Keywords

1 Introduction

The identification of the syntactic status of hé in the subject position in Mandarin is a controversial issue. It falls into three basic types, which are referred to as the context-deterministic account, the multi-categorizer account, and the preposition-taking-all account. Firstly, it could be a preposition or a conjunctor and the specific identification of it is context-dependent [1,2,3]. The second typical view takes this syntactic object as a special hybrid category. To be specific, hé is taken to be a prepositional conjunction [1] or a conjunctive preposition [4, 5].Footnote 1 Recently, there is a third view which takes all occurrences of hé in the subject position to be prepositional [6, 7]. Specifically, there is no conjunctive hé in the subject position in Chinese, and there is a subject-object asymmetry given that the specific instantiation of hé in the object position is exclusively conjunctive. In other words, the categorial identification of hé is a function of the syntactic position it occurs, namely when it occurs in the subject position it is a preposition, and it is a conjunctor when occurring in the object position. Therefore, there is not any corresponding counterpart of the English-type conjunctor and in Mandarin. Given this, [7] proposes that hé-DP2 is a secondary predicate (SP) with hé as the relator, and that hé projects a maximal projection SP with PRO as its specifier according to [8]. Furthermore, [7] argues that SP is a zhuijia chenshuFootnote 2 to the first conjunct DP1. Syntactically, it is a secondary predicate to DP1; semantically, SP yields a conjunctive reading between PRO and DP2 which is originally thought to be the result of DP1-hé-DP2, and this is done by co-indexing PRO and DP1.

This article revisits the syntactic status of hé from both the theoretical and the empirical perspectives, and argues that the conjunctive hé is indeed present in Mandarin on the basis of a bunch of syntactic and semantic evidence. Furthermore, the evidence provided in [7] in arguing against the conjunctive status of hé is not as valid as they appear to be. Thus, the prepositional re-identification of the conjunctive hé in the subject position needs to be reexamined. Then we point out some of the advantages of assuming a preposition-conjunctor dichotomy.

2 A Preposition or a Conjunctor: Is This a Question?

The nominal hé-structure can appear in argument positions. When it occurs in the object position, the syntactic status of hé is a conjunctor. However, when it occurs in the subject position, its syntactic status is controversial. As introduced above, the identification of it is either context-dependent or multi-categorial. Take (1a) as an example. As indicated in (1b-c), it is ambiguous between two different readings: it refers to either the case where I watched a movie with her, or the case where both she and I each watched a movie. The two readings of (1a) can be disambiguated from each other by some syntactic or semantic diagnostics. For instance, the item hé in (1b) is a conjunctor which indicates the case where there are two different movies watched. On the contrary, the insertion of “zuowan (yesterday)” before “hé ta (with her)” in (1c) suggests that the syntactic status of hé is prepositional, as indicated by the English translation. However, when the DP1-hé-DP2 structure occurs in the object position, hé is unanimously identified to be a conjunctor [1, 4].

figure a

Instead of keeping a conjunctor-preposition dichotomy, [7] argues that hé is not grammaticalized into a conjunctor yet, but at the same time departs from its verbal origin. Thus, they argue the only possible reading for hé is a preposition. Given the conjunctive reading of hé in the object position maintains, [7] claims there is a subject-object asymmetry. One of the benefits of this idea lies in its ability to formulate a specific condition to distinguish the conjunctive hé from the prepositional one; to be specific, it takes the category identification of hé as a function of its syntactic distribution: when it appears in the object position, it is a conjunctor, whereas the occurrence of hé in the subject position yields a prepositional reading only. This is called the preposition-taking-all account. In what follows, we will review the main claims and argumentations in [7].

[7] compares the differences between the subject hé construction with the object hé construction from the following four perspectives, namely, the tolerance of CSC (Coordinate Structure Constraint, CSC, [9]) violation in topicalization, focalization only of the first conjunct, the negation and A-not-A reduplication of the conjunctor. We have demonstrated that all these exceptions can be accounted for by mechanisms that are motivated independently, so that there is no need to assume the prepositional status of hé. They observe that the subject hé construction tolerates all these four violations. For the sake of space, this paper only focuses on the topicalization of the first conjunct, as indicated in (3), with (2) as the baseline example, which are all cited from [7].

figure b

(3a) shows that the topicalization of the first conjunct is immune from the CSC, whereas (3b) demonstrates that the same cannot be applied to the second conjunct and the object hé-constructions. [7] argues that given CSC is a universal constraint that could not be violated; therefore, the grammaticality of (3a) suggests hé in the subject position is not a conjunctor, but a preposition.

In this paper, we will show that the CSC violation exceptions can be accounted for by independently motivated mechanisms without assuming the prepositional status of hé and the preposition-conjunctor dichotomy should be maintained.

