[xmca] Book review

From: Phil Chappell (philchappell@mac.com)
Date: Sat Oct 01 2005 - 20:56:04 PDT


  AUTHOR: Geoffrey Richard Sampson
  TITLE: The 'Language Instinct' Debate
  SUBTITLE: Revised Edition
  PUBLISHER: Continuum International Publishing Group Ltd
  YEAR: 2005
  Announced at http://linguistlist.org/issues/16/16-1915.html

  LIU Haitao, Applied Linguistics Department, Communication University of
  China

  The book under review is a revised and expanded edition of a book with
the
  title "Educating Eve" (London: Cassell, 1997). In this edition, Sampson
  revisits and strengthens his original arguments against linguistic
  nativism using fresh evidence. [See
http://linguistlist.org/issues/9/9-734.html
  for a review of the previous edition. -- Eds.]

  "The 'Language Instinct' Debate" consists of six chapters with a
foreword
  by Paul M. Postal, who not only clearly presents the origin of the
  problem, but also quotes some of Chomsky's more tendentious arguments
for
  nativism, such as the following:

  "To say that 'language is not innate' is to say that there is no
  difference between my granddaughter, a rock and a rabbit. In other
words,
  if you take a rock, a rabbit and my granddaughter and put them in a
  community where people are talking English, they'll all learn English.
If
  people believe that, then they believe that language is not innate. If
  they believe that there is a difference between my granddaughter, a
rabbit
  and a rock, then they believe that language is innate." (p. viii, the
  quotation is from Chomsky 2000: 50)

  "The telephone exchange, for example, has 'heard' much more English
than
  any of us, but lacking the principles of universal grammar (inter
alia) it
  develops no grammar of English as part of its internal structure." (p.
x
  the quotation is from Chomsky 1981: 8)

  According to Sampson, one can believe that there is a difference
between
  Chomsky's granddaughter and rock, while at the same time, not believe
that
  language is innate. He opens the first chapter of the book with the
  assertion: "This book is written in order to establish that human
beings
  have no language instinct." (p.1) Instead of assuming that we are born
  with complex features of linguistic structure encoded in our genes,
  Sampson maintains the "common sense" position that languages are
  institutions like country dancing or the game of cricket: cultural
  creations which individuals may learn during their lifetimes. Sampson
  elaborates on this view by describing a character in Willy Russell's
  play "Educating Rita", which provides a vivid metaphor for the growth
of
  human knowledge. The story tells us that we don't inherit knowledge but
  rather the ability to gain knowledge. Substituting "Eve" for "Rita", we
  not only get the title of the first edition of this book but also the
  conclusion of the first chapter, namely that "Eve was not a born
know-all.
  She was ignorant. But she was a good learner." (p. 25).

  The nativist view is neatly expressed by Pinker's comparison of a
human's
  acquisition and use of language to a bird's nest-building or a dog's
habit
  of burying bones -- behavior programmed into the respective organisms'
  DNA. However, "The Original Arguments for a Language Instinct" (the
title
  of chapter two) are due to Chomsky, which Sampson summarizes as follows
  (pp. 30-36):
  (C1) Speed of acquisition: Children learn their first language
remarkably
  fast.
  (C2) Age-dependence: Language acquisition in childhood works quite
  differently from language acquisition in later life.
  (C3) Poverty of data: The child must induce the general rules
underlying
  the linguistic behavior of his elders from individual examples of that
  behavior - children are usually given little or no explicit instruction
  about the structure of their first language.
  (C4) Convergence among grammars: The various children in a language
  community all acquire essentially the same language as one another and
the
  same language that their elders speak.
  (C5) Language universals: All languages that are or have been actually
  used by human beings resemble on another with respect to a number of
  structural features.
  (C6) Non-linguistic analogies: Occasionally Chomsky refers to other
human
  cognitive achievements as resembling in being uniform across the
species.
  (C7) Species-specificity: Members of species other than Homo sapiens do
  not master human-like languages even when given access to experience
  comparable to that available to human children.

