with the inanimate subject referents and the habitual meaning is the only possible interpretation, as in (3). More examples are given in (5).

Chinese miànfěn ‘flour’ in (5a), Thai ˈtóʔˈmáay ‘wooden table’ in (5b), and Lao kʰɨ̄aːŋhɨ́aːn ‘furniture’ in (5c) are all inanimate subject referents which cannot act either as an experiencer of liking or loving emotions or as an actor of the action or event designated by the second verb. Thus, only the habitual meaning obtains.

However, in Vietnamese, Khmer, and Burmese, the LIKE/LOVE constructions are incompatible with inanimate subject referents, as illustrated by (6) in Khmer.

With regard to the bridging context (Stage II), these six languages vary as to whether they have fully completed this stage. In Chinese, Thai, and Lao, the LIKE/LOVE constructions are fully developed as they can occur in contexts where the meaning of LIKE/LOVE is in contradiction with what is described by the event concerned, as in (7). ‘Get seasick’, ‘stumble while walking’, and ‘get arrested’ are all undesirable, out of the control of the subject referents’ free will, and highly unlikely to be liked or loved. Thus, the habitual meaning is the most plausible interpretation, yet their lexical source meaning cannot be completely ruled out as the meaning of LIKE/LOVE is still possible if a certain occasion arises (e.g., a child may like to get seasick because in this way she will get her parents’ attention). This is one of the crucial differences between the bridging context (Stage II) and the switch context (Stage III), since at Stage III, the original source meaning is completely excluded.

Unlike Chinese, Thai, and Lao, the constructions in Khmer, Burmese, and Vietnamese are incompatible with undesirable events that the subject referents have no control of, as in (8).

Nevertheless, they are compatible with events that the subject referents have control of, be it desirable or undesirable, as in (9). The habitual meaning is present in all three sentences. In (9a), the statement ‘The rich like to look down on the poor’ has an inference that they do it often; in (9b), ‘I like singing’ implies that I sing very often; in (9c), ‘hitting people’ is most probably not liked by people; thus, the habitual meaning becomes more plausible than the LIKE meaning.

(9)

a.

Khmer
អ្នកមាន	ចូលចិត្ត	មើលងាយ	អ្នកក្រ
neakmien	coolcət	məəlŋiey	neakkrɑɑ
rich.people	like	despise	poor.people
(i) ‘The rich like to look down on the poor.’[8]
(ii) ‘The rich often look down on the poor.’

b.

Burmese
ʧəma̰	θaʧʰí̃	sʰòtà	ʧaiʔtè
1sg.f	song	sing.nmlz	like.nfut
(i) ‘I like singing.’
(ii) ‘I sing very often.’

c.

Vietnamese
đứa-trẻ	này	thích	đánh	người
child	this	like	hit	People
(i) ‘This child often hits people.’
(ii) ‘This child likes hitting people.’

The stages of grammaticalization of the constructions across the six languages in MESEA are summarized in Table 2.

Heine’s (2002) four-stage scenario provides a very good explanation for what happens to a linguistic item/construction on the way to grammaticalization, but Heine is also fully aware of the fact that his four-stage scenario “rests on a simplification of the facts considered. First, what it suggests is that there are four discrete stages, while in fact we are dealing with a continuum leading from stage I to stage IV and beyond” (Heine 2002: 86; emphasis mine). Similarly, Traugott (2003: 626) points out that, “[g]rammaticalization phenomena are essentially gradient and variable. They proceed by minimal steps, not abrupt leaps or parametric changes”. This is particularly true concerning the development of the LIKE/LOVE constructions from Stage I to Stage II. It is impossible for a lexical item to jump abruptly into a grammatical meaning that is more favored than the lexical one. Certain intermediate small steps must be involved. Then the question arises: what intermediate steps are relevant for the grammaticalization of a lexical LIKE/LOVE verb into a habitual marker? This question will be discussed in Section 2.3.

2.3 Intermediate steps at the initial stage of grammaticalization

The previous section is devoted to an overall account of the different stages of grammaticalization for the LIKE/LOVE constructions across the languages in MESEA. This section will focus on the relevant intermediate steps of development from Stage I to Stage II.