3 The Potential Problems of the Prepositional Account of Hé

This section examines the potential problems of the prepositional account from both the theoretical and empirical perspectives, even though the preposition-taking-all account as proposed in [7] could capture the asymmetries in (3).

3.1 Theoretical Problems

Firstly, syntactic operations like passivization have changed the category of the subject and the object due to the different categorial identifications of hé. If hé in the subject position is prepositional, but a coordinator in the object position, then we predict that the demotion of the subject and the promotion of the object during the process of passivization as shown in (4a) to (4b) should not alter the syntactic category.Footnote 3 However, this is not as predicted as the account claimed in [7] given that the prepositional hé originally contained in the subject position is altered to a conjunctive instance; similarly, the conjunctive hé generated originally in the object position is changed to a prepositional one. Evidently, this goes against the structure-preserving principle.

figure c

Secondly, hé-DP2 as a SP is a zhuijia chenshu to the first conjunct DP1, thus it is non-at-issue [10], which resembles the comitative structure, not coordination. Put differently, the prepositional account seems to target a comitative construction, not a coordinative structure.

figure d

In (5a), the agent of eating is Zhangsan and Lisi; in (5b), it could only be Zhangsan, and Lisi is just a participant accompanying the eater. We may wonder whether the prepositional account could capture this difference given that they are both comitative in nature according to [7]. The entity denoted by DP2 does not appear to be the agent of the eventive predicate since SP and the predicate are not in a syntactic predication relation. The zhuijia chenshu function of SP constitutes another piece of collaborating evidence. The conjunctive reading forces SP to be an argument of the eventive predicate via predication; if so, the assumed null element PRO is also taken to be one of the participants of the eating event. However, this line of reasoning has three problems: Firstly, SP can only be an argument when it is identified to be nominal, but in fact it is a maximal projection headed by a prepositional phrase in the analysis of [7]. That is to say, it is difficult for the prepositional SP to form a predication relation in terms of syntax with the eventive predicate. Secondly, the eventive predicate is predicated of the same entity as its argument in figuring out the right semantics of (5b): one is PRO and the other is its controller—the first conjunct DP1. If SP is the source of plurality, then it seems possible to equally take all these three elements, namely the first conjunct DP1, PRO and the second conjunct DP2, to be the eventive agents. Clearly, it is not so. Thirdly, Zhangsan and Lisi in the conjunctive construction share the same theta role. According to Uniformity of Theta Assignment Hypothesis (UTAH) as proposed in [11], Zhangsan and Lisi should occupy the same syntactic position, namely the subject position. However, according to the SP analysis, Zhangsan and Lisi are assigned divergent theta roles while both holding the subject position. If UTAH is equally applied in Chinese, then the phenomenon that two DPs with different theta roles are assigned to the same syntactic position turns out to be mysterious.

Thirdly, the distribution of PRO awaits further explanation. Following [8], [7] assumes the subject of the secondary predicate turns out to be PRO. That is to say, they end up claiming that PRO could be the null subject of a prepositional phrase. The cross-linguistic research on the distribution of PRO demonstrates that it distributes almost exclusively in non-finite clauses. Along with [7], we are led to take the prepositional phrase to be non-finite. However, this inference is unreasonable. Moreover, the topicalization of the first conjunct indicates that the DP1-hé-DP2 structure is contained within a finite clause since the domain that A’-movement crosses over can only be finite in nature [12]. Therefore, the observation of topicalization of the first conjunct in the first place invalidates the existence of PRO.

Fourthly, the interpretation of PRO is mysterious. Even if the SP analysis is reasonable, whether PRO is obligatorily controlled or not is still an issue. For this, we need to evaluate the satisfaction of the OC Signature [13], namely locality and PRO as a variable. As indicated in (6a), PRO as a variable is bound by the closest antecedent Zhangsan, whereas (6b) shows that PRO as a variable could also be controlled by Zhangsan, a further DP crossing over Lisi. Thus, we are now in a dilemma. On the one hand, the OC Signature is met, on the other, it is not. Furthermore, [6] takes “Zhangsan, Lisi” in (6b) as a real conjunction in Chinese, which is expected to control PRO as a single unit, yet this is not possible since Zhangsan could control PRO independently. What’s worse, Chinese is a pro-drop language, thus the subject can be silent. In this case, the real subject according to the SP analysis is the element before SP, namely Zhangsan in (6a) or “Zhangsan, Lisi” in (6b). When the subject is dropped, PRO ends up being controlled by a silent element per se, which itself needs to be identified by an entity in the discourse to begin with. This further confirms the non-obligatoriness of PRO construal in (6b) which reaches a conflicting conclusion as compared with (6a). Besides, the topicalization of the first conjunct leaves PRO to be bound by a lambda operator, which is rarely seen in the study of control.

