16ku,   PDF
Kucsera Márton

Preposed headed relatives in Hungarian

0. Introduction This paper discusses the Alignment-based analysis of Hungarian headed restrictive relative clauses which precede their head noun. First the data itself is introduced in detail as it has not part been subject to much earlier research; then, an Alignment Syntactic analysis is outlined that proposes to explain the problem by referring to language change as constraint re-ranking. The analysis makes use of the strictly alignment-based syntactic system proposed by Newson (2000) and a late vocabulary insertion mechanism (cf. Newson, 2010). It is shown that Hungarian preposed relatives are an example of sideswitching phenomena that are handled particularly well by alignment constraints.

1. The data Headed relative clauses in Hungarian are traditionally argued to follow the head noun. There are, however, cases when relatives can appear in front of their lexical head nouns. The phenomenon was first described by Nádasdy (2011) while neither Kenesei et al. (1998) nor É. Kiss (2003) notice this fact. This section introduces and expands on the data provided by Nádasdy (2011). The term “preposed relative” is used throughout the paper in a narrow sense as there are many different cases when a relative clause can be on the left side of its head. Consider the examples in (1): (1) a. Amit Péter mondott, az furcsa volt. a-what Péter said dem strange was What Péter said, was strange. b. Amit Péter mondott az a beszéd érdekes volt. a-what Péter said dem art speech interesting was The speech that Péter gave was interesting c. Engem meglepett amit Péter mondott beszéd me surprised a-what Péter said speech The speech that Péter gave surprised me. From this point on, the term “preposed relative” will only be used to describe constructions like (1c). There are a number of reasons to treat this construction The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 58

differently from the other cases. First, notice that in (1a) and (1b), the relative clause is a topic in the matrix clause whereas in (1c) it is not. Also, the first one seems to be more closely related to a free relative, as it clearly involves a universal interpretation, and both (1a) and (1b) comply with Lipták’s (2009) definition of correlatives; therefore, it is not surprising that the relative clause should surface at the (left) edge of the main clause. Second, the demonstrative pronoun az is obligatory in both (1a) and (1b), which is, again, expected of correlatives. While the (1c) only shows a preposed relative in a sentence final position, they can be found a wide a variety of positions where ordinary postnominal relatives can also appear. As can be seen from (2), preposed relatives can be sentence medial (a), be both topicalised (b) and focused ©, or appear in a PP (d): (2) a. János elmondta a János told

tegnap

amit

hallott viccet Péternek

the yesterday a-what heard joke

Péter-DAT

János told Péter the joke he heard yesterday. b. A

Péter amit

mondott beszédet János hallgatta

the Péter a-what said

speech

János listened

meg prev.

It was János who listened to the speech Péter gave. c. János a

Péter amit

mondott beszédet hallgatta meg

János the Péter a-what said

speech

listened

prev.

It was the speech Péter gave that János listened to. d. A

Péter amit

mondott beszéd után mindenki hazament

the Péter a-what told

speech after everyone home-went

Everyone went home after the speech Péter gave. Preposed relatives also differ from regular relative clauses both in word order and in the range of possible positions. In preposed relatives, it is possible to place a topic in front of the relative pronoun while postnominal relatives must be introduced by a relative pronoun. In fact, according to the judgments of native speakers, instances with a topic preceding the relative pronoun are significantly better than the topic following the relative pronoun as in (3b).

The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 59

(3) a. Engem meglepett a beszéd, amit Péter mondott ? b. Engem meglepett amit Péter mondott beszéd c. * Engem meglepett a beszéd Péter amit mondott d. Engem meglepett Péter amit mondott beszéd Another significant difference is that in postnominal relatives, there might be intervening elements between the head noun and relative clause. However, such intervention is not possible in preposed relatives. (4) a. A könyv nekem is kell amit tegnap vettél. art book me-dat too need a-what yesterday bought-2sg I also need the book that you bought yesterday.

• Amit tegnap vettél nekem is kell a könyv.

It is also important to notice that in preposed relatives, there is no demonstrative pronoun referring to the head noun while in (1b) it is obligatory. (5)

• Amit Péter mondott, a beszéd furcsa volt.

As far as the range of possible determiners preceding a preposed relative is concerned, the situation is fairly complex. If the preposed relative begins with a relative pronoun there may be no determiner preceding it. (6)

• Elköltözött az aki

kutyákat tartott ember. moved art a-who dogs kept man The man who keeps dogs has moved away.

