2.1. A NAP and its anchor neither are merged in the same DP nor are in the same clause

In this section, I show that a NAP and its anchor are not in the same DP or same clause.

First, as reported by Doron (Reference Doron, Buchalla and Mitwoch1994: 56), Heringa (Reference Heringa2011: 95–101, 180–182), and Onea & Ott (Reference Onea and Ott2022: 361), an adverb may occur between a NAP and its anchor, as seen in (3).

An adverb, by definition, occurs in a non-nominal context. Since no adverb may occur between two nominals that form a bigger nominal (e.g., a drug (*once) addict), the examples in (3) show that a NAP and its anchor do not form a nominal constituent. In such examples, as stated by Heringa (Reference Heringa2011: 180), “the presence of adverbials reveals the predicative or even clausal status of the apposition.”Footnote ³ This adverb fact challenges Pott’s (Reference Potts2005: 97) and de Vries’s (Reference Vries2006: 238, 2012: 154) claim that a NAP and its anchor form a nominal constituent.

Also, in examples like (4), the anchor and the NAP are not next to each other. If they were base-generated in a definite DP, their separation would violate the definite DP island.

We thus get the following conclusion:

(5) implies that a NAP is not a nominal conjunct in any assumed nominal coordination (contra de Vries Reference Vries2006). An anonymous reviewer challenges the ungrammaticality of (5) by the apparent possibility of its topicalization and passivization. Later, in (50) and (52), I show that the adjacency of an anchor and a NAP in passive and topicalization constructions does not challenge (5) and explain where the illusion comes from.

Second, the formal properties of a NAP are invisible to the clause where it surfaces. As pointed out by Heringa (Reference Heringa2011: 9), “if some element in the matrix agrees with the appositional construction, it only agrees with the anchor.” In (6a) (Onea & Ott Reference Onea and Ott2022: 362), only the singularity of the anchor the loot, but not the plurality of the NAP fourteen pure diamonds, triggers the number agreement on the auxiliary in the AnC. (6b) (adapted from Potts Reference Potts2005: 124) shows the same point (also see Burton-Roberts Reference Burton-Roberts, Asher and Simpson1994: 185).

Third, neither it-cleft (Heringa Reference Heringa2011: 110) nor pseudo-cleft may target a NAP. See (7).

Fourth, according to Onea & Ott (Reference Onea and Ott2022), a NAP and the AnC are independent illocutionary objects. In (8a), the matrix clause is a question but the p-NAP is not the part of the question (Onea & Ott Reference Onea and Ott2022: 361). In (8b), the AnC is not a question, but the p-NAP is.

Different illocutionary forces are represented in different clause types in syntax. Therefore, a NAP and its anchor cannot be hosted in the same simple clause.

The above three points further show that a NAP and its anchor are not base-generated in the same simple clause. We thus get a more general constraint:

2.2. A NAP and its anchor do not form a clause-containing DP

It is also unlikely for an assumed NAP-containing clause and the anchor to form a complex DP. In this subsection, I show that the form in (10) is also ill-formed.

First, the relation between an anchor and its NAP is not that between a relational noun and its licensor, and thus a NAP cannot be contained in a complement clause of an anchor.

Second, a NAP functions as a reduced appositive clause. But this clause is associated with an expression bigger than its associated DP (or called Head), i.e., the anchor. In (11), if the clause containing the NAP [a married woman] were merged with the anchor Mary alone, forming a complex subject, it would be in the scope of believed. But this is not the intended reading. More plausibly, (11) should be interpreted as the combination of two clauses: the AnC [John believes that Mary is still single] and another clause where the NAP [a married woman] appears as a predicate and its null subject is anaphoric to Mary in the AnC. See Potts (Reference Potts2005: 114f) for more discussion of the fact that a NAP is not in the scope of any AnC element.

Third, the Head of a restrictive relative c-commands the relative, but that of an appositive clause does not (Emonds Reference Emonds1979); like the latter, an anchor does not c-command its NAP. In (12a), the quantifier Head every assistant can bind his in the restrictive relative; but in (12b), the same Head cannot bind his in the appositive clause (Emonds Reference Emonds1979: 236). Similarly, in (12c, d) (Potts Reference Potts2005: 122), if the clause containing the NAP [{an/the} experienced adventurer] were merged with the anchor every climber, forming a complex DP, it would be c-commanded by the anchor; and the NAP should be qualified as a variable. But this is not the case.

Fourth, in (13b), for example, the NAP and its anchor co-refer. If they were in the same complex DP where the anchor c-commands the NAP in the reduced appositive clause, Condition C would be violated.

