The Development of [Patient-Subj V-Qilai AP] as a Middle Construction in Chinese

Abstract

The middle construction (MC) is a term originally used to account for derived intransitives in the generative tradition and is well-documented in many Indo-European languages. While diverse views exist on the Chinese MC, some scholars have argued that Chinese [patient-subj V-qilai AP] sentences share traits with the English MC. Although diachronic literature on the development of the Chinese MC is limited, examining the grammaticalization of directional verb phrases like V-qilai provides valuable insights into its evolution. Building on previous analyses, I identify [patient-subj V-qilai AP] as the Chinese MC with V-qilai as its marker. Through an analysis of Classical Chinese data and the approach of constructional assemblies, I propose that the Chinese MC originated from assembled constructions comprising a subject, a directional verb phrase, and a descriptive construction. Over time, these assemblies evolved, with shifts in subject roles, motion interpretation, and descriptive functions, resulting in the emergence of the MC. Quantitative analysis of Classical data indicates the absence of the MC before 1900, suggesting its development through a gradual increase in the frequency of interconnected assemblies. By tracing the development of the Chinese MC, this research enhances our understanding of how middle voice mechanisms emerge across languages.

1. Introduction

The middle construction (MC) is a term originally used in the generative tradition to account for derived intransitives. O’Grady (1980) was one of the first linguists to focus on this issue1, identifying derived intransitives as instances of the MC in English. Since then, the MC has been attested in other Indo-European languages, such as French (Fellbaum & Zribi-Hertz, 1989), German (Fagan, 1992), Italian (Bentley, 2004), and Spanish (Felíiu, 2008). However, whether an MC exists in Chinese has been a matter of debate among scholars (cf. Cheng & Huang, 1994; Sung, 1994; W. He, 2004; Z. Xiong, 2004; Cao, 2004, 2005; Furukawa, 2005; Ting, 2006; Han, 2007; Yu & Si, 2008; Shen & Sima, 2010; Tao, 2010; Yan, 2011). Among the various accounts of the Chinese MC, Cao (2004, 2005), among others, argues that Chinese [patient-subject V-qilai AP] sentences, as exemplified in (1), exhibit properties characteristic of the English MC and should therefore be considered equivalent to it.

(1)	这首诗读起来很顺口
	zhe	shou	shi	du-qilai	hen	shunkou
	this	CL	poem	read-up	very	easy
	This poem reads very easily.

From a diachronic perspective, there is limited literature that systematically examines the development of the Chinese MC. However, the grammaticalization of directional verb phrases such as V-qilai has long been a central topic in studies of grammatical change. Therefore, discussions on the evolution of these related directional phrases may offer valuable insights into the emergence of the MC as an independent sentence pattern.

In this paper, I follow Cao (2004, 2005) and others in identifying [patient-subject V-qilai AP] as the Chinese MC, with V-qilai serving as its marker. Drawing on extensive Classical Chinese data and the concept of constructional assemblies (Petré, 2019; Budts & Petré, 2020), I argue that the [patient-subject V-qilai AP] construction originated from a set of assembled constructions—specifically, a subject, a directional verb construction, and a descriptive construction. Each of these components can function independently, but they are often combined within a single utterance. Over time, particular assemblies and their preferred contextual uses began to shift. Starting in the 13th century, in certain contexts, the subject shifted to a patient role, the directional motion ceased to be intended, and the description was increasingly realized through adjectival phrases. This demotion of motion and agency within the assembled constructions ultimately led to the emergence of the MC in the early 20th century. Quantitative analysis of Classical Chinese data suggests that the MC did not exist prior to 1900, and that the simultaneous rise in frequency of certain assemblies may have facilitated their interconnection, eventually giving rise to the MC as a distinct construction.

The organization of the paper is as follows: Section 2 outlines the key properties of the MC, laying the groundwork for the discussion of the Chinese MC. Section 3 reviews major accounts of the Chinese MC in previous studies and proposes that there may be two distinct forms: the easy–difficult MC and the V-qilai MC. Section 4 provides a concise overview of the data and introduces the concept of constructional assemblies. Section 5 examines pivotal examples that illustrate the historical evolution of the V-qilai MC and offers a constructional account of its development. Finally, Section 6 presents the concluding remarks.

2. The Properties of the Middle Construction (MC)

In a paper on derived intransitives, O’Grady (1980, p. 60) states that “derived intransitives constitute one of the least studied verb constructions in English, and they are considered to be somewhat idiosyncratic and marginal because of their curious syntactic properties”.

(2)	a. This oven cleans easily.
	b. The clothes iron well.

O’Grady views (2a) and (2b) as examples of derived intransitives, which he terms the middle construction (MC) in English. In an earlier work, Grady (1965) referred to this construction as medio-passive. Fellbaum (1985) later described it as the patient-subject construction. Quirk et al. (1985, p. 735) characterize it as “an agentive form to denote passive meaning”.

Subsequently, this construction became a central topic in lexical semantics, especially in efforts to explain the non-agentive and non-passive nature of the middle voice (cf. Klaiman, 1991; Kemmer, 1993; Steinbach, 2002; Davidse & Heyvaert, 2007; Kaufmann, 2007). These studies primarily investigate how the semantics of the middle voice corresponds to syntactic structures across languages. Kemmer (1993) argues that the middle voice represents a coherent—albeit complex—semantic category that is grammatically instantiated in many languages.2

Some Indo-European languages express reflexive meaning through verbs marked with specific morphemes. Such verbs are known as middle verbs, and the associated markers are referred to as middle markers (with middle markers italicized in the examples, which are cited from Kemmer (1988, pp. 22, 76, 78):

(3)	French: se boutonner—“button one’s clothes”.
	Spanish: se peinar—“comb (one’s hair)”.
	Turkish: giy-in—“dress (oneself)”.
	German: sich hinlegen—“lie down”.
	Hungarian: emel-ked—“get up”.
	Guugu Yimidhirr: daga-adhi—“sit down”.