In cases when there is a topicalised noun in front of the relative pronoun, there may be a determiner preceding the relative. Then, the situation is similar to possessive construction where a single preceding determiner may belong to either the possessor or the possessum. Similarly, in (7a), the definite article cannot belong to the noun Róma as that proper noun never appears with an article elsewhere in Hungarian. However, in (7b), the article must belong to angol filmeket as the verb szereti requires a definite object. (7) a. a Rómát aki jól ismeri idegenvezető art Rome a-who well knows tour guide the tour guide who knows Rome well

The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 60

b. az angol filmeket aki szereti art English films a-who likes the reviewer who likes English films

kritikus reviewer

Finally, preposed relatives are only possible in the case of restrictive relatives as the ungrammaticality of (8b) shows. (8) a. anyám aki tanár mother-SG1 a-who teacher My mother, who is a teacher b. *aki tanár, anyám To summarise, preposed relatives are constructions which precede the head noun and often have a topic preceding the relative pronoun, but lack many of the features standardly associated with correlatives.

2. Alignment-based analysis This section lays out the basics of an analysis that uses only linear alignment constraints to account for preposed relatives. The aim of this analysis is to show how the re-ranking of certain constraints can lead to the emergence of preposed relative clauses. The mechanics of the alignment-based analysis follow the analysis given for English restrictive relatives by Kucsera (2014). This section aims at presenting the rankings which make preposed relatives the most harmonic candidate while the following section discusses the possible implementation of these rankings in an Optimality Theoretic grammar. There are two important expectations that such an explanation must meet. First, it must not assume any sort of hierarchical structure in the output candidates. Second, the proposed change in the ranking of constraints must be kept to a minimum as a radical reordering would most probably have repercussions in the grammar that are too far-reaching. 2.1. The relevant constraints Apart from the basic precedence and adjacency constraints, I also make use of domains as described by Gáspár (2005). Furthermore, the idea of domain adjacency (put forward by Newson and Maunula, 2006) is going to be crucial. The reason for this is that such a constraint (provided that it is ranked high enough) can model side switching phenomena quite well. In this particular case, the analysis makes use of the following domains and constraints. There are two overlapping domains: the relative domain (R), and the domain of the information structure of a proposition (I). The crucial The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 61

difference between the elements of the two is that the former does not contain the root corresponding to the head noun, but the latter does. The involved functional Conceptual Units are, in turn, |rest|, which is responsible for the restrictive interpretation, and |top|, which provides the topic interpretation. The constraints used are listed in (9). (9)

List of constraints a. √ ¬a R the head noun either precedes or follows every element of R b. √ a |rest| the head noun is adjacent to the restrictive FCU c. |rest| p R the restrictive FCU precedes every element of R d. |top| p I the topic FCU precedes every element of I

It must be noted that (9a) differs significantly from the notation used by Newson and Maunula (2006). In fact, what is proposed here as an antialignment constraint would be defined as alignment constraint in their system. However, Mark Newson (p.c.) points out that the earlier definition of the alignment of a CU with respect to a domain is not very consistent with the notion of alignment. He therefore suggests that the most felicitous way to formalise that a CU is required to be on either side of a domain is by saying that it should not be aligned with it (i.e. aligned with as few domain members as possible). This can be achieved by placing it on either side of the domain. 2.2. Possible rankings for headed relatives First, the tableau in (10) shows what ranking of the constraints in (9) produces the postnominal relative clause. (10) Input: |rest|, |top|, √, … √ ¬a R ☞ √ |rest| |top| … √ |top| |rest| …

!*

!*

√ a |rest|

!*

!*

!*

In this competition, the winner only violates the alignment constraint targeting the topic as the others do not conflict if the relative clause is behind the head noun.

The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 62

Then, simply by re-ranking the topic-alignment constraint a situation emerges where it is more important for the topic to be the first than for the

(11)Input: |rest|, |top|, √, …

☞

√ |rest| |top| … √ |top| |rest| …

√ ¬a R

!*

!*

!*

!

