On the syntax of wan ‘finish/complete’ in Mandarin Chinese

ABSTRACT

In this study, I revisit the linguistic properties of wan ‘finish/complete’ in Chinese and provide a syntactic account. I demonstrate that wan is undergoing a process of grammaticalization from being a lexical item to a functional one. I also argue that because of this process, wan is found in different places in syntax. As a main predicate, it is projected in the head of the VP; in contrast, while functioning as a telic morpheme, it is in the head of the Inner Aspect Phrase between the vP and VP. This study sheds light on the Chinese aspectual system and suggests that Chinese uses at least two different mechanisms to mark telicity: resultative complements and overt telic morphemes.

KEYWORDS:

• Inner Aspect Phrase
• Chinese V–wan compound
• Chinese V–V resultatives
• grammaticalization

1. Introduction

The main function of wan ‘finish/complete’ in Modern (Mandarin) Chinese has been identified as a resultative complement, appearing after the first verb in a V–V compound (e.g. Li & Thompson, 1981; Wu, 2004). Observe the three examples in (1) below:¹

Table

(1)	a.	Lisi	ca-ganjing	beizi	le.
		Lisi	wipe-clean	glass	LE
		‘Lisi has wiped the glass(es) clean.’
	b.	Lisi	xie-wan	zuoye	le.
		Lisi	write-complete	homework	LE
		‘Lisi has finished writing the homework.’
	c.	Lisi	kan-wan	na	bu	dianying	le.
		Lisi	watch-complete	that	CL	movie	LE
		‘Lisi has finished watching the movie.’

The sentence in (1a) exemplifies a typical V–V resultative compound in which the compound appears before the object and the result-denoting element predicates over the object. Similarly, wan in (1b) predicates over the object zuoye ‘homework’, expressing that the homework has been completed, which would be the result of Lisi’s writing. Finally, in (1c), although the completion of the movie is not directly due to Lisi’s watching, the result state can be somewhat identified. As illustrated, the first element of the compound in all three examples denotes an activity that is atelic. In addition to describing a result state, the second element turns the first one telic. Owing to the morphosyntactic and semantic properties of wan presented in the examples above, it is not surprising that wan has been categorized as a type of resultative complement.

However, a sentence containing a V–wan compound shows several properties that differ from a typical V–V resultative compound. For example, wan in a V–wan compound does not always specify what the result state is. In contrast, in a typical V–V resultative compound, the result state is used to describe the state of either the subject or the object. Previous studies, such as Chao (1968), Li and Thompson (1981), and Sybesma (2017), have noticed this difference. Li and Thompson (1981), for example, indicated that wan simply expresses “something more like the type of action described by the first verb or the degree to which it is carried out than the result” (p. 65). Sybesma (2017) also noticed the variations shown in a V–wan compound and claimed that wan may offer a dual function and does not always function as a resultative complement. For example, the author indicated that a typical regular V–V resultative is compatible with the progressive aspect, but not a V–wan compound.

Moreover, in a sentence containing wan and a VO verb, such as chi-fan ‘eat-rice: eat’, wan can occasionally appear after the VO cluster, which is not generally observed in a V–V resultative. For example, the resultative compound ca-gan beizi ‘wipe-dry glass’ can never be rephrased as ca-beizi gan ‘wipe-glass dry’. Nevertheless, the word order of VO–wan can be found in corpora such as the Academia Sinica Balanced Corpus of Modern Chinese (2013) as well as the Baidu Search Engine (2020). Examples include jie-hun wan ‘knot-marriage finish: finish getting married’, shua-ya wan ‘brush-tooth finish: finish brushing teeth’, chi-fan wan ‘eat-rice finish: finished eating’ and several others. Furthermore, the results of a judgement test also suggest the acceptance of the cluster of VO–wan by native speakers.²

Finally, another morphosyntactic variation between the two types of compounds can be observed when they interact with a durative phrase, as (2) and (3) illustrate:

Table

(2)	Lisi	zuotian	ba	qian	men	da-kai-le	shi	fenzhong.	(V–V resultative)
	Lisi	yesterday	BA	front	door	hit-open-Prf	10	minute
	‘Lisi had the front door open for 10 minutes yesterday.’

Table

(3)	*Lisi	zuotian	kan-wan-le	xiaoshuo	shi	fenzhong.			(V–wan compound)
	Lisi	yesterday	read-wan-Prf	novel	10	minute

Both (2) and (3) are perfective and contain the durative phrase shi fenzhong ‘(for) 10 minutes’. The durative phrase in (2) is compatible with a typical V–V resultative compound and is used to measure the temporal duration of the result state. In contrast, the same phrase is not compatible with a V–wan compound, as (3) shows, given that the compound does not express a salient result state that the durative phrase can measure.³ If we assume that different types of durative phrases are sensitive to syntax, as suggested by Li (1987), Liao (2015) and Lin (2008), the examples in (2) and (3) seem to suggest that a V–wan compound is also morphosyntactically different from a typical V–V resultative compound. Thus, the main purpose of this study is to provide a syntactic analysis that can account for the semantic differences shown in a sentence containing wan. As previous studies such as Borer (2005), Ritter and Rosen (2005), Slabakova (2001) and Travis (2005, 2010) have suggested, in addition to the resultative construction, languages use other mechanisms, such as telic morphemes (overt or covert) to mark telicity; it is thus interesting to investigate the Chinese data to see if this language also uses a mechanism other than resultative complement to compute telicity.

Two main issues will be addressed in this paper. First, I will investigate the historical linguistic development of wan by comparing its usage in both Classical and Modern Chinese. Following Wu (2004), I will show that wan is transforming from a pure lexical item to a functional morpheme whose main function is to signal telicity; that is, it is undergoing a grammaticalization process. The second issue that will be addressed is the syntactic distribution of wan. According to Roberts and Roussou (2003), grammaticalization can be realized in syntax and frequently results from the movement of a lexical element (or a functional one) into a functional structure that immediately dominates the lexical element.⁴ If the proposal of wan’s development into a telic morpheme is on the right track, then we would expect the syntactic movement of wan to be from a VP to a functional phrase that immediately dominates VP. In line with Xuan (2008) and Sybesma (2017), I demonstrate with additional evidence that the functional phrase is the Inner Aspect Phrase projected between the vP and the VP. Syntactically, this phrase immediately dominates the VP in which wan is originally base-generated. This fulfils the requirements of the process of grammaticalization. Semantically, this functional phrase encodes telic information and is consistent with the function of wan.

The rest of this paper is organized as follows. I present the linguistic properties of wan in detail in §2. Section 3 gives a syntactic account of the linguistic properties of wan. In §4, a detailed discussion of the consequences of the new proposal is presented. Finally, I conclude the paper in §5.

2. The linguistic properties of wan

This section briefly introduces some linguistic properties of a V–V resultative compound as well as a V–wan compound. The focus is to demonstrate that wan behaves differently from many resultative complements, not only semantically but also morphosyntactically.

2.1 The V–wan compound and V–V resultative construction

The sentences in (4) give two additional examples of a typical V–V resultative construction:

Table

(4)	a.	Lisi	ca-gan-le	bolibei.
		Lisi	wipe-dry-Prf	glass
		‘Lisi wiped the glass(es) dry.’
	b.	Lisi	pao-lei	le.
		Lisi	run-tired	LE
		‘Lisi has become tired as a result of running.’

As illustrated, the second element in a typical V–V resultative construction can be used to specify either the state of the direct object or the subject of a sentence (Hoekstra, 1988; Huang, 2006; Huang et al., 2009; Sybesma, 1999; Tang, 1997). For example, the predicate gan ‘dry’ in (4a) is used to specify the state of the direct object as a result of the subject’s action. In contrast, the word lei ‘tired’ in (4b) is used to describe the state of the subject, as it contains an intransitive verb. Similarly, wan may also appear immediately after a verb to form a compound, as illustrated in (1b–c). However, in a sentence containing a V–wan compound, the result state is not always clear.

