Niuean raising is copy-raising

As mentioned earlier, there is a tension between treating Niuean raising as a movement operation, on the one hand, and treating ergative case as the spell-out of a case feature valued in competition with a lower DP. In this Appendix, we propose a detailed analysis of Niuean raising that reconciles the full array of facts with the results obtained thus far.

To illustrate the particular challenge our analysis faces in the context of Niuean raising, recall that we are taking morphologically ergative and absolutive/genitive case to be the spellout of a valued and an unvalued case feature, respectively. As we discussed in the context of conversion to genitive, this association of ergative with case valuation predicts that once a DP has been assigned ergative case (had its case feature valued), it should be impossible for that DP to surface as an absolutive/genitive, independent of the particular operations it undergoes. This captures the fact that ergative subjects never undergo conversion to genitive, despite being accessible to the φ-probe on n and thus capable of moving into the nominal domain (of course, we have yet to establish that φ-Agree in Niuean can target marked case). This same inability to transition from marked to unmarked case, however, presents a problem in the context of raising. Recall that raised DPs associated with an transitive subject gap surface in absolutive case, not the ergative they would be expected to receive in the embedded clause subject position. This is despite the abundance of evidence, reviewed in Section 2 and fully documented by Seiter (1980), that the raised DP is an argument of the lower predicate. This is especially problematic given that we have argued that subject case assignment occurs in situ and is uninfluenced by the presence or character of the functional structure above vP. The challenge we face can be summarized thus: raised underlying subjects have a clear syntactic salience in the lower clause, yet their case marking is inconsistent with them appearing in the embedded subject position.

The solution we propose is that Niuean raising is really an instance of copy raising (Rogers 1971; Ura 1994; Potsdam and Runner 2001; Asudeh 2004; 2012; Rezac 2004, a.o.). Following Potsdam and Runner (2001) and Rezac (2004), in particular, we take copy-raising to be a construction where a A-chain is formed between a higher DP and a lower coindexed pronoun. Crucially, the relevant chain must show canonical A-properties, chief among which is that only the lowest link in the chain occupies a semantic argument position (the links in the chain share a θ-role). As an example, a variety of perception verbs in English license this construction, as diagnosed by i) the presence of a lower pronoun coindexed with the relevant DP, and ii) the ability of idiom chunks and expletives to participate, indicating that the higher position in the chain is not a semantic argument.

The relevant dependency is crucially limited by familiar locality constraints, so that objects and more deeply embedded subjects may not participate. Thus, while examples like (82) involving coindexed object or embedded subject pronouns are possible, as in (83), there is an absence of familiar A-chain behavior in such cases (see 84)), suggesting a different construction is at stake.

Returning to cases like (82), despite the A-chain like behavior, movement is not involved, so that the higher chain link is never present in the lower clause. In English, this primarily manifests in an absence of reconstruction effects (cf. raising and copy-raising in (85-a,b), (86-a,b)).

Returning to Niuean, the raising construction shows the key characteristics of copy-raising outlined above. First, a Niuean raising chain may involve an optional downstairs pronoun in the gap site. This pronoun must crucially be from the personal series, not the separate resumptive series. The optionality of the relevant pronouns follows from the general ability to drop third person animate pronouns in Niuean when they are sufficiently salient in the discourse (Massam 2001: 236).

Similar pronouns are not licensed in other filler-gap situations, for instance, in argument relative clauses.

Next, as we have already seen, the raised DP in the construction under consideration is not a semantic argument of the higher predicate, which is an indication of an A-chain. We repeat the crucial data involving idiom chunks below.

Finally, the raised DP appears to have never been present in the embedded clause. The ability to “raise” ergative DPs to absolutive subject position constitutes one of the main arguments for this point, although there is independent evidence from the absence of reconstruction effects.;

With the feasibility of a copy-raising treatment of Niuean raising thus established, the final task is to provide an analysis that is consistent with our previous findings. To this end, we adopt the φ-Agree-based approach to copy-raising developed by Rezac (2004). The account explicitly links φ-Agree with the establishment of a derived predicate, thus allowing the higher DP to be interpreted as the semantic argument of the lower predicate despite never appearing in the clause. To accomplish this, Rezac posits an “index feature” that is unvalued on φ-probes but valued and interpretable on DPs (like number). Index features are interpreted at LF as inducing λ-abstraction via a modified version of the predicate abstraction rule (Heim and Kratzer, 1998), as below.^[31]

(92)

Predicate Abstraction:

Let α be a tree immediately dominating two sub-trees, β and γ, such that β has an interpretable index feature [ix=1]. Then for any variable assignment a,

(Rezac 2004: 296)

Rezac then assumes the following condition on merge, which ensures that the interpretable index feature on a DP matches the Agree-valued index feature identifying the variable to be bound in its sister subtree.

CR is then derived in this system as follows: a pronoun bearing an interpretable index feature is merged as the subject of the embedded clause. A φ-probe (with an unvalued index feature) associated with the functional structure of the raising verb agrees with the subject pronoun, and the index feature is copied to the associated head. A DP whose φ-features and index feature match those on the relevant probe then merges into the structure, ensuring the establishment of an appropriate chain. The result is then interpreted via the modified predicate abstraction rule above, triggered by the index feature on the highest DP. Given that we have independently posited the existence of φ-features on Niuean v, we take it that this is the locus of copy-raising in the language, so that the ultimate derivation and LF are as below.^[32]

(94)

The association of copy-raising with φ-Agree ensures that this analysis is compatible with our previous findings. Finally, because we are assuming φ-Agree targets an underlying ergative pronoun, we posit that φ-Agree in Niuean may target DPs bearing dependent case, as in, for example, Nepali (Bobaljik 2008). This assumption is also necessary if we are to maintain our treatment of conversion to genitive, which holds that the syntactic operations responsible for the conversion – movement to Spec(nP) – are available for both ergatives and absoltuives, with the overt manifestation of this movement, namely genitive case, blocked for ergatives by other mechanisms.

(95)