Adjunction, Condition C, and the Background Adjunct Coreference Principle

        In this paper I investigate different types of adjunction and discuss cyclic and acyclic
approaches to adjunct merger.
        It is a well-known fact that there exists an adjunct-argument asymmetry with respect
to reconstruction and Binding Principle C (1). I show that adjuncts cannot be treated
uniformly. Based on data from Czech, I argue that in addition to the standardly accepted
distinction between arguments and adjuncts with respect to Condition C, it is necessary to
introduce the distinction between clausal and non-clausal adjuncts. Non-clausal adjuncts (no
matter whether adnominal or adverbial) reconstruct regardless of their referential and
information-structural status or the status of the phrase containing them and always induce a
Condition C effect (2).
        Both adnominal and adverbial clausal adjuncts can obviate a Condition C violation
under certain conditions. These are determined by the presuppositional and information-
structural status of the clausal adjunct or the element containing it and by the information-
structural status of the relevant r-expression. Clausal adjuncts embedded in a non-
presuppositional element (3a) or spelled-out in a position c-commanded by the coindexed
pronoun (3b) induce a Condition C violation. If the backgrounded clausal adjunct (4) (or the
presupposed element containing it (5)) precedes the pronoun, a Condition C effect appears
(4a), (5a) or does not (4b), (5b). For coreference to be possible, the r-expression must be
backgrounded (not focused) in the clausal adjunct, as in (4b) and (5b). Therefore I propose the
following principle:
    Background Adjunct Coreference Principle
    Coreference between an r-expression within an adjunct clause and a pronoun in the
    matrix clause is possible only if the r-expression is backgrounded in the adjunct clause.
        I will show that this principle is indeed limited to adjunct clauses (not complements)
and that it also holds for non-subject r-expressions and I will also show that the pronoun does
not have to be the subject of the matrix clause.
        Given the data, I argue that it is not the time of adjunction (e.g. Lebeaux 1988,
Stepanov 2000, 2001) or the type of merger (e.g. Chomsky 2001, Rubin 2003) that is relevant
with respect to Condition C in adjunction - both clausal and non-clausal adjuncts are merged
cyclically and are `c-command-visible' for the relevant pronoun in their non-spelled-out
positions - but the depth and type of embedding of the r-expression in the quantificational and
information structure of the sentence. The depth of embedding is crucial in the distinction
between clausal and non-clausal adjuncts because clausal adjuncts, in contrast to non-clausal
ones, have their own information structure. The type of embedding (Background Adjunct
Coreference Principle) is crucial in the distinction between focused and backgrounded r-
expressions in the clausal adjunct; only elements embedded in the background domain of the
adjunct clause do not have to reconstruct.
        I further argue that the application of Condition C must be capable to wait until the
semantic interface (of the highest phase in the sentence, consider (5b)) and see the whole
derivation. Since the piece of information about the semantic interface of the adjunct clause
(about the r-expression) is accessible in the course of the computation, the derivation cannot
forget the preceding phases.

(1)    [wh Which [[picture [of Bill1]] [that John2 liked]]] did he*1,2 buy twh?
                                                                              Chomsky (2001, 15)

(2) a. *Kolik        knízek na Pavlov1 polici pro1 pecetl?
       how many      books on Pavel's shelf         read
       `How many books from Pavel's shelf did he read?'
    b. *Kterou z     knízek na Pavlov1 polici pro1 pecetl?
       which of      books on Pavel's shelf           read
       `Which of the books from Pavel's shelf did he read?'
    c. *Tu knízku     na Pavlov1 polici pro1 v pátek       pecetl.
       the bookACC    on Pavel's shelf        on Friday read
       `The book from Pavel's shelf, he read on Friday.'

(3) a. *Kolik     argument, které Pavel1          pednesl, pro1 zuiv      bránil?
       how many arguments which PavelNOM gave                   furiously defended
       `How many arguments that Pavel gave did he defend like a fury?'
    b. *pro1 políbil       Marii,    nez     Pavel1   odjel.
               kissed      MarieACC before PavelNOM left
       `Before Pavel left, he kissed Marie.'

(4) a. *Nez odjel           Pavel1,      pro1   políbil    Marii.
       before left          PavelNOM            kissed     MarieACC
       `Before Pavel left, he kissed Marie.'
    b. Nez     Pavel1      odjel, pro1 políbil Marii.
       before PavelNOM left            kissed MarieACC
       `Before Pavel left, he kissed Marie.'

(5) a. *Který argument, který pednesl Pavel1, pro1 zuiv           bránil?
       which argument which gave          PavelNOM furiously defended
       `Which argument that Pavel gave did he defend like a fury?'
    b. ?Který argument, který Pavel1      pednesl, pro1 zuiv bránil?
       which argument which PavelNOM gave                furiously defended
       `Which argument that Pavel gave did he defend like a fury?'



References

Chomsky, N. (2001): Beyond Explanatory Adequacy. MIT Occasional Papers in Linguistics
20. Cambridge, Massachusetts: MITWPL.
Lebeaux, D. (1988): Language Acquisition and the Form of Grammar. Ph.D. dissertation.
University of Massachusetts at Amherst.
Rubin, E.J. (2003): Determining Pair-Merge. Linguistic Inquiry, 34, 660-668.
Stepanov, A. (2000): The Timing of Adjunction. In: M. Hirotani, A. Coetzee, N. Hall, J. Kim
(eds.), Proceedings of NELS 30, University of Massachusetts at Amherst, 597-611.
Stepanov, A. (2001): Late Adjunction and Minimalist Phrase Structure. Syntax, 4, 94-125.