Summary of Clancey, Chap. 9
Transformational Processes, Grammars, and Self-Organizing Strings
- This chapter tries to relate conceptual coordination to Piaget's and
of learning, as well as to Langacker's theory of Cognitive Grammar.
Problem is one of relating two types of development processes :
- Biological (molecular genetics/mechanistic issues)
- Cognitive (knowledge)
We have to view coordination processes as both the
- carrier (vehicle) and
- constructor of memory,
i.e. content and assembly, in order to disentangle the dichotomy.
Use Kauffman's theory of self-organizing processes to relate the two?
- Piaget's transformational Processes:
- Reflective Abstraction, and
are relevant to conceptual coordination. Reflective abstraction in
particular can be shown to have parallels to conceptualization, as shown in
Table 9.1, pg. 224.
The second process, abstraction, fits (C's) claim that every
categorization is a generalization. Piaget's main point:
The ability to comprehend, focus, make sense in logical ways (...)
develops in experience.
[I omit further
of Piaget because I don't have a way to summarize it.]
Mental Physiology: provides another, competing? opposing? account of
| Note: these green (deleted from print version)
blocks will be my way of injecting my own opinions about things I don't
agree with, or have a critique of. |
- First problem is that
I think that Clancey bases his understanding of Chomsky on older
versions of Chomsky's theory, going back to 1965, 1981, etc. when now many
things have changed.
- Second problem I have is that Chomsky's
view of what goes on in language is always focused on what he calls the
ideal speaker-hearer who never makes a mistake, never has a doubt
about his/her intuitions about language, always knows what is grammatical
and what is not, never is concerned about the social appropriateness of
utterances, never thinks about grammaticality beyond the level of the
sentence, and so on. Such persons, of course, do not exist. This
perception of language is known as ``autonomous" theory, i.e. that
language structure is independent of personal and social variables. But
among `autonomous' theorists there is a tendency to lapse into treating
oneself as the ideal speaker-hearer, i.e. the intuitions
of the `speaker who doesn't
exist' can be found by consulting one's own intuitions. One doesn't go
out and get data, one sits in one's office in the linguistics department. and
consults one's own brain.
Result: Chomsky's perception of language is not based on empirical
observation of language behavior, or even empirical observation of neural
activity, but on his own (admittedly brilliant) ideas and intuitions about
how language must be structured.
- Because Clancey only consults the earlier writings of Chomsky, there
overemphasis in Clancey on the notion that grammar is transformational
when in fact the earlier emphasis on this notion has been largely
abandoned. What happened was that everybody and his brother went out and
began proposing (a) elaborate "deep structures" for the simplest
and (b) elaborate transformations to derive these deep structures and
convert them into (c) surface structures. People, Chomsky included, began
to have doubts about these elaborate abstract structures, wondering (e.g)
young children could even begin to learn them. The elaboration
of vast complicated deep structures and transformations also meant that the goal of
proposing some kind of underlying universal grammar became more
elusive: languages were just too different.
- Result: elaborate deep structures and transformations were abandoned, and
the goal was to find a way to get to the complexity and great difference
that we find on the `surface' of natural languages, but derive it somehow
from some finite, small number of universal features.
Some people (former transformational/generativists, such as Langacker)
abandoned this line of theorizing altogether, but among the true believers
(loyal to Chomsky), the emphasis became one of proposing generative
grammars, that generated all the grammatical sentences of the
language, but also generated some that were ungrammatical. These needed
to be filtered out, so what had to be proposed them was a filtration of
interpretive system, which were different for different
languages, and would get rid of the weird sentences that the grammar would
generate, but were not appropriate. (This was also thought to take care
of the social and other 'appropriateness'
- Cognitively, this interpretive component bothers many people (incl.
me) because it means that we are constantly judging the grammaticality or
appropriateness of a vast number of ungrammatical sentences generated by
our internal grammars. Many people feel there must be some other way to
Clancey's view of what Chomsky is saying is given
in the first quote on pg. 226. (About the only part
of this that most people would still accept is the idea that nouns, verbs,
etc. are `likely to prove universal deep-syntactic categories of human
languages'). Another claim is that grammars etc. consist of or are
based on meaning-independent processes, i.e. that they operate at
a deep level without meaning, and meaning is assigned later by
the interpretive component. [This is also problematical for many.]
Clearly Chomsky's theory won early adherents among the techie community
because (as Clancey says) "it was determined that transformational
grammars could be formulated as augmented transition networks (ATNs)." The
generativity part of this is still accepted by many, but not the emphasis
on transformational grammar. (See fn. 2, pg. 226).
Not clear to me whether Clancey is waffling about Chomsky on pg. 227, when
he says that Chomsky's view is controversial (true) and maybe misunderstood
(also possibly true, but maybe not always). In the opposing view (e.g.
Langacker 1986, theory of cognitive grammar ) patterns are
representations of experienced expressions:
Grammar is not a distinct level of linguistic representation, but reduces
instead to the structuring and symbolization of conceptual content...