Pragmatic inference is one of the driving forces of grammaticalization. As observed by Bybee et al. (1994: 286), “[a] semantic change can take place when a certain implication commonly arises with a certain linguistic form. That implication can be taken as part of the inherent meaning of the form, and can even go so far as to replace the original meaning of the form”.^[9] The languages of MESEA are well known for serializing verbs in one clause without overt marking of coordination or subordination. When the LIKE/LOVE verb takes another verb as its complement, the pragmatic inference of habitual aspect commonly arises, especially when the action falls into the category of daily activities that one tends to do very often, e.g., reading, exercising, etc. This semantic extension is quite natural since if one likes/loves to do something, it is probable that one will do it frequently. This tendency is given in (10).

Exercising in (11a) and going shopping in (11b) are daily, recurring activities. This naturally gives rise to a habitual inference as the participant enjoys doing it.

This is the first step for the habitual meaning to arise. Then over time, this pragmatic inference evolves into the second step, which involves the extension from desirable actions into something generally considered to be socially undesirable, as in (12). Actions such as seeking revenge in (12a) and cursing in (12b) are actions with typically perceived negative features, which are less likely to be liked or enjoyed by people other than the subject referent. Thus, the interpretation of habitual aspect is more prominent than the original lexical LIKE/LOVE meaning.

(12)

a.

Chinese
他	爱	打击报复
tā	ài	dǎjíbàofù
he	love	take.revenge
(i) ‘He often seeks revenge.’
(ii) ‘He likes seeking revenge.’

b.

Burmese
θù	lù	sʰétà	kò	ʧaiʔtè
3sg.m	person	curse	nmlz	like.nfut
(i) ‘He curses a lot.’
(ii) ‘He likes cursing.’

While the second step involves undesirable actions that the actors have control/preference over, the third step involves the contextual extension to the kind of undesirable events that the human participants have no control over, as in (7) above. More examples are given in (13). Getting seasick (7a), stumbling while walking (7b), getting arrested (7c), not catching the bus (13a), and being forgetful (13b) are all events that the human participants cannot control, do not expect to happen, and are thus in contradiction with the LIKE/LOVE meaning. However, the original lexical meaning still cannot be completely ruled out, as some people may take delight in unfortunate events. Then at this step the lexical source meaning becomes marginal with the habitual meaning being dominant. This third step is what Heine calls the stage of the bridging context (Stage II).

This context-induced three-step development from the lexical source meaning to the habitual meaning at Stage II is summarized in Table 3.

As seen from Table 3, the development from Stage I to Stage II involves three gradual steps, at which contextual extensions play a vital role. As remarked by Traugott (2003: 624), “early in grammaticalization, lexemes grammaticalize only in certain highly specifiable morphosyntactic contexts, and under specifiable pragmatic conditions”. The extension from desirable events (Step 1) to socially undesirable controllable events (Step 2), and then to undesirable uncontrollable events (Step 3) witnesses an increase in the degree of contradiction with the original LIKE/LOVE meaning, which is paralleled with an increase in the prominence of the grammatical meaning and a decrease in the presence of the lexical source meaning. As a result, the defeasibility of the habitual meaning from Step 1 to Step 3 also diminishes. As the grammaticalization process is still ongoing in MESEA, this step-by-step contextual extension is observable in synchronic data. For instance, in Chinese as in (14a), at Step 1, the habitual inference can be readily canceled, leaving LOVE as the only reading (desirable activity), whereas at Step 2, as in (14b), the habitual meaning can hardly be canceled (socially undesirable controllable behavior), and at Step 3, it cannot be canceled at all as in (14c) (undesirable uncontrollable event). In both (14b) and (14c), it is either very odd or semantically unacceptable for the construction to have the LOVE meaning only.

If a meaning cannot be canceled, that means that it is not pragmatically inferred, but entailed and semanticized as part of the inherent meaning of the item/construction. However, as seen in many examples in this paper, this semanticization is limited to the contexts at Step 2 and Step 3 as well as at Stage III (switch context). This might indicate that the grammaticalization process is still ongoing; nevertheless, it does not necessarily follow that the process is new and recent. For instance, the Chinese ài construction already had a habitual use even dated back to the Five Dynasties (907–960 AD) as pointed out by Ota (2003: 190).