!*

√ a |rest|

This account has a considerable advantage over the ones that are grounded in constituent structure. In the latter, it would require quite fundamental changes to alter the direction of adjunction and such a change would not only affect relatives, but it would have quite far-reaching consequences all over the grammar. Also, notice that this analysis excludes the possibility of head noun, topic, relative pronoun sequences (which are, in fact unattested in this case) as that candidate is harmonically bound by the winner. Finally, only one more change in the ranking is needed to arrive at the preposed relative introduced by the topic. In fact, that change involves a single re-reranking in the hierarchy. Again if the topic-alignment constraint is moved further up the hierarchy we get exactly this result as (12) shows. (12) Input: |rest|, |top|, √, … √ ¬a R

☞

√ |rest| |top| … √ |top| |rest| …

! !*

!*

!*

!*

√ a |rest|

This third tableau illustrates that by promoting the same constraint again placing the topic first becomes even more preferable; so much so that the topic is now placed in front of the restrictive element. Therefore, another advantage of the outlined analysis is that it accounts for this particular change only by changing the place of a single constraint in the ranking. The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 63

In addition, the crucial components of the change are the head of the relative clause and the restrictive element; therefore it has no effect on either free or non-restrictive relatives as both of those lack one of the two elements. Also, the proposed analysis assumes that faithfulness constraints play no part in this phenomenon for the following reason. Deletion would actually give rise to either postnominal relatives with the topic at the front or just regular postnominal relatives as deleting domain features would mean getting rid of the asymmetry between the elements of R and I.

3. Directions for further research There are two main directions in which the present analysis can be furthered. First, any future research on this topic requires a wider range of data as preposed relatives are generally underdocumented at present. Also, the primary theoretical problem to overcome is that of definite determiners. In particular, an explanation must be found for the fact that two definite determiners cannot appear preceding a relative. However, as a very similar phenomenon is exhibited by possessives, it is very likely that the solution lies not in the content of relative clauses, but rather in how vocabulary insertion works. Apart from these theoretical issues, another important question is the implementation of a system that is able to produce these different rankings. In the case of preposed relatives, it could not be argued that they are only observed in a given a dialect or age group as they are quite widespread. Also, as both postnominal and preposed relatives are produced by the same speakers, it is not possible to hypothesize two separate grammars that differ only on this point. One direction to explore is to employ some probabilistic version of Optimality Theory such as Boersma’s (1999) stochastic implementation of OT. In this particular case, this is would involve the free re-reanking of the constraints (9b), (9c), and (9d).1 Thus, the analysis has to account for 3!=6 combinations. Section 2.2 has shown the result of three possible rankings, and as the reader may verify, the places of √ a |rest| and |rest| p R may be swapped in each of these cases, so the most harmonic candidate in each case is a grammatical sentence.

1

This is a necessary of consequence of assigning a weight to |top| p I such that it is sufficiently likely to end up either above or below or inbetween both √ a |rest| and |rest| p R. This, in turn will mean that the latter two are so close in rank to each other that either may end up dominating the other.

The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton

Preposed headed relatives in Hungarian 64

References Boersma Paul. 1999. Optimality-Theoretic learning in the Praat program. ROA report #380. É. Kiss Katalin, Kiefer, Ferenc and Siptár, Péter. 2003. Új magyar nyelvtan. Budapest: Osiris. Gáspár, Miklós 2005. Coordination in Optimality Theory. PhD dissertaion. Eötvös Loránd University. Lipták, Anikó. 2009. The landscape of correlatives. In: Anikó Lipták (ed). Correlatives cross-linguistically. Amsterdam: John Benjamins. 1-48. Kenesei, István, Vago, Robert M., Fenvesi, Anna. 2002. Hungarian. London: Routledge. Kucsera, Márton. 2014. Restrictive relative clauses in Alignment Syntax. The Odd Yearbook 9. Nádasdy, Ádám. 2011. Balról korlátozó mellékmondatok. Lecture. HAS Research Institute for Linguistics. Newson, Mark. 2000. A restrictive theory of alignment and some of its consequences. Manuscript. Eötvös Loránd University. Newson, Mark. 2010. Syntax first, words after: A possible consequence of doing Alignment Syntax without a lexicaon. The Even Yearbook 9. Newson, Mark and Maunula Vili. 2006. Word order in Finnish: Whose side is the focus on?. The Even Yearbook 7. Kucsera Márton Department of English Linguistics, Eötvös Loránd University kucseram@inf.elte.hu

The Even Yearbook 12 (2016), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2016, Kucsera Márton