With regard to middle constructions in their modern sense, there are considerable cross-linguistic differences. These constructions can be broadly categorized into two types: morphological patterns and syntactic analytical patterns. The morphological pattern refers to middle markers that appear as affixes or clitics attached to verbs. Such patterns are widely attested in Indo-European languages. The following serve as examples:

(4)	German
	Der Stoff wäscht sich gut.
	“The fabric launders well”.
	(wäscht = “launders,” sich = middle marker)
	(Fagan, 1992, p. 22)
	French
	Un veston de laine se lave facilement.
	“A wool jacket washes easily”.
	(se = middle marker)
	(Wehrli, 1986, p. 266)
	Russian
	Dver’ legko otkryvaet-sja.
	“The door opens easily”.
	(-sja = middle marker)
	(Greenspon, 1996, p. 24)
	Icelandic
	Bókin selst vel.
	“The book sells well”.
	(selst = verb with middle marker)
	(Valfells, 1970, p. 556)

Unlike German and French, middle constructions in English are of the syntactic analytical type—they do not involve any morphological markers. The same applies to Dutch, where middle constructions are also unmarked and similar to English in structure. The following serves as an example:

(5)	Dutch
	Het brood snijdt gemakkelijk
	“The bread slices easily”.
	(Greenspon, 1996, p. 23)

Based on the cross-linguistic data, many scholars (cf. Keyser & Roeper, 1984; Fagan, 1992; Stroik, 1992; Ackema & Schoorlemmer, 1994) characterize the MC according to the following properties (a–e):

a. The MC is derived from a basically transitive verb (O’Grady, 1980). The verb retains active morphology, and the subject corresponds to the logical object of the verb: the deep-structure (d-structure) direct object assumes the surface-structure (s-structure) subject function (Stalmaszczyk, 1993, p. 133). Accordingly, (6b), as an example of the MC, is derived from its transitive d-structure in (6a).

(6)	a. John cleaned the oven.
	b. This oven cleans easily.

b. There are constraints on the types of objects that can become subjects in the MC: the MC is only possible when the internal argument of a transitive verb is affected by the event or action (Fellbaum & Zribi-Hertz, 1989). This hypothesis accounts for the grammaticality of the sentences in (7):

(7)	a. The ice breaks easily.
	b. *The Eiffel Tower sees easily.

c. The MC is theme-oriented: any agent can perform the action because of the inherent or invariant properties of the theme (Fellbaum, 1985; Fagan, 1988; Ackema & Schoorlemmer, 2003; Marelj, 2004). This means that the action described by the verb is interpreted as dependent on the properties of the theme and subject. For instance, (7a) can be rephrased as follows: “It is easy for anyone to break the ice due to its inherent properties, such as being thin or about to melt”.

d. The MC typically conveys a non-eventive, generic, habitual, or potential interpretation (Ackema & Schoorlemmer, 2003; Lekakou, 2005). This semantic profile makes the MC incompatible with imperatives (see 8a). Syntactically, the MC tends to occur in the present or habitual tense and is incompatible with progressive aspect (see 8b).

(8)	a. *Break easily, the ice!
	b. *The ice is breaking easily now.

e. A modifier is required (Keyser & Roeper, 1984; Fellbaum, 1986). This modifier is typically adverbial and reinforces the interpretation of an inherent or potential property, rather than describing a specific temporal event.

3. Previous Studies on the Chinese MC

Compared with Indo-European languages, the Chinese MC is considerably more complex and has been the subject of extensive discussion since Sung (1994). Sung (1994), followed by Cao (2004, 2005), W. He (2004), Han (2007), J. Xiong (2017), and others, identifies V-qilai as the marker of the MC and characterizes the structure as “NP + V-qilai + AP”. In contrast, other scholars have proposed different analyses: some regard unmarked (notional or pseudo) passives as instances of the MC (Cheng & Huang, 1994; Ting, 2006); others argue that -gei passives should be considered the MC (Shen & Sima, 2010). Additional accounts suggest that easy–difficult constructions (J. Xiong, 2017; Cai & Shi, 2023) and certain modal constructions (Furukawa, 2005; Y. He, 2010; Tao, 2010) also fall under the category of the MC. As a result, there is currently no consensus on what constitutes an MC in Chinese.3

In the following section, I will review these major accounts from previous studies on the Chinese MC one by one.

3.1. Unmarked Passive Middles

Unmarked passives, also referred to as notional passives or pseudo-passives, are illustrated in example (9):

(9)	米饭煮焦了
	mǐfàn	zhǔ	jiāo	le
	rice	cook	burned	CRS
	“The rice was cooked burned”.

Cheng and Huang (1994) propose that Chinese has an MC derived from the ergative construction. They distinguish between surface and deep ergatives. Deep ergatives, as exemplified in (10a), represent legitimate ergatives, while surface ergatives, as seen in (10b), do not involve agentivity. Middles, according to Cheng and Huang, are surface ergatives resulting from the suppression of their agents, whereas deep ergatives maintain ergative properties both in underlying and surface structures.

(10)	a. 门自动开了
	mén	zìdòng	kāi	le
	door	by-itself	open	CRS
	“The door opened by itself”.
	b. *米饭自动煮了
	mǐfàn	zìdòng	zhǔ	le
	rice	by-itself	cook	CRS
	“The rice cooked by itself”.

Ting (2006) adopts a presyntactic approach and follows Cheng and Huang in claiming that the Chinese MC is a “notional passive”. A key issue with this account is its exclusive focus on the formal aspects of the MC, without sufficiently addressing its semantic implications. While “unmarked passives” or “notional passives” indeed reassign the logical object to the subject position, semantically, they continue to denote an event, rather than the inherent property encoded in the MC (as outlined in properties c and d in Section 2). As a result, it can be argued that unmarked passives do not fully capture the meaning inherent in the MC.

3.2. Gei-Middles

Shen and Sima (2010, p. 232) define the MC in which the subject is not the external argument of the agent, but semantically the agent exists, and there must be a structural marker to show the semantics of the external agent, as well as the restriction of the syntax.

(11)	孩子给吃饱了
	háizi	gěi	chī	bǎo	le
	kid	GEI	eat	full	CRS
	“The kid was fed full”.

