Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Linguistics. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 27 March 2023

Nominal Ellipsis in Chinesefree

Nominal Ellipsis in Chinesefree

  • Audrey Yen-hui LiAudrey Yen-hui LiUniversity of Southern California
  •  and Ting-chi WeiTing-chi WeiNational Kaohsiung Normal University


Chinese prominently allows its subjects and objects (arguments) or parts of arguments to be omitted (labeled as nominal ellipsis). This partially contributes to the widespread perception that Chinese is a discourse-oriented or topic-prominent language and that discourse/context or pragmatics are responsible for leaving elements unsaid and for interpreting them; what will be understood from context can be omitted. This article emphasizes the fact that, even though context can be a factor, the acceptability of nominal ellipsis and the different interpretive possibilities for null subjects and null objects must follow grammatical constraints. Such restrictions are not predicted or accommodated by approaches that analyze null subjects and null objects in the same way or by a deletion mechanism applying to the Phonological Form (PF-deletion). Available proposals that treat null subjects and null objects differently are evaluated and it will be shown that null subjects and null objects can be distinguished without stipulations. A null subject can be an empty pronoun but not a null object. A null object is a placeholder that is empty in content. Such an empty element is also responsible for the fact that noun phrases restrict their subparts that can be missing. Missing contents are filled via copying of lexical materials at the Logical Form (LF-copying).


  • Syntax

Elliptical constructions in languages offer an amazing window to understanding language users’ linguistic competence. Speakers can leave parts of their intended expressions unsaid; yet hearers have no problems interpreting what is inaudible or invisible. For instance, the parts in parentheses in the following English sentence need not be said, but the sentence must always be interpreted as if the parenthetical parts are present.


Chinese goes beyond the level of omission as illustrated by (1). It has been noted for “freely” allowing arguments (subjects and objects) or parts within arguments to be missing—nominal ellipsis.1

The pervasiveness of empty or non-overt subjects and objects in Chinese, in contrast to languages like English generally requiring their subjects and objects to be realized with overt lexical materials, partly contributed to the typological classification of Chinese being a topic-prominent language or discourse-oriented language vs. English being a subject-prominent or syntax-oriented language (Li & Thompson, 1976, 1981; Tsao, 1977, among others). Empty arguments can be of different types. In the framework of government and binding theory (Chomsky, 1981), four types of empty categories can be distinguished. Their status in Chinese has been explored in great depth, beginning with Huang (1982). The four types of empty categories are A-traces (NP-traces) left by movement to A(rgument)-position, A’-traces (variables) left by movement to A’(non-argument)-position, PRO, and pro. The last two differ in where they occur (essentially, PRO as subject of an infinitival clause and pro in other positions). Huang’s contributions include uniting PRO and pro in Chinese, a language with few morphological clues to clause types (e.g., clauses being finite/tensed or not). Both PRO and pro must be identified by the closest c-commanding antecedent—the generalized control rule (Huang, 1982, 1984, 1989). In addition, he suggests that Chinese allows null topics, which can bind variables in argument positions. The study has inspired much more extensive and detailed investigation of the issues up to the current time. Putting aside A and A’-traces, this section focuses on null arguments that have been the target of study in the literature on “(radical) pro-drop”.2

Null arguments can be illustrated by the following examples, with the missing parts in parentheses. The content of the missing argument is supplied by the corresponding subject or object in the preceding antecedent clause, which can be a noun phrase or a clause.


Examples like these raise many questions, such as how an indefinite noun phrase like ‘a boy’ in (2a) and the pronoun ta in (2b) supposedly contained within the missing object should or can be interpreted. A related question is what if an indefinite noun phrase appears in the subject position, considering the fact that Chinese largely, but not always, avoids using an indefinite noun phrase in the subject position. The relevant literature is enormous and the issues raised are multipronged; see, for instance, a recent handbook on ellipsis by van Craenenbroeck & Temmerman (2019), or the edited volume by Güneş & Lipták (2022; especially the substantial literature overview in chapter 1 by the editors), among numerous others, as well as many other handbooks or encyclopedic collections containing ellipsis chapters such as Li & Wei (2014), Takahashi (2014), Saito (2017), and so on. Special issues of journals have also been devoted to the topic, such as Lingua 151 (2014), among others. To make this article manageable and comprehensible, we will focus on the following issues (references to appear in relevant parts in the following sections):


We will present the main empirical generalizations that need to be captured by any account in section 1. Section 2 reviews the pros and cons of the main lines of analysis in the literature, followed by a summary and brief discussion of the remaining issues in section 3. Section 4 extends the empirical coverage to null elements within noun phrases. Section 5 concludes.

1. Possible and Impossible Null Arguments

1.1 Possible Null Arguments

In principle, the subject and object of a sentence can be null, as in (2a–b) above. Null objects can be direct or indirect objects, illustrated below.


In such cases, the first object c-commands the second object according to scope and binding facts (see Aoun & Li, 1993; Soh, 1998, among others).4 Moreover, an object can be followed by and c-command a duration or frequency phrase postverbally. However, only the postverbal object can be missing, not the c-commanded duration/frequency phrase.5 That is, (5) cannot be interpreted as ‘I scolded him for two hours’ or ‘I scolded him three times’, if the duration/frequency phrase is not overtly present.


The O of [V+O] idioms can be missing in some cases, as below.


Moreover, the verbs of the antecedent clause and the clause with missing objects can be different:



1.2 Impossible Null Arguments

Nonetheless, not all arguments can be missing. For instance, in contrast to (6), the following case is not acceptable with the O of the V+O idioms missing (see more on idioms in section 2 and notes 9 and 13).


Many other constructions disallow null objects. Y.-H. A. Li (2005, 2007, 2014) and Aoun & Li (2008) list the following restrictions.

1.2.1 Affected Outer Object

The construction in question contains an “outer” object that is assigned an affected theta role by the verb and its direct (inner) object (see Huang, 1982, 2007; Lu, 2002; Thompson, 1973, among many others).



In these instances, the second object is the one normally subcategorized for by the verb—the ‘inner object’. The first object, the one right after the verb, is the ‘affected’ or ‘outer object’. Huang (1982, 1997, 2007) notes that the additional outer argument is possible (receiving “affected” theta role) when the verb and the inner object can combine to denote an action that affects another entity.


Huang further notes that the inner object (the theme) cannot be missing in these cases (cf. Pylkkänen, 2008). This is true, even in parallel structures—the context that facilitates missing elements the most:


1.2.2 Affected Argument as Subject: Unaccusative verb

Chinese also allows an unaccusative verb and its internal argument to combine and take an additional argument. This additional argument can surface as the subject of the sentence, interpreted as the experiencer of the event. That is, the construction has two arguments, occupying the subject and object positions respectively [SubjectExperiencer + VUnaccusative + ObjectTheme], with [VUnaccusative + ObjectTheme] assigning a thematic role to [SubjectExperiencer], like what we saw in (12). This construction does not allow the object to be missing.


