21.2.1 Titles, Professions, and Locations

In Modern Chinese, a surname must precede the title, such as Zhāng Xiānshēng “Zhang Master” and Wáng Nǚshì “Wang Madam,” the order of which is the reverse of the English expressions. However, there were two alternatives in Old Chinese; for instance, in Analects (c. 500 BC), the honorific title zǐ “Master” could either precede the surname (e.g. Kǒng Zǐ “Kong Master” and Zēng Zǐ “Zeng Master”) or follow it (e.g. Zǐ Gòng “Master Gong” and Zǐ Zhāng “Master Zhang”).

Similarly, as with professional designations, in Contemporary Chinese, surnames always precede professional titles, as in Wáng Jiàoshòu “Wang Professor” and Lǐ Lǎoshī “Li Teacher.” In contrast, in Old Chinese, the constituent order was the opposite. For example, in Zhuangzi (c. 300 BC) there were Páo Dīng “Butcher Ding,” Yì Qiū “Chess-player Qiu,” and Jiàng Shí “Craftsman Shi,” which have the same order as in English. Note that, if the modifiers were adjectives, the order was always the opposite: Adj. + professional title, as illustrated in (9):

If the modifier was a surname, the order was “professional title + surname”; if the modifier was an adjective, the order was “Adj + professional title.” The same contrast is found in English, e.g. Professor Smith and an excellent professor.

Moreover, the same constituent order occurred in naming locations. In Modern Chinese, place names always precede political units, such as Běijīng shì “Beijing City.” However, in Old Chinese, the order was the opposite. For example, in the texts of the Zuo Zhuan (550–400 BC), there were Chéng Yǐng “City Ying” and Chéng Pú “City Pu.” Similarly, in Old Chinese, we find expressions such as hé zhōng “river middle” and gǔ zhōng “valley inside,” in which normal nouns precede locative nouns. In Contemporary Chinese, the constituent order is exactly the opposite.

Apparently, the order of the modifier and the head noun in Old Chinese was different from that in Contemporary Chinese. It remains unclear what motivated the change in the above constituent orders within nominal phrases.

21.2.2 Head Noun and “Num + CL” Phrases

In Contemporary Chinese, the “Num + CL” phrase always precedes the head noun, and a proper classifier must be used to connect the number to the head noun, such as sān běn shū “three Cl book” and wǔ gèrén “five Cl person.” However, this grammatical structure was not firmly established until the thirteenth century AD. According to Reference ZhangZhang (2001), in Early Old Chinese the number modifier generally followed the head noun instead, as illustrated in (10):

In Old Chinese, the situation changed: numbers generally preceded the head noun; although classifiers were mostly not needed, if there was a classifier, the “Num + CL” phrase had to occur in postnominal position, namely following the head noun:

Classifiers grammaticalized from ordinary nouns over time – a change that can be traced back to as early as the first century BC (Reference WangWang 1989: 18). It was in postnominal position that ordinary nouns gradually developed into classifiers, and, once they were fully grammaticalized, “Num + Cl” phrases started to occur before the head noun (Reference ShiShi 2016: 676‒697).

21.2.3 Adjectival Modifier and Head Noun

When used as a modifier, in Contemporary Chinese, the adjective must precede the head noun, regardless of the structural complexity (i.e. weight) of the adjectival phrase. If the adjective is simple, the associative morpheme de is optional in connecting the modifier and the head, as in dà shù “big tree” and dà de shù “big ASSOC tree.” If the adjective is quantified or intensified by reduplication, degree words, and the like, the associative morpheme de becomes obligatory to connect the adjectival modifier and the head, as illustrated in (12):

1. (12)

   好好的衣服hǎohǎo de yīfú	*好好衣服	*hǎohǎo [ ] yīfú
   很貴的車hěn guì de chē	*很貴車	*hěn guì [ ] chē
   “quite good clothes”
   “very expensive car”

However, there were two alternative orders in Old Chinese: adjectival modifiers could either precede or follow the head noun, as illustrated in the following examples. In these cases, the adjectival phrases in postnominal position were often marked by the relativizer zhě, signaling that Old Chinese treated complex adjectival modifiers as a type of relative clause.

When adjectives preceded the noun, grammatical markers were not needed to associate them. When adjectives followed the head noun, the morpheme zhī optionally occurred between them, as illustrated in (15), but the morpheme zhě was necessary to mark the postnominal adjectival modifiers, as illustrated in (16). In Old Chinese, the morpheme zhī was the most widely used grammatical marker in adjectives, genitives, and relative clauses – a very common cross-linguistic phenomenon (Reference AristarAristar 1991). Note that the morpheme zhī would occur between the head and the modifier regardless of whether it was “modifier–head” or “head–modifier” order, although the former was more common than the latter. According to Reference DryerDryer (1992), adjectives across languages are treated as relative clauses; thus the morpheme zhě was also an obligatory marker for postnominal relative clauses – a point that we discuss in great detail in Section 21.3.