Example (11) conveys that someone (unspecified) fed the child until they were full. The child is not the agent of the action, but the patient. Gei here functions similarly to a passive or causative-passive marker, somewhat akin to bei in standard passive constructions, but with a more informal, often emotional or emphatic tone. According to Shen and Sima, gei in (11) is a structural marker to show the semantics of the external agent, “the one who fed the kid,” and to restrict the syntax of the external agent at the subject position. Based on their definition, (11) is an example of the MC. Once more, the analysis of gei middles reveals a limitation: it focuses solely on the formal aspects, neglecting the nuanced meanings embedded in these constructions. Furthermore, it is essential to note that gei is generally perceived as an informal passive marker rather than the neutral structural marker suggested by Shen and Sima. Consequently, akin to the limitations observed in “unmarked passive” middles, gei-middles also fail to fully encapsulate the intended linguistic nuances of the MC.

3.3. Easy–Difficult Middles and Modal Middles

Example (12a) is a sentence that is considered an easy–difficult middle involving the morphemes hao (“easy”), rongyi (“easy”), and nan (“difficult”) (Furukawa, 2005; Y. He, 2010; Tao, 2010; J. Xiong, 2017). J. Xiong (2017) adopts a generative syntactic approach and argues that, morphologically, hao functions as a prefix, whereas rongyi and nan are free morphemes. Syntactically, the [NP_patient hao-V] construction is unergative and qualifies as a middle construction. In contrast, the [NP_patient rongyi/nan V] structure displays variable syntactic behavior, which Xiong attributes to two distinct underlying structures: [NP_patient rongyi/nan (…) V_process], representing a middle construction, and [NP_patient rongyi/nan (…) V_rvc], representing a tough construction4. The key distinction lies in the separability of rongyi/nan from the verb.

(12)	a. 这本书容易/好/难读
	zhè		běn		shū	róngyì/hǎo/nán				dú
	this		CL		book	easy/difficult				read
	“This book reads easily/difficultly”.
	b. 这个花瓶打得破
	zhè	gè		huāpíng			dǎ	de	pò
	this	CL		vase			hit	DE	broken
	‘This vase can be broken”.

Xiong also examines the generic [NP_patient V de A] and [NP_patient V bu A] constructions, such as (12b), treating them as middles as well. Although these two constructions contrast semantically in terms of polarity—de being affirmative and bu negative—she argues that they are not syntactic counterparts but rather structurally distinct, yet both align with the broader class of easy–difficult middles.

Tao (2010, p. 105) points out the easy–difficult (including the generic V-de/bu) sentences are akin to the MC in terms of the following properties:

(a)

The patient surfaces as the subject.

(b)

The verb does not carry a passive marker.

(c)

The agent does not appear in overt syntax but remains perceivable.

(d)

Such constructions represent generic statements, expressing an inherent property of the referent of the subject.

Tao (2010) goes on to claim that examples like (13) parallel middles:

(13)	这种水可以重复使用
	zhè	zhǒng	shuǐ	kěyǐ	chóngfù	shǐyòng
	this	CL	water	can	repetitively	use
	“This kind of water can be used repetitively”.

Examples like (13) are modals, and they are considered modal middles because of the following properties (Tao, 2010, p. 116):

(a)

The patient of Vt surfaces as the subject of the sentence.

(b)

The agent of Vt is not overtly realized but implied.

(c)

Vt does not bear passive morphology.

(d)

Such sentences are generic statements, ascribing some inherent property to the referent of the subject.

While sentences with a modal structure, such as example (13), convey a generic inherent property associated with the subject referent, syntactically, they deviate by employing a modal verb rather than an adverbial modifier. Consequently, they fall short of meeting the syntactic criteria prescribed by the MC (as shown in properties e in Section 2).

3.4. NP+V-qilai+AP Middles

Sung (1994) proposes that only some NP + V-qilai + AP sentences qualify as the MC in Chinese. See the examples in (14):

(14)

a. 她笑起来挺好看

tā

xiào-qǐlái

Tǐng

hǎokàn

she

smile

Very

beautiful

“She smiles very beautifully”.

b. 这本书看起来不错

zhè

běn

shū

kàn-qǐlái

bùcuò

this

CL

book

Look

good

“This book looks good”.

c. 这首歌听起来好听

zhè

shǒu

gē

tīng-qǐlái

hǎotīng

this

CL

song

sound

beautiful

“This song sounds beautifully”.

d. 这首诗读起来很顺口

zhè

shǒu

shī

dú-qǐlái

hěn

shùnkǒu

this

CL

poem

Read

very

easy

“The poem reads very easily”.

e. 这个字写起来不方便

zhè

gè

zì

xiě-qǐlái

bù

fāngbiàn

this

CL

character

write

not

convenient

“The character writes inconveniently”.

Sung (1994) excludes (14a) from the category of the MC because the subject does not correspond to the patient or logical object of the verb. W. He (2004) and Han (2007), among others, follow Sung in identifying (14b–e) as examples of the Chinese MC. In contrast, Cao (2004, 2005), and Yu and Si (2008) propose a narrower account, arguing that only (14d–e) qualify as MCs. They suggest that although all the examples involve a patient-subject, in (14b–c), the sentences remain acceptable even if V-qilai is removed. In contrast, removing V-qilai from (14d–e) renders the sentences semantically awkward. This suggests that in (14b–c), the adjective phrase (AP) modifies the subject directly, rather than being predicated via V-qilai as the core predicate, and thus these do not meet the syntactic criteria of the MC.

Tao (2010, p. 13) argues that NP + V-qilai + AP sentences should not be considered middles because the verb in V-qilai forms part of an adverbial phrase rather than the main predicate. According to Tao, V-qilai functions as a modifier of the AP, rather than serving as a main verb. Based on syntactic diagnostic tests—such as the behavior of the reflexive ziji, the A-not-A construction, the distribution of the emphatic marker dou, the focus marker shi, and the adverbial queshi (“indeed”)—Tao concludes that V-qilai does not act as the core predicate. However, Tao’s analysis applies more directly to the (14b–c) type of sentences, in which the AP encodes attributes of the subject, rather than to the (14d–e) type.

J. Xiong (2017) also argues that (14d–e) type of sentences should be considered as the MC. Her analysis of [NP_patient V-qilai AP] demonstrates that the complex predicate [V-qilai AP] is unergative, meaning the subject NP_patient is base-generated in the subject position rather than derived from an object position.

