610: LING Semantics and Generative Grammar
Angelika Kratzer Relatives and correlatives, part 2 November 7 & 9, 2007 1. (1) Dayals structure for the left-adjoined relative clause a. b. c. jo laRkii khaRii hai vo lambii hai. which girl standing be-PR she tall be-PR jo laRkii khaRii hai vo laRkii lambii hai. which girl standing be-PR that girl tall be-PR jo khaRii hai vo laRkii lambii hai. who standing be-PR that girl tall be-PR
Which girl is standing, she/that girl is tall.
Dayal (1996) posits a structure like (2) for Hindi left-adjoined relative clauses: TP (2)
CP
TP
jo
laRkii khaRii hai standing is
vo
laRkii lambii hai. tall is
Which girl
that girl
1
Rather than the adjunction structure in (2), the relative clause might occupy the specifier position of a Topic projection, as in (3):
TopicP (3) Topic CP TP
Topic
jo
laRkii khaRii hai
vo
laRkii lambii hai. tall is
Which girl standing is
that girl
2.
The left-adjoined relative clause: a first proposal
According to Dayal, the left-adjoined relative clause is interpreted as a definite description. She herself interprets it as a quantifier phrase, but well implement the proposal by giving the relative clause a denotation of type e. In our example, the definite description presupposes that there is exactly one girl who is standing. A first proposal T(jo) = PQ v1(P(v1) & Q(v1))
2
(4) CP
NP
jo
laRkii 5
TP
5 Exercise 1
khaRii
hai
Step by step, translate (4) into a well-formed expression of ETL. This should be an expression in MEe. Add non-logical constants as needed (remember to declare their types).
3.
The left-adjoined relative clause: a revised proposal
Do we really have two different jos, one for DP-internal relatives, and one for correlatives? Our current proposal says that in a correlative, jo has the job of mapping a pair of properties of individuals to the unique individual that has those two properties (if there is one). This cannot be the job of jo in DP-internal relative clauses, though, since those relative clauses need to have denotations of type <et>. A revised proposal Suppose correlatives (and free relatives) are really DPs headed by a feature [definite], which is not only interpreted like the definite article, but also projects the same structure as the definite determiner. According to Rajesh, some dialects do indeed have an overt definite article in this position, and this is also the case in German (see data below). Neglecting the feature [singular], we might have the following structure: 3
DP (5) NP CP (N) DP TP
[def]
[wh]
laRkii
5
5
khaRii
hai
The features [def] and [wh] (and possibly an additional gender feature) seem to be spelled out together, surfacing as jo. In German free relatives, this spell-out configuration seems to produce a remarkable range of variability, as shown in (6): In 6(a), {[def], [wh]} is spelled out as wer; in 6(b), the same feature complex is spelled out as der; in 6(c), each feature is spelled out separately, giving rise to two instances of der (d-pronouns are used as demonstratives and relative pronouns in English). (6) a. Wer das gesagt hat, (der) who that said muss sich entschuldigen.
has, that one must refl. apologize
Who said that, that one must apologize. b. Der das gesagt The one that said c. Der, hat, (der) muss sich entschuldigen.
has, that one must refl. apologize muss sich entschuldigen.
der das gesagt hat, (der)
The one who that said has, that one must refl. apologize
4
Exercise 2 Step by step, interpret structure (5), assuming that [wh] has been checked (possibly against a matching feature on C) and is not visible to the semantic interpretation component. Likewise, nodes that dominate no lexical material could be ignored by the semantics.
4.
Deriving DP-internal relative clauses via head raising
A DP-internal relative clause can now be derived from (5) via some version of head raising, as proposed in Bhatt (2002), among others. DP (7) NP CP N DP TP
[def]
[wh]
laRkii
5
5
khaRii
hai
One argument for such a head raising analysis for DP-internal relative clauses (even in languages like German or English) comes from so-called Funktionsverbgefge in German: (8) Der Vorteil, der uns daraus erwachsen ist, war beachtlich. The advantage that us from.that up.grown is was significant The advantage that accrued to us from that was significant. Der Weg, den er eingeschlagen hat, ist riskant. The path that he in.hit is has risky. 5
(9)
The path he took is risky. (10) Der Vorschlag, den sie uns unterbreitet hat, war unrealistisch. The proposal that she us under.spread has was unrealistic. The proposal that she made to us was unrealistic.
German has hundreds of so-called Funktionsverbgefge, where the verb in a noun-verb combination can only be interpreted together with that particular noun. This shows that even during the derivation of a DP-internal relative clause, there must be an interpretable stage where the head of the relative clause is within the relative clause. Movement of the head of the relative clause could potentially take place after semantic interpretation. More about the head raising analysis of relative clauses: BHATT, RAJESH. 2002. The Raising Analysis of Relative Clauses: Evidence from Adjectival Modifiers. Natural Language Semantics, 10. 43-90.
Major consequence We predict that, in contrast to correlatives, DP-internal relative clauses should not be able to appear with internal heads.
5.
Interpreting the matrix clause
Paul Postal has argued that pronouns are in fact Ds, a position that has recently been revived by Paul Elbourne. In many languages, definite determiners are homophonous with demonstrative pronouns. This is the case in Hindi, too (Dayal p. 179): vo chiiz / ye chiiz vo/ye ve chiizeN / ye chiizeN ve/ye us chiiz / is chiiz us/is that thing / this thing (s)he, it those things / these things they that thing / this thing (s)he, it 6 (Oblique) (Nominative) (Nominative)
un chiizoN / in chiizoN un/in
those things / these things they
(Oblique)
Rajesh told me that the pronouns in the table above are in fact the only 3rd person pronouns in Hindi, and they can be quite generally used as bound pronouns. They are not truly demonstrative pronouns, then, which tend not to have bound variable readings. Dayal proposes that the correlating DPs in the matrix sentence of a correlative construction can be interpreted as bound. Here is a proposal for the translation of vo into ETL: T(vo5) = P v1 (P(v1) & v1 = v5 ) For more motivation for this particular analysis of definite determiners, see Elbourne (2001): On the semantics of pronouns and definite articles. In Leora Bar-el & Karine Megerdoomian (eds.) Proceedings of the 19th West Coast Conference on Formal Linguistics, Somerville (Cascadilla Press), 2001, 8396. Also his Situations and Individuals, Cambridge (The MIT Press), 2005.
TopicP (11) Topic DP TP
Topic5
jo laRkii khaRii hai Which girl standing is
vo5 laRkii lambii hai. that girl tall is
7
Exercise 3 Step by step, translate the tree in (11) into ETL. Open questions Our analysis so far provides ...