Semantic scope ambiguity in gapping and non-constituent coordination: a generative analysis

Abstract

Coordination has been thought to be a reliable test of constituency; thus all the cases of apparent non-constituent coordination (noncanonical coordination) were assumed to be derived via reduction: movement or ellipsis. This view has been challenged by facts from the semantics of non-canonical coordination, particularly scope ambiguity in gapping and non-constituent coordination. I provide here an analysis that accounts for this type of ambiguity. I propose that the ambiguity that arises in non-canonical coordination is structural; that is, the cases of coordination are derived from two sources (a vP source and a CP source), where each source is derived via AT B movement or ellipsis. I spell out an analysis in terms of left-to-right syntax, in which copying of displaced elements is allowed to be minimal under some circumstances, which facilitates the wide scope reading of scope-taking elements in non-canonical coordination. The analysis confirms the assumptions about constituency and structure in phrase structure grammars, such as Generative Grammar, by providing a purely syntactic analysis of the scopal peculiarities of non-canonical coordination. This result has implications on how syntactic chunks are produced and processed in the human brain, which can in turn benefit fields, such as psycholinguistics, neurolinguistics, or even computational linguistics.

Keywords:

• Coordination
• ellipsis
• movement
• gapping
• non-constituent coordination
• left to right syntax

REVIEWING EDITOR:

• Jeroen van de Weijer
• Shenzhen University
• China

SUBJECTS:

• Grammar, Syntax & Linguistic Structure
• Syntax

1. Introduction

A long-standing assumption in the literature on syntax is that coordination operates on constituents and coordination has been long used as a constituency test. However, there are at least two major cases in which it appears that non-constituents are conjoined: gapping and non-constituent coordination (NCC). In (1a), the non-initial conjunct involves gapping where the verb goes missing; thus, it appears that non-constituents are conjoined. Similarly, in (1b), the non-initial conjunct does not constitute a full-fledged constituent (ie clause or TP) because it appears that phrases are being juxtaposed and then conjoined to the initial conjunct.

• (1). a. Some have served mussels to Sue and others swordfish. (Johnson, 2009, p. 289, (1a)) (gapping)

• b. Mary caught a fish on Monday with a fly rod and on Tuesday with a spear. (Dowty, 1988) (NCC)

Following the literature, I distinguish NCC, where apparently different non-continuous parts within a non-initial conjunct are missing, and gapping where it is obvious that at least one verbal element is missing from the surface structure. The standard assumption in the syntactic literature is that the coordinate constructions such as the ones in (1) are not really non-canonical; they involve coordination of full-fledged constituents at an abstract level, where the surface structures are derived via some form of reduction, which can be movement or ellipsis (eg Bruening, 2015; Chaves, 2014; Johnson, 2009; Potter et al., 2017; Sailor & Thoms, 2013). However, facts from scope ambiguity exemplified in (2) challenge this assumption.

• (2). a. Mrs. J can’t live in Boston and Mr. J ø in LA. (Kubota & Levine, 2016, p. 109, (5a))

• b. Sam sent no girl chocolates today and flowers yesterday. (Osborne & Gross, 2017, p. 669, (50a); adapted)

These data are claimed to suggest that a reduction analysis makes wrong predictions about the semantics of the structures. As exemplified in (2a), the scopal elements that appear outside the coordination, namely modality and negation, seem to take both a wide scope reading, and a distributive scope reading with respect to coordination. In the non-elliptical counterparts of these cases of coordination, only the latter reading is available. This poses a major challenge to the analyses which take these forms of coordination to have been derived via reduction (movement or ellipsis).

This paper contributes to the debate on this issue by providing an account of the puzzle of the scopal ambiguity as an argument against the assumption that coordination targets constituents. In particular, I will show that the scope facts in non-canonical coordination as presented above can be accounted for in left-to-right/top-down syntax, where copies of moved elements are restricted by economy (cf. Al Khalaf, 2019; Al Khalaf & Mashaqba, 2023 for floating quantifiers and Al Khalaf, 2017 for negative polarity items), and should therefore be minimal under particular circumstances, as an effect of processing and linear distance between the constituent that is merged first and the lower copies that are merged later in the derivation. This will account for the wide scope reading of semantic operators. The rest of the paper is organized as follows. To provide a theoretical background of the paper, Section 2 presents the facts about scope ambiguity in non-canonical coordination, highlighting the puzzle of semantic scope with relation to constituency in coordination. Section 3 cites evidence that coordination targets constituents. This evidence is crucial to the analysis provided as it constitutes the basis of this analysis. Section 4 discusses evidence that non-canonical coordination can be derived from two sources: a CP source and a vP source. This two-source analysis will be integrated with the left-to right derivation in Section 5 to present an analysis that explains the peculiarities of semantic scope ambiguity non-canonical coordination. Section 6 is a conclusion.