As discussed in Section 3.1, Section 3.2, Section 3.3 and Section 3.4, there are numerous accounts of the Chinese MC. Upon scrutinizing these accounts, it becomes evident that unmarked (notional/pseudo) passive middles and gei- middles do not express an inherent or potential property of the subject referent; instead, they encode temporally bounded events. Formally, modal middles include a modal verb rather than an adverbial modifier, which disqualifies them from being genuine modifiers in the syntactic sense. For certain NP + V-qilai + AP sentences (e.g., 14b–c), the adjectival modifier targets the subject rather than the verb predicate. Accordingly, these so-called middles should not be classified as instances of the Chinese MC.

Consequently, what remain are the easy–difficult middles—including V-de/bu constructions (as in 12)—and a subset of NP + V-qilai + AP sentences (14d–e). In alignment with the syntactic and semantic properties associated with middle constructions, as proposed by scholars such as Keyser and Roeper (1984), Fagan (1992), Stroik (1992), and Ackema and Schoorlemmer (1994), and as discussed in Section 2, I propose that these two sentence types may be considered valid manifestations of the Chinese MC. Unlike the English MC, which has a single form, the Chinese MC appears to take two structural forms, which can be schematized as follows:

(15)	a. [patient-subj AP V]/[patient-subj V de/bu A] (the easy–difficult MC, e.g., (12)).5
	b. [patient-subj V-qilai AP] (the V-qilai MC, such as (14d–e)).6

In this study, I hypothesize that the two forms of the Chinese MC do not share the same base, nor do they originate from the same historical source. This study focuses on the V-qilai MC and its development in the history of Chinese.7

3.5. Previous Research on the Development of the V-qilai MC

In contrast to the abundant synchronic research on the Chinese MC, studies on its diachronic development remain relatively scarce. A few scholars (cf. Xu, 2006; Lin, 2006; Huang & Ma, 2008; Cai, 2015; Yang, 2021) have explored the origins and evolution of the V-qilai MC and generally agree that V-qilai initially functioned as a directional verb phrase, appearing as early as the Southern Song period (1127–1279), as illustrated in (16).

(16)	公若知得放下不好，便提掇起来，便是敬。
	Gōng	ruò	zhī de	fàngxià	Bù	hǎo,
	You	if	know	put-down	Not	good,
	biàn	tíduō	qǐlái,	biàn	Shì	Jìng
	then	lift	up,	then	COP	Respect
	“If you know that putting it down is not good, then just lift it up; that is perfectly acceptable”.
	Zhuzi yulei (1270)

Around the end of the Southern Song period, the sequence of [subj, V-qilai, VP] began to appear in texts. In such structures, the subject, V-qilai, and the VP are loosely sequenced in the discourse, often separated by phonological pauses or other discourse elements. As shown in (17), the logic and implicit subject, “I”; the verb phrase, shuo-qilai; and the VP, “scare you eight times,” are loosely sequenced. Cai (2015) hypothesizes that this loose sequence of [subj, V-qilai, VP] may constitute the diachronic origin of the V-qilai MC.

(17)	你不知道俺哥哥的名儿，若说起来，唬你八跌
	Nǐ	bù	zhīdào	Ǎn	gēge de	míngr,
	you	not	know	my	brother’s	name,
	ruò	shuō-qǐlái,	hǔ	nǐ	bā	Dié
	if	say-up	scare	you	eight	Fall
	“You don’t know my brother’s name—if I told you, it’d scare you so bad you’d fall eight times!”
	Nansong Huaben (ca. 1279)

Huang and Ma (2008) suggest that the immediate source of the V-qilai MC is the use of V-qilai as a parenthetical element within the sequence [subj, V-qilai, VP], which later underwent processes of expansion and coalesce. Similarly, Xu (2006) and Lin (2006) also view V-qilai in this sequence as functioning parenthetically. As illustrated in (18), shuo-qilai is used to insert an evaluative remark about the mountain, functioning more as commentary rather than as part of the core predicate.8

(18)	三众前来，不上半日，果逢一座高山，说起来，十分险峻。
	Sān	Zhòng	qián	lái,	bú	shàng
	Three people		front	come,	not	up to
	bàn rì,	Guǒ	féng	yī	zuò	gāo
	half a day	as expected	come across	one	CL	high
	shān,	shuō-qǐlái,	shífēn	xiǎnjùn.
	mountain,	say-up,	very	dangerously steep
	“The three people came, and within half a day, they came across a high mountain, which (come to think of it), was dangerously steep”.
	Xi Youji (1542)

Yang (2021) further argues that the V-qilai MC evolved from the topic structure [subj, V-qilai, VP], where V-qilai functions as a topic marker.

(19)

a. 说起来，我府中冠、服、衣、靴、履、袜等件，各自派一个养娘分掌。

Shuō qǐlái,

wǒ

fǔ

zhōng

guàn,

fú,

yī,

Speak-up,

my

household

in

hat

robe

clothing

xuē,

lǚ,

wà

děng

jiàn,

gè

zì

boot

shoe,

sock

so on

CL

each

one

pài

yī

gè

yǎngniáng

fēn

zhǎng

assign

one

CL

nursemaid

individually

manage

“Speaking of it, in my household, each item—hats, robes, clothing, boots, shoes, socks, and so on—is assigned to a separate nursemaid to manage individually”.

Yuandai Huaben (ca. 1368)

b. 看起来，这件事有点儿难。

Kàn

qǐlái,

zhè

jiàn

shì

yǒu

diǎnr

nán.

Look-up,

this

CL

thing

EXIST

a bit

Difficult

“It seems that this matter is a bit difficult”.

(Yang, 2021, p. 107)

c. 这个大棚搭起来很高。

Zhè

ge

dàpéng

dā qǐlái

hěn

Gāo

This

CL

greenhouse

set-up

very

Tall

“This greenhouse sets up very tall”.