The second element of a V–V resultative compound, as well as wan, can also function as a one-place predicate. The examples in (5) both have an inchoative reading:⁵

Table

(5)	a.	Bolibei	gan	le.
		glass	dry	LE
		‘The glass is dry (now).’
	b.	Qizhong	kaoshi	wan-le	yihou,	ni	zai	lai	jian	wo.
		Midterm	exam	complete-Prf	after	you	then	come	meet	I
		‘Come meet me after the midterm exam is over.’

In addition, wan also has a transitive usage, found in Classical Chinese, as (6) illustrates:

Table

(6)	a.	…  bei	bianjing,	wan	yaosai.
		prepare	border	complete	fortress
		‘ … to prepare for the border and complete the fortress.’
		(吕氏春秋 Lü Shi Chunqiu, Mengdong ji, sec. 15)
	b.	Wo	shi	Zhang	jia	zisun,	lishang	xu	xian	wan	Zhang
		I	be	the	Zhang’s	descendant	courtesy	must	first	finish	Zhang
		jiade	shi.
		family’s	business
		‘I am descendant of the Zhang family and out of courtesy, I must first finish the Zhang family’s business.’
		(初刻拍案驚奇 Chuke Pai’an Jingqi, sec. 38)

As the two examples depict, wan has a transitive usage in both (6a) and (6b), more specifically, a causative usage meaning ‘to complete’ or ‘to finish’, followed by the direct objects, yaosai ‘fortress’ and shi ‘tasks/business’, respectively.

Nevertheless, in Modern Chinese, wan does not generally have a causative usage, as illustrated in (7a). It can, however, combine with a nonreferential object to form a VO compound, as shown in (7b):

Table

(7)	a.	*Lujun	wan	yaosai	le.
		army	complete	fortress	LE
		Intended reading: ‘The army has completed the fortress.’
	b.	Zhangsan	he	Lisi	wan-hun	le.
		Zhangsan	and	Lisi	finish-marriage	LE
		‘Zhangsan and Lisi have married (/have finished getting married).’

A comparison of (6) and (7) suggests that while the causative usage of wan in Classical Chinese is not uncommon, it is not the case in Modern Chinese. As presented in §1, the more common usage of wan in Modern Chinese is to signal telicity, specifically when the word appears immediately after a verb to form a V–wan compound. Given that a V–wan compound shows the same morphosyntactic properties as a V–V resultative and these two types of constructions are similar in that the compounds as a whole denote a telic event, it is not surprising that the V–wan compound has been traditionally grouped into a subtype of the V–V resultative construction. In the following subsection, nevertheless, I will illustrate some variations shown by a V–wan compound.

2.2 Linguistic variations of wan

As presented in §2.1, one of the reasons why a V–wan compound has been identified as a type of V–V resultative relates to its morphosyntactic property regarding a verb in the same sentence; both wan and the resultative complement appear immediately after the verb and before the object (if any). However, a crucial element that is lacking in many sentences with a V–wan compound is a clearly specified result state. This raises the question of whether wan should always be considered a resultative complement when it is attached to a verb, as many previous studies have argued that languages generally use more than just resultative complements to mark telicity (Borer, 2005; Slabakova, 1997, 2001; Sybesma, 2017; Travis, 2005, 2010).⁶

Previous studies on wan have noted the difference between a V–wan compound and a typical V–V resultative compound. For example, Chao (1968) called wan a “phase complement” whose main function is to signal only the completion of an event. Similarly, Li and Thompson (1981) indicated that this morpheme belongs to a special type of resultative complement: a phase resultative complement.⁷ In addition to wan, several other complements belong to the same group. Some examples of phase complements from Li and Thompson (1981) are presented in (8):

Table

(8)	a.	chang-wan	‘sing-finish = finish singing’
	b.	cai-zhao	‘guess-be on target = guess (something) right’
	c.	zuo-hao	‘do-complete task = complete the task of doing’
	d.	kan-dao	‘see-arrive = succeed in seeing’
		(Li & Thompson, 1981, p. 65 [101])

These phase complements are similar in that they all appear after a verb denoting activity and have the same distribution as a V–V resultative. Semantically, most of these complements contain different idiosyncratic information, but they all express telic information.

Lu et al. (2019) and Sybesma (2017) also observed the variations that wan shows in a V–wan compound, and both studies have pointed out that in a V–wan compound wan may have a dual function and does not always function as a resultative complement. Observe both examples taken from Lu et al. (2019) in (9a) and (9b) below:

Table

(9)	a.	Ta	he-wan-le	ni	zuotian	gei	ta	zuo	de	tang.
		he	drink-finish-Prf	you	yesterday	for	him	make	DE	soup
		‘He finished the soup you prepared for him yesterday.’
		(Lu et al., 2019, p. 297 [32a])

Table

	b.	Keting	wo	zao-jiu	ca-wan-le.
		Living.room	I	long-since	wipe-finished-Prf
		‘The living room I finished cleaning a long time ago.’
		(Lu et al., 2019, p. 297 [31a])

According to Lu et al. (2019), in (9a) wan has a full lexical meaning and functions as a resultative complement that predicates over the direct object ni zuotian gei ta zuode tang ‘the soup you prepared for him yesterday’. In contrast, the authors indicated that in (9b), although wan also appears immediately after a verb, it cannot be analyzed as a resultative complement. First, the object keting ‘living room’ cannot be predicated over by wan, leading to an interpretation like ‘there was a cleaning event, and the result was that the living room was completed’. According to the authors, the meaning of (9b) is more appropriately expressed as ‘I’m done cleaning the living room’, and wan is used to express the successfulness or completion of the cleaning event. In this case, wan does not express the result state but simply provides telic information to the first verb of the compound. These studies have pointed out the dual nature of wan in a V–wan compound and have suggested that the nature is because wan is undergoing a grammaticalization process.⁸

Another variation shown in a V–wan compound is that unlike many V–V resultative compounds, it is not compatible with the progressive aspect. The example in (10) illustrates the difference:

Table

(10)	a.	Ta	(zheng)	zai	ca-gan	boli.
		he	just	be.at	wipe-dry	glass
		‘He is wiping the glass dry.’
		(Sybesma, 2017, p. 190 [11a])
	b.	*Wo	(zheng)	zai	sao-wan	keting.
		I	just	Prog	swipe-finish	living.room.

According to Sybesma (2017), a V–V compound such as ca-gan ‘wipe-dry’ in (10a) is compatible with the progressive aspect. This suggests that the compound is an event denoting accomplishment. In contrast, the fact that the V–wan compound in (10b) is not compatible with the progressive aspect indicates that a V–wan compound is an event denoting achievement. Although the difference between these two types of telic events concerning their compatibility with the progressive aspect is more complicated (see Basciano (2019) and Tham (2013) for more details), the crucial point here, based on Sybesma (2017), is that a V–wan compound is generally incompatible with the progressive aspect. Sybesma (2017) then claimed that wan is projected in a different place from a typical resultative complement when it is used only to signal telicity information. A simpler version of Sybesma’s (2017) proposed structure is presented in (11):

Table

(11)	… [vP	[AspP1	[AspP2	[AspP3	[VP]]]]]
		-wan	lei ‘tired’	zou ‘walk’

As (11) shows, there are three functional phrases related to the situation aspect syntactically projected between vP and VP. According to Sybesma (2017), a typical resultative complement such as lei ‘tired’ in zou-lei ‘walk-tired’ is in the lowest aspectual phrase; the morpheme wan, on the other hand, is projected in the middle one. Sybesma stated that the word order of a V–V resultative compound is a result of the main verb moving up to Asp3 and that of a V–wan compound is a result of the verb moving to Asp2. However, in Mandarin Chinese, AspP2 and AspP3 cannot be realized simultaneously; therefore, a compound such as zou-lei-wan ‘walk-tired-finish’ is ungrammatical.