Instead of the complex DP analysis, if an appositive clause and its associated DP are each hosted in a separate clause, and the two clauses form a bigger clause (Ross Reference Ross1967; Emonds Reference Emonds1979; or any other versions of a bi-clausal analysis), and if a NAP is analyzed as a reduced appositive clause, the above challenges disappear.Footnote ⁴

Fifth, in Onea and Ott’s (Reference Onea and Ott2022) well-argued bi-clausal analysis, it is an AnC, not an anchor alone, that denotes an implicit question, which is answered by the NAP. From this perspective, it is also unlikely for an anchor and its NAP to form a complex DP. I have thus explained why (10) is ill-formed.

A NAP is independent of the other parts of the NAP construction. Syntactically, a parenthetical is not an argument of any element of the AnC. It is thus “invisible” to the internal structure of the AnC (cf. de Vries Reference Vries, Dehé and Kavalova2007: 207). The relation between a NAP and its AnC is like that between two sentences in discourse (cf. Ott Reference Ott2016: 488). The possible prosodic breaks next to a NAP are expected from this relation. As claimed by Kayne (Reference Kayne, Epstein and Seely2002: 139; also his Reference Kayne2026: 3), in a discourse example such as John is famous. He’s smart, too, “the two sentences in question form a single syntactic entity, akin to coordination.” In §3 through §6, I probe the issue of how to form a p-NAP construction, which is a complex CP, not a complex DP or simple CP, and the issue of how to exclude unacceptable forms of such constructions in syntax.

4.1. Background: a generalized non-argument-taking structure

Zhang (Reference Zhang2023) proposes a general structure for a non-argument-taking relation, covering both coordination and modification relations. I introduce this structure in this subsection.

In the minimal form of coordination, there are two conjuncts. One of them is the complement of a functional element, called Junct (J). J can be realized by a coordinator. J has no categorial features. But like all other functional heads, it must have a complement. The combination of J and its complement is called J-set, which also has no syntactic category, since the complement of an element does not project. In this respect, both J and a J-set are like a root. Then, it is the other conjunct, which is the sister of J-set, that decides the category of the whole coordinate complex. The conjunct that is the complement of J is called internal conjunct, and the higher one is called external conjunct. Like some other functional heads, J can be null.

In this structure, the J-set empirically functions as an adjunct of XP, although a well-recognized adjunct or modifier has its category, but a J-set does not. Moreover, J does not get a category from YP, since the latter is its complement. If J got its category from its complement, XP would be merged with YP (Chomsky’s Reference Chomsky2013 Pair Merge). To solve the labeling problem of the assumed Pair Merge, neither of the two ways proposed by Chomsky works here, because the two phrases do not have to share features and neither moves away (see Zhang Reference Zhang2023: 10 for a comment on Chomsky’s analysis of coordination). In the J-theory, a generalized categorization occurs, instead of Pair Merge.Footnote ⁵

The main arguments for this J-theory of coordination are the following. First, a coordinator must occur with at least two conjuncts. This is shown in (24). This follows the obligatory co-occurrence of the complement of J and the categorizer of the J-set. Second, the combination of a coordinator and a conjunct is syntactically optional. This is shown in (25). In the J-theory, a J-set depends on its categorizer, but not vice versa. Third, unlike arguments, conjuncts can be unlimited in number, syntactically. This is shown in (26). In the J-structure, multiple layers of J-set are possible, before the complex J-set is categorized by another element. Fourth, it is the highest conjunct, rather than the coordinator or any other conjunct, that decides the category of the whole coordinate complex. This is shown in (27) (cf. Sag et al. Reference Sag, Gazdar, Wasow and Weisler1985). In (27a), the high conjunct (i.e., the external one) is a nominal, which satisfies the c-selection of the preposition on. (27b) shows that on does not take a declarative clause as its complement. (27c) shows that if the high conjunct is a declarative clause, the c-section of on cannot be satisfied and thus the construction is not acceptable.

For a modification relation, a modifier has the same syntactic position as an internal conjunct, and the modified element has the same syntactic position as an external conjunct. The above arguments for the J-theory are also found in modification constructions: the occurrence of a modifier depends on the occurrence of a modified element; a modifier is optional syntactically; unlike arguments, modifiers can be unlimited in number; and it is the modified element that decides the category of the whole modification construction.

J can also be realized by a modification marker, which is found in some languages, but not in languages such as English (Rubin Reference Rubin1994, Reference Rubin2003). In the two Mandarin Chinese examples in (28), de is a modification marker, glossed as MOD (the different written forms, 的 and 地, are contextual variations of the same functional element; see Zhang Reference Zhang2023: 34).