2. Scope ambiguity in gapping and non-constituent coordination

In this section, I will present the problem of scope ambiguity that poses a challenge to analyses that assume that coordination conjoins full-fledged constituents even in cases that indicate otherwise. To begin with, Scopal ambiguity in gapping is one source of such problem as has been show in much of the previous literature (eg Johnson, 2004; Kubota & Levine, 2016; McCawley, 1993; Oehrle, 1987; Potter et al., 2017; Siegel, 1984, 1987). Consider the examples in (3) below:

• (3). a. James can’t order caviar and Mary chili. (Potter et al., 2017, p. 1125, (4a))

• b. Mrs. J can’t live in Boston and Mr. J ø in LA. (Kubota & Levine, 2016, p. 109, (5a))

The negation and the modal can have a wide scope reading in (3a) and (3b), respectively, where they outscope the coordinate structure, or a distributive scope reading in which they scope locally within each conjunct. In (3a), the sentence, informally, can have two interpretations. The first is that a situation in which James orders caviar and Mary orders chili is not possible. That is, they cannot order these two dishes at the same time. The second interpretation is that it is not possible for James to order caviar and it is not possible for Mary to order chili. That is, scopal elements either scope above the coordinate phrase or below it. Note that in the non-elliptical version of the sentence (ie Mrs. can’t order caviar and Mary can’t order chili.), only the second reading is available. Thus, it is not clear how (3a) is derived from a source where full-fledged constituents have been coordinated. The same can be said about sentence (3b), (though the relevant operator here is negation rather than modality).

Similarly, scopal ambiguity was observed in non-constituent coordination as was discussed in much of previous work (Kubota & Levine, 2015; Osborne & Gross, 2017; Sailor & Thoms, 2013).

• (4). a. Sam sent no girl chocolates today and flowers yesterday. (Osborne & Gross, 2017, p. 669, (50a); adapted)

• b. John didn’t speak to Mary on Thursday and Bill on Friday. (Sailor & Thoms, 2013, p. 367, (29))

In (4a), the negative operator introduced by no girl can have a wide-scope reading in which the sentence means: it is not the case that Sam sent some girl chocolates today and sent the same girl flowers yesterday, a reading that is unavailable in the non-elliptical counterpart of the sentence (Sam sent no girl chocolates today and Sam sent no girl flowers yesterday), where the negative operator (ie no) scopes locally in each conjunct. In this case, the sentence means: it is not the case that Sam sent some girl chocolates today and it is not the case that Sam sent (another) girl flowers yesterday. The same can be said about (4b). Obviously, this fact is problematic to analyses that take NCC to be derived from larger categories.

Thus, as can be seen from the discussion above, assuming that non-canonical cases of coordination are derived from a more abstract level of representation where constituents are conjoined is challenged by facts from semantic scope. The cases of non-canonical coordination allow more readings that the non-elliptical versions or the full-fledged coordination versions. This is problematic to the assumption that coordination operates on constituents, which in turn challenges the concept of constituency in syntax.

3. Coordination targets constituents

In this section, I will briefly go through evidence that in non-canonical coordination, constituents are being conjoined at a more abstract level.

To begin with gapping, and to take (5) as an example, it is more plausible to assume that the gapped conjunct has a more articulate structure where the verb is syntactically projected at a more abstract level. In particular, since the gapped conjunct has participants in some action/event/state (thematic roles), these participants are associated with some predicate (ie a verb). Here, the gapped conjunct James a book has two thematic roles, AGENT (James) and THEME (a book). These thematic roles participate in some action/event represented by a predicate, which is understood to be PUBLISHED. Crucially, the verb in the non-gapped conjunct cannot serve as the predicate of the thematic roles of the gapped conjunct because sharing predicates is not allowed, as there should be a one-to-one relationship between a predicate and a thematic role such as the AGENT role.

• (5). Sally published a paper, and James a book.