(Yang, 2021, p. 107); Yongzheng Jianxiatu (1920)

In (19a), shuo-qilai functions as a topic marker, which originally appeared in sentence-initial position. According to Yang (2021), the grammaticalization of the V-qilai MC involved a shift in the position of the topic marker from sentence-initial to medial, along with the elimination of the phonological pause between the topic marker and the following clause. At the same time, the subject was extended to a patient role, the verb category expanded from speech, perceptual, and mental verbs to include more transitive verbs, and the post-V-qilai verb phrase was reanalyzed as an adjectival phrase (AP). As shown in (19b), When used as a topic marker, kan-qilai does not contribute to the propositional content of the sentence and therefore occupies a syntactically flexible position; it can be omitted without altering the core meaning. In contrast, in the MC, as in (19c), da-qilai contributes to the propositional meaning and can no longer be omitted or displaced.

As shown above, previous studies have investigated the origins and evolution of the V-qilai MC. However, there is no consensus on the developmental trajectory of the Chinese MC, and much research tends to conflate the functions of V-qilai as a middle marker and as a discourse device. Moreover, many studies focus on the grammaticalization of a single lexical item—either the verb or V-qilai—without situating it within the broader constructional context. This paper adopts a constructional perspective, focusing on how a set of constructions gradually assembled, interconnected, and ultimately led to the emergence of the MC.

4. Data and Framework

In this section, I provide a brief examination of the underlying theoretical framework and the concept of constructional assemblies in 4.1, while offering an overview of the data for this study in Section 4.2.

4.1. The Constructional Framework

Scholars such as Barðdal et al. (2015) and Fried (2008) have noted various interpretations of Diachronic Construction Grammar. This paper adopts the constructionalization framework proposed by Traugott and Trousdale (2013) and further developed by Petré (2019). This perspective considers constructions as symbolic and conventionalized pairings of form and meaning9 (Croft, 2001; Goldberg, 2006; Bybee, 2010; F. Zhan, 2022; among others). These constructions form a structured inventory or network that includes both vertical links (taxonomic and meronymic) and horizontal links connecting different constructions. Vertical links connect lower-level constructions to their higher-level schematic parents, while horizontal links connect sister constructions that are members of the same constructional schema or constructions that co-occur in actual discourse. This kind of syntagmatic knowledge of co-occurrence relations is well-documented in synchronic research on collocates and, specifically, within Construction Grammar, in the study of collostructions (Gries & Stefanowitsch, 2004).

Diachronic Construction Grammar understands language change as involving modifications within a network of constructions. These changes may encompass the creation of new nodes—leading to constructionalization—as well as the reorganization of the constructional network and other possible transformations (Perek, 2020; Smirnova & Sommerer, 2020).

Petré (2019), drawing on Langacker (2009), introduces the concept of an assembly—a meaningful compositional configuration of existing constructions and their recurrent co-texts. He proposes that a key factor in the emergence of a new construction lies in the regular use of such assemblies, which may shift over time as speakers begin to prefer certain configurations in specific contexts. Assemblies consist of a set of constructions that co-occur and are functionally integrated. For instance, rather than positing a single BE going to V “motion with a purpose” construction around 1600, Petré (2019, p. 166) identifies an assembly comprising three distinct constructions:

(20)	a. The progressive construction [[BE Ving] * [ongoing activity]];
	b. the verb [[GO] * [“go”]] construction;
	c. the purposive construction [[to INF] * [intended activity]].

According to Petré (2019), and Budts and Petré (2020), each of these assemblies can function independently, yet they are frequently combined within a single expression. This combination forms a recognizable pattern, though it does not qualify as a conventionalized construction for the following reasons: (i) the combination is fully compositional and lacks idiosyncrasy; and (ii) it remains phonologically intact—never reduced to forms like gonna—indicating it is not processed as a single lexical unit. In the 17th century, such combinations were often employed in contexts where motion was only vaguely specified, particularly in topicalized or passivized structures that downplayed the notions of motion and agency.

By the early 18th century, instances began to emerge in which there was no overt agent, and references to “motion with a purpose” were absent. In these contexts, the BE going to V structure began to exhibit semantic bleaching and idiosyncrasy, ultimately becoming non-compositional. Over time, it came to align more closely with other future-tense expressions (e.g., will, shall, and be to), and was eventually subject to phonological reduction in spoken language, yielding forms such as gonna and its variants.

Petré (2019, p. 165) argues that when multiple assemblies increase in frequency simultaneously, they may begin to coalesce, leading to the emergence of a new construction. This process can be interpreted as follows: when speakers or writers repeatedly produce these assemblies within a specific time frame, their patterned usage fosters the conventionalization of a construction. This emergence is gradual and cumulative. Importantly, passive and topicalized examples from the 1620s—which predate the constructionalization of BE going to V as a future marker—do not unambiguously encode future meaning. Instead, they constitute part of the enabling conditions that paved the way for the eventual rise of the construction.

4.2. Data

This study draws on data retrieved from the online searchable version of the Chinese corpora maintained by the Center for Chinese Linguistics (CCL) at Peking University (W. Zhan et al., 2003) and most recently updated in 2024. The current version of the CCL corpora comprises approximately 5.84 billion characters and includes both Modern and Classical Chinese texts. All data used in this paper were manually curated and selected from these corpora. The CCL Classical Chinese Corpus contains texts spanning Old, Middle, and Early Modern Chinese10.

Within the CCL Classical Chinese Corpus, there are 13,925 tokens of qilai, with a frequency of 2.4 per million characters (pmc). The compound qilai is attested as a motion verb beginning in Middle Chinese (see example (21)), and V-qilai is observed functioning as a directional verb phrase from as early as the 10th century (see example (22)). Notably, there are no instances of the MC involving qilai recorded before 1900.

This paper hypothesizes that the constructionalization of the Chinese V-qilai MC arose from assemblies comprising a subject, a directional verb construction, and a descriptive construction. While each of these constructions may function independently, they frequently appear together within a single utterance. Over time, particular assemblies and their favored contextual uses began to shift. From the 13th century onward, a key development was the shift in subject roles—from agents to patients—resulting in a reduced emphasis on motion and agency, as seen in example (25). This evolution ultimately led to the emergence of the MC in the early 20th century (see example (26)). An analysis of Classical Chinese data reveals that the increased frequency of these assemblies likely facilitated the rise of the V-qilai MC.

The following section provides a detailed account of the historical development of the V-qilai MC in Chinese.

5. The Development of the Chinese MC

In the preceding section, I elaborated on the V-qilai MC in Modern Chinese. The subsequent part will delve into the genesis of the V-qilai MC (Section 5.1), as well as the constructional modeling of the emergence of the V-qilai MC (Section 5.2).

