LINGUIST List 4.952

Fri 12 Nov 1993

Disc: Psycholinguistic evidence in linguistics

Editor for this issue: <>

Directory

1. Anjum Saleemi, Re: 4.934 Psycholinguistics
2. , Re: 4.934 Psycholinguistics
3. Steven Schaufele, linguistics as psychology, as anthropology, as ... ?

Message 1: Re: 4.934 Psycholinguistics

I'd like to contribute once again to the discussion that (I think) was started
by Mike Hammond's comments about the increasingly greater acceptance of
experimental results in mainstream linguistics. It seems to me that my earlier
remarks about the validity of introspective/intuitive judgements of
grammaticality have been partly misunderstood, as some of the responses to
those remarks (those sent to me directly as well as those sent to LINGUIST)
obviously indicate. Below I'll make an effort to respond to some of the
reactions.

(a) Re Esa Itkonen's remarks: I'm well aware that the debate regarding the
relationship between linguistics and psychology is not new; the debate,
however, remains unresolved, and therefore, deserves to be reopened. By the
way, it's not a question of the new generation reinventing the wheel, as
no wheel was invented in the first place: the best we can do, as I said
earlier, is to reopen the debate, probably in terms not quite the same as those
prevalent in the seventies, since so much water has flown under the bridge.

(b) The distinction between "internal" and "external" evidence (Wendy Sandler,
Arnold Zwicky cited by Vicky Fromkin) is not quite clear to me. I think ALL
evidence is external, whether introspective, experimental or naturalistic, as
they're all based on performance (as Fromkin points out), though one type of
evidence might be less biased, less affected by the various confounding factors
than the other types; so Mike Maxwell (direct response) argues that intuitive
judgements constitute better linguistic evidence than the other varieties of
such data. Now if experimental "psycholinguistic" (i.e related to processing)
and naturalsitic (e.g. some acquisitional) evidence are inferior to
introspective data, there is still no guarantee that the latter is the best
type of evidence we can ever get: if the typical psycholinguistic evidence is
not too reliable, and introspective data are indeed more reliable, then that
doesn't mean that EVEN BETTER data can't be obtained by means of some other
(possibly new) techniques for the verification of facts.

(c) To use a concrete example: Hindi-Urdu has split agreement, and in some
cases object-verb agreement is possible. Some recent studies (e.g. the very
insightful work of Mahajan) claims that object-verb agreement is forced by
specific interpretation, i.e. the object agrees with the verb only if it has
a specific referent. Now many speakers of the language, including myself, don't
get the specificity effect at all! Is it there, or not there? Is it there for
some speakers, and not for others? The problem is that an analysis such as
the one I have mentioned assumes that the specifity effect is universal, so
the individual variation shouldn't exist. But it does! How do we know that what
we're encountering is a case of I-language variation, and not the result of
someone's theory colouring his judgements of grammaticality? The point I'm
trying to make is that so long as robust data are concerned, the question of
employing any non-introspective evidence is not very relevant, but the moment
we begin to handle non-robust data, problems arise, and one would like to know
if any different techniques for the verification of facts (I mean other than
the "psycholinguistic" and introspective ones) are available that would help
one settle the issues one way or another. In this context, it appears to me
that us linguists have done very little to improve our methodology: if the
tools that we borrowed from behavioural psyhcology don't work, we're happy to
declare that the best method is still the good old intuitive method, and
nothing better will ever be devised.

(d) Re Edith Moravcsik's comments: I don't I disagree with her all, though
probably something I said made her think that she had to disagree with me. To
be more specific, I never said that "the more traditional structural approach"
should be or can be replaced by what are presumably the less traditional
non-structural methods; in fact I have no trouble with accepting that the two
approaches must be held to be "complementary," or that structural analysis
provides "a necesary basis for the psychological study of language." What does
bother one, of course, is the fact that non-structural evidence is often
readily accepted by structural linguists (as Fromkin points out) if it confirms
their results, but is conveniently ignored otherwise, providing support to
the claim (re Sam Wang) that the competence-performence distinction is often
used as "an easy escape from counter-evidence in experimental results." (Wang)

Let me point out, before concluding, that I'm NOT and experimental
(psycho)linguist, and therefore have no axe to grind.

Anjum Saleemi
National University of Singapore
ellapsnusvm.bitnet OR ellapsleonis.nus.sg

Message 2: Re: 4.934 Psycholinguistics

The current discussion of psycholinguistics contains a near concensus on
the idea that it is wrong for a linguist to accept evidence from
psycholinguistic experimentation only when it supports his/her theory.
The appropriate anecdote here concerns the apocryphal student who
objected to a professor's theory by shouting "But this theory of yours
-- it's unfalsifiable!" To which the professor replied "Well I
certainly hope so!"

All in all, the professor had, I think, the right attitude. If we had
some psycholinguistic paradigm P whose interpretation in terms of
property-of-human-language-use L were not in doubt, then it would be
irresponsible of a linguist presenting a theory about L to ignore data
from P. But I don't know of any such paradigm at the moment. Instead,
we have the more usual situation in science, where results from a
variety of domains are converging in various places (in a quite exciting
fashion), but without an absolutely perfect fit. One valid program of
research is to continue to work on results in the individual areas,
taking comfort from convergence when it is evident, and putting aside
divergence when it looks irrelevant to present tasks. That's what we
always do when we have more data floating around us than we can handle
with our theories -- internal to traditional linguistics or in the
broader world of language sciences. Another valid program of research
is to study the divergences directly, and perhaps ultimately develop the
Holy Grail P described above.