Importantly, it is not the thematic structure [SubjectExperiencer + V + ObjectTheme] that means it is unacceptable for the object to be missing. If both experiencer and theme arguments are subcategorized for by a verb, the theme object can be deleted, as illustrated in Example (15).

Experiencer Subject—Theme Object (subcategorized for by the verb)


Examples (14) and (15) have the same thematic roles for their arguments—experiencer subjects and theme objects. They are only distinguished by the generation of the arguments. In the latter set, the verbs are two-argument verbs, whereas those in the former are one-argument verbs (cf. (12)). The latter allow missing objects, not the former.6

1.2.3 Result Argument of Cause-Result Verb

The construction illustrated below contains a subject that is a causer and an object that is the argument bearing the result of the causing effect. In this construction, the causer argument is available only when the result argument is present (e.g., Xu, 2003).



1.2.4 Direct Object Followed by Secondary Predicate

An object followed by a secondary predicate cannot be missing (cf. Huang, 1987 on existential constructions). In the following examples (18a and b), the predicate following the object is predicated of the object (secondary predicate).


1.2.5 Clausal Object

Y.-H. A. Li (2005, 2007) observes the following generalization:


(19a) is illustrated by (20a-b)—verbs allowing nominal and clausal objects also accept null objects:


(19b) is illustrated by (21a-b). The adverb zhe/na-me/yang ‘so’ appears in cases of missing clausal objects:7


1.2.6 “Referential” Constraint on Null Objects

Additionally, Landau (2022) notes an interesting constraint on the type of objects that can be missing—only referential objects (those that can be replaced by pronouns) can be missing, and nonreferential ones cannot take the null form.8 The latter include measure phrases, predicate nominals, names in the cases of naming verbs, and some quantificational expressions.9 Similar constraints seem to hold in Chinese: measure phrase in (22), predicate nominal (23), and name of a naming verb (24).




In brief, what we have seen so far is that objects in Chinese cannot be omitted at will.

1.3 Differences in Interpreting Null Subjects vs. Null Objects

It is generally accepted that the interpretation of missing objects can include the strict reading and the sloppy reading (e.g., Huang, 1991).10


Importantly, null objects and null subjects have different interpretive possibilities (see Aoun & Li, 2008; Cheng, 2013; Huang, 1982; H.-J. G. Li, 2002; Y-H. A. Li, 2005, 2007, 2014; Miyagawa, 2010; Sato, 2014, 2019; Tomioka, 2014; among many others). Null subjects are interpreted as PRO/pro, identified by the closest c-commanding antecedent, or variables bound by an overt or empty topic. When variables are derived by A’-movement, locality conditions on movement are relevant. Null objects are not interpreted as PRO/pro. We briefly present below differences between interpreting null subjects and null objects.

Consider the following sentences, containing the verb ‘like’ in the relative clause, which takes human subjects and objects equally well. However, the empty object in the relative clause, not the empty subject, can be coindexed with a topic phrase across island boundaries and across the subject of the higher clause.11


Indeed, missing objects are quite liberal in their choice of antecedents. The antecedent can be an A or A'-phrase and no locality conditions apply. To illustrate, (27)–(28) below show that the missing object inside an adjunct clause and a complex NP can be coindexed with a subject across the island boundary one or two clauses/islands up:



The availability of such a wide range of interpretive options does not hold with empty subjects. The indexing in the following examples (29–30) shows that the empty subject of the adjunct clause must be interpreted as coindexed with the subject of the immediately higher clause and cannot be interpreted with a prominent entity in the discourse or a subject beyond the immediately higher clause.



The generalization is that an empty subject must be coindexed with the closest c-commanding nominal. This is the case even when the context strongly favors a different interpretation. Thus, despite the fact that the empty subject should clearly refer to Miss Li in (31b) according to the discourse context, the reading is not available when the subject is missing.


In addition, there exists a contrast between null subjects and null objects in terms of the possibility of omitting the indefinite noun phrase, illustrated by the following pair of sentences (31a–b) (see Y.-H. A. Li, 2007, 2014).


Similarly, the pair in (33a–b) below shows that the sloppy interpretation is available to a null object, not a null subject. In contrast to the availability of the sloppy reading to the null object in (32b) (the pronoun contained in the missing object coindexed with the matrix subject), the empty subject itself in (33a) is interpreted as coindexed with the matrix subject. That is, it is interpreted as ‘Zhangsan1 is happy because self's1/his1 son has taught math; Lisi2 is proud because Lisi2 has taught linguistics’, not ‘Zhangsan1 is happy because self's1/his1 son has taught math; Lisi2 is proud because self’s2/his2 son has taught linguistics’.


Tomioka (2014) uses cases with ‘every. . . but’ to demonstrate that even pragmatics cannot force the sloppy reading for null subjects in Chinese. He observes that in English, singular personal pronouns typically do not permit sloppy interpretations, but such an interpretation is rendered somewhat easier to obtain when it is preceded by an expression such as everyone but X. This point is illustrated by the following sentence (33), which allows the she in the second clause to refer to Johnny’s mother in the sloppy interpretation.


Importantly, the sloppy interpretation is not available in Chinese even in this context. The Chinese counterpart is given below, showing that the sloppy interpretation is achieved only when the relevant position is overtly filled by a lexical subject, zijide laoshi ‘self’s teacher’.12


In brief, empty subjects and empty objects are interpreted differently. The latter allow indefinite or sloppy interpretations, and their antecedents can be across island boundaries in A or A'-positions or prominent entities in the discourse. Missing subjects are more restricted in their interpretive possibilities.

2. Analyses

How have the facts briefly listed in the previous sections been accounted for? It seems that all the logical possibilities have been explored. Due to limited space, we will focus on the points raised in the more recent literature that have not been (adequately) responded to and discuss the most appropriate candidate to capture the relevant empirical generalizations. We will also put aside those analyses that treat null subjects and null objects as the same entity, due to the differences in interpreting null subjects and null objects described in section 1.3; such analyses include the proposal that all null arguments are the equivalent of empty pronouns, or empty concept nouns (e.g., Hoji, 1998; H.-J. G. Li, 1998; but see H.-J. G. Li, 2002 for arguments against this analysis), or NP-deletion stranding a null determiner (e.g., Tomioka, 2003; see Y.-H. A. Li, 2007 for arguments against such an analysis).

2.1 Subject–Object Asymmetry

Representative analyses considering differences between null subjects and null objects will be briefly listed here for reference. Readers are referred to the respective works for their mechanisms and elaboration on the relevant concepts. Cheng (2013) argues that the availability of argument ellipsis is tied to the lack of a Determiner Phrase (DP) in a given language. He proposes that deletion at the Phonological Form (PF) only allows complements of a phase head to be deleted. The lack of DP in languages such as Chinese and Japanese (NP languages) makes the VP a phase. Therefore, the direct object of a verb, being the complement of a phase head V, can be elided in the PF component – this is known as argument ellipsis resulting from PF deletion. However, argument ellipsis is not available to subjects, because a subject is not a complement of a phase head and fails to undergo PF-deletion. Null subjects in these languages are cases of deep anaphora. Nonetheless, Sato (2019) argues against Cheng’s analysis and claims that the deep anaphora analysis for null subjects cannot stand, because a linguistic antecedent is required for a null subject in Japanese. Y.-H. A. Li (2007, 2018) argues that Chinese is a DP language. Moreover, Chinese, not Japanese, has the subject/object differences described in section 1.3.