The fact that the result state is not specified in many sentences having a V–wan compound has some significant consequences, one of them being the difference in the interaction between a V–wan compound and a durative phrase such as wu fenzhong ‘(for) five minutes’. The examples in (12) and (13) illustrate the variation:

Table

(12)	a.	Lisi	ba	chuanghu	da-kai-le	wu	fenzhong.
		Lisi	BA	window	hit-open-Prf	5	minute
		‘Lisi opened the window for five minutes.’
	b.	Lisi	ba	chuanghu	da-kai(-le)	wu	fengzhong	le.
		Lisi	BA	window	hit-open(-Prf)	5	minute	LE
		‘It has been five minutes since the window was open (by Lisi).’
(13)	a.	*Lisi	ba	zuoye	xie-wan-le	wu	fenzhong.
		Lisi	BA	homework	write-finish-Prf	5	minute
	b.	Lisi	ba	zuoye	xie-wan(-le)	wu	fenzhong	le.
		Lisi	BA	homework	write-finish(-Prf)	5	minute	LE
		‘It has been five minutes since Lisi completed the homework.’

Both (12a) and (12b) are grammatical, and the durative phrase wu fenzhong ‘(for) five minutes’ in both examples is used to measure the length of the result state. The only difference is that the result state still holds at the speech time in (12b). If wan behaved like a typical resultative complement, we would expect it to act in the same way while interacting with the same durative phrase. However, this is not the case, as (13a) is ungrammatical. The sentence becomes grammatical if we add sentence-final le, as illustrated in (13b), and the durative phrase measures the length between the end point of the main verb and the speech time.⁹ As I will demonstrate later in §4, the variation illustrated here is due to the lack of a resultative state of a V–wan compound.

A sentence containing a V–wan compound shows another variation while interacting with a locating temporal adverbial, such as zuotian ‘yesterday’, along with a durative phrase, as shown in (14):

Table

(14)	a.	Lisi	zuotian	tiao-dao-le	shui	li		(V–V resultative)
		Lisi	yesterday	jump-arrive-Prf	water	inside
		wu	fenzhong.
		5	minute
		‘Lisi jumped into the water (and stayed there) for five minutes yesterday.’
	b.	Qu Yuan	zuotian	zhenghao	tiao-dao	shui	li	(V–V resultative)
		Qu Yuan	yesterday	exactly	jump-arrive	water	inside
		2296	nian.
		2296	year
		‘It had been exactly 2,296 years yesterday since Qu Yuan jumped into the water.’
	c.	Lisi	zuotian	zhenghao	nian-wan-le	daxue		(V–wan compound)
		Lisi	yesterday	exactly	study-finish-Prf	college
		san	nian.
		3	year
		‘It had been exactly three years yesterday since Lisi finished college.’

Both (14a) and (14b) contain a typical V–V resultative, more specifically, a goal V–V compound, (14c) displays a V–wan compound, and they all include the locating temporal adverbial yesterday that appears before the compound. However, the function of the adverbial varies. First, in (14a), the adverbial modifies both the event and the result state; that is, both Lisi’s jumping into the water and his being in the water for five minutes occurred yesterday. The adverbial can have another function, as shown in (14b). With a slight modification of the subject, zuotian ‘yesterday’ in (14b) does not mean that Qu Yuan’s jumping into the water (and being in the water) occurred yesterday.¹⁰ Rather, the adverbial is used as one of the boundaries of the interval being measured, and the durative phrase is used to specify the temporal duration between the main event and the locating adverbial. The examples in (14a–b) suggest that a locating temporal adverbial such as zuotian ‘yesterday’ may be used to modify both the main event and the result state as well as a reference point for the measurement of a durative phrase when a sentence contains a V–V resultative compound. In contrast, the same adverbial in (14c) cannot be used to modify the main event (i.e. finishing college). The sentence does not mean that the main event occurred yesterday. Rather, the temporal adverbial is used as one of the boundaries of the interval being measured. The phenomenon again suggests that a V–wan compound is different from a typical V–V resultative one.

Finally, another variation that can be found in a sentence containing a V–wan compound is wan’s interaction with a VO verb. According to studies such as Cheng and Sybesma (1998) and Li and Thompson (1981), Chinese has a group of verbs, particularly activity verbs, that must always have a complement, which can be a nonreferential object. This type of verb contains several special linguistic properties. First, when used in a V–V resultative construction, there are some morphosyntactic constraints, as illustrated in (15) and (16):

Table

(15)	a.	Lisi	chi-fan	chi	bao	le.
		Lisi	eat-rice	eat	full	LE
		‘Lisi has become full as a result of eating.’
	b.	Lisi	chi-bao	fan	le.
		Lisi	eat-full	rice	LE
		‘Lisi has become full as a result of eating.’

Table

(16)		*Lisi	chi-fan	bao	le.
		Lisi	eat-rice	full	LE
		Intended reading: ‘Lisi has become full as a result of eating.’

As (15a) shows, the verb chi ‘eat’ is duplicated and appears after the VO verb and before the resultative complement, bao ‘full’. The resultative construction indicates that the subject being full is a result of the eating action. Second, a different way to express the same meaning is to put the nonreferential object after the V–V compound, as (15b) shows. Finally, as (16) illustrates, the resultative complement cannot appear after the nonreferential object.

The morphosyntactic constraints observed in (16), however, do not seem to be as strict in a sentence containing a V–wan compound. See the examples taken from the Academia Sinica Balanced Corpus of Modern Chinese in (17) and from the Baidu Search Engine (Baidu knows) in (18):¹¹

Table

(17)

a.

#

shua-ya

wan,

chuan-hao

yifu,

chi

zaocan.

brush-teeth

finish

wear-done

clothes

eat

breakfast

‘ … finished brushing teeth, done with putting on clothes, and then ate breakfast.’

Table

b.

#

… youshihou,

mama

jiu

shi

shang-cecuo-wan

yihou …

… sometimes

mom

exactly

be

up-bathroom-finish

after

‘ … sometimes exactly after Mom finishes using the bathroom … ’

Table

(18)

a.

#

Shua-ya

wan

neng

he

suannai

me?

brush-teeth

finish

may

drink

yogurt

Q

‘Can I have yogurt after brushing my teeth?’

b.

#

Weishenme

chi-fan

wan

hou

ganjue

hen

lei,

quan

shen

wuli?

why

eat-rice

finish

after

feel

very

tired

whole

body

exhausted

‘Why do I feel tired, and the entire body is exhausted after having a meal?’

The example in (17a) contains two VO verbs: shua-ya ‘brush teeth’ and chuan-yifu ‘put on clothes’. These phrases are also modified by two words that express telic information, wan and hao ‘complete’ respectively. However, as shown, the former appears after the object, while the latter appears before the object. Similarly, wan in (17b) also appears after the object without verb duplication.¹² In a different corpus, some occurrences are also found in the usage of VO–wan, as shown in (18). Although this type of sentence is rare in Modern Chinese, it nevertheless suggests that wan acts differently from a typical resultative complement.

3. A syntactic account

In this section, I provide a syntactic analysis for the multifunctional properties of wan.

3.1 Theoretical background

3.1.1 Grammaticalization and syntax

The current analysis of the diachronic development of wan is inspired by studies such as Roberts and Roussou (2003) and Wu (2004). One of the main goals of Roberts and Roussou’s (2003) study was to address the issue of syntactic change in the context of the Minimalist Program to provide a general analysis of grammaticalization. They used grammaticalization as a probe to explore the nature of functional categories such as Tense and Determiner. Each functional head is argued to be associated with a phonetic form realization that can be achieved by either Merge or Move, and language change comprises change in the PF realization of functional heads. The change occurs when the trigger experience for a parameter setting is ambiguous. The authors also provided evidence for the idea that grammaticalization involves a reanalysis of functional categories. The principal idea is that the reanalysis of functional heads always involves the reanalysis of movement. For example, they investigated the development of English modals and argued that English modals like may are found to be a reanalysis of a lexical verb (a VP element) to an auxiliary.