21.2.4 Adverbial Phrases in Sentence-Final Position

In Modern Chinese, adverbial phrases always precede the matrix verb, e.g. bái lái-le “in-vain come PERF” and fēicháng xǐhuān “extremely love.” However, in the history of Chinese, adverbial phrases appeared relatively freely in sentence-final position, as illustrated in the following examples:

In Contemporary Chinese, however, adverbial phrases are not allowed to appear in sentence-final position, and they must occur somewhere in preverbal position, either at the beginning of a sentence or between the subject and the predicate.

21.2.5 The Prepositional Modifier and Its Head Noun

A PP can be used as a modifier to delineate the referents of the NP or the VP. In English, PPs can only follow the head NP, such as the chair in the classroom or the cook in the kitchen. In contrast, in Contemporary Chinese, the PP always precedes the head;Footnote ⁹ when modifying a nominal head, PPs behave like relative clauses, which must be marked by the relativizer de, as illustrated in (19):

In Old Chinese, however, the PPs used as modifiers usually followed the head noun, which was marked by the relativizer zhě – the same marker for postnominal relative clauses (we discuss this issue in detail in Section 21.3):

That is, when used as modifiers, PPs and relative clauses were treated as having the same grammatical status at the time, both following the head noun and being marked by the relativizer zhě. From the Modern Chinese period to the present day, both have been restricted to prenominal position.

In summary, the constituent orders in Old Chinese were: (a) noun–adjective, (b) verb–adverb and (c) verb–PP; historically, more importantly, there was also noun–relative clause, as discussed below. That is, the constituent orders of many structures in Old Chinese were systematically different from those in Contemporary Chinese. Considering all four of the major VO and OV language features listed in Table 21.1, it is clear that the Chinese language historically was much more consistent with VO languages than Contemporary Chinese is. In other words, the grammatical system of Contemporary Chinese fundamentally deviates from the typical properties of SVO languages. The focus of this chapter is on the motivation of the historical typological change from “NP-REL” to “REL-NP.”

21.3.1 Positions of Relative Clauses

The above two examples actually represent two subtypes of relative clause in Old Chinese for which distributions differed. The type of relative clause in (22) took the order in (24a), but the type in (23) adopted the order in (24b):

1. 1. (a)

      NP + REL + zhě,

   2. (b)

      REL + zhī + NP.

In the pattern of (24a), relative clauses followed the noun, which was necessarily marked by the clause-final relativizer zhě. In the pattern of (24b), relative clauses instead preceded the noun, which was necessarily associated with the marker zhī. There was a basic division of labor between the two patterns: roughly, the relative clause in (24a) expressed a real event that had already happened, and that in (24b) referred to a statement about functions, properties, or other characteristics. Therefore the two types of relative clause, as exemplified in (22) and (23), respectively, actually corresponded to two different structures in Old Chinese: (22) corresponded to (24a), and (23) corresponded to (24b).

The major division of labor between the two relativizers zhī and zhě was that the head noun of the zhě clause had to be the subject of the clause, which is typically a human noun, but the head noun of the zhī clause could be the object of the clause, which might be a patient argument, a temporal expression, a locative argument, an instrument argument, and so on. According to Reference CroftCroft (1990), many languages, including English, have nonfinite forms (participles) that function as relative clauses, allowing only subject NP heads, as in the child playing in the bedroom. The zhě clause in Old Chinese behaved like these nonfinite relative clauses in English.

The distinctive distributions of the zhī and zhě clauses were related to their lexical sources. Both zhī and zhě were demonstratives in Old Chinese, but zhī always preceded a head noun and zhě typically followed a head noun. According to Reference Heine and KutevaHeine and Kuteva (2002: 113‒115), relative clause markers are also frequently derived from demonstratives. There is a cross-linguistic tendency: if a language has a noun before a demonstrative, then it has a noun before a relative clause (Reference HawkinsHawkins 1983: 84). This tendency implies that, if a language has a noun after a demonstrative, then it has a noun after a relative clause, which explains why the two relativizers differed syntactically.

As (25) shows, the demonstrative zhī always preceded its head noun, but the demonstrative zhě often occurred after its head noun when used as an anaphor coindexing the preceding noun. The example in (26) is the typical structure of copular sentences in Old Chinese, in which the demonstrative zhě was used as an anaphor to coindex the subject (Reference WangWang 1989: 65). These distinctive contexts for their grammaticalization into relativizers were responsible for the different distributions, prenominal or postnominal.

In Early Old Chinese, there were no overt relativizers to mark a relative clause and a VP was simply placed prior to the head noun to delineate the referents of the noun (Reference ChenChen 1956: 133, Reference GaoGao 1987: 283, Reference SerruysSerruys 1981: 356, Reference ZhangZhang 2001):Footnote ¹⁰

New patterns for relative clauses came into being around the fifth century BC. At that time, the morpheme zhě could function as a complementizer to introduce an object clause, as illustrated in (28):

The morpheme zhě in Old Chinese behaved exactly like that in English, which has multiple functions, such as demonstrative, relativizer, complementizer, and anaphora; a comprehensive discussion of its various usages is beyond the scope of this chapter. In what follows, I concentrate on the syntax of its relativizer use.