In (15), the two clauses are in parallel structure, and each word or phrase in the first clause is symmetric in meaning to the counterpart(s) in the second clause. Ài is presented to have a similar meaning to pín ‘frequently’. Although the Chinese ài construction had already developed a habitual meaning even more than a thousand years ago, it is highly likely that the co-existence of the LIKE/LOVE meaning and the habitual meaning will continue for a long time. This might be due to two main reasons. One is that habitual expressions have a relatively low frequency of occurrence as compared to progressive or anterior (Bybee et al. 1994: 159), which may slow down their evolutionary advancement; the other is that, as in many languages of the world, a grammaticalization process may retain the lexical use of the linguistic item even when it is highly grammaticalized. It is comparatively common for the traces of diachronic change to be observable synchronically.

Earlier discussions in this section have shown that the constructions in Vietnamese, Khmer, and Burmese have not reached Step 3 whereas those in Chinese, Thai, and Lao not only reached it but went beyond to arrive at the stage of the switch context. This distribution is given in Table 4.

One of the possible speculations for this variation is that at Step 1 in the bridging stage, the constructions in Vietnamese, Khmer, and Burmese do not usually co-occur with the activities that people tend to perform very often or can occur in contexts that do not support a habitual reading, and thus give rise to other meanings, which consequently results in the grammatical meaning of habitual aspect not being inferred frequently enough in the language to prompt its full development at the stage of the bridging context. This may have to do with the fact that, in addition to the LIKE/LOVE meaning, the constructions in Vietnamese, Khmer, and Burmese can have the volitional meaning ‘prefer’ as in (16).

It seems also quite natural for an enjoyment meaning to develop into a meaning of volition. If one enjoys (likes/loves) doing something, naturally one would want/prefer to do it. Presumably, the multiple meanings of the constructions in Vietnamese, Khmer, and Burmese could give rise to multiple pathways of grammaticalization, which may somehow hinder their further development of the habitual meaning. However, this is usually not the case in Chinese, Thai, and Lao. For instance, as in (17), it is ungrammatical to use the ài construction in the context that facilitates a ‘prefer’ reading.^[11] Undoubtedly, given the intense language contact in this region, this is a rather complex issue. Historical evidence and research are needed to see how this has evolved diachronically.

2.4 Divergences at the later stage of grammaticalization

2.4.2 Stative verbs

Based on the data samples, in general, all the LIKE/LOVE constructions have a preference for dynamic verbs, as in (18).

(18)

a.

Khmer
គាត់	ចូលចិត្ត	លេង	ទូរ	ទូរស័ព្ទ
koat	coolcət	leeŋ	tuu	tuː.rĕəʔ.sap
3sg.m	like	play	compare	cell.phone
(i) ‘He likes to play on the phone.’
(ii) ‘He plays on the phone often.’

b.

Chinese
这	种	木头	很	爱	裂
zhè	zhǒng	mùtou	hěn	ài	liè
this	type	wood	very	hab	crack
‘This type of wood tends to crack easily.’

c.

Burmese
ʧəma̰	yízá	eiʔ	kʰàní	yèiʧʰótà	ma̰	ʧaiʔ	bú
1sg.f	boyfriend	sleep	before	shower	neg	like	neg
(i) ‘My boyfriend doesn’t like to take a shower before going to bed.’
(ii) ‘My boyfriend often does not take a shower before going to bed.’

Stative verbs such as ‘know’ and ‘resemble’ differ from other Aktionsart types of verbs in that the de-compositional analysis of a stative verb results in the homogeneity of subevents (Pustejovsky 1991). The combinability of the constructions in the languages of MESEA with stative verbs varies depending on their degree of grammaticalization. The more advanced in their degree of grammaticalization, the more compatible they are with stative verbs. The LIKE/LOVE constructions in Vietnamese, Khmer, and Burmese are less grammaticalized and thus incompatible with stative verbs as illustrated by Vietnamese in (19).

(19)

Vietnamese
*Tôi	thích	tin	rằng	bạn	không	thích	tôi
1sg	like	believe	that	you	neg	like	1sg
‘I tend to believe that you don’t like me.’