According to Roberts and Roussou (2003) and Wu (2004), grammaticalization generally results from the movement of a lexical (or a functional) element α into the functional structure that immediately dominates α. For example, in Stage 1, as in (19a), α is in its merged place, which is dominated by a functional structure headed by β.¹³ However, due to some syntactic operations, such as head movement, α may move to β (which can be overt or covert), as illustrated in (19b). Finally, α may come to be reanalyzed as instantiating and being directly base generated in F1 so that a new lexical element δ can exist in α's original syntactic position, as shown in (19c).

(19)

We can use the English modal may as an example to show how the grammaticalization process may work. Observe (20) below:

(20)

According to Roberts and Roussou (2003), (Middle) English modal verbs such as may used to function solely as lexical verbs and were base-generated in the head of a VP, as shown in (20a).¹⁴ Nevertheless, because V-to-T movement was observed in this language, may constantly moves to T (as in (20b)). This movement, in fact, can be observed in a negative sentence where may precedes clausal negation. Finally, during the early sixteenth century, the operation of the V-to-T movement was lost, and modals such as may were reanalyzed and began directly being merged in the head of the TP, as seen in (20c).

In line with Roberts and Roussou (2003) and Wu (2004), I argue that a similar grammaticalization process can be used to explain the multifunctional characteristics of wan. I claim that wan is in the process of being reanalyzed from being a pure lexical item to a functional one that immediately dominates the VP from which wan originates. A more detailed discussion is given in §3.2.

3.1.2 Computation of telicity

Following Smith (1997), I assume that the aspectual meaning of a sentence conveys two kinds of information: viewpoint (or outer) and situation (or inner) aspects. The former presents a situation with a particular perspective or focus and comprises three major types: perfective, imperfective and neutral. The latter concerns different types of verbs or situations based on their temporal properties. In line with Vendler (1967), I assume that verbs can be categorized into different groups based on their semantic properties; telic verbs include accomplishments and achievements, and atelic verbs include activities and states. The computation of the situational aspect includes more than just the verb itself. Other elements, such as the semantic properties of the direct object (e.g. the quantity of the object), telic morphemes (e.g. preverbs in Russian), goal prepositional phrases (e.g. push the cart to the park) and resultative complements may also contribute to the computation. According to Slabakova (2001), telic constructions subsume two large categories: those that are accomplishment- or achievement-denoting predicates and those that contain resultative or dative elements.

I also assume that the viewpoint aspect is computed in the functional domain above vP (Huang et al., 2009; Travis, 2010) and the situation aspect and other telic-related constructions (e.g. resultative complements) inside the lexical domain. However, to distinguish further among the different types of telic-related constructions, I follow Slabakova (1997, 2001) and Travis (2010) and assume that language uses different mechanisms to compute telicity and that telicity markers may be encoded in different places within the lexical domain: (a) the head of the vP; (b) the head of the Inner Aspect Phrase; and (c) the complement of VP.¹⁵ I focus the discussion on (b) and (c), as I argue that Chinese uses these two mechanisms to compute telicity. I also follow studies such as Ritter and Rosen (2005) and Travis (2005, 2010), which have provided a syntactic account of the telicity of different types of verbs. Despite some differences in their analyses, these scholars generally agree that there is at least one functional phrase in the lexical domain that handles the computation of telicity. However, to explain some semantic properties related to telicity of predicates denoting state and achievement, I follow Slabakova (2001) and assume that these predicates are specified for the plus or minus value of the feature [telic] in the lexicon.

Let us begin the discussion with the use of the Inner Aspect Phrase by using some English examples. The telicity of different types of verbs in English mainly depends on the quantity or specificity of the direct object, illustrated in (21):

Table

(21)	a.	John ate the apple in an hour.	(telic)
	b.	John ate three apples in an hour.	(telic)
	c.	John ate apples for an hour.	(atelic)

With a quantized object or an [+SQA] object (cf. Verkuyl, 1993), (21a) and (21b) have a telic reading. In contrast, without a quantized object, as in (21c), the sentence has only an atelic reading. According to studies such as Tenny (1987, 1994), objects that delimit an event must be affected and undergo some sort of change of state over the course of the event. According to Slabakova (2001), English dynamic predicates (i.e. accomplishments and activities) have the [α telic] feature that needs to be valued, and these predicates take their telicity value from the nominal feature of the direct object. When the object has a [+SQA] value, the predicate is valued as [+telic], as in (21a) and (21b). In contrast, when the object has a [−SQA] feature, the predicate has a [−telic] feature, as in (21c). The feature checking between the object and predicate is based on studies that have observed the semantic resemblances between the nominal and verbal domains. Krifka (1992), for example, indicated that an atelic event predicate (e.g. run) and a mass noun (e.g. wine) are both cumulative; in contrast, a telic event predicate (e.g. run a mile) and a countable noun (e.g. a glass of wine) are quantized.

Slabakova (2001) further claimed that the computation is under a spec–head relationship between the predicate and the object inside the Inner Aspect Phrase; therefore, for the predicate to receive a value, the object moves to the specifier, and the predicate moves to the head of the Inner Aspect Phrase. The predicate then moves up to a null CAUSE or DO morpheme situated in vP. The tree of the derivation is depicted in (22):

(22)

Nevertheless, there are languages in which the quantity or specificity of the direct object does not directly contribute to the telicity of the event. Rather, an overt telic morpheme, responsible for the telic reading of a sentence, may directly merge in the head of the Inner Aspect Phrase. For example, Travis (2010) argued that a language such as Tagalog has overt telic morphemes that are directly merged in the head of the Inner Aspect Phrase between the vP and VP. Nossalik (2009) demonstrated that in several Slavic languages, preverbs signalling telicity are syntactically situated in the head of the Inner Aspect Phrase. In these cases, telicity is determined by the preverbs. Observe the Russian examples in (23) from Nossalik (2009):

Table

(23)	a.	Petja	ubežal		(telic)
		Petja	u-run-Prf
		‘Petja ran away.’
		(Nossalik, 2009, p. 159 [81a])
	b.	Petja	priš’ol		(telic)
		Petja	pri-walk-Prf
		‘Petja came.’
		(Nossalik, 2009, p. 159 [81b])
	c.	Petja	El	(sup).	(atelic)
		Petja	ate-Imp	soup
		‘Petja was eating (the) soup.’
		(Nossalik, 2009, p. 158 [79a])
	d.	Petja	s’el	*(sup).	(telic)
		Petja	ate-Prf	soup
		‘Petja ate the soup.’
		(Nossalik, 2009, p. 158 [79b])

Examples (23a–b) contain an intransitive verb and have a telic reading via the preverbs. Given that there is no affected object that undergoes a change of state, the only argument that can be affected over the course of the event is the surface subject. Both verbs in the sentences are motion verbs. Nossalik (2009) argued that the surface subjects in both sentences are arguments that undergo a change (of location). In contrast, both (23c) and (23d) have a direct object. Nevertheless, their interpretations differ. In (23c), without a preverb, the sentence is only atelic, and the direct object can be interpreted as either definite or indefinite. In contrast, (23d) contains a preverb and has a telic reading. The direct object can have only a definite reading. The above examples show that a sentence containing a preverb always has a telic reading regardless of the properties of the direct object (if any). This indicates that a sentence with an overt telic morpheme does not require the appearance of an object that delimits the event in Russian.