Another line of research is to relate the absence of agreement to the possibility of argument ellipsis, as found in Saito (2007), Miyagawa (2010), Şener & Takahashi (2010), Takahashi (2014), Sato (2015), Sato & Karimi (2016), and so on. These works claim that agreement is relevant to subjects in Chinese, not objects. Therefore, only the latter allow objects to undergo deletion. Such an analysis has been challenged by Simpson et al. (2013) and Y.-H. A. Li (2014), among others.

A third approach to the observed subject/object differences is built on the assumption that subjects in Chinese must be definite (Sato, 2019; also see Takahashi, 2019). Unfortunately, this assumption does not reflect the fact that subjects can be indefinite (cf. (32a–b). Also see Lee (1986), Li & Wei (2014), and Paul & Whitman (2017), among others.

Finally, an approach that has been widely considered is to subsume null objects under VP-ellipsis (stranded-V VP-ellipsis, e.g., Goldberg, 2005; Huang, 1991; H.-J. G. Li, 2002; Otani & Whitman, 1991, and more recently, Liu, 2014; Simpson, in press; Wu, 2016; Ye, 2020, among others; cf. Pan, 2019 allowing both VP-ellipsis and argument ellipsis). The limited space in this article necessitates the omission of detailed arguments against such an analysis. We will just point out these major points: (a) some impossible null objects in section 1.2 demonstrate that the properties of the constituents within VPs such as objects need to be considered, which would be unexpected according to the VP-ellipsis approach; (b) the verbs in the antecedent clause and the clause containing a null object can be different, such as (6–7) in section 1.1; and (c) additional mechanisms such as those used in Ye (2020) to solve some of the problems facing the VP-ellipsis approach are not supported empirically; for example, the proposal that the Object2/Duration/Frequency c-commanded by Object1 in the construction [V + Object1 + Object2/Duration/ Frequency] must be focused and raised, right-adjoined to a verbal projection containing [V + Object1]. However, there is no evidence for the raising of the Object2/Duration/ Frequency phrase to the right of the verbal projection or for its being a focus (for instance, focus can be on another phrase in the sentence instead).

2.2 Constraining Null Subjects and Null Objects: True Empty Category

Focusing on the facts described in sections (we will return to the constraint in section 1.2.6 shortly), Y.-H. A. Li (2005, 2007, 2014) and Aoun & Li (2008) distinguish between null subjects and null objects. They adopt Huang’s (1982) claim that empty categories recognized in the government and binding framework are also available to Chinese—NP-traces, variables, PRO, or pro. The latter two are not differentiated in their need to be identified by the closest c-commanding antecedent: This is the generalized control rule (GCR). The GCR coupled with the disjointness requirement on pronouns (Binding Principle B) rules out a PRO/pro in the object position. According to the GCR, an object PRO/pro needs to be identified by the subject of the clause, being the closest c-commanding antecedent for an object in the same clause. On the other hand, Binding Principle B requires an object PRO/pro to be disjoint from the subject of the clause. This amounts to saying that when a null object is not derived by movement (NP-trace/variable), it cannot be any one of the four recognized empty categories. It is a “true empty category” (TEC)—a placeholder in the null-argument position in the syntactic structure when PRO/pro is not available, acquiring its contents via copying at the Logical Form (LF-copying). Indeed, such a true null element is necessary for any approaches that adopt the mechanism of LF-copying to null arguments.13 Li proposes that such a TEC has nothing but categorial features (D, due to subcategorization requirement of V) and noninterpretable Case features (Y.-H. A. Li, 2008, 2013; cf. Duguine, 2013). These features are necessary to check off the selection features of the verb in order for merger to succeed and be visible for lexical materials to be copied at LF.14

Such a TEC approach captures the subject/object difference in interpretive possibilities (section 1.3). A null subject, when not derived by movement (variable/NP-trace), is a PRO/pro, subject to the GCR. However, the PRO/pro option is not available to null objects, due to the conflicting requirements between the GCR and Binding Principle B mentioned above—and is a TEC as a result. A TEC obtains content through LF-copying. It is a last-resort placeholder, when none of the recognized empty categories (NP-trace, variable, PRO, pro) are possible for the position. It is devoid of content other than the formal categorial and Case features. It contains no lexical materials; that is, it has no thematic features (cf. “empty noun” in Panagiotidis, 2003). Therefore, it cannot assign or participate in the assignment of thematic roles to an additional argument (sections and or the licensing of a secondary predicate (section 1.2.4). The Case requirement on TEC makes it impossible to have the null counterpart of an embedded clause not Case-marked by the selecting verb (section 1.2.5; see Y.-H. A. Li, 2011, 2013). The true emptiness of a TEC captures the restrictions on the distribution of null objects described in sections 1.2.1.–1.2.5. In addition, the verbs of the antecedent clause and the one containing null arguments need not be identical, as argument ellipsis is not VP-ellipsis (cf. Ye, 2020).

A TEC account captures almost all the generalizations in section 1 through contextual definition of empty categories in the government and binding framework. Crucially, the null-argument position only has a placeholder syntactically without lexical materials. LF-copying provides the contents (see Y.-H. A. Li, 2005, 2007, 2014; Oku, 1998; Saito, 2004, 2007, among others).15 There is further evidence for the emptiness of the null object—extraction from within the null object is not possible because the object does not have full structures with lexical materials. The following example shows that Chinese does not allow extraction to originate from within a null argument.16


This example does not have the interpretation of ‘apples that he bought [a box of __]’—the relativized ‘apple’ cannot be a subpart of a null object containing a missing yi-he ‘a box’.

Extraction from within a null clause is not possible either.17


To summarize briefly, the TEC account with the contents of the placeholder filled at LF captures the extraction fact above, as well as all the facts in sections What is left is the referential restriction in section 1.2.6. This peculiar restriction is an important contribution in Landau’s work, and he captures the restriction by incorporating a “referential” feature in his account. The TEC approach does not predict this. Nonetheless, it can be amended by adding a referential requirement, similar to what Landau proposes. We can require that only referential antecedents can be copied to the TEC at LF, thereby accommodating the fact in section 1.2.6. That is, by adding a referential requirement to how the TEC can obtain its contents at LF, we can accommodate all the facts presented in this article.18

3. Summary on Null Arguments

Arguments cannot be missing in certain constructions (section In addition, null subjects and null objects differ in interpretive possibilities (section 1.3). The TEC/LF-copying approach captures these empirical generalizations in sections The observation noted in section 1.2.6 can be accommodated if a referential requirement is added to the type of elements that can serve as antecedents for the TEC.