In the case of more grammaticalized constructions in Chinese, Thai, and Lao, they also vary as regards their combinability. The Chinese ài construction usually does not co-occur with stative verbs, yet there are marginal cases in which the co-occurrence is acceptable under certain circumstances. Zhī-dào ‘know’ is stative as in (20a). However, native speakers find it odd when it combines with ài as in (20b), which is only acceptable when ài is emphasized and has the LOVE meaning only. Rèn-wéi ‘believe’ is also stative, yet it is compatible with the ài construction as in (20c), in which the presence of ài indicates a series of bounded states of believing, instead of a single, continuous state.

(20)

Chinese

a.

我们	知道	你	会	同	我们	一起	去
wǒmen	zhīdào	nǐ	huì	tóng	wǒmen	yiqǐ	qù
1pl	know	2sg	fut	with	1pl	together	go
‘I know that you will go together with me.’

b.

?我们	爱	知道	你	会	同	我们	一起	去
wǒmen	ài	zhīdào	nǐ	huì	tóng	wǒmen	yīqǐ	qù
1pl	love	know	2sg	fut	with	1pl	together	go
‘I love to know that you will go together with me.’

c.

四

和

五

年级

的

同学,

平时

就

爱

sì

hé

wǔ

niánjí

de

tóngxué,

píngshí

jiù

ài

four

and

five

grade

attr

student

usually

then

love

认为

自己

长大

了

rènwéi

zìjǐ

zhǎngda

le

believe

oneself

grow.up

perf

(i) ‘Fourth and fifth grade students usually love to believe that they have grown up.’

(ii) ‘Fourth and fifth grade students usually often believe that they have grown up.’

While the combination of stative verbs with ài is marginal in Chinese, it is more common in both Thai as in (21) and Lao as in (22).

(21)

Thai
ฉัน	ชอบ	คิด	ว่า	เธอ	ไม่	รัก	ฉัน
ˈcʰǎn	ˈcʰɔ̂ɔp	ˈkʰít	ˈwâa	ˈtʰəə	ˈmây	ˈrák	ˈcʰǎn
1sg	like	believe	say	2sg	neg	love	1sg
‘I tend to believe that you don’t love me.’

(22)

Lao

a.

Context: the same speaker is describing what local people think about burial forests and cemeteries. (Enfield 2007: 222)
juu 1	bὸὸn 1	nan 4	hanø	mak 1 -caø	thùù 3	vaa 1	man 2
be.at	place	dem.nprox	top.dist	like-irr	believe	comp	3.b

mii 2 -

mii 2 -

mii 2

phii 3

there.is

there.is

there.is

spirit

‘In those places, (people) often believe that there are- there are there are spirits.’

b.

Enfield (2007: 520)
khacaw 4	bὸø	khὸὸj 1	mak 1	juu 1	hùan 2	hanø	naa 3
3pl.p	neg	much	like	exist	house	top.dist	fac.fillin
‘They (the boss and her colleagues) didn’t tend to be at home, you see.’

In (21), lexically, ˈkʰít ‘believe’ denotes an unbounded state. When it combines with ˈcʰɔ̂ɔp, this unbounded single state refers to a series of episodic situations. The single state of ‘believe’ becomes recurrent on a series of occasions. The same is true for (22) in Lao, in which the combination of mak ¹ with thùù ³ ‘believe’ in (22a) and the combination with juu ¹ ‘exist’ in (22b) result in a shift in meaning from denoting a single unchanged state to denoting a series of states.

According to Heine (2003: 580), during the process of grammaticalization, “extension obtains when a linguistic item can be used in new contexts where it could not be used previously”. If all the constructions are heading towards the same direction of change, and constructions diverge in whether they are extended to certain contexts (in this case, stative verbs), then we may see this extension as a sign of further development along the common grammaticalization path. This means that, regarding the combination with stative verbs, among the three most advanced constructions, the constructions in Thai and Lao are most likely to be more advanced in grammaticalization than that in Chinese.

Among the studies on habituality, whether a habitual marker can combine with stative verbs is still under debate. On the surface, it seems that the LIKE/LOVE constructions have undergone a similar grammaticalization process to English used to, as remarked by Bybee et al. (1994: 156): English used to “generalized from use with only human subjects to use with subjects of all types, and in addition it has come to be employed with stative verbs, giving the sense of a past state” as in (23).