I have presented illustrations on how telicity is computed in the Inner Aspect Phrase. For example, English accomplishments receive a [+telic] value that is implemented by the [+SQA] feature of the direct object. Russian, in contrast, uses overt telic morphemes to mark telicity. In Chinese, the direct object with the [+SQA] does not seem to always be able to delimit an event. For example, Tai (1984) indicated that an object with a [+SQA] feature in Chinese generally does not delimit an event, as illustrated in (24a); most Chinese telic predicates include both a verb denoting activity and a resultative complement, as in (24b):

Table

(24)	a.	Zhangsan	xie-le	na	feng	xin,	keshi	mei	xie-wan.	(atelic)
		Zhangsan	write-Prf	that	CL	letter	but	not	write-finish
		#‘Zhangsan wrote a letter but did not finish it.’
	b.	*Zhangsan	xie-wan-le	na	feng	xin,	keshi	mei	xie-wan.	(telic)
		Zhangsan	write-finish-Prf	that	CL	letter	but	not	write-finish
		*‘Zhangsan finished writing a letter but did not finish it.’

The examples shown in (24) seem to suggest that the mechanism used to account for the computation of accomplishments in English would not account for the Chinese data. Nevertheless, it is not the case that no object with the [+SQA] feature can delimit an event in Chinese.¹⁶ The two examples in (25) offer an illustration:

Table

(25)	a.	*Ta	zuo-le	yi	ge	dangao,	keshi	mei	zuo-hao.
		he	make-Prf	one	CL	cake	but	not	bake-finish
		#‘He baked a cake, but did not finish baking it.’
	b.	*Zhangsan	xie-le	wu	feng	xin,	keshi	mei	xie-wan.
		Zhangsan	write-Prf	five	CL	letter	but	not	write-finish
		*‘Zhangsan wrote five letters but did not finish them.’

Both examples contain a creation verb and an object with the [+SQA] feature, and they must have a telic reading. According to Soh and Kuo (2005), a sentence such as (25a) can have only a telic reading because of some semantic properties of the direct object. Similarly, Huang (2015) indicated that a sentence like (25b) only has a telic reading when the direct object is quantified. Given that the two examples in (25) demonstrate that the [+SQA] feature contributes to the computation of telicity, I would like to suggest that the mechanism presented in (22) can be applied to sentences such as (25a) and (25b). Thus, a null telic morpheme is situated in the head of the Inner Aspect Phrase between the vP and VP in Chinese.

Finally, in this subsection, I include a brief discussion on the syntax of the resultative construction, as wan has been categorized as a type of resultative complement in several previous studies. However, given that there are numerous studies on this topic and because of the scope of the present study, I focus on studies that have adopted approaches arguing for a null telic morpheme below the VP that handles the telic reading of a resultative construction. Snyder (1995), for example, argued that the English resultative construction contains a null telic morpheme projected below VP. According to Snyder, English allows a null aspectual morpheme (ϕ _TELIC) that takes an event and a predicate of events and ensures that the predicate is true only at the natural end point of that event. If a language allows this morpheme to be projected in syntax, it is possible to convert an activity into an accomplishment through the addition of the telic morpheme and a predicative complement to that morpheme. It has been argued that Chinese also has a null telic morpheme below the VP that handles the telic reading of a resultative construction. For example, Tang (1997) claimed that the matrix verb in a V–V resultative is unbound and is not compatible with the result state. Thus, there is a null telic morpheme whose main function is to close off the open range of the matrix verb and denote the endpoint in temporal extension. According to Tang (1997), this telic morpheme is situated in a functional phrase that is projected between the VP and the result-denoting phrase, and for morphological reasons, the resultative complement raises and is right-adjoined to the null telic morpheme to check the strong aspectual feature. Finally, Zhang (2017) provided a similar account to the structure of the resultative construction in Chinese. Following Ramchand (2008), it is assumed that in a V–V resultative compound, the VP selects a resultative phrase (RP) as its complement, the head of which is the element that handles the telic reading. Zhang (2017) also argued that for the construction to derive the V–V (+object) word order, the head of the RP must first move to R, and the cluster further moves to the matrix V and is (right) adjoined to the V.¹⁷

The argument that elements expressing telic information are encoded in different places within the vP domain provides the current study with valuable insights. As I will show in the next subsection, wan is undergoing a grammaticalization process and has become a telic morpheme situated in the head of the Inner Aspect Phrase.

3.2 Proposals and analyses

3.2.1 On the grammaticalization process of wan

In the spirit of Roberts and Roussou (2003) and Wu (2004), I propose that wan is in a process of being reanalyzed from a lexical item to a functional one. The grammaticalization process is illustrated in (26) below:

(26)

First, in Stage 1, as a full lexical item, wan is directly merged in the head of a VP. In this stage, as a full lexical item, wan functions as both a causative and an inchoative predicate. Recall that according to Roberts and Roussou (2003) and Wu (2004), grammaticalization frequently results from the movement of a lexical element (or a functional one) into a functional structure that immediately dominates the lexical element. The syntactic movement of wan is, therefore, from the VP to a functional phrase that dominates it.

As I discussed in §3.1.2, the computation of the situation aspect is in the Inner Aspect Phrase. I showed that in a language such as English, the semantic properties of a noun phrase, such as the direct object, determine the telicity of verbs denoting activity and accomplishment, which contain an [α telic] feature that needs to be valued. In syntax, the main verb moves to the head of InAspP, and the direct object moves to the specifier of this functional phrase. Under this spec–head relation, the verb is assigned a telic value when the direct object contains the feature [+SQA]. Nevertheless, predicates denoting achievement and state are marked telic and atelic, respectively, in the lexicon. I follow Slabakova (2001) and assume that Inner Aspect Phrase is where telicity is computational calculated and is always projected in syntax; therefore, predicates whose telicity are specified in the lexicon still move to the head of InAspP for the computation of telicity.

The movement of a verb to Inner Aspect Phrase provides a way for wan to reanalyze. Syntactically, the functional phrase immediately dominates the VP in which wan is originally merged. This fulfils the syntactic requirement of the grammaticalization process (Roberts & Roussou, 2003; Wu, 2004). As a result, in Stage 2, wan continuously moves to InAspP, which is a functional phrase that dominates the VP. Nevertheless, one may wonder why wan is the only verb that is experiencing grammaticalization, as other types of verbs also move to InAspP for the telicity to be computed. I claim this is due to the core meaning of wan: to complete or finish. As argued by previous studies, such as Ritter and Rosen (2005) and Travis (2010), the Inner Aspect Phrase is a functional phrase that encodes telic information. The core meaning of wan exactly matches the function of the Inner Aspect Phrase; therefore, different from many other verbs, it has been reanalyzed as a telic marker and begun to merge directly in the head of the Inner Aspect Phrase (in Stage 3, as shown in (26c)) and is able to assign a telic value to verbs containing an [α telic] feature without the help of a noun phrase that delimits an event. For example, an ergative verb such as pao ‘to run’ can form a compound with wan and thereby be assigned a telic feature. The use of an overt telic morpheme such as wan in the Inner Aspect Phrase indicates that Chinese uses at least two mechanisms to compute telicity: the use of a DP that can measure out an event and the use of overt telic morphemes.

The reanalysis process can be chronologically observed in the development of the Chinese language. In Classical Chinese, wan mainly functions as a lexical verb base-generated in the head of a VP and can freely function either as a causative or an unaccusative predicate. Owing to grammaticalization, it has gradually lost its lexical status. More specifically, causative usage is slowly disappearing, as illustrated in (7) above. A verb that contains wan and a dummy object, such as wan-hun ‘finish-wedding: finished getting married’ or wan shi ‘finish-thing: to complete a task’, may still maintain some causative meaning; the objects, however, are nonreferential. Another phenomenon that suggests the grammaticalization process of wan is its dual function in Modern Chinese. In addition to functioning as an unaccusative predicate, wan in Modern Chinese has two more functions (Lu et al., 2019; Sybesma, 2017): as a resultative complement, as in (9), or as a telic morpheme, as in (10). One may wonder why we can still find different types of instances in Modern Chinese. In line with Wu (2004), I argue that in the course of grammaticalization, some earlier usages of wan have been retained in the process. Thus, we can still find sentences in which wan functions as the main predicate, as a resultative complement or as a pure telic marker.

