The article that wouldn’t go away

In the previous sections I discussed cases in which the definite article was expected to overtly surface in the noun phrase but it didn’t. Let us now turn our attention to cases in which its appearance is obligatorily triggered by another nominal modifier. The relevant modifiers are in-flecting demonstratives and dative possessors. Inin-flecting demonstratives always appear with the definite article, and in definite possessive constructions so do dative possessors.³⁹ This is so even if the head noun is a proper name that otherwise cannot appear with the definite article (i.e. if it spans the D position).

(153) a. (*a)

As (i) shows, two adjacent non-homophonous articles are perfectly grammatical. However, center-embedding be-comes consistently unavailable when it would result in adjacent homophonous articles.

(iii) [t-´ees

‘of the dominion of the city’ (Gloston, 1995, p. 353., ex. 34. a.) (iv) *[t-´ees

‘of the dominion of the city’ (Gloston, 1995, p. 353., ex. 34. b.) Ancient Greek The OCP rules in Ancient Greek and Hungarian, however, use different strategies to avoid article repetition. An-cient Greek has an independently available alternative word order, which becomes obligatory in the relevant cases.

Hungarian, on the other hand, PF-deletes one of the articles. It also appears to be the case the the Ancient Greek article deletion is sensitive to phonological rather than featural identity. I refer the reader to Gloston (1995) for a detailed exposition of the facts.

39In possessive constructions, the article can be missing if the possessee is indefinite:

(i) L´attam

‘I saw three movies of David Fincher.’

c. *Ma´o-nak Mao-dat

K´ın´a-j´a-t China-poss-acc

‘Mao’s China’ (Alberti and Balogh, 2004, ex. 3. d. )

The relevance of these cases for article haplology was first pointed out in Alberti and Balogh (2004). They observe that (153) and (154) are problematic for Szabolcsi’s (1994) haplology analysis, as the article ends up being adjacent with the proper name but it cannot undergo deletion. It is therefore a significant advantage of the current proposal that it correctly predicts the pattern in (153) and (154). I will discuss these examples in turn.

Let us begin with inflecting demonstratives. As already discussed in this chapter, the main-stream position is that inflecting demonstratives are base-generated in their surface position, that is, in spec, DP. It is only Ihsane and Pusk´as (2001) and the present proposal that argue for the surface position being a derived one. The idea that inflecting demonstratives are merged lower and reach their surface position via movement turns out to be crucial in the explanation of (153).

In the beginning of this chapter I proposed that inflecting demonstratives are base-generated in spec, DemP, and they subsequently raise to spec, DP. Consequently, every noun phrase containing an inflecting demonstrative also contains a Dem head.

(155) [DP infl. dem. [D^′ D [DemP tinf l.dem [Dem^′ Dem [ . . . N ]]]]]

Proper names do not have an indexical meaning component, they do not express either the proximal or the distal variety of the Dem head. This means that proper names do not spell out Dem⁰. This immediately derives (153): the Dem head acts as an intervener between N and D, and forces these two positions to be spelled out by two separate morphemes.

So far I motivated the raising approach to inflecting demonstratives on semantic grounds only.

The argument was that specific interpretations are tied to specific projections in the phrase marker, and if a constituent is interpreted as a demonstrative, it must have passed through spec, DemP.

The pattern in (153) provides syntactic evidence for the correctness of this position. Thus again, we see how a lexicalization algorithm may have an impact on how we set up the functional sequence:

spanning helps to detect traces.

Let us now turn to dative possessors. On the model advocated here, the pattern in (154) is most naturally captured in the same way as (153): the article appears overtly due to an intervening head. As (154) is a possessive construction, it necessarily contains such a head. As I will explain in detail in Chapter 7, possessive constructions are built with the help of PossP. PossP is situated right abovenP; its head is spelled out overtly by the possessedness suffix-ja/-je/-a/-e (glossed as poss), and its specifier hosts possessors themselves in the beginning of the derivation.

(156) [DP D . . . [P ossP possessor [P oss^′ Poss (-ja/-je/-a/-e) [ . . . N ]]]]]

Proper names do not have a possessive meaning component, thus they are not exponents of the Poss head. This is corroborated by the fact that the Poss head has its own spellout. (In example (154), possis fused with first person singular possessive agreement and has the form-om.) This, in turn, means that the presence of the Poss head in possessive constructions causes a disruption effect. It prevents proper names from spanning N and D, thereby it results in the separate lexicalization of N and D.

In conclusion, the pattern in (153) and (154) provides one of the subtlest and most relevant sources of evidence in favour of the whole approach presented here.

4.3 Summary

Let us recapitulate the main points of this chapter. I showed that the functional sequence of the DP between numerals and case markers contains quantifiers, participial relatives, Nominative and Dative possessors, inflecting and non-inflecting demonstratives as well as the definite article in the following order:

(157) dat. poss > infl. dem. > def. art. > nom. poss. > part. rel. > infl. dem. > part. rel.

> quantifier

I argued that these constituents provide evidence for the functional sequence in (158).

(158) DP > Poss2P > pRelClP > DemP > pRelClP > QP > NumP > pRelClP

In this bit of the functional sequence, too, there reside elements that instantiate the lexicaliza-tion problem. Non-inflecting demonstratives, some quantifiers and proper names may appear on their own, without the definite article. In this case they appear to be the source of definiteness themselves, and appear to be in D(P) on the surface. Under the right circumstances, however, they co-occur with the definite article. In this case the article is the contributor of definiteness, and non-inflecting demonstratives, some quantifiers and proper names appear lower than D. These alternations don’t affect the shape of the lexical items involved. I argued that the Superset Prin-ciple offers a genuine insight into how these alternations happen; it provides constrained theory of the way such polysemy can arise.

The way structure and lexicalization interact in the higher portion of the Hungarian DP sup-ports a non-terminal approach to spellout. The well-known observation that an intervener forces determiners and proper names to co-occur with the article receives a natural account in terms of spanning. The unified analysis of intervention phenomena also offered an insight into the repre-sentation of participial relatives, for which we had no clear clues otherwise.

Finally, in a detailed comparison of Szabolcsi’s haplology analysis and my spanning proposal I concluded in line with ´E. Kiss (2002) that haplology does play a role at PF, but its role is much more limited than suggested in Szabolcsi (1994). In particular, haplology comes into play only when two definite articles end up being adjacent in the DP. All the other cases of apparent article deletion were given a principled structural account.