LINGUIST List 2.96

Monday, 25 Mar 1991

Disc: Cognitive Linguistics

Editor for this issue: <>

Directory

1. Vicki Fromkin, Re: Responses on Cognitive
2. Scott Delancey, autonomous linguistics
3. AVERY ANDREWS, Autonomy/Cognitive Linguistics
4. John Coleman, Comment
5. "Bruce E. Nevin", connectionist/cognitivist TR

more grist for the mill department:

There are two, not one modularity questions: (1) language (or language faculty)
is independent, autonomous, genetically pre-wired, non derivative of non-
linguistic cognitive abilities, and (2) within the grammar, components are
independent of each other, i.e. syntactic principles or constraints are
distinct from semantic, phonological etc, in organization, structure, function
and processing.

My last remarks referred to the 1st question. Both are empirical questions
requiring evidence but not necessarily the same evidence. One hypothesis
can be true while the other is false.

Vicki Fromkin

A couple of earlier contributions to the "cognitive linguistics" debate
left me fighting the temptation to submit a symmetrical but probably
not terribly productive rejoinder along the lines of "when these formalists
have something useful to say about really interesting problems like
split ergativity, free word order, grammaticalization, or the alternation
of full NP's and anaphoric devices in connected discourse, they'll
have something worth taking seriously ..." Fortunately, some recent
contributions seem to be probing at the same point less contentiously.
Poser and Everett, among others, seem to treat "cognitive" and "functionalist",
in this context, as synonymous. This is both correct and incorrect (ass
any functionalist or cognitive grammarian would predict, of course; this
is how catgorization works); the essential point in which it is correct
is that all of the various research programs (there are at least three
clearly distinguishable ones) which fall under one or the other of these
terms share an unwillingness to accept a priori the assumption that
significant aspects of morphology and syntax (phonology is likely a
different story) are to be explained only in terms of language-specific
formal priniciples. Obviously researchers with this set of background
assumptions will not feeled inclined to spend (i.e., from this perspective,
waste) much time or effort looking for or at the sort of formal conditions
on grammars which Pesetsky or Poser are more interested in. I think
there is a widespread sense within this community that many of these
phenomena (e.g. locality conditions) are artifacts of the formal approach:
if you approach the data expecting to find phenomena which have purely
structural explanations, that's likely to be what you end up seeing
in the data.
 In this connection I have a comment on someone's suggestion
(Manaster-Ramer, I think--or maybe it was that dog?) that if we all
just pursue our own research programs, the truth will out, and those
who were pursuing false leads will see their error and return to the
light. I hope, and half believe, that as a practical proposition this
is probably true, but it certainly doesn't logically have to be. One
of the arguments for the functionalist research program (articulated
in print by Givon in On Understanding Grammar) is that it is indeed
subject to such disconfirmation, while the structuralist (i.e. formalist)
program is immune. If we start with the assumption that most of the
interesting facts about linguistic structure can be explained in terms
that are based in general cognition or communicative function, and we
are wrong, then as research proceeds we will be left with a more and
more obvious and irreducible body of facts for which no such explanations
can be found, and will finally be forced to conclude that these
data reflect a set of purely linguistic principles of structure. But
if we assume from the beginning that our data will be susceptible of
purely structural explanation, there is no incentive to look for more
"explanatory" explanations (in the sense that George Miller refers to
in the June 1990 number of Language), and no inevitable crisis in the
research program that will force linguists to consider alternative
possibilities.

Scott DeLancey
University of Oregon

Three bits' worth on autonomy/cognitive linguistics:

1. Things that look pretty autonomous now might look a lot less
 so if we knew more about how action is coordinated & parsed.
 I would expect NL syntax to be more closely related the low-level
 organization of action than to higher-level strategy & tactics.
 I'd also suggest that the facilities that pre-hominids must have
 had for parsing the behavior of their associates, predators &
 prey (if any) would be the most likely pre-adaptation for
 syntactic parsing.

2. Autonomous adaptations for efficient language-learning via a
 constrained UG have a problem in getting started, in that they
 won't confer a selective advantage on their possessors unless these
 happen to be members of a population using languages that already
 happen to be appropriately constrained, which isn't likely to happen
 often, if at all. So I'd suggest that the quirky features of UG
 are due to inherent & perhaps accidental properties of the facilities
 that were co-opted to support language use, rather than UG being
 the way it is in order to make languages more learnable (in slogan
 from, language is the way it is because that's what happens to be
 easily learnable by humans, not in order to make it learnable). The
 language-learning adaptations would consist in a greater allocation
 of resources to the facilities being used for language, including
 practice time, physical space, etc.