21.3.2 The Syntax of the Zhě Relative Clause

The syntax of relative clauses in any language can be examined from two perspectives: first, what role the head noun plays in the clause, such as subject or object, and second, which noun of a declarative sentence can be relativized. As mentioned previously, in Old Chinese, the head noun can only be the subject of the relative clause. However, various nouns in different syntactic positions can be relativized. Reference Keenan and ComrieKeenan and Comrie (1979) presented the well-known accessibility hierarchy as follows:

1. (29) Subject > Direct Object > Indirect Object > Oblique Object > Genitive Noun > Object of Comparison.

The above formula indicates that in any language the primary relative clause-forming strategy must apply to the top segment of the hierarchy. In other words, if it applies to any position in the hierarchy, it must also apply to all of the positions above this position. In Old Chinese, the postnominal relative clause marked by zhě could apply to the nouns in all of the syntactic positions listed in (30) (Reference ZhuZhu 1983, Reference WangWang 1989).

(a) Subject. This structure is the most common “NP + REL” structure from Old to Medieval Chinese. In the following examples, the parts in bold font are the subjects, and the underlined parts are the relative clauses marked by the relativizer zhě. Since these clauses describe a real event that has already happened, they can occur only in postnominal position, marked by zhě. In these cases, the prenominal relative clause marked by zhī could not be used since it could not introduce a relative clause portraying a concrete and real event, as mentioned previously. We saw one such example in (5); two more examples follow:

Throughout the history of the Chinese language, there has been a special type of sentential structure in which a “subject + predicate” phrase can be used directly as the predicate of a higher-level sentence without being connected by any verbal element, such as a copula, a structure that is rarely attested in other languages (Reference ZhuZhu 1982: 100). In (32), mù “tree” is the main subject, which is modified by the relative clause shí fán zhě “whose fruits are too numerous,” and the predicate is another clause consisting of a subject (i.e. zhī “branch”) and a predicate (i.e. bì pī “must crack”):

(b) Object. According to my investigation, a relative clause that did not describe a concrete event could also be used in the postnominal position, marked by the relativizer zhě; for instance, the relative clause in (33) states the possibility “who is likely to give birth to a boy.” In other words, the scope of the zhě relative clause is broader than that of the zhī one: the former can introduce either a real clause or a subjunctive clause in postnominal position, but the latter can introduce only a subjunctive clause in prenominal position. Occasionally, the possessive marker zhī can occur between the head noun and the zhě relative clause, forming a pair of markers “zhī … zhě” for the relative clause, as illustrated in (34) and (35), respectively. This is further evidence of the cross-linguistically intrinsic relationship between genitives and relative clauses (Reference AristarAristar 1991).

Historically, a noun that is the complement of a copular structure can also be modified by a zhě relative clause, as illustrated in (36). Note that there was no copular verb in Old Chinese; hence the sentence-final particle yě was obligatory to mark the copular structure. Between the subject and the complement was a pause, and the subject was often coindexed by the anaphoric pronoun zhě (i.e. one of the multiple functions of the pronominal element zhě at that time). The copular verb shì, which is between the subject and the complement in Contemporary Chinese, did not come into being until the first century BC (Reference WangWang 1989: 183).

(c) Oblique object. It seems that the zhě relative clause can be applied to any position in terms of the accessibility hierarchy identified by Reference Keenan and ComrieKeenan and Comrie (1979). The noun in PP, namely the object of a preposition, can also be modified by a zhě relative clause. In (37), for instance, the head noun bīnkè “guest” is the object of the preposition jí “with,” which is modified by the zhě relative clause “who knew the mission”:

In (38), the zhě relative clause modifies a genitive phrase that is the object of the preposition zhū “among.” It reveals another important rule governing the application of the zhě relative clause: if the head noun was preceded by another modifier, such as an adjective, a possessor, or the zhī relative clause, only the postnominal relative clause marked by zhě was allowed. The head in (39) is a genitive phrase and hence cannot be modified by a preverbal relative clause; more examples of this type are analyzed as follows.

21.3.3 Usage of the Zhě Relative Clause

In Contemporary Chinese, the relative clause marked by de is restricted to occur in prenominal position, regardless of whether the head noun has other attributives, as illustrated in (39). In fact, all attributives must precede the head noun, unlike the grammar of Old Chinese, which allowed various types of modifier to follow the head noun, as discussed in Section 21.2.