  In the remainder of this chapter, Sampson attempts to refute all the
  arguments for nativism based on these observations except for (C5),
which
  is taken up in chapter 5. Concerning (C1), Sampson contends that the
  argument has to be built on precisely observed data, "it is senseless
to
  claim that acquisition is in general 'remarkably fast'." (p. 37)
  Concerning (C2), Sampson seems more inclined to believe Bruner's
argument
  that "any subject can be taught effectively in some intellectually
honest
  form to any child at any stage of development." (p. 38) Concerning
(C3),
  Sampson denies that it is true, and also claims that Chomsky's argument
  from poverty of data to innate knowledge of language is self-refuting.
(p.
  43)

  If language isn't innate, "how [do] people really speak"? This is
focus of
  the chapter three. Chomsky's original arguments appealed to grammatical
  evidence, and grammar continues to form the central battleground for
  nativists and their opponents today. It might be expected that these
  arguments are characterized by copious references to examples of
real-life
  grammatical usage. Unfortunately, it isn't easy to find such examples
from
  nativist's works. According to Sampson, "This is because of a strange
  disdainful attitude on many nativists' part towards observable data."
(p.
  72)

  The first example Sampson discusses is from Carlson and Roeper's (1980)
  contention that "addition of prefixes to verbs rules out non-nominal
  complements." (p. 71) Sampson responds by listing eight
counterexamples he
  found using Google. Further on in this chapter, Sampson uses the
British
  National Corpus (BNC) as a test-bed to check claims about what
structures
  people do and do not use when speaking naturally. According to
Sampson, if
  the speakers in the BNC regularly use English in ways that a theory
  predicts no one uses, then the theory is wrong. (p. 74) Sampson devotes
  the next 15 pages to demonstrate that some structures in English aren't
  so rare as nativists have claimed. When some nativists said that the
  construction "modal + perfective + progressive" is rare, he found 61
clear
  examples in BNC/demographic corpus. In the same way, to some question
  structures considered by nativists are rare, Sampson also tries to
debate
  them using the data from corpus, he firstly classifies it into two
  subclasses 'S-within-MS questions' (subordinate clause within
main-clause
  subject, e.g. Will those who are coming raise their hands?) and
'S-before-
  MS questions' (subordinate clause before main-clause subject, e.g. If
you
  don't need this, can I have it?), and searches for them in the BNC. 23
  examples of S-before-MS questions are found. It is worth noticing that
  Sampson didn't find examples of S-within-MS in the spontaneous speech
of
  the BNC/demographic, but he interprets the absence of S-within-MS as
the
  difference between question patterns in speech and writing, in other
  words, he considers that the phenomenon that S-within-MS questions are
  missing from speech though not from written language can be made by
  literacy rather than innate knowledge. Sampson concludes: "The
linguistic
  data available to a young child are not poor. They are very rich." (p.
79)

  I believe that more and more linguists including nativists are
accepting
  the viewpoint that corpora are useful to linguistic research. As
  Meurers (2005: 1619) observes: "Theoretical linguistics requires
example
  sentences both as empirical basis for the construction of theories and
as
  counterexamples to previous generalizations. In addition to obtaining
such
  examples by introspection, electronic corpora can be used to search for
  examples which are relevant for a particular theoretical issue."

  In chapter four, Sampson discusses arguments for nativism from more
recent
  literature, including Bickerton (1990), Jackendoff (1993) and Pinker
  (1994), which Sampson considers representative of a new wave of
nativist
  thinking. Bickerton contends that there is a sharp discontinuity
between
  adult human language and various language-like systems, but Sampson
finds
  only smooth transitions bridging the "immense gulf" that Bickerton
  perceives. To Jackendoff's argument for nativism from the fact that
  children can readily learn American Sign Language, Sampson replies:
"these
  kinds of evidence do nothing to show that children have knowledge
built-in
  rather than seek it by research" (p. 109). Also: "If human beings are
born
  with a rich body of detailed knowledge of language, it is surely true
that
  we would expect to find some identifiable brain module housing or
  embodying that knowledge. But the fact that the brain has modules does
not
  in itself imply that innate knowledge of language occupies one of
them."
  (p. 108) Finally, although Pinker (1994) uses many of Chomsky's old
  arguments for nativism, Sampson considers them worth reexamining,
because
  Pinker develops them in more persuasive ways than his predecessors, as
in
  his summation: "Language is not a cultural artifact that we learn the
way
  we learn to tell time or how the federal government works. Instead, it
is
  a distinct piece of the biological makeup of our brains." (1994: 4) The
  central issue for Pinker is that young children seem just too good at
  language learning, essentially (C1) above. Part of Pinker's argument
  depends on assuming the concept of Mentalese, the one specific language
  that all humans are presumed to be born knowing: "Knowing a [specific]
  language, then, is knowing how to translate mentalese into strings of
  words and vice versa." (Pinker 1994: 73) Sampson replies: "If human
beings
  were all born knowing a single, universal Mentalese language, one might
  wonder why separate human communities would have developed separate
spoken
  language." (p. 131)

  Language universals often are considered as providing important support
  for nativism. Chapter five introduces presumed universals in sections
  entitled "words grow on trees", "the architecture of complexity",
"evolution
  everywhere", "trees and nothing but trees", "chunks and islands", "true
  and false universals", and "which way to the X-bar?" Sampson concludes:
  "yes, there are universal features in human languages, but what they
  mainly show is that human beings have to learn their mother tongues
  from scratch rather than having knowledge of language innate in their
  minds." (p. 166)

  In chapter six "The Creative Mind", Sampson systematically analyzes the
  Popperian view of human nature. This chapter seems to have been
prepared
  for the possible nativists who perhaps will criticize Sampson's
empiricist
  view as incoherent.