3.2.2 A sentence having a V–wan compound with an object

In this subsection, I discuss a sentence with a V–wan compound that contains a direct object. The examples in (27) are all perfective, and wan expresses telicity. I propose that these sentences have the syntactic structure in (28):

Table

(27)	a.	Lisi	ca-wan-le	san	ge	beizi	le.
		Lisi	wipe-finish-Prf	three	CL	glass	LE
		‘Lisi has finished wiping three glasses.’
	b.	Lisi	dasao-wan-le	na	ge	fangjian	le.
		Lisi	clean-finish-Prf	that	CL	room	LE
		‘Lisi has finished cleaning that room.’
	c.	Lisi	wan-wan-le	diandong	le.
		Lisi	play-finish-Prf	game	LE
		‘Lisi has finished playing the video game(s).’

(28)

To account for the telic reading of the sentence, I assume that dynamic verbs in Chinese also contain the feature [α telic] that must be valued. As seen in (28), wan is situated at the head of the Inner Aspect Phrase projected between the vP and VP. It contains the feature [+telic] and indicates that an event has come to an end. However, the telicity value of the [α telic] on V is strong, and the checking relationship must be local; therefore, the verb first moves to InAsp and is adjoined to wan. As wan is [+telic], the verb is valued as [+telic]. Recall the agreement relationship between a dynamic verb and its object: the computation is in the Inner Aspect Phrase under a spec–head relation. Therefore, the object moves to the specifier of the Inner Aspect Phrase. The next step is that the cluster V–wan moves to little v, which is situated with a null light verb CAUSE or DO morpheme. The movement of the cluster is because little v bears a strong categorical feature, namely, the InAsp.

To further explain the movement of both the predicates and the objects in the examples in (27a–c), I would like to first discuss some linguistic properties of the direct objects. The examples have different types of objects: the object has a numeral form in (27a), the determiner na ‘that’ in (27b) and a bare form in (27c). Nevertheless, these sentences all have a telic reading regarding the form of the object. The definite reading of the object with the bare form is particularly interesting given that Chinese bare nouns can have different readings (e.g. a definite, a generic, a specific/definite reading). This phenomenon is like the Russian examples discussed in §3.1, where the direct objects must have a definite reading when the verb is modified by a preverb. I would like to suggest that wan and possibly other phase complements discussed by Li and Thompson (1981), are merged in the head of the Inner Aspect Phrase.¹⁸ When the object contains a [+SQA] feature, the dynamic verb has a [+telic] feature. Thus, when the verb has a [+telic] feature like a V–wan compound, the object must have a [+SQA] feature; therefore, the interpretation of the bare noun in (27c) is definite.

Finally, I also follow studies such as Huang et al. (2009) and Travis (2005, 2010) and assume that the viewpoint aspect (or outer aspect) is syntactically projected above the little vP. Given that the examples in (27a–c) all have a perfective reading, I assume that OutAsp contains the value of [+perfective] and is realized as the verbal suffix -le. As verbs do not move out of the vP, as suggested by studies such as Gu (1995) and Huang et al. (2009), I assume that the word order of [V–wan–le] is formed due to affix hopping.

Empirical evidence supports the claim that like English, Chinese dynamic predicates also have the [α telic] feature that needs a value. First, as seen in (24) and (25), the telicity of a sentence containing a dynamic verb depends on the semantic features of the direct object. For example, when the object is quantized, the sentence must have a telic reading. Second, the examples in (29) adopted from Wu (2002) also suggest that a dynamic verb, such as pao ‘run’, may have different aspectual values.

Table

(29)

a.

Libai

pao-le

yi

ge

xiaoshi.

Libai

run-Prf

one

CL

hour

‘Libai ran for an hour.’

(Wu, 2002, p. 151 [21a])

b.

(Nei

zuo

jianyu)

(zai)

yi

ge

xiaoshi

nei

pao-le

wu

ge

fanren.

that

CL

prison

in

one

CL

hour

in

run-Prf

5

CL

criminal

‘(lit.) The jail made five criminals run away in one hour.’^19

‘Five criminals escaped from the jail within an hour.’

(Wu, 2002, p. 151 [22a])

According to Wu (2002), (29a) has an atelic reading, as the dynamic event can be modified by the temporal adverbial yi ge xiao shi ‘(for) one hour’, which is compatible with only atelic predicates; in contrast, in (29b), the event is modified by a different adverbial, yi ge xiaoshi nei ‘within one hour’, which is compatible only with a telic event. Wu concluded that predicates in Chinese can have different telic values based on the linguistic environment in which they appear. The conclusion confirms the argument that dynamic predicates can have either a telic or atelic reading based on the elements that assign values to them.

3.2.3 A sentence having a V–wan compound without objects or with a nonreferential object

Given that a sentence containing a telic morpheme such as wan always has a telic reading regardless of the property of the direct object, the same mechanism can be used to account for a sentence with a nonreferential object or without an object. The sentences in (30a–b) provide two additional examples, and (31) presents their syntactic structure:

Table

(30)	a.	Lisi	pao-wan	bu	le.
		Lisi	run-finish	step	LE
		‘Lisi has finished running.’
	b.	Lisi	ku-wan	le.
		Lisi	cry-finish	LE
		‘Lisi has finished crying.’

(31)

In (30a) the object bu ‘step’ is nonreferential and is not affected. Without the object, the meaning does not change. The sentence has a telic reading thanks to the telic morpheme wan. According to Her (1996), VO compounds can be divided further into three different groups. In the second group, the verb and the nonreferential object may be separated by other elements, such as the perfective suffix -le or an adjunct modifier. In the same vein, I suggest that a telic morpheme such as wan may also appear between the two elements, as it is also a type of aspectual marker. Thus, we see the word order change to V–wan O in (30a). Similarly, (30b), without an object, also has a telic reading. The derivations of the two sentences are the same as those demonstrated in (31). The verb having the feature [α telic] is first adjoined to wan to receive the value [+telic] since the telicity value of the [α telic] on V is strong, and the checking relationship must be local. The cluster of V–wan further moves to little v and the movement is because little v bears a strong InAsp feature. Finally, perfective -le is base-generated in the head of the Outer Aspect Phrase above the vP, and due to affix hopping, the suffix appears after the V–wan cluster.

4. Consequences

The argument that wan does not always function as a resultative complement has several consequences. The first related issue is the linear order of VO–wan (e.g. shua-ya-wan ‘brush-teeth-finish’ and xi-zao-wan, ‘wash-bath-finish’) found in the two corpora. As discussed, this linear order was accepted by several native speakers. I provide a possible explanation here; however, it is only suggestive. As seen in (30a), the word order of [V–wan O] is due to the movement of the verb to wan, so its [α telic] can be valued. The tree in (31) gives a more detailed representation of the verb phrase functioning as the complement of wan. Following Huang et al. (2009), I assume that the verb and the object form a V’; however, only the head itself moves and is adjoined to wan, as shown in (32a). To account for the order of VO–wan, I argue that the V’ that contains the verb, and the object may be reanalyzed as a V. Therefore, the entire V with the form of VO moves up to InAsp, and wan is attached to the nonreferential object, as demonstrated in (32b).²⁰

(32)

The next issue I address is the interaction between V–wan and V–V resultative compounds and durative phrases. Recall that the two types of compounds show some variations when interacting with durative phrases, as illustrated in (12) and (13) above. According to Liao (2015), there are three different types of durative phrases in Chinese, two of which are used to modify durations that are related to the resultative construction. The examples in (33), taken from Liao (2015), illustrate the three types of durative phrases:

Table

(33)	a.	Zhangsan	tiao-dao	shui	li	wu	fenzhong.
		Zhangsan	jump-to	water	inside	5	minute
		‘Zhangsan jumped into the water (and stayed in the water) for five minutes.’
		(Liao, 2015, p. 304 [10a])
	b.	Zhangsan	nian-le	wu	nian	gaozhong.
		Zhangsan	study-Asp	5	year	high.school
		‘Zhangsan attended high school for five years.’
		(Liao, 2015, p. 304 [10b])
	c.	Zhangsan	nian-wan	gaozhong	wu	nian	le.
		Zhangsan	study-finish	high.school	5	year	LE
		‘Zhangsan has completed high school for five years.’
		(Liao, 2015, p. 304 [10c])

