LINGUIST List 5.1469

Sun 18 Dec 1994

Disc: Trends in linguistics

Editor for this issue: <>

Directory

1. Mark Liberman, Housen on trends
2. Annabel Cormack, RE: 5.1452 Sum: Trends and developments in Linguistics

In LINGUIST 5.1452, Alex Housen summarized responses from a number
of correspondents, including me, by saying:

 )Two general themes emerged from these responses:
 )1. As far as theory-building is concerned, the main areas of interest and
 )development for the future are believed to lie in the field of cognitive
 )linguistics, i.e. the relation of linguistic structure to cognition as
 )envisaged by R Langacker, G Lakoff, and others.
 )2. In terms of methodology, a stronger emphasis on corpus based
 )research, made possible by the ongoing micro-electronic revolution. The
 )increasing application of computers in linguistic research is expected to
 )bring about some major theoretical revolutions/paradigm shifts (cf. PDP,
 )connectionism).

What I sent Housen was a slightly edited version of the introduction
to a sort of review document that the Penn Linguistics department
produced, at the request of our administration, a couple of years ago.
This document is too long to burden the readers of LINGUIST with.
However, I would like to observe that it does not actually support
Housen's conclusions, which are his own; I disagree with them,
and I believe that my colleagues in the department here do as well.

Our review began

 Looking first within linguistics itself, we will identify nine
 subdisciplines that are useful in discussing our program, and briefly
 consider the intellectual trends within each of them. Following this,
 we will discuss two strong current national trends that involve
 cross-disciplinary research.

The bulk of the review was thus devoted to trends within areas such as
pragmatics, syntax, phonology, historical linguistics, etc.; within
each area, we discussed theoretical trends and our own program's
outlook. At the end, we added a discussion of two factors that cut
across traditional subdisciplinary boundaries: the rise of the work
that goes under the name of Cognitive Science, and the trend towards
computer-assisted study of large databases of text and speech to
provide broader empirical grounding. I am afraid that our discussion
of these points may have misled Housen.

With respect to the first point, there is an unfortunate confusion of
nomenclature. "Cognitive Science," as represented by the NSF Science
and Technology Center in cognitive science here at Penn, or the
cognitive science programs in quite a few other universities, is not
at all the same as "Cognitive Linguistics." With respect to the second
point, I do not believe that the rise of corpus-based research is
leading towards connectionism; it simply provides a valuable and
increasingly convenient source of empirical evidence for linguistic
research of whatever kind.

Let me offer one concrete example that makes both points. Lila
Gleitman (co-director of the Institute for Cognitive Science here at
Penn) and I are editing a multi-author volume intended for use in
undergraduate courses in cognitive science. One of its chapters, by
Steven Pinker, reviews a wide range of empirically-grounded studies of
children's acquisition of English past tense morphology, and offers a
carefully-reasoned argument against connectionist (i.e. analogical)
accounts of children's error patterns, and in favor of a (mainly)
rule-based account. This chapter is an excellent example of cognitive
science, in my opinion, although it has nothing to do with "cognitive
linguistics," and indeed seems contrary to some of the important
themes of "cognitive linguistics" as I understand them. The research
that Pinker surveys is corpus-based---indeed the documentation for the
CHILDES corpus is in the "suggestions for further reading"---but he
specifically argues for the empirical inadequacy of connectionist
models of the phenomena he treats.

 Mark Liberman mylunagi.cis.upenn.edu
 University of Pennsylvania

Message 2: RE: 5.1452 Sum: Trends and developments in Linguistics

I didn't summon the energy to reply to you r first query, but given what
looks like a somewhat unbalanced pair of conclusions, I am adding some not
 unbiassed comments of my own.

Theory: Where theory ought to be going, and has been, slowly, is towards a
reconciliation between different theories. It is anomolous in a scientific
discipline for there to be rival theories on such a scale as there are in
syntax, for instance. The other necessary reconciliation is between syntax
and various forms of semantics: formal semantics both lexical and
compositional, so-called cognitive semantics, and the sort of inferential
semantics needed for pragmatics.

I doubt whether any reconciliation between
computional methods based on probabilistic or PDP style methods, and formal
theory, is possible yet, but it has to happen one day. Corpus-based work is
a necessary preliminary to this, as is a good computational pragmatics.

Corpus based work should also provide useful input to construction of a
serious representation of the mental lexicon. Most theories put more
emphasis on the lexicon than a decade ago. However, I feel strongly that
more theoretical work on the syntax/semantic interface is needed. We also
need more work on parametric differences between languages in relation to
the lexicon, and need to take account of theories of mind from other
disciplines, and from psycholinguistics itself. Work on a large variety of
languages will and must continue to inform theory (and challenge
computatonal modelling).

Annabel Cormack.