In (39), all three modifiers – an adjectival phrase, a numeral phrase, and a relative clause – precede the head noun shū “book,” but this structure was not allowed in Chinese before the fifth century AD. Once there was another modifier in prenominal position, regardless of the grammatical category to which it belonged, the relative clause marked by zhě was the only remaining choice; in other words, the prenominal relative clause marked by zhī was inapplicable in this context. This rule can be illustrated by the following examples. First, in (40) the head noun dú “calf” was already modified by an adjective, so a prenominal relative clause marked by zhī could no longer be used, and the zhě relative clause in postnominal position was the only option:

Second, in (41), the head noun fù “woman” was preceded by a verbal modifier (equivalent to the past participle in English), so the relative clause could be introduced only by the relativizer zhě in postnominal position:

Third, if the head noun was preceded by a possessor modifier, once again, the zhě relative clause in the postnominal position was the only option for relativizing the head noun, as illustrated in (42):

Fourth, if there was already a relative clause in prenominal position, another relative clause could be used only in postnominal position, which was marked by the relativizer zhě. In (43), a relative clause (i.e. Shí Huáng suǒ fēng “that the emperor wrote”) precedes the head noun (i.e. shū “the imperial edict”); hence another clause (i.e. cì gōngzǐ Fúsū zhě “to bestow upon Prince Fusu”) can only follow the head noun:

In fact, neither the zhě relative clause nor the zhī relative clause could be recursive within the NP; that is, neither of them could be used twice there. In Contemporary Chinese, in contrast, the de relative clause can be recursive in prenominal position, indicating that it can be used more than once within an NP, as illustrated in (44):

From the point of view of language universals, we can see that Old Chinese actually obeyed the heavy-constituent principle, by which longer, more complex modifiers often follow the head, whereas other lighter modifiers precede the head. Reference HawkinsHawkins (1983: 98‒106) offered a similar pragmatic explanation for the heaviness serialization principle, which defines a preference for “lighter” modifiers to occur leftward in their respective phrases, while “heavier” modifiers tend to be placed rightward. Preposed long modifiers make it more difficult to determine the head; once the head has been identified, a subsequent long modifier (e.g. relative clause or adjectival phrase) is more easily processed. Thus category harmony and heaviness shift aid language users in unpacking the constituent structure. This concept resembles Greenberg’s concept of dominance in its effect of complementing harmony: heavier modifiers follow the noun even if the harmonic order is modifier–noun. The rank of heaviness is as follows: relative clause > genitive noun > adjective > demonstrative/pronoun. That is, the grammar of Chinese historically exhibited more consistency with certain language universals.

The robust empirical evidence provided above is sufficient to demonstrate that the postnominal relative clause not only was well attested but also played a key role in the grammar. In the next section, I attempt to find the answer to why this type of relative clause disappeared.

21.4.1 The New Information-Organizing Principle

As we saw in Chapter 7, the structures of declarative sentences underwent a fundamental change in Medieval Chinese, which caused the redistribution of PPs (i.e. adjuncts). It is observed in the following two syntactic configurations:

1. (46) Subj + PP_{non-resultative} + VP > Subj + PP_{non-resultative} + VP.

The motivation for this schematic change was the establishment of the resultative construction, which in turn triggered the occurrence of the new principle of information structure (for a fuller discussion, see Section 7.9). Under this principle, only PPs denoting resultatives, such as the end point or recipient, remained in postverbal position, whereas PP adjuncts expressing non-resultative features could no longer stay in the original postverbal position and thus became extinct. Note that a group of verbs grammaticalized in the first verbal position of a serial verb structure and assumed the function of the old grammatical markers. That is, the schematic change did not involve any movement at all. This process started as early as 100 BC and was not completed until AD 1800, lasting nearly 2,000 years.

The position changes of these PPs in sentence-final position to delineate the preceding VPs is critical for the present analysis. In Old Chinese, as in English and other VO languages, PPs usually appeared in sentence-final position to modify the preceding VPs in various ways. As typological studies (Reference Greenberg and GreenbergGreenberg 1966a, Reference ComrieComrie 1981: 86‒103, Reference CroftCroft 1990: 208‒213) have shown, it is universal that SVO languages allow PPs to occur freely in sentence-final position, modifying the predicate. That is, the grammar of Old Chinese was perfectively consistent with this language universal. The PPs could express the agent for passives, the standard for comparison, the instrument for action, the location for events, and so on. However, all of these PPs could appear only in preverbal position except for those expressing some types of resultatives of verbs (for a fuller discussion, see Section 7.9).

Obviously, PPs in passives, comparatives, instrumentals, and locatives do not express any resultative senses for the verb. Under the impact of the new information-organizing principle, all of these functional constructions were grouped together to form a construction schema and underwent the same path of change (for details, see Chapter 7). However, only a few PPs that express the resultative state of the action remain in postverbal position (Reference Chao and ShuxiangChao 1979: 176‒180), as illustrated in (50):

In example (50), “Mary” is the destination (end point) of the “mail” action, a type of resultative state. Thus the structure in (50) remains grammatical in Contemporary Chinese.

Although some locative PPs can occur in either the preverbal or postverbal position, they produce two very different interpretations as a result of the analogizing effects of the resultative construction. The distribution of locative PPs is an ideal window through which to see how the information-organizing principle works in Modern Chinese grammar. Some locative PPs can occur in either preverbal or postverbal position, but they produce completely different interpretations.

