CLIN 2005 Abstracts
  • Semantic construction for relative clauses
    Markus Egg (Rijksuniversiteit Groningen)
    For surface-oriented syntactic frameworks like HPSG, which do not rearrange constituents on some syntactic level (e.g., by `move-alpha'), crosslinguistic semantic construction for relative clauses poses a challenge. This challenge emerges if we accept the following standard linguistic tenets:

    (a) the only operation of semantic construction is functional application (b) functional application presupposes that the involved constituents are syntactic sisters in the underlying (binary branching) syntactic structure (c) semantic contributions of DP constituents have the same types across languages. : $«e,t>,«e,t>,t»$ for determiners, $<e,t>$ for nouns, $«e,t>,<e,t»$ for nominal modifiers; whole DPs are semantically of type $«e,t>,t>$.

    Consequently, noun and modifier semantics must be combined first, the result is then combined with the determiner semantics. But then the derminer must appear peripherally; the ordering Mod-Det-N or N-Det-Mod should be ruled out.

    While this prediction agrees with English DPs as (1), it does not hold good for its Turkish analogue (2), where the relative clause (literally, `of my loving') usually precedes, and the noun follows the determiner:

    (1) every girl that I love
    (2) sev- dig- -Im her kIz
    love nominalisation my every girl

    I will show that analyses for well-known English syntax-semantics mismatches can be used to capture the challenge in (2) without relinquishing the surface- orientation of the syntactic framework. Tenets (b) and (c) can be upheld, but I will argue against tenet (a). This analysis extends naturally to other tricky interface issues in Turkish which are due to its agglutinative nature.