There are many structural similarities between a coordinator and a modification marker. For example, neither has a theta relation with another element; both are a closed set; both must be associated with at least two elements (one introduces a modifier to a modified element, and the other introduces one conjunct to another conjunct); neither has category features; both can be silent in a language or in certain constructions of a language; both occur in a linker position morphologically, e.g., if there are only two elements, they occur between them (Zhang Reference Zhang2023: 20); and when they occur with more than two linked elements, they both must be next to the right-edge element. The last point is seen in the coordination construction in (29) and the modification construction in (30) (Zhang Reference Zhang2023: 35).

Zhang (Reference Zhang2023) proposes a general syntactic structure for a non-argument-taking relation.

In this J-structure, the J-set, as well as J, is categoryless. Like a root, it must be categorized. It is XP that categorizes the J-set and decides the category of the whole construction. It is also the category of XP that satisfies the c-selection of any higher element that takes the construction as its argument. This analysis explains why an internal conjunct needs an external conjunct, and why a modifier needs a modified element, syntactically.

In (31), the category of the complement of J (YP), as well as that of the categorizer of the J-set (XP), is underspecified. The exact choice depends on the constructions.

Moreover, the relation between the complement of J (YP) and the categorizer of a J-set (XP) is not argument-taking. But what kind of non-argument-taking relation it should be is underspecified. The exact relation depends on the context and J’s realization. For example, if XP and YP are both DPs and J is realized by and, the whole complex is a DP coordinate complex, which can trigger a plural agreement. But in other situations, the whole complex does not need to be a coordinate complex, and does not have to trigger a plural agreement.

Furthermore, in this theory, what is crucial is the hierarchy. At each level in (31), the surface order of the sisters varies. For example, a modifier can either follow or precede its modified element, as seen in (32a) and (32b), respectively. If a modifier is contained in a J-set, such examples show that a J-set can follow or precede its categorizer.

However, the position of the element that realizes J is always between the complement of J and the sister of the J-set. Generally, coordinators occur between conjuncts. Zhang (Reference Zhang2024: 49) uses the Mandarin Chinese examples in (33) to show the position of modification markers. In (33a), the modifier hěn kāixīn ‘very heartily’ follows the modified expression xiào ‘laugh’, but in (33b), the same modifier precedes xiào; however, the enclitic modification marker de always occurs between the modifier and the modified element.

The J-structure introduced in this subsection paves a way for us to analyze the syntactic relation between a NaC and its AnC, to be discussed in the next subsection.

4.2. A NaC as the complement of J

I claim that a NaC, like a modifier or an internal conjunct, is the complement of J, and an AnC is the categorizer of the J-set. Then, syntactically, the J-set that contains a NaC is dependent on the AnC. This explains why a NAP needs an anchor, syntactically. Moreover, a NaC can be treated as a reduced appositive clause (cf. §2.2). Such a clause cannot be Headless (Emonds Reference Emonds1979: 231). Appositives need their Heads, and similarly, NAPs need their anchors.

This analysis explains the possible conjunct property and possible modifier property of a NAP. In the J-structure, whether an element is an internal conjunct or a modifier makes no structural difference: it is just the complement of J. On the one hand, a NAP may occur with a coordinator, or a cluster that is composed of a coordinator and a particle, as seen in (34).

In (31), a coordinator realizes J. The coordinator or the cluster of a coordinator and a particle in (34) can be a realization of J. The existence of the constructions in (34) shows the conjunct property of a NaC. Note that as pointed out by a reviewer, while the conjunction of two DPs can trigger a plural agreement (John and Pete are/*is…), the occurrence of a coordinator between a singular anchor and its NAP never triggers a plural agreement (the teacher, a genius, is/*are; the teacher, John, is/*are). In my analysis, this is because the NAP is hosted in a clause, and the anchor and its NAP do not form a DP (see (5)). On the other hand, although a NaC is not an argument of any lexical element of the AnC, it enriches the meaning of the whole construction, and it usually rejects a coordinator, as a modifier does. From the perspective of this J-theory, a NaC can be viewed as a conjunct, if it may occur with a coordinator; and it can also be viewed as a kind of modifier. In either way, it is just the complement of J.

To understand the syntax of NAP constructions, where NaCs are conjunct-and-modifier-like elements, we can consider the syntax of a kind of conjunct that functions as a modifier, the conditional conjunct, such as the left conjunct in (35).

