A spanning account of non-inflecting demonstratives and -ik quan- quan-tifiers

From Num to D

4.2 Lexicalizing the D position

4.2.2 A spanning account of non-inflecting demonstratives and -ik quan- quan-tifiers

Recall from the theoretical discussion in Chapter 2 that in a framework employing non-terminal spellout, movement potentially changes the spell-out possibilities of a structure. A lexical item that is associated to the features A, B and C can spell out all of these features at once only if they appear in a contiguous sequence in the structure, as in (87).

(86) LI1 ⇔{A, B, C}

(87) A

B C

LI1

On the other hand, in a structural conﬁguration where C, for instance, is moved away, it is not possible to spell out all of A, B and C byLI1anymore (traces don’t count for spell-out purposes).

(88)

CP . . . C. . .

X A

B C

Recall further that disruption eﬀects breaking the contiguity of certain head sequences can arise not only as a result of movement, but also as a result of base-generation. If the functional sequence

21The translations of the two determiners in the running text are based on Zepeda (1983). I kept Szabolcsi’s glosses in (85) (which she took over from Abney, 1987).

deﬁnes the hierarchy A > X > B > C, then in order to useLI1 as an exponent of A, B and C, X must not be generated in the tree. In case X is present in the structure, it acts as an intervener and results in A being lexicalized separately from the other features. If the language has no lexical item that could spell out only B and C, then by the Superset PrincipleLI1shrinks down to B and C. This is illustrated in (89).

(89)

A LI2 X

LI3 B C LI1

In sum, the conditions under which contiguity is broken and which force A, B and C to be spelled out in two distinct bits are i) linear separation by an intervener and ii) movement. In Szabolcsi’s Haplology rule in (83), these are precisely the conﬁgurations that require the overt, separate spell-out of the deﬁnite article.

I suggest that the pattern under consideration is best seen as an eﬀect of a structural intervener on a spanning lexical item. I propose that non-inﬂecting demonstratives and quantiﬁers triggering article deletion (-ik quantiﬁers, minden, valamennyi) span up to the D position.²² Nominative possessors and participial relatives make the co-occurrence obligatory because they are both intro-duced in the speciﬁer of a specialized functional head. These cut the sequence of heads lexicalized by demonstratives or quantiﬁers into two. By virtue of their position between D and Dem/Q, they eﬀectively separate D from the other heads it could be spelled out together with. That is, instead of a surfacy Haplology rule, I suggest capturing the pattern in terms of the size of the lexical items involved.

Crucial for this analysis is that both Nominative possessors and participial relatives are intro-duced by a functional head of their own (otherwise they could not function as interveners). This is entirely uncontroversial for Nominative possessors, as we have seen. I called the relevant func-tional head Poss2. For participial relatives, we have seen a number of diﬀerent possible approaches that could all capture their surface distribution, but we have not had clear evidence for any of them. Given the way spanning works, i.e. that only heads but not adjuncts or speciﬁers count as interveners for spanning lexical items, from my analysis of intervention eﬀects a number of things follow for the representation of participial relatives.

Firstly, they cannot be adjuncts, because in an adjoined position they could not prevent span-ning of D and Dem or D and Q, and the pattern in (79) through (82) could not arise. Secondly, contra Kenesei (2006), they cannot sit in the speciﬁer of the functional head that introduces non-inﬂecting demonstratives.²³ If participial relatives were sitting in this position, then the D and Dem heads would still be contiguous in the sequence of heads, and inﬂecting demonstratives should be able to spell out the D position even in the presence of a high participial relative. As (79) shows, this is not possible.

For participial relatives to function as interveners in (79) – (82), they must be introduced in the speciﬁer of a specialized functional head that is diﬀerent from Dem. That is, the sequence of functional heads in (79) through (82) must be D > X > Dem, where X intervenes between D and Dem and prevents their joint spellout. That the X head cannot be Poss2 is shown by examples that contain both a Nominative possessor and a high participial relative.

(90) az the

´en I

[tegnap yesterday

el˝oadott]

presented

b´armelyik any

javaslat-om proposal-poss.1sg

‘any proposal of mine presented yesterday’

This means that the sequence of functional heads in (90) is D > Poss2 > X > Dem, where X is the head responsible for the introduction of non-ﬁnite relatives. I call this head pRelCl for participial relative clause. Participial relatives have been independently argued to occupy a speciﬁer projection in various works, including Ouhalla (2004) and Cinque (2010). The analysis presented here supports their conclusions.

22This will be qualified for quantifiers below.

23I call this projection DemP, Kenesei calls it DetP, but this is only a terminological difference.