There is one more construction that is closely related to the null-object construction discussed in this article and that is harder to define—null complement anaphora (see Hankamer & Sag, 1976 on deep anaphora, Partee, 1989 on implicit variables, etc.).19 A position on null complement anaphora taken in many relevant works is not to postulate a silent object in syntax at all. For instance, English, despite being a language that does not delete its objects, allows some verbs, such as know, understand, and find out, to occur by themselves without an overt object. They are intransitive verbs grammatically. However, they are interpreted as if they have an object, whose content is determined according to the context. It is “inferable” (Prince, 1981) or “accommodated” into the domain of discourse “(Condoravdi & Gawron, 1996)” (Williams, 2012, p. 127).

Chinese has verbs that require (clausal) objects such as ta yiwei/cai *(ta bu hui lai) ‘He thought/guessed he would not come.’ Nonetheless, there are also verbs that do not require objects but interpreted as though they have objects (null complement anaphora), such as ta zhidao/liaojie ‘He knows/understands’. Verbs allowing null complement anaphora might not be the same in different languages (lexical idiosyncrasy; see Wei, 2022). Potential tests to distinguish the two sets of verbs—those with and without syntactic presence of a null object—might be those based on mixed readings for the (un)availability of true sloppy readings or locality conditions, as in H.-J. G. Li (2002). Unfortunately, judgments in such cases are generally subtle and difficult to make. Clearly distinguishing the two via large-scale experimental study on speakers’ judgments while teasing out all potential intervening factors can take the study of null objects much further (see, for instance, Hoji, 2015; Plesniak, 2021 for possible experimental designs on sorting out factors affecting speakers’ judgments).

Finally, a remaining question concerns crosslinguistic variation: When are null arguments possible and when are they impossible in specific languages? Landau (2020, pp. 360–361) listed some parametric options but commented that none of them captured the relevant facts. The parametric options he mentioned are (a) the “antiagreement parameter” (ellipsis of an argument is allowed if and only if no functional head agrees with that argument; Saito, 2007; Şener & Takahashi, 2010; Takahashi, 2008, 2014), (b) the presence/absence of scrambling (Oku, 1998), (c) the absence of a DP layer in the nominal projection (Cheng, 2013), (d) the agglutinative nature of the nominal-morphology (such as Simpson et al., 2013; Otaki, 2014), or (e) the property of radical pro-drop (Sakamoto, 2017). Importantly, Landau observed that none of these options capture the fact that languages like Hebrew, Russian, and Portuguese allow argument ellipsis. He concluded that no currently existing proposal can do justice to the typological diversity of AE languages. Unfortunately, we do not have answers to complete this complex picture either; so it just remains for us to end this section with more speculations. First of all, if we clearly identify the properties of individual languages, we might have better hopes of identifying a parameter. For instance, Chinese was categorized as an NP language like Japanese, and that created problems (section 2.1). Secondly, the types of empty element allowed might vary with the nominal structures in different languages, such as Tang’s (2011) comparison of Cantonese, Mandarin, and other languages. Other possible factors can include variation in formal features, such as Case (Duguine, 2013) or choices of lexicon/feature-based parameters (Roberts, 2019). We leave the issue to be explored in future research.

4. Null Elements Within Noun Phrases

Before concluding this article, we would like to briefly present further empirical advantages of a TEC approach to nominal ellipsis; this approach not only captures the facts described so far but also accounts for why null elements within noun phrases are possible only in certain cases.

Chinese noun phrases can have the form [Dem + Num + Cl + N] or [XP + de + N(P)], with the XP being a possessor, a modifier, or a complement to the following N(P). Classifiers and the marker de license the following N(P) to be null. Nonetheless, such “nominal ellipsis” is constrained. As observed in Aoun & Li (2003, chapter 4, pp. 95–130), two types of N(P) cannot be missing. These are: (a) a noun phrase modified by a relative clause containing a null relative operator; and (b) a head noun assigning thematic roles. The first case can be illustrated by adjunct relativization in (37), and the second in (38).20



According to Aoun & Li (2003), an adjunct relative clause contains a null relative operator which needs to be licensed and interpreted by the NP modified by the adjunct relative clause. When the modified NP is null, it has no contents to license the null relative operator. This captures the unacceptability of cases like (37a–b). For (38a–b), a null head noun has no thematic features and fails to assign a thematic role to its complement, and as a result it is unacceptable for the head noun to be null. In other words, the null noun (phrase) in these cases is a TEC, empty in content. Patterson (2020) extends the null operator analysis to all the cases of so-called gapless relative clause illustrated below (cf. Aoun & Li, 2003, chapter 4; Huang, 2016, Pan, 2022, among others).21


Patterson analyzes gapless relative clauses as adjunct relative clauses containing null relative operators. A null relative operator in this construction is ruled out for the same reason as in (37a–b): the null relative operator fails to be licensed by a contentless null noun modified by the gapless relative clause.

5. Conclusion

Chinese prominently allows ‘nominal ellipsis’, which has partially contributed to the widespread perception that Chinese is a discourse-oriented or topic-prominent language, and that discourse/context or pragmatics is responsible for leaving elements unsaid; that which can be understood from the discourse/context can generally be left out. This article emphasizes the fact that the acceptability of null arguments and their interpretive possibilities are subject to grammatical constraints (section 1), as is the possibility of leaving portions of noun phrases unsaid (section 4). We have argued that such constraints can be best understood if we recognize the existence of a true empty category (TEC), whose contents are filled via LF-copying. We hope our work has added more empirical coverage and provided further factors and options to consider in the pursuit of a greater understanding of the general properties of, and variation in, nominal ellipsis across languages.


