CSLI Publications logo
new books
contact us
for authors
LFG Proceedings
CSLI Publications

The Hungarian Operator Raising Revisited

Peter Szűcs


The Hungarian construction type traditionally referred to as Operator Raising (OR, see É. Kiss 2002, Gervain 2004, Coppock 2003) is a configuration in which a constituent bearing some grammatical function semantically belongs to a subordinate clause but surfaces in the matrix sentence. The examples in (1) show various constituents involved in OR: for instance in (1a) and (1a') the subordinate subject is "raised" (and, as these examples demonstrate, the case of the raised constituent may be either accusative or nominative), (1b) exemplifies a "raised" oblique. Objects may also be "raised" as in (1c). (1c') shows that with raised objects, the main verb may either have definite or indefinite conjugation

(1) a János-t mondtad, hogy jön.
John-ACC said.2SG that comes
'It is John who you said that he is coming.'
  a' János mondtad, hogy jön.
John.NOM said.2SG that comes
'As for John you said that he is coming.'
  b Londonba mondtam, hogy megyek.
London.INTO said.1SG that go.PRES.1SG
'It is to London that I said I would go.'
  c Kutyá-t mondtad, hogy veszel.
dog-ACC said..2SG.DEF that buy.PRES.2SG
'It is a dog that you said you are buying.'
  c' Kutyá-t mondtál, hogy veszel.
dog-ACC said.2SG.INDEF that buy.PRES.2SG
'It is a dog that you said you are buying.'

Coppock (2003) gives an LFG account of the phenomena. She claims that the fronted element may be an athematic object of the main verb and the identification with the lower grammatical function could be either anaphoric or functional in nature. Anaphoric control must be allowed since if the raised element bears accusative case, it is a subject in the embedded clause and includes a quantified element, then the lower verb can either have 3rd person singular or 3rd plural agreement (this was first reported by Gervain 2002):

(2) Az összes lány-t mondtad, hogy jön/jönnek.
the all girl-ACC say.PAST.2SG that come.PRES.3SG/come.PRES.3PL
'It is all the girls that you said that they are coming.'

Since the athematic object with anaphoric identification leads to the violation of Semantic Coherence Condition (Dalrymple 2001), Coppock (2003) claims that this condition is just violable OT-style constraint.

I have two goals. First, I aim to argue that there is no need for the demotion of Semantic Coherence. I claim that the fronted element is a thematic argument of the main verb, so (2) (and 1a, 1c') is an instance of prolepsis, and akin to structures attested for instance in Madurese (Davies 2005) or Greek (Kotzoglou & Papangeli 2008). I summon arguments from the empirically tested availability of resumptive pronouns (3), idiom distribution (4), island-insensitivity (5) and from the fact that the verbs which allow OR may not be bridge verbs (6 is a canonical raising construction). Such behaviour is typical of prolepsis, as is argued by Davies (2005).

(3) ?János-t mondtad, hogy ő jön.
John-acc said.2sg that he comes.
'It is John who you said that he is coming.'
(4) #Szöge-t mondtál, hogy ütött a fejedbe az ötlet.
nail-ACC said.2SG that hammered.3SG the head.2SG.INTO the idea
'It is a nail that you said that the idea hit into your head.' Intended: 'You said that the idea is stuck into your head.'
(5) Az elnök-öt mondtad, hogy hallottad a hír-t, hogy megérkezett.
The president-ACC said.2SG that heard.2SG the news-ACC that arrived.3SG
'It is the president that you heard the news that he has arrived.'
(6) *A busz-t közeledni mondtad. / A busz-t közeledni vélted.
The bus-ACC approach.INFINITIVE said.2SG the bus-ACC approach.INFINITIVE believed.2SG
'*You said the bus to approach' / 'You believed the bus to approach.'

My second goal is to present some empirical data about OR and to show that discourse-structure (King & Zaenen 2004, Gazdik 2011) may provide a useful tool for describing some experimental results. I will argue that empirical data shows that in the all the shown sentences except for 1a', the discourse function of the fronted constituent may be either focus or contrastive topic, but focus is preferred. I seek to explain this by claiming that informants disprefer contrastive topic because the corresponding discourse structure would be larger than with one focus, as contrastive topic involves "splitting up the original question into implicit subquestions and answer (some of) them" (Gazdik 2011:169, quoting Büring 2003). So if the fronted element is focus in 1a it may correspond to discourse-tree like figure 1, if the fronted elememt is contrastive topic, it may correspond to a discourse tree like figure 2. Such discourse trees may be neatly formalized in LFG, as is proven by Gazdik (2011).

Figure 1.
JÁNOST-focus mondtad, hogy jön.
Question: Who (question-focus) did you say that comes?
Answer: John
Figure 2.
Jánost-contrastive topic MONDTAD, hogy jön
Question: Who(question-focus) did you say that comes?
/ | \
Subquestion1: Subquestion2: Subquestion3:
Did you say that Peter comes? Did you say that John comes? Did you say that X comes?
| | |
Answer: No. Answer: Yes. Answer: ...

So it could be inferred that the LFG principle Economy of Expression could be interpreted at the d-structure level as well. This is in line with various Optimality-theoretic approaches which penalize large structures (see for example Bresnan 1998).

Interestingly, in 1a' the preferred discourse function is contrastive topic and not focus. I assume that it is because that in 1a' the "raised" element is subject (unlike in the other sentences) and the subject function is prototypically associated with topics (see eg. Gregory & Michaelis 2001, Lambrecht 1994). Speaking in Optimality-theoretic terms, this constraint outranks the previously mentioned principle about small discourse trees.


pubs @ csli.stanford.edu 
CSLI Publications
Stanford University
Cordura Hall
210 Panama Street
Stanford, CA 94305-4101
(650) 723-1839