In addition, there are good reasons to believe that the gapped conjunct has an abstract level of representation, in which a verb is syntactically projected. For example, verbs that require plural objects, such as pair provide evidence for the existence of a null copy of the verb in the gapped conjunct. In (6a), the verb does not allow a singular object. In (6b), where gapping has applied, the gapped conjunct is only grammatical with a plural object. The explanation here is that there is a verb at an abstract level that forces the selectional restrictions on the object.

• (6). a. He paired {*a sheet/two sheets of different materials}.

• b. Sally paired {*one sheet/two sheets}, and John {*a portrait/two portraits}.

Turning to NCC, Bruening (2015) shows that NCC is derived from larger categories, citing empirical evidence. For instance, he shows that NCC displays plural agreement (Beavers & Sag, 2004; Bruening, 2015) and licensing of items sensitive to plurality, like plural floating quantifiers, as shown in (7). That verbs show plural agreement with subjects with a NCC structure and that they may associate with floating quantifiers that require plurality show that NCC is actually a conjunction of constituents. Note that one might question the validity of using number agreement to argue for a particular syntactic configuration for coordination and argue that number agreement is determined by the speaker’s perspective and discourse. Still, examples such as (7) can be used to argue that coordination targets constituents given the use of both and different which suggest particular constituencies within the coordinate phrase. The role of pragmatics and discourse in number agreement cannot be ignored, though, but that does not imply that number agreement cannot be determined by syntax, as well.

• (7). (Bruening, 2015, p. 2, (6a,b))

• a. Bill catching a fish on Monday with a fly rod and on Tuesday with a spear were both surprising.

• b. Micah claiming he was an astronaut to impress Bill and a spy to impress Bob were different events.

In addition, medial conjuncts where it appears that a non-constituent is conjoined with other conjuncts is a puzzle to analyses that do without a reductionist analysis of coordination. As shown by Beavers and Sag (2004), examples like (8) is evidence that NCC involves coordination of a more articulate structure at a more abstract level; otherwise, it is hard to explain the fact that full-fledged conjuncts are conjoined with the NCC conjunct. Coordination is generally known to respect parallelism (eg Al Khalaf 2015, 2018, 2021, 2022). These facts suggest that NCC is derived from larger categories that have undergone some form of reduction.

• (8). Jan [[travels to Rome tomorrow], [[to Paris on Friday], and [will fly to Tokyo on Sunday]]]. (Beavers & Sag, 2004, p. 54, (10); adapted)

Thus, the facts above suggest that intuitively gapping and NCC are derived from larger structures that have undergone some form of reduction. Facts from thematic roles, agreement and reduction in medial conjunct lead to the conclusion that in cases where it appears that non-constituents are conjoined in fact involve coordination of full-fledged constituents.

4. A two-source analysis of non-canonical coordination

In this section, I will go through evidence that non-canonical coordination can have two sources: a vP source and a CP source. These sources, when undergoing syntactic operations such as movement or ellipsis, give rise to a single surface from that can have two readings in cases of semantic ambiguity as shown in Section 2.

Potter et al. (2017) argue that the source of ambiguity in gapping is structural. In particular, they propose that gapping can be derived from a vP source or a CP source. The former gives rise to a wide scope reading of scope-taking elements, whereas the latter gives rise to a distributive scope reading. They provide independent evidence for this proposal by presenting cases of gapping that are structurally conditioned to be only derived from a vP source and others which are structurally conditioned to be derived from a CP source. Then they show that these cases are not ambiguous and can only give one of the two possible readings. To clarify, consider the examples below:

• (9). (Potter et al., 2017, p. (39))

• a. To Mary, James didn’t give a cupcake or Bill chocolates.

• b. With only ten dollars between them, James could get a sandwich, and Mary a bowl of soup.

• c. Often, James orders mussels, and Mary shrimp.

• d. Never does Kim play bingo or Sandy chess.

As can be seen in these examples, a constituent appears in the left periphery of topic phrase in examples like (9a,b) (Rizzi, 1990) and in the left periphery of a focus phrase in examples like (9c,d) (Haegeman, 2000; Rizzi, 1990). Thus, the gapping here can only be derived from a CP source because it is impossible for a topicalized/focused phrase to scope over coordination unless the coordination is a combination of two clauses. Thus, this fact suggests that a subset of gapping constructions have a CP source.