  • Adams, M. (1987). From Old French to the theory of pro-drop. Natural Language and Linguistic Theory, 5, 1–32.
  • Aoun, J., & Li, Y-H. A. (1993). Syntax of scope. Cambridge, MA: MIT Press.
  • Aoun, J., & Li, Y.-H. A. (2003). Essays on the representational and derivational nature of grammar: The diversity of wh-constructions. Cambridge, MA: MIT Press.
  • Aoun, J., & Li, Y.-H. A. (2008). Ellipsis and missing objects. In R. Freidin, C. Ostero, & M.-L. Zubizarreta (Eds.), Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud (pp. 251–274). Cambridge, MA: MIT Press.
  • Borer, H. (2005a). In name only: Structuring sense (Vol. 1). Oxford, UK: Oxford University Press.
  • Borer, H. (2005b). The normal course of events: Structuring sense (Vol. II). Oxford, UK: Oxford University Press.
  • Cheng, H.-T. (2013). Argument ellipsis, classifier phrases, and the DP parameter [Doctoral dissertation, University of Connecticut].
  • Chomsky, N. (1981). Lectures on government and binding: The Pisa lectures. Dordrecht, The Netherlands: Foris Publications.
  • Chomsky, N. (1993). A minimalist program for linguistic theory. In K. Hale & S. Keyser (Eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger (pp. 1–52). Cambridge, MA: MIT Press.
  • Chung, S., Ladusaw, W. A., & McCloskey, J. (1995). Sluicing and logical form. Natural Language Semantics, 3, 239–282.
  • Condoravdi, C., & Gawron, J.-M. (1996). The context-dependency of implicit arguments. In M. Kanazawa, C. Pinon, & H, de Swart (Eds.), Quantifiers, deduction, and context (pp. 1–32). Stanford, CA: CSLI Publications.
  • van Craenenbroeck, J., & Temmerman, T. (Eds). (2019). The Oxford handbook of ellipsis. Oxford, UK: Oxford University Press.
  • Duguine, M. (2013). Null arguments and linguistic variation: A minimalist analysis of pro-drop [Ph.D. dissertation, University of the Basque Country UPV/EHU & Université de Nantes].
  • Fiengo, R., & May, R. (1994). Indices and identity. Cambridge, MA: MIT Press.
  • Fillmore, C. (1986). Pragmatically controlled zero anaphora. Berkeley Linguistics Society (BLS), 12, 95–107.
  • Gauker, C. (2012). What Tipper is ready for: A semantics for incomplete predicates. Noûs, 46(1), 61–85.
  • Goldberg, L. (2005). Verb-stranding VP-ellipsis: A cross-linguistic study [Doctoral dissertation, McGill University].
  • Güneş, G., & Lipták, A. K. (2022). The derivational timing of ellipsis. Oxford, UK: Oxford University Press.
  • Hale, K., & Keyser, S. J. (1993). On argument structure and the lexical expression of syntactic relations. In K Hale & S. J. Keyser (Eds.), The view from building 20: Essays in linguistics in honor of Sylvain Bromberger (pp. 51–109). Cambridge, MA: MIT Press.
  • Hankamer, J., & Sag, I. (1976). Deep and surface anaphora. Linguistic Inquiry, 7, 391–428.
  • Hicks, G. (2009). The derivation of anaphoric relations. Amsterdam, The Netherlands: John Benjamins.
  • Hoji, H. (1998). Null object and sloppy identity in Japanese. Linguistic Inquiry, 29, 127–152.
  • Hoji, H. (2015). Language faculty science. Cambridge, UK: Cambridge University Press.
  • Huang, C.-T. J. (1982). Logical relations in Chinese and the theory of grammar [Doctoral dissertation, MIT].
  • Huang, C.-T. J. (1984). On the distribution and reference of empty pronouns. Linguistic Inquiry, 15, 531–574.
  • Huang, C.-T. J. (1987). Existential sentences in Chinese and (in)definiteness. In E. Reuland & A. ter Meulen, (Eds.), The representation of (in)definiteness (pp. 226–253). Cambridge, MA: MIT Press.
  • Huang, C.-T. J. (1989). Pro drop in Chinese: A generalized control approach. In O. Jaeggli & K. Safir (Eds.), The null subject parameter (pp. 185–214). Dordrecht, The Netherlands: D. Reidel.
  • Huang, C.-T. J. (1991). Remarks on the status of the null object. In R. Freidin (Ed.), Principles and parameters in comparative grammar (pp. 56–76). Cambridge, MA: MIT Press.
  • Huang, C.-T. J. (1997). On lexical structure and syntactic projection. Chinese Languages and Linguistics, 3, 45–89.
  • Huang, C.-T. J. (2007). Hanyu dongci de tiyuan jiegou yu qi jufa biaoxian [The thematic structures of verbs in Chinese and their syntactic projection]. Linguistic Sciences, 6, 3–21.
  • Huang, C.-T. J. (2016). The syntax and semantics of prenominals: Construction or composition? Language and Linguistics, 17, 431–475.
  • Johnson, K. (Ed.). (2008). Topics in ellipsis. Cambridge, UK: Cambridge University Press.
  • Landau, I. (2020). On the non-existence of verb-stranding VP-ellipsis. Linguistic Inquiry, 51, 341–365.
  • Landau, I. (2022). Argument ellipsis as external merge after transfer. Natural Language & Linguistic Theory.
  • Lee, H.-T. T. (1986). Studies on quantification in Chinese [Doctoral Dissertation, University of California at Los Angeles].
  • Li, C. N., & Thompson, S. A. (1976). Subject and topic: A new typology of language. In C. N. Li (Ed.), Subject and topic (pp. 457–461). Austin, TX: University of Texas Press.
  • Li, C. N., & Thompson, S. A. (1981). Mandarin Chinese: A functional reference grammar. Berkeley, CA: University of California Press.
  • Li, H.-J. G. (1998). Null object and VP-ellipsis in Chinese. In H. Lin (Ed.), Proceedings of the 9th North American Conference on Chinese Linguistics (pp. 155–172). Los Angeles, CA: GSIL, University of Southern California.
  • Li, H.-J. G. (2002). Ellipsis constructions in Chinese [Doctoral dissertation, University of Southern California].
  • Li, Y.-H. A. (1990). Order and constituency in Mandarin Chinese. Dordrecht: Kluwer Academic Publishers.
  • Li, Y.-H. A. (2005). VP-deletion and null objects. Linguistic Sciences, 15, 3–19.
  • Li, Y.-H. A. (2007). Beyond empty categories. Bulletin of the Chinese Linguistic Society of Japan, 254, 74–106.
  • Li, Y.-H. A. (2008). Case, 20 years later. In M. K. M. Chan & H. Kang (Eds.), Proceedings of the 20th North American Conference on Chinese Linguistics (pp. 41–68). Columbus, OH: Ohio State University.
  • Li, Y.-H. A. (2011). Looking into clauses. In H. L. Soh, L. Tao, & X. Yun (Eds.), Studies in Chinese linguistics (pp. 27–58). London, UK: Cambridge Scholars.
  • Li, Y.-H. A. (2013). Case, tense and clauses. In H. Chappell, R. Djamouri, & T. Wiebusch (Eds.), Breaking down the barriers: Interdisciplinary studies in Chinese linguistics and beyond. Le Centre de recherches linguistiques sur l'Asie orientale, l'infrastructure de l'Ecole des Hautes Etudes en Sciences Sociales (EHESS), Paris. Taipei, Republic of China: Academia Sinica.
  • Li, Y.-H. A. (2014). Born empty. Lingua, 151, 43–68.
  • Li, Y.-H. A. (2018). Argument ellipsis and structures of noun phrases. The 53rd Annual Meeting of the Chicago Linguistic Society (pp. 195–214). Chicago: University of Chicago.
  • Li, Y.-H. A., & Wei, T.-C. (2014). Ellipsis. In C.-T. J. Huang, Y.-H. A. Li, & A. Simpson (Eds.), The Wiley–Blackwell handbook on Chinese linguistics (pp. 275–310). Malden, MA: Wiley–Blackwell.
  • Lin, T.-H. (2001). Light verb syntax and the theory of phrase structure. (Doctoral dissertation, University of California, Irvine, CA)
  • Liu, C.-M. L. (2014). A modular theory of radical pro drop [Doctoral dissertation, Harvard University].
  • Lu, J. (2002). Zai tan “chi-le ta san-ge pingguo” yilei jiegou de xingzhi [On chi-le ta san-ge pingguo type of constructions]. Zhongguo Yuwen, 89, 317–325.