This proposal sheds light on the way a lexicalization algorithm interacts with the functional sequence. From the way lexicalization is set up, viz. that spanning items spell out a contiguous head sequence, it follows that only heads can have a disruption eﬀect. Therefore any element that causes the disruption eﬀect involves the merger of an intervening head, and these elements spell out either the head itself or its speciﬁer. In essence, spanning allows us to distinguish adjuncts from functional heads and their speciﬁers.

Let us now turn to a detailed exposition of how spanning works for non-inﬂecting demonstra-tives. The relevant portion of the functional sequence is repeated below.

(91) [DP inﬂ dem [D^′ def. art [P oss2P pronom Nom poss [pRelClP part. rel. [DemP inﬂ dem [Dem^′ non-inﬂ dem ]]]]]]

Demonstratives are merged in DemP. Inﬂecting demonstratives are phrases that move from spec, DemP to spec, DP. Non-inﬂecting demonstratives, on the other hand, are in the head of DemP, and they do not co-occur contiguous to the deﬁnite article because they span the D position themselves. Their lexical entry thus contains two categorial features: one for Dem and another for D.

(92) non-inﬂecting demonstrative⇔{Dem, D}

Non-inﬂecting demonstratives are not speciﬁed either for the Poss2 or the pRelCl head. When these heads are not present in the syntax, then D and Dem form an uninterrupted sequence and they can be spelled out together by the non-inﬂecting demonstrative.

(93) (*az) the

eme this

javaslat proposal

‘this proposal’

(94) DP

D DemP

Dem . . . eme

The reason they have to be spelled out together, that is, why the article appears to be obligatorily deleted, is the economy principle Maximize Span/Union Spellout Principle discussed in Chapter 2 (also known as Minimize Exponence in DM).

(95) Minimize Exponence

The most economical derivation will be the one that maximally realizes all the formal features of the derivation with the fewest morphemes. (Siddiqi, 2009, p. 4.)

Spelling out D and Dem with just the non-inﬂecting demonstrative is more economical than using the deﬁnite article for D and shrinking down the demonstrative to Dem.

However, when one or both of the Poss2 or pRelCl heads appear between D and Dem, spanning is not an option any more. In this case the non-inﬂecting demonstrative must shrink down to Dem and D is spelled out by the deﬁnite article, as depicted in (97).²⁴

24One question that may arise here is why Dem/Q to D spanning is allowed with a propossessor. As is well known, Hungarian is aprodrop language and it allows not only its subjects but also its possessors to be covert.

Pronominal subjects and possessors are overt only if emphasized or contrasted.

(i) a. az

the

´en I

¨ otlet-em idea-poss.1sg

‘my idea’

b. az the

¨ otlet-em idea-poss.1sg

‘my idea’

If the possessee is modified by a non-inflecting demonstrative or spanning quantifier and the possessor ispro, then the definite article does not appear on the surface.

(96) *(az) the

´en my

tegnap yesterday

el˝oadott presented

eme this

javaslat-om proposal-poss.3sg

‘this proposal presented yesterday’

(97) DP

D az

Poss2P

´en

Poss2 pRelClP

tegnap el˝oadott

pRelCl DemP

Dem eme

. . .

Let us now turn to quantiﬁers. Minden ‘every’,valamennyi ‘each’ and -ik quantiﬁers exhibit the same distribution with respect to the deﬁnite article as non-inﬂecting demonstratives do. In the most parsimonious account, they should receive the same analysis, too. However, the exact same analysis will not work here because quantiﬁers are phrases. In spanning, lexical items are able to spell out multiple head positions but they are not able to span a mixture of head and speciﬁer positions. Therefore it is technically impossible for these quantiﬁers to span the D position.

While the exact same analysis used for non-inﬂecting demonstratives is inapplicable to the relevant quantiﬁers, it is possible to preserve the spirit of the account and capture the behaviour of quantiﬁers in an analogous fashion. Recall from Chapter 2 that the Exhaustive Lexicalization Principle is an important constraint on lexicalization.

(98) Exhaustive Lexicalization Principle:

Every syntactic feature must be lexicalized. (F´abregas, 2007)

The reason the Exhaustive Lexicalization Principle is relevant in the present context is that the Q head hosting quantiﬁers does not have an overt exponent. As according to (98) Q⁰ must be associated to a piece of phonology, the only possibility is that it receives a phonologically zero spellout.

(99) ∅ ⇔{Q}

The article deletion data with quantiﬁers can be captured in the same fashion as those with non-inﬂecting demonstratives if the Q head appearing with the relevant quantiﬁers spans the D position.

Thus the lexical entry of the morpheme spelling out the Q head appearing with these quantiﬁers is as in (100).