21.4.2 The Grammaticalization of the Marker De

Motivated by the emergence of the classifier system, the demonstrative dǐ grammaticalized as an associative particle to connect modifiers to its head noun (for details, see Chapter 20).Footnote ¹¹ In addition to “Num + CL” phrases, in Chinese, the modifiers of nouns could be adjectives, genitive nouns, and relative clauses. As mentioned previously, in Old Chinese, numbers directly preceded their head nouns without any intervening classifiers. When, in Medieval Chinese, the classifier system emerged, some of which came into being as early as Late Old Chinese, a proper classifier had to be used to associate a number with its head noun. Consequently, the structure of numeral NPs underwent the following change:

1. (51) Num + N > Num + CL + N.

In Late Medieval Chinese, by analogy, the demonstrative dǐ was triggered to develop into a morpheme to connect all of the other types of attributive, except for numbers, to the head noun. This new morpheme, de (a phonologically reduced form of the demonstrative dǐ), took over the functions of both zhī and zhě, but it was restricted to occurring in prenominal position. The earliest examples show that the new marker de first connected a relative clause with its head noun, as illustrated in (52) and (53), and then was extended to introduce other types of attributive:

Beyond the usages in nominal phrases, the marker de could also connect a complex adverbial phrase, such as a reduplicated form, to its verb head (i.e. predicate), as illustrated in (54) and (55).Footnote ¹² This indicates that in Chinese the structures “Modifier + Head,” regardless of whether the head is nominal or verbal, are treated the same, at least morphologically.

According to Reference HeHe (2007), in the texts of Early Modern Chinese (roughly from AD 1200 to 1400), the morpheme de had to be used to associate the adverbial phrase with the predicate verb when the adverbials were a reduplication of adjectives or a complex VP, as illustrated in (56) and (57), respectively. In the following examples, the underlined parts are the adverbial phrases and the parts in bold font are the predicates, with the former parts functioning to modify the latter:

More importantly, in Contemporary Chinese the rule regarding the usage of the marker de is equally applied to both nominal and verbal modifiers (Reference LüLü 1999). For example, the morpheme de is optional when the nominal and verbal modifiers are simple adjectives, but it is obligatory when the modifiers are “degree word + adjective” phrases, regardless of whether the head is a noun or a verb (cf. Chapter 20), as illustrated in (58) and (59), respectively:Footnote ¹³

In traditional Chinese linguistics, PPs are regarded as adverbials because they share the functions and distributions of ordinary adverbs (e.g. Reference Chao and ShuxiangChao 1979: 330, Reference ZhuZhu 1982: 151, Reference LüLü 1979: 64). Indeed, PPs used as adverbials are not marked by any morphological forms. Undoubtedly, however, the change in PP-adverbials from the postverbal (sentence-final position) to preverbal position was parallel to the change in relative clauses from postnominal to prenominal position. This diachronic evidence proves that the relationship between the noun and relative clause is the same as that between the verb and the PP-adverbial; in other words, the relative order of O and V bears no direct relationship to the order of the noun and relative clause, as many typological linguists have assumed (Reference Greenberg and GreenbergGreenberg 1966a, Reference DryerDryer 1992, Reference CroftCroft 1990: 47‒49). That is, my analysis reveals that there is no correlation between the two constituent orders: VO and N + REL, and, at least in Chinese, the typological change from “V + PP” to “PP + V” motivated the shift from “N + REL” to “REL + V,” an issue to which I return in subsequent sections.

21.4.3 Consistency of Old Chinese with VO Languages

According to Reference Dryer, Thurgood and LaPollaDryer (2003), the “REL N” order in VO languages is exceedingly rare cross-linguistically; the only attested instances are Bai and the Chinese language, both of which are Sino-Tibetan (p. 26).Footnote ¹⁴ Bai is a minority language spoken in the province of Yunnan in China. It is reasonable to assume that the “REL N” order in Bai arose out of intensive language contact with Modern Chinese. If the Chinese and Bai cases are not considered, therefore, we have identified the only exception-free universals among all forty-four of the Greenbergian universals:

1. (60) If VO, then N REL.

Due to a series of internal changes over history, Chinese gradually deviated from the most typical features of VO languages. This fact calls into question the assumption that language contact is the only possible factor causing deviation from the so-called universals (Reference LehmanLehman 1973, Reference Lehman and Lehman1978).

In the above analysis, we also briefly discussed the comparative structure in Old Chinese, which was also consistent with other VO languages, as predicted by Reference Greenberg and GreenbergGreenberg (1966a). In the literature, once again, the comparative structure of Modern Chinese creates the only counterexample for the related Greenbergian Universal 23. This deviation is in fact also due to the diachronic change that was responsible for the order change of noun and relative.