As seen in the examples above, there is no morphological element that distinguishes these three types of durative phrases. However, they are used to modify different temporal intervals. In the spirit of Parsons (1990), Liao (2015) argued that the first type (Type 1) is the ‘result-related’ durative phrase, which measures the temporal duration of a result (or target) state that is directly caused by an event, as illustrated in (33a). According to Parsons (1990), there are two types of states that come into existence as a result of a culminative event. The first is the resultant state and the second is the target (result) state. Parsons states that “for every event e that culminates, there is a corresponding state that holds forever after” (p. 234). This state is the ‘Resultant State of e’. For example, if Mary eats lunch, then there is a state that holds forever after: the state of Mary having eaten lunch. In contrast, the target state (or the result state in Piñón’s (1999) concept) is the regular result state of an event. For example, in the sentence I throw a ball onto the roof, the target state is the ball being on the roof. As stated by Kratzer (2000), these two types of states differ in that the result state is reversible in principle, whereas the resultant state holds forever. Liao further argued that a Type 1 durative phrase is admissible only for lexical content that generates a result state. For example, the durative phrase wu fenzhong ‘(for) five minutes’ in (33a) measures the temporal duration of the result state of Zhangsan being in the water for at least five minutes. The second type (Type 2) is the ‘process-related’ durative phrase, which measures the duration of the event itself. For example, in (33b), the event of Zhangsan attending high school lasted for five years. Finally, the third type (Type 3) is the ‘result time-related’ durative phrase. This type of durative phrase measures the length of two specific points on the time axis (Liao, 2015). The first point is the salient part of an event, and the second is the reference time (cf. Klein, 1994).²¹ Take the sentence in (33c), for example. Different from (33a), where there is a result state, the salient part of the event is the point at which the main event ended, which is signalled by wan in (33c). As for the reference time, it is equivalent to the speech time ‘now’, which is signalled by the sentential le (as suggested by Liao (2015)).

Considering the discussion of the durative phrases above, we can now analyze the variations shown in (12) and (13), beginning by looking at the two examples in (34a = 12a) and (34b = 12b).

Table

(34)	a.	Lisi	ba	chuanghu	da-kai-le	wu	fenzhong.		(Type 1 durative phrase)
		Lisi	BA	window	hit-open-Prf	5	minute
		‘Lisi opened the window for five minutes.’
	a’.	--------------[++++++++++++]---------|--
		A        B    now
		{  5 minutes  }
	b.	Lisi	ba	chuanghu	da-kai(-le)	wu	fengzhong	le.	(Type 3 durative phrase)
		Lisi	BA	window	hit-open(-Prf)	5	minute	LE
		‘It has been five minutes since the window was open (by Lisi).’
	b’.	--------------[++++++++++++]------------
		A        now
		{  5 minutes  }

The phrase wu fenzhong ‘(for) five minutes’ in (34a) is a Type 1 durative phrase, which is used to measure the duration of the result state of the window being open and the duration that the state may be true at any given time before the speech time. The nature of a durative phrase is that it is used to measure a specific temporal duration of an atelic (sub)-event (i.e. an activity or a state). In other words, the event being measured must last for a certain period. A result state can always be measured because it contains a durative feature (Smith, 1997). Thus, a Type 1 durative phrase can be used to measure the length of any two points (or boundaries) where the result state remains consistent on the time axis. Resultingly, the windows could have been open for five minutes yesterday, two weeks ago or three months ago. The length of the result state in (34a) was five minutes. This can be represented by the two points A and B on the time axis in (34a’). As illustrated, ‘A’ represents the time at which the window became open, and ‘B’ is the time at which the window was closed. The length between ‘A’ and ‘B’ is five minutes. On the other hand, the durative phrase in (34b) is a Type 3 durative phrase because the reference time is clearly stated in the sentence (i.e. now). In other words, the durative phrase in (34b) indicates that the window has been open until now for at least five minutes.

When a sentence contains a V–wan compound, such as (35a = 13a), a Type 1 durative phrase cannot be used; however, if we add a sentence final le to the same sentence, it becomes grammatical, as (35b = 13b) shows. In other words, a sentence with a V–wan compound cannot be modified by a Type 1 durative phrase but can be modified by a Type 3 durative phrase.

Table

(35)	a.	*Lisi	ba	zuoye	xie-wan-le	wu	fenzhong.		(Type 1 durative phrase)
		Lisi	BA	homework	write-finish-Prf	5	minute
	a’.	--------------[++++++++++++
			C
			{
	b.	Lisi	ba	zuoye	xie-wan(-le)	wu	fenzhong	le.	(Type 3 durative phrase)
		Lisi	BA	homework	write-finish(-Prf)	5	minute	LE
		‘It has been five minutes since Lisi completed the homework.’
	b’.	--------------[++++++++++++]------------
			C	now
		{  5 minutes  }

Recall that some V–wan compounds do not have a resultative state, and the morpheme simply signals completion of an event. In other words, on the time axis, there is not a period that a Type 1 phrase can measure; there is only a single point (represented by ‘C’ in (35a’)) at which the main event is finished (i.e. the end point of the event). Without the second point on the time axis, the durative phrase cannot be used, and therefore, the sentence is ungrammatical. Differently from (35a), however, the sentence-final le in (35b) provides the second temporal point on the time axis (i.e. now). Thus, the durative phrase in (35b) is a Type 3 durative phrase used to measure the length between the endpoint of the homeworking writing event (represented by ‘C’ in (35b’)) and the speech time.

5. Concluding remarks

I have revisited the linguistic properties of wan ‘complete’ in Mandarin Chinese. As discussed, in Classical Chinese, wan is used freely either as a causative or inchoative predicate that semantically functions like the English complete or finish. In contrast, in Modern Chinese, it mainly appears immediately after a verb and before the object (if any) and functions as a telic morpheme. This study demonstrated that the multifunctional characteristics of wan arise because the word is undergoing a process of grammaticalization from a pure lexical item to a functional morpheme.

The grammaticalization process can also be observed in the syntax. Although a V–wan compound has been traditionally grouped into a special type of V–V resultative because of its morphosyntactic properties, I have demonstrated several differences between these two types of constructions. For example, I showed a significant difference between these two types of constructions when they were modified by a durative phrase. Therefore, in a sentence containing a V–wan without a result state, wan syntactically differs from other resultative complements in that it is base-generated in the head of the Inner Aspect Phrase between the vP and VP based on its semantic property; that is, it simply signals telicity. The analysis provided here is consistent with previous studies that have argued that grammaticalization is frequently a result of the movement of a lexical item α into the functional structure that immediately dominates α.

Acknowledgements

I would like to express my sincere gratitude to the two anonymous reviewers for their insightful comments. I would also like to thank the editor for the valuable suggestions. Any remaining mistakes are mine.

Data availability statement

The data that support the findings of this study are openly available in Academia Sinica’s Balanced Corpus of Modern Chinese (http://asbc.iis.sinica.edu.tw).

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

I-hao Woo

I-hao Woo received his PhD in Applied Linguistics at Boston University and is an assistant professor of Chinese at the University of Colorado Denver. His research has been focusing on theoretical Chinese linguistics as well as teaching Chinese as a foreign and second language.

Notes

1 The abbreviations used in glossing are BA ‘light verb ba’, CL ‘classifier’, DE ‘particle de’, Imp ‘imperfective aspect’, LE ‘sentence-final le’, Prf ‘perfective aspect’ and Pro ‘progressive aspect’.