A NAP is like such a modifier-like conjunct. Also, in English, a clausal modifier can precede or follow the matrix clause in certain constructions, as seen in (32). A modifier is the complement of J. If a NaC is also the complement of J, and the AnC is the sister of the J-set. I propose that the base-order of the two clauses of a NAP construction is like that in (32b). As shown in (36), the AnC precedes the NaC. In this new understanding of NAP constructions, a NAP is a nominal in a clause that is the complement of J. In this NaC, there is a silent pronominal (see (19)). This pronominal is anaphoric to a nominal in the matrix clause (AnC). An anchor is a nominal in the AnC that the silent pronominal in the NaC is anaphoric to. Thus, although J syntactically links two clauses together, the semantic dependency between a NAP and its anchor is represented by the anaphoric relation between the nominals contained in the two clauses.

In this structure, the pro takes a preceding DP (i.e., the anchor) as its antecedent. This is an unmarked order in pronominalization, as pointed out by a reviewer. I represent this unmarked order in both the base and the surface positions of the NAP and its anchor.

My proposal made in this section is that a NaC is the complement of J, and the J-set built is categorized by the AnC in their base-positions. The non-argument-taking J-structure explains why a NaC depends on an AnC syntactically. We move on to the issue of the surface position of a NAP in the next section.

5.1. The right-edge NAPs

A NAP may occur at the right-edge of a construction, as seen (38). Also, a NAP and its anchor do not need to be next to each other, as seen in (38b) (Burton-Roberts Reference Burton-Roberts, Asher and Simpson1994: 184, see Onea & Ott Reference Onea and Ott2022: 370 for similar examples).

Onea and Ott (Reference Onea and Ott2022: 383) claim that a construction like (38b) is possible for s-NAPs only, not p-NAPs. However, p-NAP examples like (39) are available.

The right-edge NAPs are also called Afterthoughts in Ott and de Vries (Reference Ott and De Vries2016: 643). They are different from the non-right-edge NAPs in that they can be supplied by a second speaker (Ott & de Vries Reference Ott and De Vries2016: 649, 684). In (40), speaker A’s utterance is followed by speaker B’s utterance, which functions as a p-NAP.

Because of this difference, it is unlikely for the right-edge NAP to be derived from a structure of a non-edge NAP construction by any movement operation, including the operation that is for extraposition (cf. Heringa Reference Heringa2011: 15). Instead, I propose that the NaC follows the AnC in their base-positions, and the NaC is the complement of J, as in (36). Following Ott and de Vries (Reference Ott and De Vries2016), I analyze the NaC of such a p-NAP as a reduced copula clause (deletion of a pronoun and a copula in Ott and de Vries Reference Ott and De Vries2016: 682, and a null proform and a null copula in my approach, as proposed in §3). The structure of (39) is thus (41) (I use the subscriptions a and n to label AnC and NaC, respectively).

5.2. The non-right-edge NAPs

If a NAP does not occur at the right-edge of the construction, the surface order can be derived by movement, including meaningless movement. Movement can be meaningless because it does not need to be feature-driven, as claimed by Chomsky (Reference Chomsky2013, Reference Chomsky2021), Moro & Roberts (Reference Moro and Roberts2024), Cinque (Reference Cinque2023), and Syed & Simpson (Reference Syed and Simpson2024), among others. This type of movement has been extensively argued for in Cinque (Reference Cinque2025). Cinque (Reference Cinque2025: 44) states that certain meaningless movements affecting linear order must take place in narrow syntax (I-language). I thus assume that meaningless movement can derive various NAP constructions.

I assume that the base structure of a NAP construction is (36) consistently, as in (41). I label this structure ⓪. Based on the structure, maximally two consecutive steps are necessary: ① the NaC moves to the left-edge; ② if the anchor is a clause-initial DP, it moves out of the AnC; otherwise, ③ the core structure of the (reformed) AnC undergoes a remnant movement, after all post-anchor elements are fronted in the AnC. If a coordinator occurs, as a linker, it always surfaces to the immediate left of the NaC in the final representation.

Operation ① is similar to the movement of a modifying clause, and operation ② or ③ is similar to the movement of an element out of a modified clause. If a NaC and its AnC are treated as internal and external conjuncts, respectively, operation ① looks like a conjunct movement and operation ② or ③ looks like extraction from a conjunct. Both operations violate Ross’s (Reference Ross1967) Coordinate Structure Constraint (CSC). See Zhang (Reference Zhang2023: 42–44) for a review of the literature that discusses how the CSC can be violated.

The details of the proposed derivations are below.