Reference Li, Thompson, Anderson and JonesLi and Thompson (1974) listed seven major typological features of Contemporary Chinese that are supposed to belong to OV languages. Thus far, the robust empirical evidence in the previous sections has demonstrated that Chinese historically exhibited the following three features which are typically consistent with VO languages:

To further support my finding that the Chinese language was historically much more consistent with SVO languages, we will briefly discuss the differences between Old and Modern Chinese with respect to the remaining four features of Reference Li and ThompsonLi and Thompson (1981: 18) that are not listed above, as follows.

Feature (d) listed in (8): “Postpositions exist.” The so-called postpositions refer to shàng in zài zhuōzi shàng “on the table above.” In our view, this shàng is not truly an adposition because it can always be exchangeable with the full compound noun shàngmiàn “upper surface.” Additionally, unlike other prepositions that can be used independently, the so-called postpositions are always paired with another preposition. Even a word such as shàng is viewed as a type of adposition, which did not exist until Modern Chinese.

Feature (e) listed in (8): “OV sentences occur.” The so-called OV sentences in Contemporary Chinese refer mainly to the two following patterns: the bǎ construction and the construction with a patient noun between the subject and the verb. However, these syntactic structures did exist in Old Chinese: the former emerged around AD 700, and the latter occurred around AD 1300. The earliest examples of the bǎ construction did not appear before the eighth century AD (for details, see Reference WangWang 1989: 266‒271), and one of the earliest examples is illustrated in (61). The construction with an unmarked patient noun between the subject and the verb came into being as late as the thirteenth century AD, as illustrated in (62):

Feature (f) listed in (8): “Aspect markers follow the verb.” In Contemporary Chinese, there are three aspect suffixes – -le (perfect), -zhe (progressive), and -guo (experiential) – which follow the verb to indicate the state of progression of the action of the matrix verb (cf. Chapter 13). However, all three of these aspect markers were actually introduced into the language after the tenth century AD (Reference WangWang 1989: 102). In Old Chinese, the aspect markers behaved like auxiliary verbs that preceded the verb, as illustrated in (63); the word yǒu “have” is a perfect marker, and in (64) the word cháng, which grammaticalized from the verb “taste,” is an experiential marker:

Feature (e) listed in (8): “Genitive nouns precede the head.” Throughout history, genitive nouns always preceded the head: the genitive marker in Old Chinese was zhī, and in Modern Chinese it is de, as illustrated in (65) and (66), respectively.

Thus we can say that the genitive structure is the only exception, diachronically and synchronically, if this feature is truly correlated with VO order. In fact, however, the correlation between genitive phrases and VO languages is quite weak; for example, English as an SVO language is another exception, e.g. John’s car,Footnote ¹⁵ in which the genitive noun precedes its head. Perhaps there is actually no correlation between genitive phrases and VO order.

Based on the above analyses, it is safe to say that the Chinese language was historically much more consistent with VO languages than Contemporary Chinese is. Over the past approximately 3,000 years, Chinese has undergone a global change from a “typologically consistent” language to a “typologically inconsistent” language in terms of the Greenbergian universals, and this change has many important theoretical implications (the next section is a more complete discussion of this issue).

21.5.1 Cross-category Harmony Principle

The diachronic developments in the history of Chinese discussed above might reflect the cross-category harmony principle proposed by Reference HawkinsHawkins (1983: 97), which states the following: “modifiers either all tend to precede their heads or all tend to follow, which has been related to processing, to the need to distinguish heads of phrases from non-head modifiers.” Although the changes in Chinese caused the language to systematically deviate from the features of VO languages, the effect left Chinese extremely harmonic in conformity with the order of modifier/head. Except for the sentential level VO remaining unchanged, all of the other modifiers changed from the position after the head to the position before the head, involving both nominal and verbal phrases. As pointed out previously, the modifiers included relative clauses, adjectives, “number-classifier” phrases, PPs, and adverbial phrases. The functions of the two morphemes zhī (introducing possessives, relative clauses, and adjectives in prenominal position) and zhě (introducing relative clauses and adjectives in postnominal position) were eventually taken over by the single morpheme de. Unlike zhī and zhě, the morpheme de could be recursive; it could be used more than once within a single phrase. However, respect for cross-category harmony was achieved at the expense of violating the heavy-constituent principle, by which longer, more cumbersome dependent elements are often postposed even if head-last order is expected (Reference Rankin, Joseph and RichardRankin 2003). As a result, some awkward constructions in Modern Chinese are grammatical but are not favored typologically or from the perspective of language processing.