On the other hand, with relation to NCC, Sailor and Thoms argue that NCC is derived via movement. They provide a number of arguments for movement effects. First, NCC shows sensitivity to islands; thus, NCC chunks appearing inside a movement island result in ungrammaticality. For instance, although possessive DPs can be NCC chunks, a part of a possessive DP cannot, as shown in (10b).

• (10). (Sailor & Thoms, 2013, p. 363, (13))

• a. John wrote everyone’s favorite song about football in 2001 and everyone’s favorite song about basketball in 2012.

• b. *?John wrote everyone’s favorite song about football in 2001 and basketball in 2012.

Second, immovable constituents cannot be NCC chunks. For example, as argued by Abels (2003), a TP that is a complement to an overt C cannot move, as in (11). They show with examples that it is impossible for this sort of constructions to be chunks in NCC.

• (11). * He knows Icelandic, I’m sure that. (Sailor & Thoms, 2013, p. 364, (20a))

Sailor and Thoms argue that NCC should be derived from two sources. That is, they provide evidence that there are different heights of the coordinate constructions involving NCC. One piece of evidence comes from scope ambiguity, as shown in the example below:

• (12). John didn’t speak to Mary on Thursday or Bill on Friday.

• a. that is, he spoke to neither of them. ¬ (A ∧ B)

• b. but I can’t remember which one of them he didn’t speak to. (¬ A) ∧ (¬ B) (Sailor & Thoms, 2013, p. 367, (28))

The ambiguity here, Sailor and Thoms argue, can be explained if the NCC is derived from two sources. That is, the distributive scope of negation suggests a CP coordination, whereas the wide scope of negation over coordination is explained if coordination is at the vP level.

Thus, based on the proposals above from Potter et al. (2017) and Sailor and Thoms (2013), the cases on non-canonical coordination can be derived from two sources. This assumption is crucial to the analysis to be proposed in this paper, as will be shown in Section 5.

5. Analysis

Having laid out the background and assumptions for the analysis to be presented in Sections 2 and 3, in this section, I present an analysis that overcomes the problems presented in Section 2. An important ingredient in the analysis that I will present is the type of the derivation adopted. I propose, following Phillips (1996, 2003), Bruening (2014), Osborne and Gross (2017), Al Khalaf (2015), and Bruening and Al Khalaf (2018, 2020) that syntactic structures are built from left to right. Thus, higher copies of displaced elements are merged before lower copies. I follow Bruening and Al Khalaf (2018) in assuming that lower copies can be minimal as an effect of processing: the further the copies are from the landing site of the moved constituent, the more minimal they are. By minimal I mean that only the head of the phrase is copied; other elements do not have to be copied, unless the phrase is prosodically prominent or unless some other syntactic constraint forces full copying. Note that left-to-right derivations are more plausible in accounting for across-the-board (ATB) movement in particular, as shown by Bruening and Al Khalaf (2018). In standard derivations (ie bottom- up derivations), somehow two copies have to merge into one constituent, which is hard to comprehend and explain without ad hoc assumptions. On the other hand, left-to-right derivation of ATB movement is more natural because a full copy of the constituent that has moved across-the-board is merged first and subsequent lower copies of that constituent are merged later and could be minimal as an effect of linear distance from the higher copy. This supports processing-based models in which the working memory interacts with the way structures are built.

In addition, I propose that the cases of non-canonical coordination derived from a vP source are derived via ATB movement, following Johnson’s (2004, 2009) analysis of gapping. On the other hand, the cases of non-canonical coordination that come from a CP source are derived via ellipsis, following Sailor and Thoms’s analysis of NCC. Therefore, I differ from the previous literature on gapping and non-constituent coordination in that the analysis I propose is hybrid: gapping and NCC can be derived via movement or ellipsis. Previous proposals are divided into proposals that adopt an ellipsis account, and those that adopt a movement account. I will show that this hybrid proposal facilitates a purely syntactic analysis of the semantic ambiguity in non-canonical coordination in a simpler and less stipulative way.

As presented in Section 2, scope ambiguity is a challenge to reductionist analyses. Consider the example below repeated from above:

• (13). Sam sent no girl chocolates today and flowers yesterday.

As explained previously, this sentence is ambiguous; however, its non-elliptical version can have only one interpretation, where negation is distributive:

• (14). Sam sent no girl chocolates today and (Sam) sent no girl flowers yesterday.