  • Merchant, J. (2013). Diagnosing ellipsis. In L. L.-S. Cheng & N. Corver (Eds.), Diagnosing syntax (pp. 537–542). Oxford: Oxford University Press.
  • Miyagawa, S. (2010). Why agree? Why move? Unifying agreement-based and discourse-configurational languages. Cambridge, MA: MIT Press.
  • Neeleman, A., & Szendrői, K. (2007). Radical pro drop and the morphology of pronouns. Linguistic Inquiry, 38, 671–714.
  • Oku, S. (1998). A theory of selection and reconstruction in the minimalist perspective [Doctoral dissertation, University of Connecticut].
  • Otaki, K. (2014). Ellipsis of arguments: Its acquisition and theoretical implications [Doctoral dissertation, University of Connecticut].
  • Otani, K., & Whitman, J. (1991). V-raising and VP-deletion. Linguistic Inquiry, 22, 345–358.
  • Pan, H. (2019). Null object constructions, VP-ellipsis, and sentence interpretation. In J. Hu & H. Pan (Eds.), Interfaces in grammar (pp. 283–300). Amsterdam, The Netherlands: John Benjamins.
  • Pan, H. (2022). Remarks on gapless relative clauses and complement clauses in Mandarin Chinese. Asian Languages and Linguistics, 3(1), 109–140.
  • Pan, H., & Hu, J. (2003). Head noun movement, focus, and topicalization in Mandarin Chinese. In J. Xu, D. Ji, & K. T. Lau (Eds.), Chinese syntax and semantics (pp. 119–156). Singapore: Prentice Hall.
  • Panagiotidis, P. (2003). Empty nouns. Natural Language & Linguistic Theory, 21, 381–432.
  • Partee, B. (1989). Binding implicit variables in quantified contexts. Chicago Linguistic Society (CLS), 25, 342–356.
  • Patterson, Y. (2020). A study of nominal-clausal relations in Mandarin Chinese [Doctoral dissertation, Indiana University].
  • Paul, W., & Whitman, J. (2017). Topic prominence. In M. Everaert & H. van Riemsdijk (Eds.), The Blackwell companion to syntax (2nd ed., Vol. VII, pp. 4473–4503). Malden, MA: Wiley.
  • Pedersen, W. (2011). Implicit arguments, paychecks, and variable-free semantics. Semantics and Linguistic Theory (SALT), 21, 155–175.
  • Plesniak, D. (2021). Towards a correlational law of language: Three factors constraining judgment variation [Doctoral dissertation, University of Southern California].
  • Prince, E. F. (1981). Towards a taxonomy of given-new information. In P. Cole (Ed.), Radical pragmatics (pp. 281–297). New York, NY: Academic Press.
  • Pylkkänen, L. (2008). Introducing arguments. Cambridge, MA: MIT Press.
  • Ramchand, G. (2008). Verb meaning and the lexicon. Cambridge, UK: Cambridge University Press.
  • Rizzi, L. (1982). Issues in Italian syntax. Dordrecht, The Netherlands: Foris Publication.
  • Rizzi, L. (1986). On the status of subject clitics in Romance. Studies in Romance Linguistics, 24, 391–419.
  • Roberts, I. (2019). Parameter hierarchies and universal grammar. Oxford, UK: Oxford University Press.
  • Saito, M. (2004). Ellipsis and pronominal reference in Japanese clefts. Studies in Modern Grammar, 36, 1–44.
  • Saito, M. (2007). Notes on East Asian argument ellipsis. Language Research, 43, 203–227.
  • Saito, M. (2017). Ellipsis. In M. Shibatani, S. Miyagawa, & H. Noda (Eds.), Handbook of Japanese syntax (pp. 701–750). Berlin, Germany: Mouton de Gruyter.
  • Sakamoto, Y. (2017). Escape from silent syntax [Doctoral dissertation, University of Connecticut].
  • Sato, Y. (2014). Argument ellipsis in Colloquial Singapore English and the anti-agreement hypothesis. Journal of Linguistics, 50, 365–401.
  • Sato, Y. (2015). Argument ellipsis in Javanese and voice agreement. Studia Linguistica, 69, 58–85.
  • Sato, Y. (2019). Comparative syntax of argument ellipsis in languages without agreement: A case study with Mandarin Chinese. Journal of Linguistics, 55, 643–669.
  • Sato, Y. (2020). Idioms, argument ellipsis and LF-copy. Journal of East Asian Linguistics, 29, 259–278.
  • Sato, Y., & Karimi, S. (2016). Subject–object asymmetries in Persian argument ellipsis and the anti-agreement theory. Glossa: A Journal of General Linguistics, 1, 1–31.
  • Schwabe, K., & Winkler, S. (Eds.). (2003). Interfaces: Deriving and interpreting omitted structures. Amsterdam, The Netherlands: John Benjamin.
  • Şener, S., & Takahashi, D. (2010). Ellipsis of arguments in Japanese and Turkish. Nanzan Linguistics, 6, 79–99.
  • Shi, D. (1992). The nature of topic comment constructions and topic chains [Doctoral dissertation, University of Southern California].
  • Shopen, T. (1973). Ellipsis as grammatical indeterminacy. Foundations of Language, 10, 65–77.
  • Simpson, A. (in press). In defense of verb-stranding VP-ellipsis. Syntax.
  • Simpson, A., Choudhury, A., & Menon, M. (2013). Argument ellipsis and the licensing of covert nominals in Bangla, Hindi and Malayalam. Lingua, 134, 103–128.
  • Soh, H. L. (1998). Object scrambling in Chinese [Doctoral dissertation, Massachusetts Institute of Technology].
  • Stowell, T. (1981). Origins of phrase structure [Doctoral dissertation, Massachusetts Institute of Technology].
  • Takahashi, D. (2008). Noun phrase ellipsis. In S. Miyagawa & M. Saito (Eds.), The Oxford handbook of Japanese linguistics (pp. 394–422). Oxford, UK: Oxford University Press.
  • Takahashi, D. (2014). Argument ellipsis, anti-agreement, and scrambling. In M. Saito (Ed.), Japanese syntax in comparative perspective (pp. 88–116). Oxford, UK: Oxford University Press.
  • Takahashi, D. (2019). Derivational argument ellipsis. The Linguistic Review, 37(1), 2019–2034.
  • Tang, S.-W. (2011). A parametric approach to NP ellipsis in Mandarin and Cantonese. Journal of East Asian Linguistics, 20, 107–115.
  • Taraldsen, T. (1980). On the NIC, vacuous application and the that-trace filter. Ms. Indiana University Linguistics Club, Bloomington.
  • Thompson, S. (1973). Transitivity and some problems with the ba construction in Mandarin Chinese. Journal of Chinese Linguistics, 1, 208–221.
  • Tomioka, S. (2003). The semantics of null arguments in Japanese and its cross-linguistic implications. In K. Schwabe & S. Winkler (Eds.), Interfaces: Deriving and interpreting omitted structures (pp. 321–339). Amsterdam, The Netherlands: John Benjamin.
  • Tomioka, S. (2008). A step-by-step guide to ellipsis resolution. In K. Johnson (Ed.), Topics in ellipsis (pp. 210–228). Cambridge, UK: Cambridge University Press.
  • Tomioka, S. (2014). Micro-parameters in discourse pro-drop languages: Comments on “Born Empty” by Yen-hui A. Li. Lingua, 151, 69–77.
  • Tsao, F.-F. (1977). A functional study of topic in Chinese: The first step toward discourse analysis [Doctoral dissertation, University of Southern California].
  • Wei, T.-C. (2022). Null complement anaphora in Mandarin Chinese. Ms. National Kaohsiung Normal University.
  • Williams, A. (2012). Null complement anaphors as definite descriptions. Proceedings of SALT, 22, 125–145.
  • Williams, E. (1977). Discourse and logical form. Linguistic Inquiry, 8, 101–139.
  • Wu, H.-H. I. (2016). On the (non)-existence of verb-stranding ellipsis in Chinese. International Journal of Chinese Linguistics, 3, 79–99.
  • Xu, L.-J. (1986). Free empty category. Linguistic Inquiry, 17, 75–93.
  • Xu, L.-J. (2003). Remarks on VP-ellipsis in disguise. Linguistic Inquiry, 34, 163–171.
  • Ye, K. (2020). VVPE or NOC: New arguments for the existence of VP-ellipsis in Mandarin Chinese. Contemporary Linguistics, 22, 71–86.