In (67), the head noun chē “car” is modified by a genitive phrase, an adjectival phrase, and a numeral phrase, and the morpheme de is used twice. In the attributive position of (68), the object chē “car” of the preposition zài “in” is preceded by two relative clauses that are marked by the morpheme de, but this structure was not allowed in Old Chinese. Reference DryerDryer (2005) stated that relative clauses tend not to occur between a noun and an adposition, so prepositional languages tend to have structures of the form “P + N + REL” rather than “P + REL + N.” Furthermore, Reference FrazierFrazier (1979) argued that structures of the form “P + REL + N” present a problem for the parser since “P + REL + N” might be too long a unit, and “P + REL” does not form a semantic unit. Similarly, Reference HawkinsHawkins (1983) offered a pragmatic explanation for the heaviness serialization principle that defines a preference for “lighter” modifiers to occur earlier in their respective phrases, while “heavier” modifiers tend to be placed later. Preposed long modifiers make it more difficult to determine the head; once the head has been identified, a subsequent long modifier (e.g. relative clause or adjectival phrase) is more easily processed. However, the diachronic evidence in Chinese shows that the cross-category harmony principle (reflected in Contemporary Chinese) is somehow in conflict with the heaviness serialization principle (obeyed by Old and Medieval Chinese), although they are supposed to aid people in unpacking the constituent structure.

21.5.2 Greenbergian Parameters

There are always certain exceptions, to a greater or lesser degree, to all of the so-called language universals thus far stated (Reference VennemannVennemann 1974, Reference 572Vennemann and Juilland (ed.)1976, Reference LehmanLehman 1973, Reference HawkinsHawkins 1983: 59‒60, Reference DryerDryer 1992). As we saw above, if Old Chinese is considered, it seems that the correlation “VO ⸧ N REL” is free of exceptions. However, a critical question remains unanswered: what is the intrinsic relationship between these two types of constituent order? From Old to Modern Chinese, the word order changes from harmony to disharmony with the Greenbergian universals are theoretically significant in at least the following two aspects.

Language changes are admittedly accidental and unpredictable and are thus responsible for violations of the so-called synchronic universals (Reference DryerDryer 1992). However, our research experiences show that languages develop in a highly regular and systematic fashion and serve as an ideal window for discovering and evaluating universal correlations. Reference DryerDryer (1992) correctly criticized the proposals to explain Greenbergian universals for facing a major theoretical problem, which is that there is no external motivation for asserting that languages are “consistent,” or serialize “naturally,” or are “harmonic.” There is no pretheoretical hierarchy that these descriptions reflect, and no independent principle from which they derive their putative explanatory power. All we can know about such assertions is that they are largely consistent with the data on which they are based. Our analysis shows that universal correlations can be observed to operate only in the realm of diachrony. Under these circumstances, diachronic investigation is necessary in searching for an explanation for Greenbergian universals.

Since Reference Greenberg and GreenbergGreenberg (1966a), the relative order of V and O has been treated as the most important parameter in seeking cross-linguistic correlations (e.g. Reference HawkinsHawkins 1983, Reference DryerDryer 1991, Reference Dryer and Shopen2007). The relation between the verb and the object has been generalized as “head and adjunct” (Reference ComrieComrie 1981: 96), “head and dependent” (Reference Bartsch and TheoBartsch and Vennemann 1972: 131‒9), “mother node and non-mother node” (Reference HawkinsHawkins 1984), and “operand and operator” (Reference HawkinsHawkins 1983: 136). Superficially, the verb–object order is crucial at some level, but no proposals can explain why (Reference DryerDryer 1992). Thus Reference HawkinsHawkins (1983: 70) decided to abandon verb–object orders as a prime typological indicator.

Over the past half century or so, a number of cross-linguistically recurrent correlations have been found among the orderings of different syntactic constituents. Having examined a sample of 625 languages, Reference DryerDryer (1992) found the following property clusters – languages tend to have either A or B as follows.

To my knowledge, no one has clarified the relations among the various features in each cluster. Assuming that the order of verb and object is the parameter for determining the order of other features, there are at least two possibilities for each cluster (actually many more); as for Cluster A, the possible relation might be as follows:

1. (69)

   (a)

   Verb/Object ⸧ Verb/Manner Adverb ⸧ Verb/PP ⸧ Noun/Relative Clause;

   (b)

   Verb/Object ⸧ Verb/Manner Adverb; Verb/Object ⸧ Verb/PP;

   Verb/Object ⸧ Noun/Relative Clause; Verb/Object ⸧ Verb/PP.

In (69a), there is a sequence: Verb/Object directly determines Verb/Manner Adverb and so forth; Noun/Relative Clause is determined directly by Verb/PP but indirectly by Verb/Object. In (69b), every feature is determined by Verb/Object, but they bear no relation to each other. However, the diachronic evidence in Chinese strongly suggests such a feature hierarchy:

1. (70)

   Old Chinese: Verb/PP ∩ Verb/Adverbs ⸧ Verb/Relative Clause,Footnote 16

   Modern Chinese: PP/Verb ∩ Adverb/Verb ⸧ Relative Clause/Verb.

