# Case study 2: The definiteness effect

I mentioned this argument in 2; it follows arguments from Milsark (1977).^{[1]} It follows from a discussion of what can be in the complement position of a there be existential construction. Definite descriptions behave like "strong NPs" with respect to there be contexts. This is known as the definiteness effect. NPs with proportional quantifiers such as all and most are strong.

(20) a. ** There were all cows in the field.* b.

** There was the cow in the field.*

Weak NPs, on the other hand, such as those quantified by a numeral or a(n) are permitted in the existential construction, so the examples in (21) are fine.

(21) a. *There was a cow in the field.*

b. *There were three cows in the field.*

The data generalization captured by the difference between (20) and (21) argues in favour of classing the among the proportional quantifiers, especially when it is taken together with the data in 7.1, and the material presented next in 7.3.

# Case study 3: Specificational sentences

One area of research which has been particularly intractable has been the behaviour of specificational sentences, such as the examples in (22). The examples are from Patten (2010).

(22) a. *The murderer was John.*

b. *The one that murdered Sally was John.*

These sentences have a particularly complicated set of properties: they involve focus the postverbal name being focused here and they appear to involve nonstandard predication, where the NP immediately before the copula is a predicate and the NP after is its argument. Variants of the latter claim have been made in the literature, for example in Moro (1997) and Mikkelsen (2005). Other scholars, such as Heycock and Kroch (1999) have argued that specificational sentences are a kind of equative construction where both NPs refer, and are identified as co-referential, and the focus effects are established separately from the equative analysis.

**Figure 11. ****The murderer was John**

Both approaches have their downsides. The inverse predication approach of Moro and Mikkelsen involves a considerable amount of syntactic machinery bringing about movement. The equative approach of Heycock and Kroch requires postverbal movement in the semantic representation.

The alternative approach, which Patten (2010) offers, involves no syntactic machinery, and no LF movement. Furthermore, it does not require any kind of level or representation of information structure. Patten makes two claims that in the specificational sentence in (22a), the NP *the murderer* denotes a set, and the post-verbal NP *John* is the extension of that set. The analysis is only possible in an account where the is a quantifier; without this treatment of definite descriptions, it is not possible to have the linguistic analysis. Because definite descriptions express both existential quantification and universal quantification (over a restricted set) the analysis gets both the existential commitment of specificational sentences and their exhaustiveness.10

This last fact explains the data in (23).

(23) a. *The murderer was John.*

b. *The murderers were John, Fred, and Alfie.*

c. ** All murderers were John, Fred, and Alfie.*

d. ** All the murderers were John, Fred, and Alfie.*

The question posed by the examples in (23) is this: if the involves universal quantification over a set, why are the examples in (23c,d) ungrammatical, whereas the examples in (23a,b) are not?

The answer is simple: what makes the specificational sentence specificational is not just the universal quantification over the restricted set, but is also the existential commitment. All does not involve existential commitment, and so (23c) is ungrammatical. In the case of (23d), the story is more complex because of the *the* under all. Here, we can invoke the earlier discussion of proportional quantifiers in 6 above, where I pointed out that they necessarily involve relations between sets. The phrase *the murderers* in (23d) is a subset of *all the murderers,* and it is not possible for the extension of a subset of a set to be identified in such a construction: the extension of the *all* set is the set *the murderers,* not the members of the set of murderers, and the existence of that set is not asserted by *all.*

Patten extends this account to cleft clauses, such as those in (24).

(24) a. *It was John that was the murderer.*

b. *It was Cicero who once said, "Laws are silent at times of war".*

The argument is that in (24a), for example, *it...that was the murderer* is a discontinuous definite NP, and *John* is the extension of the set that the discontinuous NP describes. The account extends nicely to the grammatical emergence, or constructionalization of the cleft construction, as described in Gisborne and Patten (2011).

As well as grammaticalization, this analysis of cleft clauses has another very simple benefit. Patten (2010) presents her analysis in construction grammar, and her account is both simple (there are no derivations) and compatible with cognitive assumptions, once it is agreed that a treatment of the as a proportional quantifier makes sense in a cognitive theory. Patten does not present a syntax of the specificational sentence, but we can assume that the lefthand NP is simply a syntactic subject (it can raise over seem as in [25a]) and that the righthand NP is simply a complement.

(25) a. *The murderer seemed to be John.*

b. *Which murderer did she think was John?*

c. ** Which murderer did she think that was John?*

Note, in fact, that the left hand NP not only raises over seem but is also subject to the THAT-trace effect, where a subject in English cannot be extracted over that.

There is no need in this analysis to posit leftward movement of the predicate NP the apparent predication falls out of the semantics. There is no need to posit a level of information structure which captures the universal quantification and existential commitment because they are in the semantics anyway. We can capture the syntactic properties of specificational sentences and their conflict with the semantics as a simple constructional mismatch.

Patten does not give a formal account of specificational sentences, which in a formal semantics would require type-raising of the definite NP to a predicate in order to get compositionality. I present a WG analysis of the structure of a specificational sentence in the next diagram.

In the diagram, the singular NP has the same structure as I provided above in Figure 8. The set treatment of the gives us the relevant semantics, therefore providing the existential commitment and exhaustively. The syntactic structure is one of complementation, not predicative complementation, and in a complementation structure with be, you expect an identity relation as in *Tully is Cicero.* However, it does not make sense to identify a set with an individual, although you can identify a member of a set with an individual. Therefore, the semantics that works is one where the member of the set is identified with the referent of the second NP.

Essentially, the analysis differs from an equative analysis only inasmuch as the first NP is given a set-based treatment as a definite NP. The rest of the structure is what we expect for be with an NP complement, which needs there to be an identity relation in examples such as this.

I finish this section by noting that the familiarity analysis cannot explain the quantifier scope interactions or the definiteness effects, and that the analysis of specificational sentences I have just offered is only possible in an account where we treat the as a proportional quantifier.

- [1] It is necessary to make sure that we agree on which subtype of the existential there construction is under discussion. There are apparent counterexamples to the generalization:
*There was the most beautiful sunset I've ever seen*and*there was the odd drop of rain.*As Kearns (2000: 8185) points out, there are there+be constructions which permit definites: the main diagnostic sentences are the basic existential there+be sentences.