5.1. The Origin of the V-qilai MC

The term qilai originated in Middle Chinese as a compound verb indicating motion. In (21), qilai functions as a compound verb signifying “rising” or “standing up” from a bed, seat, etc. While the subject of the verb compound is implicit, it can be inferred from the context to be the agent (“I”). The verb compound is immediately followed by another verb phrase, “look at the blue sky,” forming a serial verb construction. This demonstrates that from its inception, the verb compound qilai has been utilized as the initial verb in a serial verb construction.

(21)	不觉朝已晏，起来望青天。
	Bù	jué	cháo	yǐ	yàn,	qǐlái	wàng	qīng	tiān
	Not	notice	Morning	already	sunny,	get-up	look-at	blue	sky
	“Before (I) noticed, it had been sunny, and (I) got up to look at the blue sky”.
	Wei Yingwu (737–792)

Also, in Middle Chinese, qilai occurred as a complement attached to a verb indicating direction. In (22), -qilai is a directional complement attached to the preceding verb jia (“pick”), and together they form a verb phrase meaning “pick up”. The context of (19) is as follows: A little monk went to see the master at midnight. The master asked him to stir the fire in the charcoal of the stove, and the monk said, “there is no fire”. Then, the master stirred the charcoal. He saw a spark of fire and picked it up to show the little monk.

(22)	见一星火，夹起来云：“这个不是火是什摩”。
	Jiàn	yī	xīnghuǒ,	jiā-qǐlái	yún:	“Zhège	bù	shì	huǒ	shì	Shénmó”.
	See	one	spark of fire,	pick-up	say	this CL	not	COP	fire	COP	what
	“(The master) saw a spark of fire, picked it up and said, ‘if this is not fire then what is it?’”
	Zutang Ji (937–975)

As in example (21), although the subject of the directional verb phrase jia-qilai “pick up” in (22) is implicit, it can be inferred from the context to be “the master,” who is also the agent of the serial actions: “pick it up” and “say”. Following the directional verb phrase is another verb, “say,” which introduces the words spoken by the master. This again illustrates that from its earliest attested uses, the directional verb phrase V-qilai has consistently functioned as the initial verb phrase in a serial verb construction. The serial verb structure in (22) can be schematized as follows: [(agent-subj) V-qilai VP]. In this example, [agent-subj], [V-qilai], and [VP] are loosely assembled within the discourse, separated by phonological pauses and intervening elements.

In Early Modern Chinese, the directional complement -qilai began to take on a new function, marking the onset of an action or state. This semantic extension from expressing physical direction to conveying temporal initiation exemplifies a broader linguistic trend in grammaticalization, where spatial meanings shift toward abstract temporal ones (Heine & Kuteva, 2002). In example (23), while -qilai still appears to be structurally tied to the verb zuo (“do”), its original sense of physical movement has largely faded. The presence of the adverbial phrase congtou (“from the start”) highlights that -qilai now serves to signal the initiation and progression of an activity.

(23)

颜子“克己复礼”工夫，却是从头做起来。

Yánzǐ

“kè

Jǐ

Fù

lǐ”

gōngfū,

què

shì

cóng

tóu

zuò-qǐlái

Yanzi

restrain

self

Restore

ritual

effort,

but

COP

from

beginning

do

up

“Yanzi’s efforts on self-restraint and restoration to ritual nevertheless started from scratch”.

Zhuzi yulei (1270)

Example (23) highlights the semantic expansion of V-qilai from a motion expression to more bleached functions, which provides the semantic foundation for the emergence of the MC. In the examples of (24), the verb phrase V-qilai again occurs in a string [V-qilai] [VP], where clearly the physical motion meaning of -qilai is bleached, as in (23). The context of (24a) is Master Zhu Xi tells his disciples the following: “It is very important to have ideas when you engage in scholarship. If you have some, do not block them. It won’t work if the ideas are blocked. How much effort it will take to put ideas together again!” In (24b), the speaker is too embarrassed or ashamed to recount what happened.

(24)

a. 待得再新整顿起来，费多少力！

Dàidé

zài

Xīn

zhěngdùn-qǐlái,

fèi

duōshǎo

lì!

Then

again

Newly

rectify-up,

take

how much

effort

“How much effort it will take to put ideas together again!”

Zhuzi Yulei (1270)

b. “师父！莫要问！说起来就活活羞杀人！”

Shīfu!

Mò

yào

wèn!

Shuō-qǐlái

jiù

huóhuó

xiū

Master

do

not

ask

say-up

just

utterly

embarrass

shā

rén!

kill

people

“Master! Please don’t ask! Just talking about it is utterly embarrassing!”

Xi Youji (1542)

In (24a), the subject is implicit but can be inferred from context as “the ideas”. It does not serve as the agent of the verb phrase V-qilai; rather, it functions as the patient and logical object of the verb11. The two verb phrases in (24a), zhengdun-qilai “rectify-up” and fei duoshao li “take much effort,” do not constitute a serial verb construction, as they do not represent sequential actions. Instead, the second verb phrase is descriptive in nature, qualifying or elaborating on the first. Structurally, (24a) presents three components—[patient-subject], [V-qilai], and [descriptive VP]—which are loosely connected, separated by prosodic pauses or intervening material. By contrast, in (24b), although the subject remains implicit, the [V-qilai] and [descriptive VP] components are closely integrated, forming a more cohesive unit. Here, the adverb jiu as part of the descriptive VP expresses immediacy and direct consequence between shuo-qilai and the VP. This syntactic tightening signals a significant development in the emergence of the V-qilai MC.

In comparison with (24), the examples in (25) exhibit a more integrated structure comprising a patient-subject, the verb phrase [V-qilai], and a descriptive phrase, with all components tightly combined rather than loosely conjoined. Additionally, the directional complement -qilai has undergone complete semantic bleaching of its original motion meaning. In (25a), shuo-qilai forms the verb phrase following the noun phrase Xiaomei, which acts as the patient and logical object of the verb shuo (“say”). This verb phrase is followed by the negative copula bushi plus a descriptive noun phrase “a typical young girl,” which combines to form a [COP-NP (descriptive)] structure.