Construction A: The anchor is a clause-initial DP of the AnC. For example, the NAP construction in (42) is derived by the steps in (43).

In (43), neither of the movements changes the category of the hosting expression, and thus the hosting expression projects. This is because “it is impossible for Move to target K raising α, then projecting α rather than K.” (Chomsky Reference Chomsky, Campos and Kempchinsky1994 [Reference Chomsky1995: 66]; cf. Cecchetto & Donati Reference Cecchetto and Donati2015). Both AnC and NaC are CPs and the final result is also CP.

Construction B: The anchor is not a clause-initial DP of the AnC. For example, the NAP construction in (44) can be derived in the steps in (45).

To derive a little more complicated NAP construction, (46), we need first to front the adverbial today to the left of the AnC, as in (47b). Then, front the other adverbial at the pub, to the left-edge, as in (47c). Thus, all post-anchor elements are fronted in the AnC. In another working site, a NaC is built, and it is merged with J. Then, ⓪, merge the J-set with the output of (47c), as in (47d). Then, ①, moves the NaC, as in (47e); ③ part of the AnC undergoes a remnant movement, stranding at the pub today, as in (47f).

If a NAP construction starts with a C-domain element, e.g., the adverb unfortunately in (48), the element can be merged after the basic steps (e.g., after Step f in (47)).

In a NAP construction that has a coordinator, such as those in (34), the coordinator, as a J-element, is next to the surface position of the NaC in the final representation.

If an A-movement or A-bar movement occurs in an AnC, and if the left-edge DP functions as an anchor, after it moves to the left of the NaC, the construction gives an illusion that it is the combination of an anchor and a NAP that has undergone the movement. In (50), we can see how such an illusion is created for the passive construction in (49); and in (52), we can see how such an illusion is created for the topicalization construction in (51).

In deriving these various NAP constructions, no construction-specific operation or functional category is necessary. My conclusion of this section is that various NAP constructions can be derived by the normal syntactic operations.

7.1. Similar properties of s/eq-NAPs and p-NAPs

Different types of NAPs share properties. The properties reported in §2 are all found in s/eq-NAP constructions. They all show that a NAP and its anchor are hosted in different clauses, not forming a complex DP. For example, first, like a p-NAP, an s/eq-NAP may also follow a sentential adverb, but reject a non-sentential adverb (see the p-NAP examples (3) and (14)).

Second, like p-NAPs, the formal properties of other kinds of NAPs are also invisible to the clause where they surface. In (57), the feminine ihren “her” in the clause agrees with the feminine anchor die Lise, not with the neuter NAP das arme Weib “the poor wife” (see Heringa Reference Heringa2011: 9 for the source of (57)).

Third, neither it-cleft (Heringa Reference Heringa2011: 110) nor pseudo-cleft targets a NAP. Both the p-NAP examples in (7) and the s-NAP examples in (58) show this. Eq-NAPs behave the same.

Fourth, like a p-NAP, an s/eq-NAP may also be in a different illocutionary object from its AnC (Onea & Ott Reference Onea and Ott2022: 361). In (59), the AnC is a declarative but the s-NAP is in a question.

Fifth, in an s/eq-NAP construction, the NAP and its anchor co-refer; if they were generated in the same clause, Condition C would be violated. The acceptability of examples like (60) challenges the same clause analysis.

The conclusion that an s/eq-NAP and its anchor are hosted in different clauses is different from Griffiths’s (Reference Griffiths2015) claim that an anchor and its s-NAP form a coordinate nominal. It is also different from the conclusion that a NAP and its anchor form a nominal in general (e.g., de Vries Reference Vries2006).

7.2. The copular clause structures of s/eq-NaCs

Copular clauses can be predicative, specificational, and equative, seen in (61).

“The intuition about predicational clauses is that they predicate a property of the subject referent” (Mikkelsen Reference Mikkelsen, von Heusinger and Portner2011: 1807). (61a) tells us the novelist’s property of E. M. Forster. However, “The term specificational derives from the intuition that these clauses are used to specify who (or what) someone (or something) is, rather than to say anything about that person (or entity).” (Mikkelsen Reference Mikkelsen, von Heusinger and Portner2011: 1809) (61b) tells us who wrote a particular novel, not something about that person. According to Moro (Reference Moro1997) and den Dikken (Reference den Dikken2006), the post-copula nominal in a specificational clause is base-generated at an argument position, and the pre-copula nominal is a raised predicate. Heycock (Reference Heycock2025 and the references therein), however, argues that the pre-copular nominal in specificational clauses denotes an individual concept, instead of a predicate (cf. Moro Reference Moro1997). As for an equative copular clause, the pre- and post-copula nominals refer to the same individual. In (61c), the two names refer to the same famous scholar of formal semantics.