figure e

Fifthly, the semantic type of hé is dubious. Along with the prepositional account of hé and the Likeness Condition, when the first conjunct is of type e, the second conjunct should automatically be of the same type as well. As shown in (7), e-typed DP1 requires the semantic type of DP2 to be e, and SP to be <e, e> in order to yield an e-typed DP that is subsequently plugged into VP. The type of PRO is e, thus the semantic type of hé-DP2 is calculated to be <e, e, e>. Finally, the semantic type of hé is calculated out, namely <e, e, e, e>. However, this type is not the semantic type for a typical transitive preposition, which should be <e, e>. In other words, the semantic type assigned to the prepositional hé in this structure is a new type (What’s worse, we haven’t considered the case where DP is considered to be a generalized quantifier, and it will yield another different semantic type for hé). Put differently, we end up with two types of prepositional hé at least in terms of semantic type, though the conjunctive hé in the subject position and the prepositional one is reduced to a single syntactic category. Whether the reduction of syntactic category is preferred over the complication of semantic type requires further investigation.

figure f

Sixthly, the case assignment of hé poses a problem for the Chinese language. If hé in the subject position can be reanalyzed as a preposition, then it will assign an accusative case to its internal argument, and a nominative case and (together with the internal argument) an agent theta role to its subject, namely PRO. However, PRO must be ungoverned, thus it could not be case-marked, at least not a nominative case. Recently, cross-linguistic research observes that PRO could carry a morphological case in languages like Icelandic. If the distribution and the case-marking of PRO are indeed independent from each other as argued in [13],Footnote 4 then PRO could be assigned a nominative case. If so, could we claim that the overt realization of PRO in Chinese also be assigned a case? It seems reasonable to make such a claim. However, whether the DPs in Chinese could be case-marked is a controversial issue and is contradictory to Hu himself [14]. Therefore, the assumption of PRO brings in a controversial issue in Chinese which awaits further research.

Lastly, the labeling of the maximal projection SP is problematic. To project the unit that is composed of the subject and SP is another issue for [7]. For instance, when the hé-construcion is in the object position, it is a self-contained projection;Footnote 5 however, passivization leaves the label of it unspecified. This is undesirable since a syntactic operation should not change the category of a syntactic object. The prepositional analysis in [7] clearly goes against this generalization.

To sum up, the assumption of the prepositional status of hé faces a range of theoretical problems.

3.2 The Empirical Problems and the CSC Violation Revisited

For the sake of space, we only focus on the topicalization of the first conjunct.Footnote 6 We will demonstrate that the subject-object asymmetry of the CSC violation in topicalization could be accounted for without assuming the prepositional status of hé in the subject position.

Cross-linguistically, the violation of CSC is not unique to Chinese, nor is it unique to topicalization. Evidently, the prepositional analysis of hé faces empirical problems. For instance, [15] observes that the first conjunct in Serbo-Croatian as indicated in (8) is topicalized and is a clear violation of CSC, yet the remaining sentence still remains grammatical. The reason lies in the cliticization of the conjunctor i to the second conjunct filmove, leaving the trace of the conjunctor *-marked and deleted at PF. According to [15], this phenomenon could be explained by the following observation, that is, the nullification of the head deprives the islandhood of any given island. Put differently, the violation of CSC in the topicalization of the first conjunct could be saved by the head movement of the conjunctor and the deletion of the *-marked structure at PF; thus, the fact that the remaining sentence as shown in (8) is grammatical is captured.

figure g

Another important correlation is that CSC islandhood voidance occurs only in article-less languages. This implies that the scenario observed in Chinese might be treated on a par since Chinese is a language that lacks articles.

The next step is to argue that the Chinese conjunctor hé behaves like a clitic. Note that we are not trying to argue that hé is a clitic, but simply demonstrating that hé displays the properties of a clitic.

The theoretical approach adopted here is the framework of canonical typology. The canonical clitic theory as proposed in [16] claims that clitics stands between functional words and affixes. In other words, [16] argues that clitics exhibit a dual feature of these two categories: its formal properties are those of the canonical affix, but its distributional properties are those of the canonical function word. The first criterion (9i-a) is modified by [18] to accommodate the tonal Chinese languages. To be specific, the clitic should not be the same as the full-form functional item with regard to tonal patterns. Furthermore, this clitic should not bear stress or focus. According to the first distributional criterion, the clitic should be attached to its host, which is a phrase as well as a word, and it remains in the same position as its full-formed counterpart. The second distributional criterion holds that the clitic should take wide scope over conjuncts. We will go through them one by one. Footnote 7

