Corpora: Chomsky and corpus linguistics

From: Geoffrey Sampson (
Date: Wed Apr 25 2001 - 11:30:05 MET DST

  • Next message: James L. Fidelholtz: "Re: Corpora: Chomsky and corpus linguistics"

    I find Amanda Schiffrin's approach to these issues extremely congenial
    (generative v. corpus-based approaches being like top-down v. bottom-up
    approaches to the same problem, which with luck will eventually meet
    in the middle). I think there are signs that the overall discipline is
    developing in a way that makes hers an increasingly accurate description
    of the situation -- I very much hope so.

    The problem which has been aired in earlier postings is that, regrettably,
    in the past, and to a considerable extent even today, things have not
    been like that. Many generative linguists have made it clear that they
    not only were working at a distance from the empirical data but saw
    empirical data as in principle largely irrelevant. Concepts such as
    "competence" v. "performance" (and more recent generative terminology
    which I have not kept up with) were used to shield generative accounts
    from any possibility of being tested against observation. Mandy writes
    that theories which cannot be backed up by "real" evidence should be
    "discarded without a second thought", to which three hearty cheers; but
    there has been a strong current within academic linguistics of "If
    performance data seem incompatible with our theory, discard the data without
    a second thought". (I realize that these are large statements which I
    myself am making here without giving chapter and verse -- I do the latter
    in parts of my recent book _Emprical Linguistics_, apologies for the plug.)

    The attitude I am describing has been so widely and strongly asserted during
    the period of my own career that the effort of disagreeing with it has
    left me feeling slightly psychologically marked, as perhaps is apparent
    from this posting! If Amanda's comments imply that a new wave of people
    are coming forward for whom being empirical just isn't an "issue" any
    more, that is really good news.

    Geoffrey Sampson

    G.R. Sampson, Professor of Natural Language Computing

    School of Cognitive & Computing Sciences
    University of Sussex
    Falmer, Brighton BN1 9QH, GB

    tel. +44 1273 678525
    fax +44 1273 671320

    This archive was generated by hypermail 2b29 : Wed Apr 25 2001 - 11:25:53 MET DST