Lexicon, morphology, and syntax form a continuum of symbolic units,
divided only arbitrarily into separate "components"---it is ultimately
as pointless to analyze grammatical units without reference to their
semantic value as it is to write a dictionary which omits the meanings of
its lexical units. (pp. 1-2)
Clancey lists the three options as some kind of continuum or spectrum:
- Piaget's position: general, uniform learning principles;
- Chomsky's position: meaning-general, linguistic mechanism;
- Langacker's position; conceptual patterns.
Langacker's claim: linguistic structure is not autonomous or
meaning-general, and does not involve an inherited language organ
but rather is conceptually constructed as a generalization
of actual utterances. [ i.e., it is based on
experience: children develop language from the input they get. ]
"Conceptualization [...] encompasses novel conception as
well as fixed concepts; sensory, kinesthetic, and emotive experience;
recognition of immediate context (social, physical, linguistic); etc. Pg.
- grammar and lexicon on a continuum of conceptualization
- all linguistic patterns derive from (i) semantic, phonological and symbolic
structures that occur overtly in linguistic expressions; or, (ii)
that are schematic for us in (i); and (iii) categorizing relationships involving
the structures in (i) and (ii).
- This rules out arbitrary descriptive things, such as dummy
elements, contentless features, abstract underlying structures
that are very different, e.g. passives from actives.
- Clancey: Langacker's sketch fits the perspective of neural map
activation and composition. (Quote pg. 228 about tree-structures,
But here Clancey gets confused: he seems to favor what Langacker says
("relationships conceptually constructed in the past are organizing new
utterances; previous sequences are being reactivated, generalized, and are
operating on different levels of abstraction. etc. etc.") But then
Clancey says that "all is captured by transformational grammar."
Langacker would deny this, many others would deny this, and even Chomsky
would have doubts. The word is generative,
- Now Clancey goes on to characterize Chomsky's claims:
- Linguistic ability reflects an evolutionary development of the
brain and its physical properties, not just ... learning. [hs: Okay, humans are different from apes etc.]
- Linguistic ability reflects a specialization of biological
function, just as in other physical processes. [hs:
but how come this so-called linguistic "organ" doesn't show up in one
place in the brain, but in many different locations? How is it different
from growing feet instead of wings? Chomsky used to use the example of
'bipedal gait' i.e. walking upright, using same 'organs' as other animals,
but different gait. ]
- Conceptualization is more than sensorimotor coordination, and is
just linguistic; but semantics can't be understood without it.
[hs: Then quotes Chomsky as saying it should be possible for
conceptual system to be intact and functioning even if the linguistic
system is not. ... [Hello? ]
- Linguistic syntax is produced by a device ... that is general
human languages. (Chomsky 1957; could we have a
- The rules of syntax are not arbitrary descriptions of patterns, but
restricted to certain structural characteristics of phrases.
Here Clancey chides Chomsky for appearing to `accept uncritically the
notion of a stored-rule mechanism.' (Chomsky quote: "Our knowledge of
language is ... a system of rules and principles that enter into a capacity
[to do certain things]."
Here Clancey clearly favors Langacker, disfavors Chomsky;
Clancey/Langacker want a mechanism in which 'patterns are
embodied in coordinated sequences of behavior.' Thus, sequences that seem
to correspond to linguistic phrases must be integrated with multimodal
attention and physical coordination. But then Clancey states himself to
favor Chomsky's 'mental organ and mental physiology.'
- The processes of categorization, sequencing and generalization
occur uniformly in normal adult humans, thus they are universal.
- But, categorization of sequences, a.k.a. conceptualization
involves additional aspects of coordination in humans, including:
- Berwick's appeal to connectionism.
[I am not competent to discuss this.]
- Kauffman's Autocatalytic symbol strings.
[I am not competent to discuss this, either.]
Postscript: For Chapter 10, in the discussion of `center-embedded'
sentences, with the limit of three, I have been collecting examples of
'real' sentences of this sort:
- [Hopes [that the crew of a tug [that sank yesterday in the Gulf of
Alaska]3 will soon be recovered]2 are rapidly fading.]1
- [Reports [that the Van Gogh painting [that was stolen last month from
the Louvre]3 had appeared on the art
market at an asking price of $6
million]2 were declared to be false today
by the French authorities.]1
- [Doubts [that the validity of the hypothesis [that was proposed by the
author of this book]3 will be
common among theoreticians of the old school.]1
- [The accusation [that the pilot of the plane [that crashed near Hainan
Island last week]3 was a daredevil
risk-taker with a history of crazy
stunts]2 was vehemently denied by the
- [A case involving a claim [that a stolen copy of the Treaty of
Fontainebleau [which was declared missing from the French National
Archives 12 years ago]3 had been put up
for auction at Sotheby's in 1995]2 will
soon go to trial.]1
The example sentences Clancey cites [``The book [that the man [who hired
me] wrote] deals with politics"] often give the impression that even 3
embeddings are hard to decode; it seems to me that the more fully
specified they are, the easier they are to decipher. They're also
more common with psychological verbs (`doubt, hope, fear, believe') or
expressions derived from psych verbs (`doubts, hopes, accusations,
Last modified 4/9/01.