Heringa (Reference Heringa2011: §3.7) claims that different types of NAPs correlate with different types of copular clauses. The post-copula nominal in a predicational clause is not referential, nor is a p-NAP. In contrast, the post-copula nominal in a specificational clause is referential, and so is an s-NAP. Also, the post-copula nominal in an equative clause is referential, and so is an eq-NAP. Therefore, the s-NAP construction in (62) (= (1b)) contains two clauses, AnC and NaC. Similarly, the eq-NAP construction in (63) (= (1c)) also contains two clauses.

I have found one more fact to support the copular clause analysis of the s/eq-NAPs. Only the sentential adverbs that can occur in a copular clause may occur with an s/eq-NAP. This is the same constraint on p-NAPs (see §3). Adverbs such as intelligently, gradually, eventually, rarely, and frequently cannot occur in a copular clause and they cannot occur with a NAP, either. (64a) has an s-NAP and (64b) has an eq-NAP.

I discuss the integration of such NACs into the whole construction in the next subsection.

7.3. Deriving s/eq-NAP constructions

The J-structure analysis proposed for p-NAP constructions can apply to non-p NAP constructions. The possible occurrence of a coordinator in the eq-NAP examples in (65) shows that the J can be realized by a coordinator in the structure.

Like the NaC for a p-NAP, the NaC for an s/eq-NAP also has a null copula and a pre-copula null proform. In an s/eq-NAP construction, both the NAP and its anchor are referential, and the null proform is co-referential with the anchor. Thus, a general structure for the NaC of all types of NAPs is as in (36). Its simple form is (66).

We first consider the right-edge s-NAP construction (67).

Like the right-end p-NAPs, the right-end s-NAPs can also be supplied by a second speaker. The German example in (68) is from Ott and de Vries (Reference Ott and De Vries2016: 649). Speaker A utters an AnC and Speaker B utters a NaC.

The NaC exhibits the properties of specificational copula clauses. Also, all such right-edge NAPs reject adverbs that are also rejected by copula clauses (e.g., eventually). I thus assume that the syntax of the right edge s-NAPs is similar to the syntax of the right edge p-NAPs discussed in 5.2. The only difference is that the null copula in the NaC is a null specificational copula. The same analysis applies to right-edge eq-NAPs, such as those in (65).

The structure of (67) is shown in (69):

The derivation steps for the non-edge s-NAP construction in (70) are in (71), which are similar to the derivation steps in (47).

Our generalization in (54) is also true of non-p NAPs, i.e., a NAP either follows its anchor immediately or occurs at the right-edge of the construction. In the unacceptable (72), the intended anchor is an old friend and the s-NAP is Skylar Garcia; neither of the two positions of the s-NAP is next to the anchor and neither shows up at the right edge position of the construction. I have explained the generalization by my proposed analysis of p-NAPs in §6. The same explanation applies to other types of NAPs.

In the previous subsections, I have proposed the same syntactic derivation of s/eq-NAP constructions as p-NAP constructions.Footnote ⁶ This is different from a clause-copying and deletion analysis for s-NAP constructions in Onea and Ott (Reference Onea and Ott2022). See the next subsection.

7.4. Against a clause-copying and deletion analysis of s-NAPs

In Onea and Ott (Reference Onea and Ott2022), although the source (i.e., NaC) of a p-NAP is recognized as a copular clause (with an elided subject and copula), as shown in (74a) for (73a), the source of an s-NAP (their r-NAP; see footnote 1) is analyzed as a clause that is “propositionally reduplicative” (as called by Burton-Roberts Reference Burton-Roberts, Asher and Simpson1994: 187) to the AnC, followed by deletion, as shown in their (74b) for (73b) (also see Burton-Roberts Reference Burton-Roberts1975: 406). In (74b), the NaC has an elided predicate.

Onea and Ott (Reference Onea and Ott2022) use a case pattern contrast in German to argue for this clause-copying and deletion analysis (I call it “clause-copying analysis” henceforth) of s-NAPs. In (75a), the p-NAP is nominative, which is different from its dative anchor; however, in (75b), the s-NAP is dative, which matches its dative anchor (Onea & Ott Reference Onea and Ott2022: 362).

However, first, examples where a p-NAP shows the same case with its anchors are reported in Heringa (Reference Heringa2011). On the other hand, examples in which an s-NAP shows a different case from its anchors are also reported (Heringa Reference Heringa2011: 182; see (76)). The existence of such examples shows that it is risky to use the German case pattern to argue for the clause-copying analysis.