In (25b), shuo-qilai follows the patient-subject, “these words”, and is again followed by shi modified by the adverb qie (“truly”), but this time the descriptive adjective chang (“long”) is used. Compared to (25a), shi in this instance may be interpreted either as a copula or as a focus marker. When interpreted as a copula, the descriptive phrase follows the pattern [COP-AP (descriptive)]; when interpreted as a focus marker, the structure simplifies to [AP (descriptive)]. Regardless of interpretation, the phrase qie shi chang conveys a non-eventive, generic, and potential reading of the action.

In example (25c), kan-qilai appears after the patient subject “the killer”, and is followed by the negated form shi, which is preceded by the adverb ye (“also”). The phrase tu cai hai ming (“to kill for money”) serves as a descriptive expression. As with (25b), shi here can be understood either as a copular verb or as a focus marker. Regardless of this ambiguity, tu cai hai ming conveys a reading that is non-eventive, general in nature, and suggestive of possibility rather than actual occurrence.

(25)

a. 这小梅说起来不是小家儿女

Zhè

Xiǎoméi

shuō-qǐlái

bù

shì

xiǎo

jiā

ér

nǚ

This

Xiaomei

say-up

not

COP

small

family

son

daughter

“Xiaomei, spoken of, is not a typical young girl”.

Xue Yue Mei (1574)

b. 这话说起来且是长哩。

Zhè

huà

shuō-qǐlái

qiě

shì

chǎng

lī

These

words

say-up

truly

SHI

long

PTCL

“This words when spoken of are rather long (to tell)”.

Sanbao Taijian Xiyangji (1597)

c. 这个杀人的, 看起来也不是图财害命。

Zhè

ge

shā

Rén

de,

kàn-qǐlái

yě

bù

shì

tú

cái

This

CL

kill

people

NOM,

look-up

also

not

SHI

seek

money

hài

mìng.

kill

life

“This killer doesn’t seem to have done it for money either”.

Xianv Qiyuan (1821)

Examples in (25) demonstrate that the three assemblies, [patient-subj], [V-qilai], and [(COP)-AP/NP/VP (descriptive)], assembled in one sentence encoding a non-eventive, generic, and potential interpretation of the action with both motion and agency diminished. This forms a recognizable pattern exhibiting the characteristics of an MC.

However, I maintain that examples in (25) are not conventionally established as an instance of the V-qilai MC because the combination is compositional, and the explicit shi tends to mark (25) as a descriptive predicate structure with the post-shi NP/AP/VP as the main predicate. In early examples with a patient-subject, the verb phrase [V-qilai], and a descriptive phrase assembled, the descriptive phrase often involved a shi. This may be because shi serves as a syntactic bridge that facilitates the integration of the evaluative or attributive component with the preceding verb phrase. The presence of shi allows the speaker to shift smoothly from reporting an action (as expressed by [V-qilai]) to offering a characterization or assessment of the subject.

Examples of (26) emerged in later texts (early 20th c.) in which shi is excluded and the three assemblies, [patient-subj], [V-qilai], and [AP (descriptive)], closely combined into a subject–predicate structure [patient-subj V-qilai AP], resulting in a non-eventive, generic, and potential representation of the action.

(26)	a. 每句话说起来都很相投
	Měi	jù	huà		shuō-qǐlái		dōu		hěn	Xiāngtóu
	Every	CL	word		Say		all		very	Agreeable
	“Every sentence feels very agreeable”.
	Qi Hong Ting (1900)
	b. 这话说起来很长。
	Zhè	huà		shuō-qǐlái		hěn		chǎng
	These	words		Say		very		long
	“These words speak long (it’s long to say these words)”.
	Laocan Youji (1904)

In this case, the construction [patient-subj V-qilai AP] becomes idiosyncratic and non-compositional, and the V-qilai MC is constructionalized. Over time, the MC becomes increasingly expanded, with different types of verbs occurring in the construction, as in (27).

(27)	这手功夫说起来很容易，做起来很难。
	Zhè	shǒu	gōngfū	shuō-qǐlái	Hěn	róngyì,	zuò-qǐlái	hěn	nán
	This	CL	skill	say	Very	easy,	do	very	difficult
	“This skill speaks easily, but does difficultly (easy to say but difficult to do)”.
	Yongzheng Jianxiatu (1920)

The two sentences in (27) also exhibit the V-qilai MC [patient-subj V-qilai AP] with a non-eventive and generic meaning, and the verb in V-qilai expanded from speech verb shuo to action verb zuo, which demonstrates that the MC became entrenched in early 20th century. Later in Modern Chinese, more transitive verbs, such as chi (“eat”), xi (“wash”), mai (“buy”), etc., are recruited to the V-qilai MC, resulting in the MC schema being augmented and enhanced.

5.2. A Constructional Model of the Development of the V-qilai MC

As mentioned in the above discussion, the V-qilai MC originated from the assemblies of the three constructions [patient-subj], [V-qilai], and [AP (descriptive)]. Each of the constructions is capable of being used standalone, yet they are combined within a single utterance. Over time, such assemblies became idiosyncratic and non-compositional, and the V-qilai MC was constructionalized. This exemplifies a procedural constructionalization, emerging incrementally through a series of micro-steps (refer to Traugott & Trousdale, 2013). Concretely, the constructionalization of the V-qilai MC unfolds through the subsequent principal phases:

(a)

Qilai as a verb compound was used as the directional complement in a verb phrase denoting motion. It occurred in the initial part of a serial verb construction with an implicit agent subject, e.g., (22).

(b)

Semantic expansion refers to semantic shifts in Early Modern Chinese when motion and agency were reduced and sequence of actions changed into description of an action, e.g., (23, 24). This semantic expansion exemplifies the progression of the linguistic component from a tangible, spatial significance to increasingly abstract and adaptable applications. Such a phenomenon is not rare within languages, mirroring the dynamic evolution of linguistic systems over time, akin to the development seen in the English language with “Be going to (future)”.

(c)

Syntactic expansion: The patient-subject was no longer implicit, and it appeared with V-qilai as the topic, followed by the structure consisting of the copula and an adjective/NP, e.g., (25). This is an important stage because even though it may exhibit a predicate adjective structure, the three constructions were closely assembled, which sets an important context for it to change into a patient-subject V-qilai MC.

