Message from Myrna Gopnik

[Liz Bates]

At some point last week, I composed and tried to send a response to Fritz
Newmeyer¹s critique of Vargha-Khadem, but I never saw it on the net, and
several friends have commented that they didn¹t see it either.  On top of
that, I have had private inquiries from a number of Funknetters asking if I
intended to respond, so I am feeling compelled to give it one more try.
Please note that this is a response to Fritz¹s message.  I want to spend a
little more time studying the longer account that he forwarded today from
Myrna Gopnik before deciding whether I would have anything useful to
contribute there, beyond what you will find below.

Let me say first that I have been hesitant about this response, because I
found Fritz¹s message to be nothing short of astonishing, a cry of outrage
that borders on an ad hominem attack.  I don't want to add to the level of
vitriol from which we had begun to retreat. Let me see if I can add some
substance to this exchange, without insulting anyone.

The general thrust of Fritz¹s objections seems to revolve around the way
that Vargha-Khadem et al. (henceforth V-K) conducted their research, which
(he believes) led to the difference between their findings and those of
Gopnik et al.  Among other things, he insists that the differences between
the affected and unaffected members of the K family that V-K observed were
artifacts of the decision to test these people in a laboratory setting
(with ³bright lights and cameras²).  He is also particularly incensed about
the use of IQ tests, implying that the very act of testing destroys the
object of inquiry (a kind of linguistic Heisenberg principle).

Starting with the IQ issue, I agree that the theoretical importance of IQ
testing is a legitimate subject of debate.  My personal view is that IQ
tests put seventeen  disparate skills together in a Waring Blender to yield
a single number, so that most of the information that might be relevant is
lost.  However, because they have been in use for so many decades,  IQ
tests have become   benchmark variables in neuropsychological research,
i.e. background information that makes it easier to compare results across
different laboratories and different populations.   Indeed,  it is
difficult to get studies of impaired populations  published in most of the
major journals  if you cannot provide this kind of background information.
Whether or not we find them useful on scientific grounds, there is nothing
particularly frightening about  the content of IQ tests for anyone who has
gone through a Western school system, or at least, no more frightening than
any other situation in which a scientist or clinician is asking questions
(which would presumably include any of Gopnik¹s tests as well).  IQ tests
were developed originally for use with army recruits, i.e. with working
class individuals in a literate society. In this regard,  let us remember
that the subjects of  this debate are members of  a working class family in
London.  They are not members  of an illiterate Cargo Cult,  never before
exposed to paper or pencil (much less lights and cameras).   They have all
spent years in the public school system, and have all been examined by
doctors and nurses in the public health system (the British  system still
makes this possible, even for working class families....).
Vargha-Khadem¹s laboratories are part of a research center allied with
Great Ormond Street Hospital for Children.  Having visited the laboratory
myself, I can assure you that it is no more frightening that the
pediatrician¹s offices  and public schools that these people have seen for
years.  Indeed, it is likely that the affected members of the family have,
because of their affliction, spent considerably more time seeing doctors
and other professionals than their unaffected relatives.  If anything, this
fact should have reduced the difference between the affected and unaffected
family members in the V-K study.   And as for the video cameras: I am told
that Gopnik first became aware of the K family when she saw them in a short
documentary on the BBC while visiting friends in London.  Hence we may
presume that the K family was  accustomed to cameras before Gopnik ever had
a chance to test them.

Assuming for the moment that the very fact of IQ testing has not served as
a poison pill, destroying all the information that accompanies it,  these
findings were only one part of a much broader  battery of tests reported by
V-K, including 13 different tests of language.  Most of these are indeed
standardized instruments.  Contrary to rumor, I do *NOT* believe that
standardized tests are the only route to truth, but they do provide a broad
overview of individual abilities, in a context that permits comparison with
a lot of baseline data.  They therefore serve as a good beginning for a
more detailed inquiry, designed around specific linguistic issues, such as
the assessment of the ability to comprehend and produce regular and
irregular verbs.  In this regard, let us ask again why it is the case that
Gopnik (1990) report a dissociation between regulars and irregulars while
V-K et al. find absolutely no evidence for such a dissociation.  Instead,
the affected members performed reliably worse than the unaffected members
on both regulars and irregulars, with no trend, not even a hint of a double
dissociation.  Why indeed!  As it turns out, Gopnik¹s original report in
Nature was based on a very small number of items.   The V-K report in the
Proceedings of the National Academy used not a standardized test, but a
large and representative battery of items designed  specifically by Karolyn
Patterson to assess the kind of linguistics and psycholinguistic issues
that are central to the controversy about regulars and irregulars.  It is
likely, I think, that this difference in materials is responsible for the
difference between the Gopnik and Vargha-Khadem findings.  In fact, Ullman
and Gopnik have a report in the Montreal Working Papers with results very
similar to V-K¹s on regulars vs. irregulars  when a broader battery of
items is used (i.e. the dissociation seems to have gone away....).  (Let me
note here that Myrna refers in various communications to the Montreal
Working Papers as a ³publication²; however, it is not peer-reviewed in the
usual sense, not available in libraries as an archival work, and wouldn¹t
pass as a publication by the standards that NIH applies to people who want
grants from them; too bad, because we have a series called the Center for
Research in Language Technical Reports that would greatly enhance our
publication list if we were allowed to call it a ³publication²!).  My point
is that the V-K findings are based on many different instruments,  tapping
many different aspects of speech and language, in addition to the hated IQ
tests.  Those findings clearly indicate that this is NOT a disorder
specific to any single aspect of language, and may not be specific to
language at all, although language is certainly involved.