  Sampson's book is worth reading, because it provides a view of how
human
  languages work without appealing to nativist assumptions. It cautions
that
  scientific arguments should be based on reliable data and that
linguistics
  is no exception. Corpus linguistics makes available many tools for
finding
  examples needed in linguistic study, and linguists should use those
  resources in addition to introspective and experimental data. However,
  empirical data can answer some questions, but not all of them. For
  instance, even though computational linguistics demonstrates the
  advantages of basing language processing systems on empirical data
(Bod,
  Scha & Sima'an 2003), computational linguists generally believe that
the
  best solutions will combine rational and empirical elements.
Computational
  linguistics can teach us that it is not a good idea to claim something
  absolutely. Returning to the problem at hand, why can't a rock learn
  English even if it is in the same circumstance as a normal child?
Presumably
  because a child has a computational faculty in its brain, but a rock
has
  no such thing. The faculty is innate, but may not be special for
language;
  nevertheless a child can still acquire language using this faculty. If
  this is correct, then the difference between Sampson's empiricism and
  Pinker's nativism may be gradient rather than categorical. Indeed,
perhaps
  there is a middle point which almost everyone can accept. For
discovering
  the middle point, nativists will need to find more evidence from
corpora
  containing real language usage, and empiricists should ask themselves
what
  are the elements of the faculty, using how a child can learn language.

  In chapter one, Sampson argues: "Despite Pinker's verbal pyrotechnics,
  there is actually no such thing as a human language instinct. There
really
  isn't. Chomsky's arguments for it do not work; and Pinker's arguments
do
  not work either. What they are telling us just ain't so. Believe me,
it is
  not. The rest of this book is designed to convince you of that." (p.
14)
  It is interesting to compare this perfervid passage with the following
  from Pinker (1994: 4): "In the pages that follow, I will try to
convince
  you that every one of these common opinions is wrong! And they are all
  wrong for a single reason."

  It seems to me that both positions are too extreme and absolute; the
  correct solution often can be found between two extremes. Perhaps one
of
  the central tasks of linguists is to find the balancing point between
  these two extremes, where the solution may well lie.

  I have recommended Pinker (1994) to my colleagues and students, and
almost
  all of them have told me that it is one of the best books that they
have
  read about language. Sampson agrees that Pinker's book "is superbly
well
  written", but he also says "a book can be well written, and its
  conclusions quite wrong" (p. 14). I will now also recommend Sampson's
book
  to my colleagues and students, and let them judge between the two.

  REFERENCES

  Bickerton, Derek (1990) Language & Species. University of Chicago
Press.

  Bod, Rens, Remko Scha and Khalil Sima'an, eds. (2003) Data-Oriented
  Parsing. Stanford: CSLI.

  Chomsky, Noam (1981) On the representation of form and function, The
  Linguistic Review 1: 3-40.

  Chomsky, Noam (2000) The Architecture of Language, New Delhi: Oxford
  University Press.

  Carlson, Greg and Thomas Roeper (1980) Morphology and subcategorization
  and the unmarked complex verb. In Teun Hoekstra, Harry van der Hulst,
and
  Michael Moortgat (eds.) Lexical Grammar, pp. 123-164. Dordrecht:
Foris..

  Jackendoff, Ray S. (1993) Patterns in the Mind: Language and Human
Nature.
  Harvester Wheatsheaf.

  Meurers, W. Detmar (2005) On the use of electronic corpora for
theoretical
  linguistics: Case studies from the syntax of German. Lingua, 115(11):
1619-
  1639.

  Pinker, Steven (1994) The Language Instinct: How the mind creates
  language. New York: HarperCollins.

  ABOUT THE REVIEWER

  LIU Haitao is professor of applied and computational linguistics at the
  Communication University of China (CUC). His research interests include
  syntactic theory, computational linguistics and language planning.
_______________________________________________
xmca mailing list
xmca@weber.ucsd.edu
http://dss.ucsd.edu/mailman/listinfo/xmca



This archive was generated by hypermail 2b29 : Tue Nov 01 2005 - 01:00:20 PST