Second, the case contrast pattern assumed in the copying analysis is not seen in some other languages. In Russian, the case of a NAP always matches that of its anchor, regardless of the type of the NAP (Heringa Reference Heringa2011: 199). (77a), has a p-NAP and (77b) has an s-NAP. They are both dative, as their anchors. A similar pattern is seen in Czech (Heringa Reference Heringa2011: 206).

The language internal inconsistency and the cross-linguistic variation of the case patterns of various types of NAPs indicate that morphological case is not always a reliable diagnostic for syntactic structures. The case variation comes from two possible sources: the default case of the language (Schütze Reference Schütze2001) and the parallelism between a NAP and its anchor. See van Riemsdijk (Reference Riemsdijk, Heny and Richards1983: 247) for the default (or called unmarked) case analysis of case-disagreement between a NAP and its anchor in German. On the other hand, recall that a NAP is the only overt nominal in the complement of J, and it is semantically related to the anchor, which is hosted in the sister of the J-set. In my analysis, the complement of J and the sister of the J-set are the positions of two conjuncts. Conjuncts tend to show parallelism. It is thus possible for the case of a NAP to be parallel to the case of its anchor. Thus, case agreement between a NAP and its anchor might come from this parallelism. But I leave this conjecture for future research. See Heringa (Reference Heringa2011) for another account of the complexity of the NAP cases. Anyway, I do not think morphological case alone is a decisive factor for syntactic analyses.

Onea and Ott (Reference Onea and Ott2022: 362) also argue that the clause that contains an s-NAP (NaC) is parallel to the matrix clause (AnC), but such a parallelism is not seen in p-NAP constructions. Thus, they claim that clause-copying occurs in s-NAP constructions, but not in p-NAP ones. A contrast is seen between the s-NAP construction in (78a) and the p-NAPconstruction in (78b).

Their account is that if the s-NAP in (78a) comes from (79a), its acceptability is explained, since the bound pronoun his is bound by the quantifier every inmate before the deletion. But my proposed copular construction in (79b) can also be a possible NaC, where the pro, which has an E-type pronoun reading, is anaphoric to the anchor a (particular) relative in the AnC. The pro in the NaC in (78b), however, as pointed out by a reviewer, cannot get an E-type reading, since Ames is a proper name.

Thus, the acceptability contrast between (78a) and (78b) does not mean that the s-NAP must come from a copied clause. The unacceptability of the p-NAP construction in (78b) is irrelevant to the analysis of s-NAP constructions. Also, consider (80a), where the s-NAP comes from (80b) in the clause-coping analysis. (80b) violates Condition C and thus is not acceptable. The clause-copying analysis predicts that (80a) should also be bad. But (80a) is perfect. In contrast, the copular clause in (80c) is fine, and if (80c) is the NaC of the s-NAP in (80a), the latter’s acceptability is explained.

In our opinion, if a theory predicts a construction is acceptable but the construction is not acceptable, the theory is not directly challenged (since the unacceptability may come from other factors); however, if a theory predicts a construction is unacceptable but the construction is acceptable, the theory is surely challenged.

P-NAPs and s-NAPs are semantically different: the former is predicative and the latter is not. This contrast explains some contrasts stated in Onea and Ott (Reference Onea and Ott2022: 362). For example, the German adverb nämlich “namely” occurs with s-NAPs, but not p-NAPs. Different types of NAPs are compatible with different types of adverbials. Heringa (Reference Heringa2011: 56f, 147) extensively discusses the compatibility of adverbials with various types of NAPs. This compatibility contrast does not lead to a clause copying analysis of s-NAPs.

My preference for the copular clause analysis of NaCs to the clause copying analysis is for the following reasons, in addition to the avoidance of the Condition C issue in (80b). First, if a NAP, as argued by Onea and Ott (Reference Onea and Ott2022), answers a potential question raised by its AnC, it is more likely to be in the form of a copular clause than in the form of a near copy of the AnC. For example, the potential question raised by the AnC in (73b) can be Who is the brother of Mary you are talking about? One may answer He is Emre, rather than Emre has a girlfriend. The former, but not the latter, gives the right answer to the who-question. The question under discussion (QUD) of the implicit dialogue is about the property or identification of the individual denoted by the anchor alone, and thus, a NAC, as an answer to the potential question, should be in the form of a copular clause, rather than a near copy of the AnC.