In this regard, Fritz also complains about V-K¹s finding that the afflicted
members of the K family have an articulation disorder, suggesting that this
too may somehow be an artifact of the testing situation.  Here I have to
say that I have seen videotapes of the family  (recorded in an informal
context, by the way), and I have heard their speech  with my own ears.
There is absolutely no question that these people have a severe speech
production disorder, the kind that you would expect if they were (as
³standardized² tests show) suffering from buccal-facial apraxia, i.e.
difficulty with complex movements of the tongue and mouth.  Several years
ago Gopnik gave a presentation at the Wisconsin Child Language Disorders
Conference where she played audiotapes of the family (presumably these are
audiotapes that she recorded herself, in the context that Fritz recommends
for linguistically meaningful research).  I was not present, but heard from
more than a dozen speech-language pathologists who were present in that
audience that the evidence for a serious speech disorder was undeniable.
And yet this was not mentioned in the 1990 Nature letter, in the 1991
Cognition paper, or in various summaries of the Gopnik results by Pinker,
Gazzaniga and others.  No one is claiming here that the articulation
disorder *CAUSED* the grammatical deficits observed in the K family. After
all, these people do poorly on a host of receptive tests, grammatical and
lexical, and on a number of non-verbal tasks as well, so we cannot
attribute all their symptoms to this single cause.  Nor is anyone claiming
that the IQ difference is the *CAUSE* of the grammatical problem.  The
point is, simply, that this is not a specific disorder.  It is not specific
to regular morphemes, it is not specific to grammatical morphology in
general, it is not specific to grammar, it may not even be specific to
language.

Fritz complains still further that the V-K paper is very short,  only 3.5
pages long.  The Proceedings of the National Academy of  Sciences, like
Science and Nature,  requires brief reports, without a great deal of
methodological detail (recall that Gopnik¹s original letter to Nature,
which started this controversey, was less than one page long).  But the V-K
results are clearly indicated in a summary table, with detailed statistics
on each and every language and non-language measure.  To be sure, it will
be useful to learn more about their findings in subsequent papers (and such
papers exist).  However, brevity can be a virtue.  The original report by
Watson and Crick was not much longer than this, following normal practice
in the journal Nature.  The real questions are: Is it true, and if so, what
does it mean?  I am persuaded that the findings are true.  This is a
distinguished and respected research team, at a major research center,
representing the fields of neuropsychology, neurology, and developmental
psycholinguistics (e.g. Paul Fletcher, an eminent researcher in this field,
certainly not naive about language and language development).  They have
used standardized tests that are recognized in this field, together with a
new (non-standardized) measure specifically tailored (by anyone¹s
standards) to reflect the relevant facts of regular vs. irregular
morphology in English.  What do the findings mean?  Well, I agree that many
many questions remain to be answered, but at the very least these findings
mean that this is not a specific deficit.  The search for the grammar gene
must continue....