(100) ∅ ⇔ {Q, D}

(ii) a. eme this

javaslat-om proposal-poss.1sg

‘this proposal of mine’

b. valamelyik a.certain

javaslat-om proposal-poss.1sg

‘a certain proposal of mine’

Ifpropossessors are also in spec, Poss2P, then Poss2 should act as an intervener and force the appearance of the definite article. I propose that (ii) shows that propossessors are in fact in a different structural position. For both (ii-a) and (ii-b) to fall out, this position must not be between Q and D. There are two possible alternatives:

propossessors stay in the low thematic position where possessors are base-generated, and in this case Poss2P is not generated at all, or alternatively apropossessor could raise straight to spec, DP, which is also the position for R-expression Nominative possessors. C.f. also Ortmann (2011) for a proposal that these cases do not involve a covert pronoun at all.

From here, everything works in the same manner as for non-inﬂecting demonstratives. The zero morpheme in (100) cannot spell out Dem, Poss2 or pRelCl. When these heads are not projected in syntax, Q and D become adjacent and they are spanned by one formative, that in (100).

(101) (*a) the

b´armelyik any

/ /

minden every

/ /

valamennyi each

k¨onyv book

‘any / every / each book’

(102) DP

D QP

b´armelyik, minden, valamennyi

Q k¨onyv

∅

When Dem, Poss2 or pRelCl are present in the structure, however, spanning is impossible and D is spelled out separately.

(103) az the

´en I

tegnap yesterday

el˝oadott presented

minden every

javaslat-om proposal-poss.1sg

‘every proposal of mine presented yesterday’

(104) DP

D az

Poss2P

´en

Poss2 pRelClP

tegnap el˝oadott pRelCl QP minden

∅ javaslatom Thus while minden ‘every’,valamennyi ‘each’ and-ik quantiﬁers do not span D themselves, the head that introduces them does spell out D. In the rest of this chapter, I will regularly state that these quantiﬁers span D. This, however is just a convenient (albeit inaccurate) shorthand for the analysis outlined above. Its use lies in avoiding cumbersome expressions like‘quantiﬁers that co-occur with a phonologically null head that spans Q and D’.²⁵

To sum up the discussion so far, Nominative possessors and participial relatives function as base-generated interveners between D/Q and Dem. We have seen that in non-terminal spellout interveners can also arise as a result of movement. Szabolcsi argues that this is attested with article

25I note here that in a phrasal spellout model the statement that ‘the relevant quantifiers spell out D’ can be taken literally. The analysis outlined above can be rendered in constituent spellout in the following manner. The relevant quantifiers function as the exponent of all terminals Q through D, that is: the head Q, the specifier of Q and the D head, but not the part of the tree abbreviated as‘. . . ’.

deletion indeed, and that moving Det away from D makes the deﬁnite article visible. She presents data with Dative possessors as evidence of this.

Independently of the article deletion data, Szabolcsi proposes that Dative possessors are derived from Nominative possessors in the syntax. Every possessor starts out as Nominative and has a copy in the position of Nominative possessors. From this position, possessors can undergo movement to the periphery of the DP, where they are assigned Dative case by D (Szabolcsi, 1994).

The relevant data where movement from the Nominative possessor position to the Dative possessor position supposedly prevents the application of haplology are in (105). In (105-a)minden is part of the Nominative possessor and it is string-adjacent to D, causing the deletion of the article.

In (105-b) the possessor has been moved to the periphery of the DP and a and minden are not adjacent any more. This causes the article to appear overtly.

(105) a. (*a) the

minden every

feltal´al´o inventor

¨otlet-e

idea-poss.3sg

‘every inventor’s idea’

b. minden every

feltal´al´o-naki

inventor-dat

*(az) the

ti ¨otlet-e idea-poss.3sg

‘every inventor’s idea’

If this was the case indeed, it would provide a showcase of how movement aﬀects spellout. However, I do not subscribe to the view that Dative possessors are created from Nominative possessors in syntax. On the one hand, den Dikken (1999) and ´E. Kiss (2002) point out that D is generally not a Dative case assigner: is never assigns Dative (or indeed any other) case to inﬂecting demonstratives in spec, DP. On the other hand, there is also a theory-internal, Nanosyntax-speciﬁc reason not to adopt Szabolcsi’s view that Nominative possessors turn into Dative in the derivation. Caha (2009) argues that cases have internal syntactic structure: Obliques structurally contain Accusative, and Accusative structurally contains Nominative case.

(106) ObliqueP

Oblique AccP

Acc NomP

Nom DP

As syntactic trees can only be extended at the root node, it follows from Caha’s system that once a DP is merged with Nominative case (or with no case, for that matter), it cannot be assigned Dative later on. In D´ek´any (2010) I show how this system can be fruitfully applied to Hungarian long operator movements featuring case competition, and I argue that apparent instances of a constituent acquiring extra layers of case in the derivation actually involve base-generating the bigger case.