(d)

Further syntactic expansion occurred when the copula were no longer explicit and the three assemblies, [patient-subj], [V-qilai (action)], and [AP (descriptive)], were syntactically integrated into the subject–predicate structure, indicating a non-eventive, generic, habitual, potential interpretation of the action. As a result, the V-qilai MC was constructionalized, e.g., (26, 27).

These steps are summarized as follows:

(a)

[(agent-subj)], [V-qilai (motion)] [VP (action)] (ex. 22);

(b)

[(patient-subj)], [V-qilai (action)] [VP (descriptive)] (ex. 23, 24);

(c)

[patient-subj V-qilai (action)] [COP AP (descriptive)] (ex. 25);

(d)

[patient-subj V-qilai (action) AP (descriptive)] (ex. 26, 27).

The result of the sequence of micro-steps is the constructionalization of a form_new-meaning_new construction: the V-qilai MC.

5.3. Frequency

As suggested in the above discussion, the V-qilai MC came into being in the early 20th century. An extensive search in CCL Classical Chinese corpus reveals that there were no instances of the V-qilai MC before 1900, but the frequency of the assemblies of [patient-subj], [V-qilai], and [VP/AP (descriptive)] grew over time, and this may have led to their interconnection and eventually to the emergence of the V-qilai MC.

Table 1. The assemblies of [(patient-subj)][V-qilai][VP/AP (descriptive)] in CCL Classical Chinese corpus12.

Time Period	[(Agent-Subj)] [V-Qilai] [VP (Action)]	[(Patient-Subj)] [V-Qilai] [VP/AP (Descriptive)]	Total Tokens of [(Subj)] [V-Qilai] [VP/AP]
10th c. (Tang and Five Dynasties)	1 (100%)	0 (0%)	1
11th–13th c. (Song and Yuan Dynasties)	43 (68.3%)	20 (31.7%)	63
14th–16th c. (Ming Dynasty)	68 (33.3%)	136 (66.7%)	204
17th–1900 (Qing Dynasty)	158 (21.5%)	577 (78.5%)	735

shows the growing percentage of the assemblies of [(patient-subj)] [V-qilai] [VP/AP (descriptive)] among the total tokens of [(subj)] [V-qilai] [VP/AP] in CCL Classical Chinese corpus.

Table 1 demonstrates that when the assemblies of [(patient-subj)] [V-qilai] [VP/AP (descriptive)] experienced simultaneous frequency growth, their functions shifted, and the associations among these combinations strengthened. In this case, the assemblies underwent functional change from indicating a sequence of actions of an agent to encoding a non-eventive, generic, and potential interpretation of the action with the agency and motion demoted and diminished. With increased frequency, the assemblies became interconnected, eventually giving rise to the new V-qilai MC construction. The concept can be understood in this manner: when speakers or writers consistently replicate structures within a particular timeframe, it can nurture the formation of patterns and ease the establishment of a conventionalized construction. This gradual process unfolds over time. It is crucial to emphasize that semantic and syntactic expansions dating back to the 13th century, preceding the constructionalization of the V-qilai MC in the early 20th century, do not decisively signify the V-qilai MC; instead, they play a role in creating the circumstances that facilitate the eventual emergence of the new construction.

6. Conclusions

This paper addresses [patient-subj V-qilai AP] as the V-qilai MC in Modern Chinese and hypothesizes the constructionalization process of the V-qilai MC in the history of Chinese. I have argued that there are two sentence types in Modern Chinese that can be regarded as manifestations of the Chinese MC, and they are the easy–difficult MC and the V-qilai MC. In alignment with the syntactic and semantic characteristics endorsed by Western scholars, in both types of the Chinese MC, the patient is instantiated as the subject; the verb does not carry a passive marker, and it is modified by a descriptive adjective phrase. Both types of the MC indicate a non-eventive, generic, habitual, potential interpretation of the action encoded by the verb.

Diachronically, I hypothesize that the V-qilai MC originated from the assemblies of three constructions [patient-subj] [V-qilai] [AP], and the constructionalization of the V-qilai MC involved a series of micro-steps. The key stages in the development involve semantic shifts in Early Modern Chinese, where motion and agency were downplayed, and sequences of actions began to be interpreted as descriptions of actions. A major syntactic shift occurred in the 16th century, when the patient-subject, previously implicit, began to appear explicitly alongside V-qilai and a descriptive phrase in a copular sentence. These developments mark the steps leading up to the constructionalization of the V-qilai MC.

I propose that the constructionalization of the V-qilai MC was enabled by the strengthening of associations among the three assemblies. In this process, the growing frequency of the combination of [(patient-subj)] [V-qilai] [VP/AP (descriptive)] functions as a dynamic catalyst giving rise to the new construction. Following the appearance of shuo-qilai in the early 20th century, additional verbs were chosen and integrated into the collection, causing the MC schema to broaden, and the MC network was enhanced.

This study contributes to the evolving field of Diachronic Construction Grammar by illustrating how the strengthening of assemblies can give rise to a new construction. Specifically, it argues that the V-qilai MC did not simply originate from the directional motion construction; instead, it emerged through the replication and contextual assembly of multiple smaller constructions, whose gradual convergence over time resulted in the formation of the V-qilai MC. It also highlights that the emergence of a new construction is often preceded by frequency shifts within assemblies, which in turn facilitate gradual semantic change. Moreover, this paper emphasizes the importance of incorporating the MC into constructional analysis. Such an exploration of how the MC and middle voice are conceptualized in Chinese offers valuable insights into cross-linguistic typological inquiries on the evolution of voice phenomena across human languages.

This paper could be understood as a pioneering contribution that seeks to expand the descriptive and explanatory potential of the constructional approach beyond its original application domain. The Chinese MC presents a valuable testing ground for evaluating how constructional assemblies can accommodate typologically distinct structures. By engaging with data from a non-Indo-European language, this study provides novel empirical insights and opens new pathways for refining and extending the approach. Moreover, this study lays a foundation for future cross-linguistic work, inviting scholars to consider how constructional mechanisms might operate across languages—a line of inquiry that has important implications for both linguistic theory and typology.