There are other approaches to the same problem, relevant to the issues of
genetic substrates and language specificity.  A number of different
laboratories are investigating a syndrome called Specific Language
Impairment, defined to include children who are 1.5 to 2 standard
deviations or more below the mean for their age on expressive (and
sometimes receptive) language abilities despite non-verbal IQs in the
normal range (i.e. no mental retardation),  in the absence of frank
neurological abnormality (e.g. cerebral palsy, hemiplegia),  severe
social-emotional disorders (e.g. no autism), uncorrectable visual or
auditory deficits (i.e. they are not blind or deaf).  It has been known for
more than two decades that this disorder  ³runs in families², and Dorothy
Bishop¹s recent work comparing monozygotic and dizygotic twins with SLI
suggests that it has a strong heritable component.  Does this syndrome
provide evidence for the grammar gene?   Despite all these exclusionary
criteria, many different laboratories have demonstrated that children with
SLI suffer from subtle deficits in processing that are not specific to
language (e.g. aspects of attention and perception), although a few
laboratories still insist that they have found a relatively pure form of
the disorder (e.g. recent claims by Van der Lely).  With regard to the
³intra-linguistic² nature of SLI, dozens of linguistic and psycholinguistic
studies of these children lead to the conclusion that the deficit is best
characterized as one of delay (i.e. they look like younger normal children)
rather than deviance (i.e. no evidence for qualitatively different error
types or sequences of development from those observed in normal children).
A large number of studies also show that the deficit affects many different
aspects of language.  However, it has been known for a long time (since
Judith Johnston¹s work 15 years ago with Schery and Kamhi)  that
grammatical morphology is the most vulnerable domain in SLI.  Does the fact
that morphology is MORE delayed than other aspects of language constitute
evidence for a specific and genetically based grammatical disorder?  Some
investigators believe that is the case (e.g. Van der Lely; Gopnik; Rice;
Clahsen).  Others have argued instead that the grammatical deficits are
secondary to the processing problems that these children display (e.g.
Leonard; Bishop; Tallal).  Data from our research center here in San Diego
provide support for the latter view.  First, my colleagues here (e.g.
Wulfeck, Thal, Townsend, Trauner) are among those who have repeatedly found
evidence for subtle non-verbal processing deficits and/or neuromotor
defects in children who meet the above definition of SLI.  Second, studies
from my own laboratory have shown than grammatical morphology is the ³weak
link in the processing chain² even for normal, neurologically intact
college students.  When these students are tested under adverse processing
conditions (e.g. perceptually degraded stimuli, or compressed speech, or
language processing with a cognitive overload), grammatical morphemes
suffer disproportionately compared to every other aspect of the signal
(e.g. content words, word order relationships).  Taken together, these
lines of evidence provide a reasonable case for the claim that the
grammatical morphology deficits in SLI are secondary to (in this case,
"caused by") a processing deficit that is not unique to language, although
it has serious consequences for the ability to learn and process language
on a normal timetable.  This would help to explain why grammatical
morphology is also an area of selective vulnerability in Down Syndrome, in
the oral language of the congenitally deaf, in many different forms of
aphasia (not just Broca¹s aphasia), and in a range of other populations as
well.  (If anyone is interested, I have a review paper on this topic).

One might suggest that SOME forms of SLI have this non-specific base, but
others really are due to an innate, language-specific malfunction --
perhaps the genetic mutation discussed in last week¹s Newsweek.  One cannot
rule this out without further evidence.  It is interesting to note,
however, a paper a few years ago by Tallal, Curtiss and colleagues
separating their large sample of children with SLI into those with and
without a family history of language disorders.  There were no differences
between the two subgroups in the nature of the language symptoms observed
across a host of measures; however, the children with a family history were
actually MORE likely (not less likely) to suffer from deficits in
non-linguistic processing.  Finally, I would refer you to a recent study by
Wulfeck, Weissmer and Marchman, in a large SLI sample combining data from
San Diego and Wisconsin, assessing the ability to generate regular and
irregular past tense morphemes on a large and representative sample of
items.  Results clearly indicate that children with SLI are delayed in both
aspects of grammatical morphology (i.e. there is no dissociation between
regulars and irregulars), producing errors that are quite similar to those
observed in younger normal children at various points in development (i.e.
the usual story in research on SLI).

I apologize for the length of this exercise, but I think it is important to
get as many of the facts out as possible.  Using terms like ³scandalous²,
Newmeyer has implied that there is something deeply wrong, perhaps
fraudulent in the Vargha-Khadem et al. results.  He has no basis whatsoever
for a claim of that kind.  Their research project is state-of-the-art in
the field of neuropsychology.  Newmeyer may want to respond by arguing that
all of neuropsychology is irrelevant to this debate, that linguists are the
only ones who know how to assess language properly, that standardized tests
are useless (as opposed to merely insufficient), and that research
conducted in a laboratory setting is invalid precisely because it was
conducted in a laboratory setting.  If he is right, we are in dire straits,
because 98% of our knowledge about language disorders and the brain bases
of language in normal and abnormal populations falls into this category.
The recent exchange on Funknet leads me to the gloomy conclusion that some
of our colleagues in the field of linguistics espouse exactly this belief.
I would prefer to re-issue the argument that got me into this exchange: we
need all the methods, all the constraints, all the information that we can
possibly find to build a reasonable theory of the human language faculty.

-liz bates