  • 1. The term “ellipsis” or “deletion” in the literature often just refers to the unsaid/invisible part of an utterance and does not always mean the grammatical mechanism of deletion. To avoid confusion, in the rest of this article we will limit these terms to the formal mechanism of deleting lexical items, while using the theory-neutral terms “null/empty/missing” for the unsaid/invisible parts.

  • 2. Chomsky (1981) discusses “null subject’ and the term “pro-drop” was used, which is mainly tied to rich morphological agreement in the relevant languages; this is known as Taraldsen’s Generalization (Adams, 1987; Rizzi, 1982, 1986; Taraldsen, 1980; among many others). Because such an agreement and null argument correlation is not found as such in languages like Chinese, which has no agreement but widely uses null arguments, terms such as ;radical pro-drop’, ‘rampant pro-drop’, or ‘discourse pro-drop’ are used (see, e.g., Neeleman & Szendrői, 2007).

  • 4. Scope interaction between quantificational expressions and referential dependency relations governed by the binding principles (Binding Principle A for anaphors, B for pronouns, and C for names) are tests to determine hierarchical relations in phrase structures, specifically, the c-command relation.

  • 5. See Landau (2020) for tests on object ellipsis excluding the c-commanded duration/frequency phrases in East Asian languages. However, see Simpson (in press) on the nuances and challenges.

  • 6. This distinction suggests that thematic contributions by lexical items should be recognized, unless the two patterns with the same experiencer-theme arguments have different event structures, which would seem to require ad hoc stipulations. In an approach that takes event or aspectual structures as basic and lexical items as being simply roots, without subcategorization properties, lexical information is irrelevant in syntax (see Borer, 2005a, 2005b; Huang, 1997; Lin, 2001; Ramchand, 2008, among many others. Also see the decomposition and hierarchical structures of lexical items in Hale & Keyser, 1993). The fact that lexical information affects the noted possibility of omitting arguments indicates that relevant lexical information should be at work in grammar. The constraint cannot be a matter of pragmatics or world knowledge.

  • 7. Questions have been raised on the validity of this generalization (see, for instance, Cheng, 2013, section, p. 237–239). The following examples are Cheng’s (28a, c) on p. 181, which seemingly allow a clausal object to be missing, even though a nominal object is disallowed.



    It is not clear whether generalizations can be built on such examples. First of all, renwei ‘think’ in (i) must take the negative form. If it is changed to the positive, tamen/Lisi ye renwei ‘they also think’, the object cannot be missing without the pro-form zheme/name ‘so’. It seems that ‘think’ and negation take a special status. A relevant case is the unique expression in English I think not, which can stand as it is; but I think is followed by an object or the pro-form so (parentheticals not under consideration). ‘Regret/Regretful’ in (ii) can be taken as an instance of null complement anaphora, not due to the absence of a linguistic object, as discussed in Cheng (2013, section 4.3.4). Note that the following sentence with a nominal topic is perfectly fine in Chinese. If we replace houhui ‘regret/be regretful’ with the verbs used in the unacceptable sentences in this section, the result is unacceptable:


    Our generalization does not apply to those verbs allowing null complement anaphora. Note that it is reasonable to separate null complement anaphora from the cases with objects missing. For instance, English does not allow objects to be deleted, but null complement anaphora is available.

  • 8. The pronoun test does not work in Chinese because Chinese has an independent constraint against using pronouns for inanimate objects (and the O of VO idioms are generally inanimate). We will not explore the exact notion of referentiality here. Readers are referred to Landau’s work. We will just present the Chinese data corresponding to the discussion by Landau (2022).

  • 9. Landau proposes that only real quantificational expressions cannot be missing—those not replaceable by pronouns. As in note 9, the pronoun test is not reliable in Chinese. We put this category aside. Landau further distinguishes the O of V+O idioms: only some V+O idioms are transparent and de-composable, and the object is referential, permitting deletion. However, it is not clear how referentiality is determined to avoid circularity (if an element is referential, then it can be null, and vice versa). Nor is it clear that the properties of being referential and decomposable are correlated with the possibility of taking the null form. Observe the following contrasts:


    The object in the VO idiom ti piqiu ‘kick ball=pass the buck’ seems not to be movable (ic–d) and its taking the null form does not sound good (ia). However, if the adverb huxiang ‘mutually’ is added before the verb, it seems acceptable for the object to be missing as in (ib), although the addition of the adverb does not rescue the unacceptable movement cases (ic–d). We leave the issue to be investigated in further research.

  • 10. As indicated by the indexing, a strict reading keeps the same referential index; a sloppy reading allows the index to vary with the subject of the sentence. The index 3 refers to some entity other than those with the index 1 and 2. H.-J. G. Li (2002, pp. 86–87), using the mixed reading test by Fiengo and May (1994), shows that the true sloppy reading for missing objects does indeed exist in Chinese.