A good example of the latter approach, in my view, is the recent
discussion in the acquisition literature of whether (Guilfoyle, Radford
et al.) child grammar circa age 2 involves phrase structures without
functional categories. This view leads to a divergence between results
from acquisition work and adult theoretical syntax, given the crucial
role of functional elements as case-theoretic and morphological glue in
many theories of the syntax of adult speech. Here, work on this
divergence has yielded an alternative theory (Wexler and Poeppel et
al.), according to whivh the phrase structures of this period are
actually adult like (w.r.t. functional elements), with some other
factor accounting for the facts that lead to the no-functional category
view. As for work going in the other direction, results on obedience
and non-obedience of children to Principle(B) may favor one view on
coreference in syntax (Reinhart's) over the more popular view, thus
deciding a linguistics question.

A good example of success in the former approach (again, my view only)
was the unwillingness of some linguists in the 1970s to be persuaded by
the lack of easy psycholinguistic evidence for filler-gap effects with
A-movement into abandoning the hypothesis that A-movement/A-trace
represents a real phenomenon. Otherwise, we would not have the results
on A-movement in, for example, Rizzi's "Chain Formation" paper,
Miyagawa's discussion of numeric quantifier stranding in Japanese, and
much else. These are results we have today only because certain
psycholinguistic data were not interpreted as falsifications of a
hypothesis from theoretical linguistics. (The opposite response, of
course, was made by Bresnan in her "Realistic Grammar" paper, which also
yielded an important program of research.) In this domain, there still
may be a divergence problem (late-1980s work by MacDonald, Bever and
others on priming by A-trace might narrow the divergence considerably),
but we also know a lot more about the character of the divergence than
we would have known if we had been too quick to cry "falsified" in the
face of divergence problem.

-David Pesetsky

In LINGUIST 4.934 Esa Itkonen <EITKONENsara.cc.utu.fi> raises the question
'Is linguistics just a subbranch of psychology?', and mentions three
possible alternatives: (1) linguists who 'systematize the intuitive notion
of "grammatical (and meaningful) sentence" while paying no attention to,
and even going against, any reasonable hypotheses about psychological
structures and/or processes and who, nevertheless, achieve exemplary
results.' (2) linguists who claim 'to be doing
psychological/psycholinguistic research' but are actually pursuing research
programs analogous to those of the first type. (3) linguists 'who both
claim to be doing and are in fact doing psycholinguistic (= preferably
experimental) research.'

I'm not sure this covers all of us. For one thing, i'm not at all sure
where i fit into this typology. Perhaps this is the time to remind
everybody that, historically, the discipline of linguistics in Europe (and,
as far as i know, elsewhere in the 'Old World') developed out of such
disciplines as classics, liturgics, and literary studies -- basically
humanistic (in a rather broad sense) type stuff. Here in North America, on
the other hand, linguistics as an academic discipline developed in
connection with anthropology; consider Boaz, Bloomfield, etc. Granted, in
the 19th century we had some American linguists trained in the philological
traditions of European linguistics, e.g. Whitney, but academic linguistics
really took off here in the early 20th century with Boaz, Bloomfield, and
that whole school of anthropological linguistics directed towards the study
of 'indigenous'=non-Indo-European languages. The notion of linguistics as
a branch of psychology is simply another point of view ('aspect' in the
literal sense) on the study of human language. As far as i know, it is
relatively new -- a product of the 'generative enterprise'. In any case it
is an enrichment of our field, but i don't think it should be allowed to
supplant the older aspects (in this respect i am in complete agreement with
Edith A Moravcsik <edithconvex.csd.uwm.edu>'s remarks in the same LINGUIST
posting).

Now, personally, i subscribe wholeheartedly to the generative agenda as far
as the goals of linguistics are concerned: i view our purpose as
illuminating an important part of human cognitive ability. In this
respect, when i am asked 'What is linguistics?' or 'What do linguists do?'
or 'What is linguistics good for?' i tend to give an answer that implies
that linguistics is a branch of psychology or, at least, that it is
primarily a 'cognitive' science. But i find, coming down to brass tacks,
that when i am actually doing research i am functioning more like an
anthropologist: collecting linguistic-behavioural data from a variety of
sources that vary along a set of axes such as ethnicity, geographical
location, chronology (both biological age and location in time, as distinct
from space), social status/function, etc. and trying to relate the observed
variation in linguistic behaviour to these extra-linguistic variables in
the manner of an anthropologist. I suppose one could start from the same
origin and address the whole business from a sociological point of view,
but it's the anthropological parallel that seems to work best for me. Note
that literary studies haven't been mentioned here, even though at the
moment i am primarily involved in the study of literary corpora; i may be
looking at the same sort of data a literary scholar might study, but i'm
treating it the way i imagine an anthropologist treats the behaviour hann
observes in the field. So although i am prepared to recognize linguistics
as certainly related to psychology, and even individual linguists as, in a
sense, basically psychologists whose area of research and expertise happens
to be human language, i have trouble seeing myself in that mold. I am
fascinated by what linguistics can tell us about human psychology and
cognition, but that's not the way i pursue linguistic research.

Perhaps this is what Esa meant by the 'HINT: accept the existence of
dissimilar objectives, but reject contradiction between words and deeds'?

Best,
Steven

------
Dr. Steven Schaufele 217-344-8240
712 West Washington Ave. fcoswsux1.cso.uiuc.edu
Urbana, IL 61801

*** O syntagmata linguarum liberemini humanarum! ***
 **** Nihil vestris privari nisi obicibus potestis! ****