(i) DP

spec D’⇒ ∅

D QP

spec Q . . .

The correct word order can be derived in the following way. First the complement of Q must be extracted such that Q, spec Q and D can from a constituent targeted by constituent spellout (recall that traces don’t count for spellout purposes). Then remnant movement must restore the original order between QP and its complement. This solution is mostly along the lines of Starke (2007, 2009a,b); Taraldsen (2009); Starke (2011). Alternatively, it can be assumed that nodes that have already undergone spellout can be ignored for further spellout. In our case these nodes would be the ones abbreviated as‘. . . ’. In this scenario∅can spell out D’ without movements taking place.

This solution is in the spirit of Caha (2009).

Tohono O’odham determiners, on the other hand, quite possibly show both the base-generated and the movement-related intervention eﬀect. The relevant example is repeated here for the reader’s convenience.

(107) a. *’am det

[g det

miisa]

table weco underneath b. ’am

det [miisa]

table weco underneath

‘under the table’

c. ’am det

weco underneath

[g det

miisa]

table

‘under the table’ (Szabolcsi, 1994, p. 212., 73.) Tohono O’odham

In (107-a) the string-adjacency of determiners leads to ungrammaticality. (107-b) suggests that this is because’am spans the position ofgand thus it is a more economical spellout. In (107-c) the extraposition of the [g miisa] constituent appears to break the adjacency and make the separate spellout of g possible. While the Tohono O’odham data are suggestive, the analysis remains tentative because I cannot verify whether’am andg are on the same projection line or not. The spanning analysis is viable only if these determiners are part of the same extended functional sequence, andg is not embedded in some speciﬁer. As I do not have a proper grasp of what the phrase structure of (107) looks like, I will not look further into the Tohono O’odham data.²⁶

In the proposed analysis of‘haplology’, non-inﬂecting demonstratives and quantiﬁers triggering the article deletion properly contain the deﬁnite article. The intuition that these elements somehow include the deﬁnite article is expressed in the literature again and again. Below I provide a selection of quotes expressing this insight.

"a hat´arozott ´es hat´arozatlan n´evel˝o nem hordoz valamif´ele csak ˝or´a jellemz˝o inform´aci´ot:

a t¨obbi determin´ans jelent´ese kb. ´ugy ´ırhat´o le, hogy‘a hat´arozott n´evel˝o vagy hat´arozatlan n´evel˝o jelent´ese + m´eg valami’. P´eld´aul: ezen h´arom piros kalap⊃a h´arom piros kalap . . . "

‘the deﬁnite and the indeﬁnite article do not carry information exclusive to them: the meaning of other determiners can be described approximately as ‘the meaning of the deﬁnite or indeﬁnite article + something else’. For instance: ezen h´arom piros kalap⊃ a h´arom piros kalap . . . ’

Szabolcsi and Laczk´o (1992, p. 223.)

"a n´evel˝ok jelent´ese r´esze a Det2-k jelent´es´enek"

‘the meaning of articles is part of the meaning of Det2-s’ (where Det2 refers to deter-miners)

Szabolcsi and Laczk´o (1992, p. 225.)

"it seems reasonable to look on determiners as having two functions: that of a subor-dinator and that of a quantiﬁer/demonstrative . . . these two functions can be either conﬂated or lexicalized separately."

(´E. D.: the signiﬁcance of this lies in the fact that Szabolcsi analyzes articles as subor-dinators)

Szabolcsi (1994, p. 218.)

26Some of the Hungarian examples above involve article deletion triggered by a determiner embedded in a Nomi-native possessor. In Section 4.2.5 I will come back to these data.

"a disztrib´uci´oja legink´abb a hat´arozott n´evel˝o´ere eml´ekeztet"

‘its distribution primarily resembles to that of the deﬁnite article’ (onminden ‘every’,

though later on in the article the claim is thatmindenfunctions as an indeﬁnite article) Kornai (1989)

"Noun phrases extended by an-ik quantiﬁer or ane/eme/ezen type of demonstrative act as deﬁnite noun phrases . . . -ik quantiﬁers and demonstratives also seem to play the role of a deﬁnite determiner."

E. Kiss (2002, p. 154.)´ In the light of this intuition, determiners instantiate the lexicalization problem: on their own, they appear to perform the function of D, and so they should spell out in D(P), but when they co-occur with the article they don’t perform the D function and appear in a lower position, too. These two uses feature the same lexical items in diﬀerent guises, though. The spanning analysis proposed here captures this intuition in a natural and straightforward way via the Superset Principle. Used on their own, determiners spell out D. They properly contain all features of the deﬁnite article and linearize in its position. When they co-occur with the deﬁnite article, determiners Underassociate

In document A profile of the Hungarian DP (Pldal 99-106)