(23)

My father used to have a car like that.

However, not all researchers agree that used to is a marker of habituality in this case. Comrie (1976: 27) defines habituality as describing “a situation which is characteristic of an extended period of time”, thus including cases such as (24) in which a stative verb co-occurs with a habitual expression to indicate a single state that lasts throughout a certain single period, without intermission.

(24)

The temple of Diana used to stand at Ephesus.

Binnick (2005: 348–349), however, argues that used to is not a habitual marker as it does not concern a recurrent series of eventualities (2005: 350). Instead, it functions as a kind of present perfect although it has past tense morphology. On the contrary, unlike used to, English would marks habituality inherently as it can refer to a series of states as in (25), which sounds odd yet means that the temple of Diana stood at Ephesus on a series of separate occasions.

(25)

The temple of Diana would stand at Ephesus (from time to time).

Whether used to is a marker of habitual aspect is still debatable, but what is true is that both used to and would can combine with stative verbs, though with different effects. The LIKE/LOVE constructions in Thai, Lao, and Chinese are like would, in that, when they combine with stative verbs, their presence turns the sentence that indicates a single state into one that indicates a series of separate states. This strongly suggests that the habitual meaning denoted by these LIKE/LOVE constructions is confined to the plurality of episodic situations, which is quite in line with the definition given by Bybee et al. (1994: 127): habitual aspect describes situations “customarily repeated on different occasions” as well as the one proposed by Verkuyl (1993: 325–327, 1995, cited from Tatevosov 2005: 193): “unbounded pluralization of temporal intervals”.

If viewed from a grammaticalization perspective, presumably, habitual expressions may be initially compatible with dynamic verbs and then are extended to co-occur with stative verbs, as dynamic events are more likely to be repetitive than states. If a habitual expression does not bring a single state into the repetition of this state when it co-occurs with a stative verb, then the question is whether this expression is a dedicated habitual marker, as seen in the debate over whether English used to is a habitual construction.

2.4.3 Future temporal context

The languages of MESEA are well known for their paucity of inflectional marking of grammatical information that is usually expressed morphologically in European languages. The grammatical information such as tense or aspect is usually not obligatorily expressed and can be inferred from context. For instance, in (26), without a specific context, its grammatical meaning of tense and aspect is indeterminate.

(26)

Chinese (Bisang 2004: 111)
(Ta)	lai.
s/she	come
‘(S/He) comes/has come/is coming/will come/etc.’

With respect to the LIKE/LOVE verbs, they are not marked obligatorily for tense, except for Burmese ʧaiʔtè, which is usually suffixed by a non-future marker as in (27).

(27)

Burmese
θù	lèĩtà	kò	ʧaiʔtè
3sg.m	lie.nmlz	nmlz	like.nfut
(i) ‘He likes telling lies.’
(ii) ‘He often tells lies.’

Hence, the temporal contexts in which the LIKE/LOVE constructions occur are usually determined either by pragmatic context or by explicit temporal expressions such as nouns or auxiliaries. The common temporal contexts in which all the constructions can occur are the present, as in (28a) and (28c), and the past, as in (28b) and (28c).

(28)

a.

Khmer (present temporal context)
ប៉ា	ចូលចិត្ត	ញ៉ាំ	ស៊ុប	ឆ្នាំង	ភ្លើង
paa	coolcət	ɲam	sup	cnaŋ	pləəŋ
father	like	eat	soup	pot	fire
(i) ‘Father likes to eat hotpots.’
(ii) ‘Father eats hotpots often.’

b.

Thai (past temporal context)
ฉัน	เคย	ชอบ	กิน	ไก่	ทอด
ˈcʰǎn	ˈkʰəəy	ˈcʰɔ̂ɔp	ˈkin	ˈkày	ˈtʰɔ̂ɔt
1sg	ever	like	eat	chicken	fry
(i) ‘I liked to eat fried chicken.’
(ii) ‘I ate fried chicken often.’

c.