2 A judgement test on the use of the VO–wan cluster was conducted in March 2020, and 30 native speakers (from Mainland China and Taiwan) were tested. The results indicated that while some speakers completely rejected the VO–wan cluster, more than half of them accepted it. The inconsistency in judgement confirms the hypothesis that wan is undergoing a grammaticalization process.

3 An anonymous reviewer pointed out that some resultative complements are not compatible with a durative phrase, as illustrated (i):

Table

(i)	*Lisi	ba	qian-men	qi-hong	shi	fenzhong.
	Lisi	BA	front-door	paint-red	10	minute
	Intended meaning: ‘Lisi painted the front door red for 10 minutes.’

However, I would like to suggest that a sentence such as (i) is ruled out for pragmatic reasons. If a proper context is provided, the durative phrase can modify the same V–V compound, as illustrated in (ii):

Table

(ii)

Lisi

ba

qian-men

qi-hong

le

san

tian,

houlai

taitai

bu

xihuan,

you

ba

men

Lisi

BA

front-door

paint-red

Prf

3

day

then

wife

not

like

again

BA

door

qi-hui

lanse

le.

paint-back

blue

LE

‘Lisi painted the door red, and it stayed in red for three days; later on, his wife didn’t like (the colour) so he painted it (back to) blue.’

4 One of the main goals of Roberts and Roussou (2003) was to address how grammaticalization is realized in the syntax in the context of the Minimalist Program. More detailed discussion on their study is presented in §3.

5 An anonymous reviewer suggested that the potential ambiguity in the use of wan as a telic morpheme (or resultative complement) and as a one-place predicate, which denotes an undesirable outcome, should be addressed. The sentence in (i), taken from the reviewer, gives an example:

Table

(i)	Wode	shiye	wan-le.
	my	career	finish-Prf
	‘My career is over.’

However, I suspect that wan in (i) is in fact a shorter form of the two-syllable verb wan-dan ‘finish-egg’, as shown in (iia). See another example in (iib).

Table

(ii)	a.	Wode	shiye	wan-(dan)	le.
		my	career	finish-egg	LE
		‘My	career	is	over.’
	b.	Ni	wan-(dan)	le.
		you	finish-egg	LE
		‘You are in a big trouble.’

Assuming that in a sentence such as (i) wan is a shorter form of a different verb, there is no potential ambiguity, and I leave the usages of wan-dan ‘finish-egg’ for future research.

6 Several different constructions related to telicity marking have been identified in the literature. Slabakova (2001) and Snyder (1995), for example, indicated that the following examples all have telic readings:

Table

(i)	a.	John ate three apples.	(accomplishment)
	b.	John reached the mountain top.	(achievement)
	c.	John painted the house red.	(resultative)
	d.	John jumped into the pond.	(goal-related construction)
	e.	John gave Mary three apples.	(double object)
	f.	John shot down the drone.	(verb-particle)

7 The term ‘phase’ used in these studies is different from the one that is used in studies on the Minimalist Program, such as Adger (2003) and Chomsky (1995).

8 An anonymous reviewer indicated that the two possible readings of a V–wan compound (i.e. a resultative reading in (9) and a completive one in (10)) are due to different types of VPs that wan is interacting with – more specifically, due to different types of direct objects. It is stated that finishing drinking a bowl of soup naturally implies the soup is gone and finishing cleaning a room does not mean the room is gone but implies the result state that the room is completely cleaned. However, I would like to suggest that the cleanness of the living room in (10) is an implicature that can be cancelled. This is different from (9), where the result state is an assertion that cannot be cancelled. Compare (ia) and (ib):

Table

(i)

a.

(Suiran)

Lisi

he-wan-le

ni

zuo

de

tang,

*keshi

tang

hai

shengxia

yi

da

ban.

although

Lisi

drink-finish-Prf

you

make

DE

soup,

but

soup

still

remain

one

big

half

‘Although Lisi finished drinking the soup you made, *more than half of it was left.’

Table

b.

(Suiran)

Lisi

ca-wan-le

keting,

keshi

ca-de

hen

bu

ganjing.

although

Lisi

wipe-finish-Prf

living.room

but

wipe-DE

very

not

clean

‘Although Lisi finished wiping the living room, it was not cleaned properly.’

As (ia) shows, the result state of the completeness of the direct object is an assertion and cannot be cancelled. In contrast, the cleanness of the direct object in (ib) is simply an implicature; therefore, it can be cancelled. The variation shown in the above two sentences suggests that wan has a dual function; therefore, I follow Lu et al. (2019) and Sybesma (2017) and assume that the ambiguity is mainly due to the status of wan although my analysis is different from theirs.

9 According to Liao (2015), the speech time is signalled by the sentence final le. As Li and Thompson (1981) pointed out, one of the functions that sentential LE has is to signal current relevance. It is then not unexpected that the speech time is used as the reference time. I would like to thank an anonymous reviewer for pointing this out.

10 Qu Yuan (343–278 B.C.) was a Chinese poet who committed suicide by jumping into Miluo River.

11 The Academia Sinica Balanced Corpus of Modern Chinese is based on Mandarin Chinese used in Taiwan (i.e. Guoyu) and the Baidu Search Engine (Baidu knows) is a web-based search engine based in Mainland China, which provides users with a query-based searchable community to share knowledge and experience. The data taken from two different types of corpora from different Chinese-speaking regions as well as the results of the judgement test suggest that the VO–wan cluster is not an isolated phenomenon. As will be discussed in §3 and §4, the variations shown in a V–wan compound are the result of the grammaticalization of wan.

12 Because the cluster of VO–wan is not acceptable by some speakers, I used the symbol ‘#’ to indicate that these sentences are marginally acceptable. A more common way to say the two phrases is to place wan immediately after the verb and before the object. Alternatively, we can also duplicate the verb, and place wan after the duplicated verb to form a VOV–wan sequence.

13 The syntactic trees in (19) are adopted from Wu’s (2004) study.

14 Roberts and Roussou (2003) used a simpler syntactic structure and assumed that TP immediately dominates VP, and V directly moves to T when discussing the issue of the V-to-T movement in Middle English. However, they also indicated that a more complex structure such as [TP … [vP … [VP]]] can be adopted as well.

15 We will see some Russian examples in which the Inner Aspect Phrase is where a telic morpheme is projected. Travis (2010) also argued that Malagasy uses InAsp to mark telicity. As for the vP, Butt and Ramchand (2005) suggested that in Hindi-Urdu, several light verbs that are projected in the head of the vP may introduce telicity and argued that v may encode telicity information in some cases. Finally, studies such as Ramchand (2008), Snyder (1995), Tang (1997) and Zhang (2017) have illustrated that there is a functional phrase below VP that is responsible for the telic reading of a resultative complement.

16 I use [+SQA] to represent the semantic features of those nominal phrases that can delimit an event for the sake of simplicity. However, this feature is a little different from that of English. See Soh and Kuo (2005) and Woo (2018) for discussions of these additional semantic features.

17 Although Zhang (2017) did not specify the morphosyntactic property of R, it has functions similar to the null telic morpheme in Tang’s (1997) study.

18 As Travis (2010) suggested, functional morphemes in the (Inner) Aspect Phrase contain some idiosyncratic information, which is what many of these phase complements contain, as indicated by Li and Thompson (1981). I suspect that these phase complements may also be undergoing a grammaticalization process. For recent discussions on related issues, see Pots (2020) and Song (2019). I would like to thank an anonymous reviewer for drawing my attention to these two studies.

19 The first translation is Wu’s (2002) and the second is mine.

20 The reanalyzing process seems rather limited given that it only occurs when the Inner Aspect Phrase contains an overt morpheme such as wan.

21 Liao (2015) did not specify what “the salient part of the event” is. In his example (represented by (33c) here), it is stated that the salient part of the event is the time of Zhangsan’s graduation from high school (i.e. the end point of the main event). However, if my interpretation is correct, then the salient part of the event can be either the end point of the main event or the beginning point of the result state (as illustrated in (34b)).