According to the proposal above, the sentence involving NCC is derived from full-fledged sources: a CP source and a vP source:

• (15). a. [_CP Sam sent no girl chocolates today] and [_CP Sam sent no girl flowers yesterday].

• b. Sam [_vP sent (no) girl chocolates today] and [_vP sent (no) girl flowers yesterday].

Note that I place parentheses around the quantifier no, as I will propose that copying of lower copies in ATB movement is minimal, where only heads are copied (see below). Each of these sources involves a different reduction process to derive the surface structure. Let’s explain how the CP source gives rise to a distributive reading of negation; that is, a reading in which the girl to which Sam did not give chocolates or flowers is not the same girl or does not have to be the same girl. According to my proposal, the CP source undergoes ellipsis. I adopt the movement analysis proposed by Sailor and Thoms (2013), where the non-initial conjunct undergoes multiple movements of the constituents surviving ellipsis, followed by ellipsis of the remnant of the movements. Thus, as illustrated in (16), the derivation of (15a) proceeds as follows. The derivation begins from left to right by building the first conjunct (I adopt a standard coordinate structure where the coordinate phrase is &P in which & projects an external argument (the initial conjunct) and an internal argument (a non-initial conjunct) (e.g, Collins, 1988):

• (16).

The derivation proceeds by building the non-initial conjunct. Here, according to Sailor and Thoms, the constituents surviving ellipsis in the non-initial conjunct undergo a focus movement followed by ellipsis of the remnant of the movement. I will remain agnostic about whether there is a focus phrase in the left periphery of the non-initial conjunct, but this is a possibility (see eg Rizzi, 1997):

• (17).

Note here that the negative operator occurs in every conjunct, giving rise to a distributive reading of negation. On the other hand, the wide scope reading of negation occurs when the reduction process is an ATB movement which applies to the vP source of non-canonical coordination according to my proposal. As assumed above, in a left-to-right derivation of ATB movement, the lower copies of the displaced constituent can be minimal as an effect of processing. Thus, the derivation of the structure begins by building the structure by merging the subject Sam then the T head. I assume, following Johnson (2009), that a PredP is projected of a Pred head that selects a coordinate vP as a complement. A verb moves out of the coordinate phrase across the board to Pred. The indirect object no girl also moves across the board to the left edge of the vP coordinate, as well. Thus copies of the displaced elements here are merged before building the coordinate structure:

• (18).

The derivation proceeds by building the coordinate phrase. As assumed above, only minimal copies of the displaced elements are merged. Thus, girl is copied into the conjuncts without the quantifier no in this case:

• (19).

As can be seen, the negative quantifier occurs outside the coordination and only minimal copies occur inside the coordination. This allows for a wide scope reading of negation. Therefore, the discussion above shows that the ambiguity in NCC is structural. NCC has two sources, each of which undergoes a different reduction process that gives rise to one of the two available readings.

Note that the analysis proposed was illustrated for semantic scope ambiguity in NCC to clarify how it operates. The same analysis applies to scope ambiguity in gapping and is equally valid in explaining the ambiguity in this case.

6. Conclusion

In this paper, I went through the major arguments leveled against a reduction analysis of non-canonical coordination. I then presented arguments that non-canonical coordination is derived from larger categories and that it can have two sources. I proposed an analysis of scope ambiguity and the facts of symmetric predicates in a left-to-right derivation. I showed that two reduction operations that apply to the two sources explain ambiguous readings. The analysis proposed here is compatible with Brueing’s proposal that NCC should be derived via deletion of a prosodic/syntactic constituent. A subset of the cases undergo ellipsis. My analysis is compatible with Johnson (2009) because a subset of gapping cases is derived via ATB movement, and it is compatible with Potter et al. (2017) that gapping has two sources. However, it differs in that the derivation of the ambiguity is hybrid: the CP source undergoes ellipsis, whereas the VP source undergoes movement. If the analysis is in the right direction, it provides support for phrase structure grammars that assume an abstract level of representation for non-canonical coordination.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The research was funded by Deanship of Scientific Research (DSR), The University of Jordan.

Notes on contributors

Eman Al Khalaf

Eman Al Khalaf is an Associate Professor of Linguistics at the University of Jordan. She earned her Ph.D. in Linguistics and Cognitive Science from the University of Delaware. Her research delves into syntax and semantics, employing advanced methods, particularly experimental techniques in linguistics, to explore and contribute to the forefront of linguistic inquiry.