3. Nonetheless, although I doubt that any strong form of autonomy
 is actually true, I think it has the good effect of encouraging
 people to look for and develop things that are bizarre and unexpected,
 and, I hope, fundamentally instructive, such as the vagaries of
 quirky case in Icelandic, or Pesetsky's LF-moved vs. LF-stationary
 Wh-words. Non-autonomist work faces and periodically succumbs to the
 threat that workers will just use language to corroborate what they
 already think they know about human nature, rather than to actually
 try to find anything out. Language being such a large and confusing
 subject, people can find pretty much anything they are looking for in
 it, but the autonomist position tells you only to look for the unexpected.

 -- Avery Andrews (ada612csc.anu.edu.au)

Bill Poser writes:
> On the general point of sophistication in this area, I'm well
> aware of the work of many AI people, and that some Natural
> Language Processing people ... are quite sophisticated
> linguistically. But the gap overall is still quite large. And I
> would say it is even larger when one looks at morphology,
> phonology and phonetics, closer to my own area. ...
> An example from phonetics is
> the statement in Douglas O'Shaughnessy's book _Speech
> Communication: Human and Machine_ (p. 62) that "Except for trills
> and ingressive sounds ... English provides good examples of
> sounds used in various languages...", which gives the misleading
> impression that the phonetic inventory of English covers most
> known speech sounds. In addition to trills and ingressives,
> English lacks glottal ejectives, rounded front vowels,
> retroflexes, uvulars, bilabial fricatives, pharyngeal fricatives,
> nasal fricatives, and pharyngealization, among others. If we
> consider distinctive oppositions we may add to this list still
> other categories, such as aspiration, nasalization, and voiceless
> sonorants. 
If we are not considering distinctive oppositions at first, then Poser's
statement that "English lacks glottal ejectives, rounded front vowels, ..."
is even wronger than O'Shaughnessy's, so long as we're talking about
phonetics, as Poser says. Trills, ingressives, ejectives, rounded front 
vowels, retroflexes, uvulars, bilabial fricatives, pharyngeal fricatives,
nasal fricatives, and pharyngealization can all be observed in the
speech of English speakers, and some of these are actually regular 
in some varieties of English.

> this statement [O'Shaughnessy's] and others like it show a very
> limited knowledge of the range of speech sounds and inventories
> in the world's languages. 
Poser may know a lot of segmental phonological analyses of 
"the sounds of the world's languages", but that is not the same
thing as "the range of speech sounds ... in the world's languages".

--- John Coleman

[A cross-post of potential interest to this distribution, from
[Neuron Digest Saturday, 23 Mar 1991 Volume 7 : Issue 14

[ FYI the digest header includes the following notice:

[Send submissions, questions, address maintenance and requests for old issues to
["neuron-requesthplabs.hp.com" or "{any backbone,uunet}!hplabs!neuron-request"
[Use "ftp" to get old issues from hplpm.hpl.hp.com (15.255.176.205).

 Bruce Nevin
 bnbbn.com

-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=- cut -=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-

Subject: TR - Connectionism and Developmental Theory
From: Kim Plunkett <psykimpaau.dk>
Date: Fri, 22 Feb 91 11:47:37 +0100

The following technical report is now available. For a copy,
email "psykloneaau.dk" and include your ordinary mail address.

Kim Plunkett


+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++


 Connectionism and Developmental Theory

 Kim Plunkett and Chris Sinha
 University of Aarhus, Denmark

 Abstract

The main goal of this paper is to argue for an ``epigenetic
developmental interpretation'' of connectionist modelling of
human cognitive processes, and to propose that parallel dis-
tributed processing (PDP) models provide a superior account
of developmental phenomena than that offered by cognitivist
(symbolic) computational theories. After comparing some of
the general characteristics of epigeneticist and cognitivist
theories, we provide a brief overview of the operating prin-
ciples underlying artificial neural networks (ANNs) and
their associated learning procedures. Four applications of
different PDP architectures to developmental phenomena are
described. First, we assess the current status of the debate
between symbolic and connectionist accounts of the process
of English past tense formation. Second, we introduce a
connectionist model of concept formation and vocabulary
growth and show how it provides an account of aspects of
semantic development in early childhood. Next, we take up
the problem of compositionality and structure dependency in
connectionist nets, and demonstrate that PDP models can be
architecturally designed to capture the structural princi-
ples characteristic of human cognition. Finally, we review a
connectionist model of cognitive development which yields
stage-like behavioural properties even though structural and
input assumptions remain constant throughout training. It is
shown how the organisational characteristics of the model
provide a simple but precise account of the equilibration of
the processes of accommodation and assimilation. The
authors conclude that a coherent epigenetic-developmental
interpretation of PDP modelling requires the rejection of
so-called hybrid-architecture theories of human cognition.