  • 11. The subject of the first clause can be the topic for the two clauses of this “topic chain”; see Tsao (1977), Shi (1992), among many others.

  • 12. Sato (2019) has examples showing that the sloppy reading is allowed for null subjects when the subject need not be definite, such as a construction that contains a hanging topic.



    We found it difficult to interpret (iib) as Lisi’s child working for Microsoft, although (ib) is acceptable with the so-called sloppy reading. To obtain the sloppy reading, (iib) needs to be changed to Lisi de (shi) wei Weiruan gongzuo ‘Lisi’s (child) works for Microsoft’. The possession marker de should be added. For (i), it is possible to use the name of a person to represent the articles that he writes: ‘Rizzi published in NLLT’. It is not clear that there is a null subject in (ib) referring to Rizzi’s articles. Other examples in Sato’s work should be reconsidered in a similar manner, either in terms of alternative explanations or different judgments.

  • 13. Specific mechanisms for LF-copying have been proposed in many works (Chung et al., 1995; Fiengo & May, 1994, Johnson, 2008; Oku, 1998; Schwabe & Winkler, 2003; Williams, 1977, and especially Tomioka, 2008, pp. 210–228, among many others). Further note that idioms do not challenge an LF-copying approach (e.g., Pan, 2019, Ye, 2020). Idioms can come together for interpretation at LF, which accounts for why idiomatic readings forces LF reconstruction in cases like the following (Chomsky, 1993, from Hicks, 2009, p. 271, (32a–b)):


    In order to obtain the idiomatic reading ‘take picture of = photograph’, reconstruction takes place, accounting for why himself can only be coindexed with the lower subject Bill in (a). The one in (b) does not have the idiomatic interpretation. Chomsky claims that on the idiomatic reading, take a picture of himself has to be interpreted at LF as a syntactic unit, so reconstruction is forced. Also see Sato (2020) for the use of different types of idioms to argue for an LF-copying approach to null arguments in Japanese.

  • 14. A TEC differs from Xu’s (1986) free empty category (FEC) in distribution and feature specifications. Xu’s FEC can be a trace/pronoun/reflexive and can appear in any positions.

  • 15. The works adopting an LF-copying approach to ellipsis constructions, such as those by Oku and Saito cited here, generally assume an “e” in the null position in the phrase structure, although attempts have not been made to clarify what this “e” is. Our TEC proposal fills the gap.

  • 16. In contrast, Landau (2022) claims that extraction from within a null argument is possible. He gives the following example from Hebrew. See the next note (18) for the contrast. His example in (71) is copied below.


  • 17. Some cases of topicalization seem to be possible, such as (i).


    Importantly, the corresponding sentence in English (ii) is also acceptable.


    English does not allow argument ellipsis. If there seems to be an object missing, it is a verb that does not select an object syntactically—the so-called null complement anaphora. No objects are present syntactically, but pragmatics makes the interpretation of an apparent object acceptable. That is, (ii) does not contain an elided clausal object. The difference between null complement anaphora and the TEC is that the latter, not the former, has a position in the syntactic structure. It is possible that (iii) are instances of null complement anaphora, with pragmatics providing interpretation without the syntactic presence of the object.

    Relativization is more consistent, disallowing any extraction from within a null argument. This raises the question of how topic structures and relativization constructions should be distinguished derivationally so that the null complement anaphora analysis is available for the topic structure (i), but not the relativization construction in (iii).


    Laudau (2022) observes that extraction from within a null object is possible. It is not clear why Chinese and Hebrew contrast in extraction possibilities. Takahashi (2019) notes that extraction from within a null phrase is possible in SOV languages, such as Japanese and Korean, but not possible in SVO languages, like Chinese or Vietnamese. He proposes a derivational ellipsis analysis (timing of ellipsis) to capture the difference. Unfortunately, Hebrew is SVO; yet it allows extraction as claimed by Landau. We leave this issue to be the subject of further research, but not without pointing out the following interesting contrast—the presence vs. absence of to in English in (iva–b) below (Merchant, 2013):


    Wh-movement is acceptable only when to is present. The example shows that the fronted wh-phrase in English must be the result of movement. This contrasts with topicalization such as that seen in (i) and (ii), which can be interpreted according to an aboutness relation that can be established pragmatically. Note that the example by Landau also involves wh-movement.

  • 18. Lin (in press) notes that objects of applicative projections are more difficult to omit. These objects include (a) unselected objects (Lin, 2001), (b) indirect objects, and (c) the objects of unaccusative verbs such as ‘die’. In contrast, it is easier for subjects to take the null form. However, it is not true that unselected objects cannot be missing, as illustrated by (i) below.


    Nonetheless, not all unselected objects can be missing. What distinguishes the acceptable and unacceptable cases requires further research to establish.

    In regard to indirect objects, Li (1990) observes that indirect objects cannot undergo movement. Therefore, they should behave differently from direct objects which can be moved. For the type involving unaccusative verbs, the inability of a missing object is captured by the emptiness of a TEC. More generally, an inner object cannot be missing when there is an outer object (which can become a subject), but the outer object can be missing when the inner object is present.

  • 19. Williams (2012, p. 126) lists some alternative terms for this construction: “null complement anaphora” (Hankamer & Sag, 1976); “definite ellipsis” (Shopen, 1973); “definite null complements” (Fillmore, 1986); “implicit variables” (Partee, 1989); “anaphoric implicit arguments” (Pedersen, 2011); and “incomplete predicates” (Gauker, 2012).

  • 20. Pan (2022) notes that adjunct relativization allows deletion of the head noun if the context is clear enough for the recovery of the missing content, illustrated by the following sentence (also see Lin, in press):


    It is not clear whether contexts are sufficient to license deletion. For instance, the examples in (3738) have clear contexts, facilitated the use of the resumptive wh within the relative clause and an antecedent clause with the same structure can be added, producing a parallel structure (such as ‘We know the reason you resigned; we don’t know the (reason) he resigned’). Yet, the addition of an antecedent does not help in these. It seems that the acceptability of a null NP such as in (i) requires its modifier not be a full clause. If an adjunct relative is clearly a full clause (embedded under a clausal taking verb), the modified NP cannot be null.


    According to the analysis of Aoun & Li (2003), it is the licensing of a null relative operator that prevents the head noun from being missing. It is possible that expressions like ‘the time of undergraduate classes’ and ‘the time of their class attending’ in (i) do not contain a null relative operator.

  • 21. Pan (2022) discusses a contrast observed in Pan and Hu (2003) as in (i) and proposes the condition in (ii) to capture the contrast.



    Nonetheless, (ib) is not fully acceptable. Bingfu Lu (personal communication) suggests that it might be an instance of parenthetical expression. It is also possible that (ib) is not a noun-complement structure but involves apposition. Clear noun-complement cases like the one below in (iii) are not acceptable.


    The questions of how appositives differ from noun-complement structures and how they can be distinguished need much more study (cf. Stowell, 1981).