Chinese (past and present temporal contexts)
我	以前	爱	喝	酒,	现在	爱	读	书
wǒ	yǐqián	ài	hē	jiǔ,	xiànzài	ài	dú	shū
1sg	past	love	drink	wine	now	love	read	book
(i) ‘I loved to drink, (but) now I love to read.’
(ii) ‘I drank, (but) now I read a lot.’

Regarding the future temporal context, however, constructions differ in whether the habitual meaning obtains although they all can occur in the future context. In Vietnamese, Burmese, Khmer, Chinese, and Thai, the habitual interpretation rarely holds in a future context. Only the LIKE/LOVE meaning remains. For instance, in Thai as in (29e), aˈnaaˈkʰót ‘future’ and kʰoŋ ja⁐ ‘will’ are two expressions that indicate the future temporal reference. In such a context, the sentence means ‘I will enjoy reading books’ and whether I will read often is irrelevant.

(29)

a.

Vietnamese
Nhất-định	người	Hongkong	sẽ	thích	xem	phim	này.
certainly	people	Hongkong	fut	like	watch	movie	this
(i) ‘Hongkong people will certainly like to watch this movie.’
(ii) *‘Hongkong people will certainly watch this movie often.’

b.

Burmese
ʧənɔ̀	naʊ̀̃ta̰kʰà	beʔsa̰kaʔbɔ́	sʰɔ̰tà	ʧaiʔmè
1sg.m	future	basketball	play.nmlz	like.fut
(i) ‘I will like playing basketball in the future.’
(ii) *‘I will play basketball often in the future.’

c.

Khmer
បងស្រី	នឹង	ចូលចិត្ត	ញ៉ាំ	ហាំប៊ឺហ្គឺ
bɑɑŋsrǝy	nɨŋ	coolcət	ɲam	ham.ɓɨː.ɡɨː
elder.sister	will	like	eat	hamburger
(i) ‘My elder sister will like eating hamburgers.’
(ii) *‘My elder sister will eat hamburgers often.’

d.

Chinese
我	会	爱	看	电视	的
wǒ	huì	ài	kàn	diànshì	de
1sg	fut	love	watch	tv	cert
(i) ‘I will love to watch TV, for sure.’
(ii) *‘I will watch TV often, for sure.’

e.

Thai
ใน	อนาคต	ฉัน	คงจะ	ชอบ	อ่าน	หนังสือ
ˈnay	aˈnaaˈkʰót	ˈcʰǎn	kʰoŋ ja⁐	ˈcʰɔ̂ɔp	ˈàan	naŋˈsʉ̌ʉ*
In	future	1sg	will	like	read	book
(i) ‘In the future I will like to read books.’
(ii) *‘In the future I will read books often.’

That the habitual meaning is unobtainable in a future context in these constructions can be confirmed by the fact that when a habitual meaning is saliently present as in the cases where the construction co-occurs with an undesirable activity (cf. Section 2.3), the sentence becomes barely acceptable in a future context. (30a) means that she is forgetful on a series of separate occasions. The presence of the future marker huì in (30b) renders the sentence rather odd. It is only acceptable under very specific circumstances such as in a community where being forgetful is favored. However, even under such a circumstance, the habitual reading is impossible.

(30)

Chinese

a.

她	爱	忘事
tā	ài	wàngshì
3sg	love	forget
‘She tends to be forgetful.’

b.

? 她	会	爱	忘事
tā	huì	ài	wàngshì
3sg	fut	love	forget
‘She will tend to be forgetful.’

By contrast, the construction in Lao can occur in a future context as in (7c) (repeated here as [31a]), and in (31b).

(31)

Lao

a

(Enfield 2008: 120)
phu 0 -ñing 2	nùng 1	sakeet 5	vêlaa 2	nan 4	han 0	mak 1	ca 0
person-F	wear	skirt	time	dem.nprox	top.part	like	irr
thùùk 5	cap 2
suffer	catch
‘Women who wore skirts at that time would tend to get arrested.’

b.

ຖ້າ	ບໍ່	ຕັ້ງໃຈ	ຮຽນ,	ມັກ	ຈະ	ໄດ້	ຄະເເນນ	ບໍ່	ດີ
tʰaː̏	bɔ̄ː	tȃŋcay	híːan	māk	cha	dai	khaeenn	bo	di
if	neg	seriously	study	like	irr	acquire	grade	neg	good
‘If (you) do not study hard, (you) would tend to get bad grades.’