In this respect, there is no contrast between a p-NAP and an s-NAP. In both cases, the proposition expressed by a NAP is associated with the anchor alone, which is an argument of an expression in the AnC, excluding the other parts of the AnC. In (73a), as claimed by Onea and Ott (Reference Onea and Ott2022) in (74a), the NaC does not encode “won the 2002 Tour de France.” Similarly, in (73b), the NaC should not encode “has a girlfriend.”

Second, if an s-NAP contains an adverb, the adverb does not scope over the non-anchor part of the AnC. The s-NAP construction in (81) entails the meaning of the AnC, and the appropriate NaC should be NaC1, rather than NaC2. Thus, the full form of the NaC does not contain “lost his wallet” at all. But in the clause copying analysis, since the adverb precedes the deleted verbal expression in NaC2, it may scope over the verbal expression. The analysis wrongly predicts that NaC2 is a possible reading for (81).

In this respect, again, there is no contrast between a p-NAP and s-NAP. In both cases, the proposition expressed by a NaC is associated with the anchor alone, excluding the other parts of the AnC. In (82), the p-NAP NaC probably an Arkansan expresses the meaning of NaC1, rather than that of NaC2. Thus, the full form of the p-NAP NaC, as claimed by Onea and Ott (Reference Onea and Ott2022) in (74a), does not contain “won the 2002 Tour de France” at all.

Third, in the clause copying analysis of the NaC of an s-NAP, the adverb with a NAP does not have to be those that are compatible with a copular clause (see (18) and (64)). Thus, my copular analysis correctly rules out eventually in (83a), as shown in the NaC in (83b), but the clause-copying analysis wrongly predicts that (83c) is a possible full form of (83a). The first clause in (83c) can be a well-formed AnC, and the second clause is its near copy, functioning as the full form of the NaC, before deleting the parts that are identical to the AnC. Then, we can see that eventually is predicted to be possible in the NAP construction, in this analysis.

Fourth, in Heringa and my copular clause analysis of the NaCs of various types of NAPs, such clauses contain a pre-copula pro, which is anaphoric to the anchor directly. In this respect, again, there is no contrast between a p-NAP and s-NAP: a pro occurs in both cases. In Onea and Ott’s (Reference Onea and Ott2022) p-NAP example in (73a), the pronominal is the deleted pronoun he, as shown in (74a). In their proposed structure in (74b) for the s-NAP example in (73b), however, there is no pronominal that is anaphoric to one of Mary’s brothers. How does the speaker express the intended coreference between the two referential expressions in the two propositions, i.e., one of Mary’s brothers and Emre? The question-answer relation alone between two clauses cannot establish a coreference relation between two referential nominals in the two clauses. The copular approach, but not the clause coping and deletion approach, explains this anaphoric relation directly.

Fifth, it is not clear how the clause-coping and deletion approach rules out the ungrammatical constructions in which the NAP precedes its anchor. In this approach, (84b) is for (84a) (also see (74b) for (73b) above).

In the assumed NaC in (84b), the pre-NAP part (I met) has undergone a forward deletion, and the post-NAP part (yesterday) has undergone a backward deletion. But it is not clear how the NaC shows up in the middle of the AnC. For a p-NAP, no such forward and backward deletion occurs. We have seen Onea and Ott’s (Reference Onea and Ott2022) (74a) above. But again, it is unclear how the NaC shows up in the middle of the AnC. Also, in this clause-copying approach, we see no clear syntactic reason to exclude the syntactically ill-positions of the NAPs in (53). Therefore, the assumed “linear Interpolation” is quite vague (also see Griffiths & de Vries Reference Griffiths and De Vries2019 for a challenge to the linear Interpolation).

Note that what I disagree with the clause-copying approach is just the form of the NaC for an s-NAP: it is a specificational copular clause in my approach, but a near copy of the AnC in the latter approach. So, even if we assume that the NaC is a near copy of the AnC, my syntactic analysis still can accommodate the assumption: a NaC is the complement of J, and the various NAP constructions can be derived in syntax, without any “linear Interpolation.” I thank an anonymous reviewer for pointing this out.

In conclusion, as claimed by Heringa (Reference Heringa2011), various types of NAPs correlate with various types of copula constructions. I have given empirical arguments to support the claim and shown problems of a non-copular analysis of s-NAPs. In this generalized copula clause analysis of NaCs, the differences between different types of NAPs come from the differences between various kinds of copula clauses. Although the syntactic interaction of an NaC (as a copular clause) with the AnC is the same, the syntactic and semantic formations of the various types of copular clauses themselves remain to be worked out in future research.