figure h

The linker hé is not stressed and is prosodically dependent on its host. [19] claims that functional elements are “generally stressless…even phonologically null.” As indicated in (10a), the cliticization of hé to the second conjunct is evidenced by the intolerance of the insertion of a plural suffix -lia, and the item could only be placed after the lexical string hé Lisi. Furthermore, since -lia could only be attached to a plurality element, which lends more support to the conjunctive reading of hé. Furthermore, (10a) and (10b) show that the host that hé is attached to is phrasal, not lexical. Finally, the scope of hé is over meiyige nanhai and meiyige nühai. As indicated in (10b), the meaning is “Every boy came and every girl came”, which could only be possible when hé takes the wide scope over the two conjuncts.

figure i

On the basis of the above analysis, the conjunctor hé could be considered as a clitic-like element. If so, the conjunctor-as-clitic movement account that is originally proposed for (8) could be extended to (3a-b), repeated here as (11a-b). This analysis could not only account for the grammaticality of (11a) but also the ungrammaticality of (11b). To be specific, the head movement of hé in the way of cliticization could invalidate the islandhood of Zhangsan hé Lisi in the subject position, and thus the extraction of Zhangsan is permitted and consequences of the CSC violation is repaired.

figure j

In contrast, the ungrammaticality of (11b) could be captured in the following way: the item hé behaves like a clitic, thus it could not be stranded from its host — the second conjunct. In this way, the obligatory extraction of the host of hé from the coordination island is strongly prohibited. This stranding prohibition account could be extended to (3d) as well, repeated here as (12a). That is, the topicalization of Lisi to the left periphery leaves hé being stranded.

figure k

The next question is how to explain the subject-object asymmetry regarding the tolerance of CSC violation in the subject hé-conjunction. That is, how to capture the grammaticality contrast between (11a) and (12b). We argue that the contrast could be attributed to the topic prominence of the Chinese languages. In other words, we could assume that there is an interpretable Topic feature on the Chinese subject. To put it more formally in syntactic terms, we assume that there is a [uToP] in the left periphery of the matrix clause that needs to be checked off, and the closest possible candidate for the valuation of this very feature is the subject that carries an [iTop] feature. It cannot be the object since it does not have a Topic feature; otherwise the locality condition of Agree is violated [20]. To be specific, if Zhangsan hé Lisi in the subject position as indicated in (11a) and (12b) is conjunctive in nature, then according to the feature percolation mechanism through spec-head agreement as implemented in [21], the [iTop] feature carried by the whole structure could be passed onto the first conjunct which occupies the spec position of this conjunctive structure, rather than the object occupying the complement position. Therefore, the [iTop] feature is transferred to Zhangsan only, not Lisi. Thus, the contrast between (11a) and (12b) is accounted for.

To sum up, the subject-object asymmetry with regard to the violation of CSC in topicalization of the first conjunct is captured without reanalyzing the conjunctive hé as a preposition. Furthermore, this paper has aligned Chinese with other article-less languages which at the same time permit the CSC violation. In this way, the apparent CSC-violation is proved to be not unique to Chinese.

4 The Merits of a Preposition-Conjunctor Dichotomy

This section points out two pieces of merits of holding a preposition-conjunctor dichotomy.

Firstly, the dichotomy avoids the problems that are brought in by reducing the two instances of hé in the subject position to a single prepositional one.

One reviewer asks whether and what kind of consequences we will have to face in Chinese language teaching or in the computational parsing of Chinese if we identify the syntactic status of hé in a wrong way. The answer is definitely yes. Let’s take the construction of Combinatory Categorial Grammar Bank as an example, the incorrect syntactic categorization will wrongly label the conjunctive hé as the prepositional one, and this will mislead us to choose and apply the wrong combinatory rules, consequently, a wrong syntactic parsing is yielded.

Secondly, holding a preposition-conjunctor dichotomy could account for the apparent CSC violation from a cross-linguistic perspective; meanwhile, it will also put Chinese into a cluster of languages that lack articles yet at the same time tolerate CSC violation. In this way, the CSC violation no longer stays as a rigid generalization and is violated as more empirical facts unfold.

5 Conclusion

This paper reexamines the syntactic status of hé in the subject position from both the theoretical and empirical perspectives. We argue that the reduction of the two instances of hé to a single syntactic category, namely preposition, could not cover all the linguistic facts that the conjunctive hé and prepositional hé pose, and might incur a range of potential problems as illustrated above. In this paper, we emphasize to return to the traditional preposition-conjunctor dichotomy. In this sense, the Occam’s Razor could not be applied freely and should be keyed to the empirical facts; otherwise, oversimplification arises which will invite more problems than solutions.