Nevertheless, this does not mean the Lao construction always has a habitual reading in a future context. There are cases in which it does not, as in (32). This shows that the extension into a future context is not yet completed for the Lao construction.

(32)

Lao
ມຶ້ໜ້າ	ຂ້ອຍ	ມັກ	ຈະ	ຕັ້ງ ໃຈ	ຮຽນໜັງສື
mu-na	khony	māk	cá	tȃŋcay	híːannăŋsɨ̆ː
future	1sg	like	irr	seriously	study
(i) ‘In the future, I will like studying hard.’
(ii) *‘In the future, I will study hard often.’

If the future temporal context is seen as a contextual extension into a new context along the same grammaticalization path, like the case of stative verbs, then the Lao construction has undergone a further step in its process of grammaticalization compared to all the other constructions at issue. This evidence somehow predicts how the other relevant constructions will evolve further.

However, this temporal extension may be true only to the evolution of the LIKE/LOVE constructions in MESEA. Cross-linguistically, habitual expressions vary greatly as to what kinds of temporal contexts they can occur in. In the past, it was believed that habituals tend to combine with past tense or occur in past contexts, as observed by Comrie (1976: 72). This is somehow echoed by Dahl (1985: 100–111), who claims that it is very hard to find the combinations of habitual markers with future markers, and habitual categories may be used secondarily in predictive contexts. However, more recent cross-linguistic investigation finds that this is not the case. For instance, Bybee, Perkins, and Pagliuca’s (1994: 151–160) corpus study reveals that habitual markers are compatible with all temporal contexts, past, present, and future. In their dataset, 17 languages have overt habituals that are not restricted by tense; 9 languages have specific past habituals; and 2 languages have present habitual. Hence, cross-linguistically, the occurrence of habitual markers in future context is by no means rare. In fact, the future tense is so closely related to the habitual marker that in many languages such as English, Welsh, Lezgian, Uralic languages, tense/aspect markers have a polysemy of habitual and future tense. This is also attested in pidgins and creoles. As observed by Maurer and the APiCS consortium (2013), 24 languages out of 76 pidgins and creoles, the same marker can express both habitual and future values.

Given the great variation in the temporal contexts that habitual expressions can occur in, more diachronic research is needed to see what kind of temporal contexts habitual expressions originally occur in and what kind of contexts are later extensions during their process of grammaticalization.

2.5 Summary

In Section 2, I have documented the grammaticalization pathway from LIKE/LOVE constructions to habituals based on the synchronic data from languages in MESEA. The LIKE/LOVE constructions in the languages of MESEA differ as to their stages and degrees of grammaticalization. In Chinese, Thai, and Lao, the grammaticalization process is relatively advanced in the sense that they have reached Stage III (switch context), whereas in Vietnamese, Khmer, and Burmese, this process is incipient in the sense that they haven’t completely reached Stage II (bridging context). As the development from Stage I to Stage II cannot be abrupt, I have postulated that three intermediate steps have to be taken to reach Stage II. At Step 1, the constructions occur in contexts of desirable daily activities. Then the constructions are extended to be used in new contexts: contexts of socially undesirable activities which the subject referents have control of (Step 2), and contexts of undesirable activities which the subject referents have no control of (Step 3). At both Step 2 and Step 3, the habitual meaning becomes indefeasible and increasingly semanticized. The constructions in question are also found to diverge at later stages of their common grammaticalization process. Although the constructions in Chinese, Thai, and Lao all have reached Stage III, the construction in Lao is more advanced than the constructions in the other two languages as it is compatible with stative verbs and future temporal context whereas the construction in Chinese is only marginally compatible with stative verbs and incompatible with future context, and the construction in Thai is incompatible with future context despite its compatibility with stative verbs. The discussions and findings are summarized in Table 5. In this table, it is just for the sake of convenience that the three extensions beyond Stage II are presented in a sequence. There is no attempt to indicate that one is the further extension of another. They are independent of each other and are most probably in parallel development.