Semantically, the relationship between verbs and adverbials is similar to that between nouns and relatives. Reference HawkinsHawkins (1983: 132) observed that diachronic links are plausible only for word order pairs that are sufficiently closely related semantically. In addition to the semantic relationship, the same morphological morpheme de has been used to associate adverbial phrases with the VP and to connect the relative clause to the head noun. According to our observations, the relative order of the verb and object has no apparent influence on the order of the other features. At least in Chinese, VO order directly determines the order of adposition and noun because all prepositions in Chinese were historically grammaticalized from ordinary verbs; even in Modern Chinese, many prepositions still can function like verbs.

Reference Keenan and ShopenKeenan (1985b) claimed that verb-final languages heavily favor preposed relative clauses. According to the grammaticalization theory of Hopper and Traugott (2003: 60), it is assumed that word order changes can have a profound effect on grammar because different phrasal orders are typically associated with VO and OV languages. VO languages tend to be prepositional; adjectives, relative clauses, and possessives follow the noun; the auxiliary precedes the main verb; and the question particle marking yes/no questions occurs in the initial position in the clause. VO languages tend to show the order in reverse. Hopper and Traugott admitted that there are no ideal VO and OV languages. Although OV instances arose in Medieval Chinese, Chinese did not undergo a word order change from VO to OV (Reference Sun and TalmySun and Givón 1985). Even if we hypothesize that Chinese truly underwent such a change from VO to OV, it fails to explain why Modern Chinese is still prepositional and the auxiliary always precedes the main verb.

21.5.3 Greenbergian Universals

In my view, the Greenbergian universals, or any other correlations in linguistic typology, can be best viewed as representing the default setting of any software (e.g. Microsoft Word), and every language can customize them in response to internal or external motivations. Once the settings have been changed, they cannot automatically return to the original state. In other words, this default setting represents the most natural state of a grammar. However, it is not the most optimal setting and does not serve as a synchronic principle to drive an “inconsistent” language to develop toward a “consistent” one because there is no such distinction between “consistent” and “inconsistent” languages. In any particular period, within a language, grammatical features are harmonic with each other in one way or another, forming an integrated system.

If a language did not change over time, its grammar would likely be closest to the natural state predicted by the universals (if the universals are correctly generalized). In reality, however, every language must change over time; hence no language is truly consistent in terms of the Greenbergian universal correlations. In other words, these so-called universals represent the starting system of a language evolution rather than the destinations toward which a language strives to develop. Some languages reserve more features of the default system, but others might dramatically deviate from them because of the particular paths of development that they have undergone. As we have seen, Old Chinese was much more harmonic with SVO languages, but it underwent a series of dramatic changes, which caused Contemporary Chinese to manifest exceptions in many aspects. The grammar of Contemporary Chinese represents another type of harmonic system, but there is no sign in Contemporary Chinese that it is returning to the state of historical Chinese. This fact challenges the hypothesis of Reference 572Vennemann and Juilland (ed.)Vennemann (1976), which defined the principle of natural serialization, stating that time provides an impetus for languages to conform to a consistent ordering of operators with their operands, regardless of the constructions that appear; hence the Greenbergian universals are viewed as the end point of language development. Exactly the opposite, however, is found in the evolution of Chinese grammar.

Compared with Contemporary Chinese, the historical Chinese language was much more consistent with typological universals. The fewer counterexamples that a universal correlation has, the more likely it is that it represents one of the default settings of a language, which defines a starting point from which a language develops. The Greenbergian universals are best viewed as the most natural states of a language, but any features can be altered due to unavoidable language development. All languages are always in the course of evolution, so no languages are completely consistent with any typological universal.

A new type of structural hierarchy has been identified in the history of the Chinese language. Although it has been assumed that VO languages strongly imply the N REL order, in fact the order of noun and relative is directly influenced by the relative constituent order of N and PP. Beyond their semantic similarity, nominal and verbal modifiers are marked by the same morphological morpheme de, providing direct morphological evidence for the present analysis. Although it has long been assumed that the relationship between the verb and the object is the same as that between the adjunct and the head, this assumption has never been seriously considered, and no scientific explorations have yet been conducted. The present analysis sheds new light on this issue. Specifically, my analysis invalidates the assumption that the two consistent word order types, adjunct–head (operator–operand, OV) and head–adjunct (operand–operator, VO), represent the most natural states for a language, and pressure to conform to these two ideal types is sufficient to initiate language change. On the contrary, in my view, the universal correlations represent only the natural states from which a language evolves, and since all languages are always in the course of change, no languages can entirely preserve the natural state. According to Reference ComrieComrie (1981: 92‒101), more than half of the world’s languages violate rules of consistency. More specifically, some languages appear to be quite stable in combinations that violate the canonical types, such as Persian, which is a head–adjunct language that nonetheless has stable final order and has been in this position for several centuries. As Reference ComrieComrie (1981: 86‒103) stated, typological theory does provide an explanation for why, given the appearance of a typological inconsistency, the language should strive to drift back into consistency, but it provides no explanation for why the inconsistency should have arisen in the first place – this fact is especially surprising given the alleged degree of pressure toward typological conformity. Thus “inconsistent” or “nonharmonic” languages would be an inevitable result of the inherent nature of change.