LINGUIST List 3.449

Tue 02 Jun 1992

Disc: Innateness, rules

Editor for this issue: <>


  1. , Rules
  2. "R.Hudson", Innateness
  3. Kean Kaufmann, Re: 3.436 Innateness
  4. , Rules

Message 1: Rules

Date: Wed, 27 May 92 16:13:20 EDRules
From: <>
Subject: Rules

Joe Stemberger <> writes:

> 1) Forms like "buyed", "goed", and "eated" may occur in child speech, but
> they are in a minority. The vast majority of attempts at irregular verbs
> are produced correctly. (Gary Marcus and his colleagues at MIT have a
> monograph on this that is not yet out.) Children DO tend to produce what
> they here, at least statistically.

This in fact confirms my claim that children follow rules.

> 2) The assumption is made, IN THE ABSENCE OF ANY DATA, that children
> hear adults produce sentences like (1b) above. This is an amazing claim,
> and I doubt that it is true.

I myself find it a very plausible claim. I suggest you try and substantiate
your doubts by providing some attested examples. The burden of proof is
clearly on those claiming that sentences like "Is the man who __ tall is in
the room?" do occur in child speech.

> 3) Even if it were true, Wyngaerd is making generalizations about

My theory of learning is that children have recourse to innate principles;
these principles determine that sentences have hierarchical structure and
that grammatical rules refer to hierarchical structure. Rules that would
refer to linear precedence are inaccessible as a matter of principle. This
explains the absence of the ungrammatical (1b) sentences in child speech.

Martti Arnold Nyman <MANYMANFINUHA.bitnet> writes:

> That "the relevant experience [is not] rich enough to
> determine the nature of the rule" echoes the well-known
> 'poverty of stimulus' argument, which has never been proven.

Has it been disconfirmed, then? Please inform me!

> In the case at hand, it is easy to conjecture what sort of
> data/experience is relevant for a child to infer that yes-no
> questions are formed in terms of hierarchical structure.
> Consider where-questions:
> Where is [X]? [X] is in Z. Is [X] in Z?
> ---------------------- = --------------------- = ---------------------
> Where is [X who is Y]? [X who is Y] is in Z. Is [X who is Y] in Z?
> In principle, analogy works here quite well: the where-question
> displays the hierarchy which can be analogically extended to other
> cases.

I fail to see how this case is different from the one not involving
'where'. Available evidence mainly consists of simple sentences (the ones
over the line in Nyman's schema); the complex cases (those below the line)
could be formed by fronting the first occurrence of 'is' or the first
occurrence following the subject (eg '*Where is [X who __ Y] is?'). The
second possibility does not occur, however. Nyman suggests analogy to
explain this: but how does the child know what (s)he has to form an analogy

> But I am not concerned with whether or not analogy works
> in this particular case. What I am concerned with is, objecting to
> wholesale innatism.

In the absence of particular cases, any discussion over innatism is bound
to be sterile. The position Nyman advocates is uninteresting, as nobody,
including me, would even care to disagree with it.

Guido J. Vanden Wyngaerd
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Innateness

Date: Thu, 28 May 92 13:26:59 +0Innateness
From: "R.Hudson" <>
Subject: Innateness

Does it really matter whether we believe in innate UG or not? I think
it does.

Like Avery Andrews, I think that a belief in innate UG can be
positively harmful in writing grammars, because it makes the
linguist's life too easy. For those of us who are sceptical about
innateness, it's really worrying if we have to postulate some very
abstract and complicated principle in our theory. Sometimes you have
to admit defeat, but you at least recognise it as defeat, and a sign
that something is badly wrong with your general theory. But if you
believe in innateness, there's no problem - on the contrary, you can
be pleased to have discovered yet more evidence that language is weird
(i.e. unique) and unlearnable.

Reversing the question, can grammarians say anything about whether
there is an innate UG or not? I agree with Avery Andrews that purely
linguistic work on language structure can't prove that there is an
innate UG (i.e. innate knowledge specific to language structure).

However I think one can go further than that: purely linguistic work
on language structure can show the difficulty of defining the boundary
around language structure (e.g. the uncertainty of the distinction
between linguistic and encyclopedic `meaning', the close connections
between words and other aspects of our behaviour in conversation,
etc.). That in itself casts doubt on the claim of innateness, because
the whole point of this claim is that knowledge of `language' is
unique, built on a unique innate framework of principles and
parameters. If borderline cases are so hard to decide on, this is
presumably because they're so similar both to the centre of language
and also to things beyond language - which suggests that language
isn't after all so different from everything else in our minds.

There may be good evidence for innateness from other kinds of research
(as Carol Georgopoulos and Avery Andrews suggest), though I think it's
still too early to decide. Until we've got much clearer independent
evidence for innate UG, and a clearer idea of what that UG actually
comprises, all theories of language structure should be judged by the
same standards.

Dick Hudson
Dept of Phonetics and Linguistics,
University College London,
Gower Street,
London WC1E 6BT
(071) 387 7050 ext 3152
home: (081) 340 1253
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Re: 3.436 Innateness

Date: Fri, 29 May 92 11:32:55 EDRe: 3.436 Innateness
From: Kean Kaufmann <>
Subject: Re: 3.436 Innateness

Vicki Fromkin writes:

>...would also like to publicly wonder why the question of
>of the innateness of specific bird species' songs is not questioned b
>by those who question are argue against built in hard-wired genetic
>language specifics. One would think that the more neurologically
>complex the species the more one would have to have such geneticallyj
>determined capabilities.

One wouldn't necessarily think so, if one viewed neurological complexity
as a necessary correlate of learning capacity: The more you're able to
learn, the less you're born knowing -- and vice versa. The fact that
humans are born more helpless than any other species, and take longer
(by orders of magnitude) to mature, would seem to indicate a tradeoff
between instinct and intelligence. From this perspective, one might
see Homo sapiens NOT as the language-acquiring animal, but as the
learning animal par excellence, the general-purpose animal; and one
would want to derive as much language function as possible from more
general cognitive abilities.

Kean Kaufmann	(
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 4: Rules

Date: Sat, 30 May 92 08:22:14 EDRules
From: <>
Subject: Rules

In Vol-3-441, Martti Arnold Nyman replies to my earlier reply to him
in Vol-3-371 (and not just to him, but to Thrainsson and Andrews
as well) on the question of normativeness. As I tried to indicate
there, it seems to me that much of the discussion on the subject,
while enlightening in itself, has not addressed the crux of the
problem that Itkonen raises. And, while I dissent from Itkonen's
conclusion, he has, I think, correctly identified the weakest
point in the armor of linguistics.

Specifically, all normal claims about language that linguists
make are about "normal" utterances. By mistake or as a joke,
people can say all sort of things. This is why, of course, many
people have pointed out that the usual form of statements and
queries re: grammaticality judgements (e.g., "Can you say X?")
makes no sense.

If water could in some cases accidentally or as a joke boil
at 110 degrees C, then physics would be in the same boat.
It is precisely because this is not so, that people like
Itkonen can, as far as I can see, argue that linguistic
claims (English NPs begin with articles, whatever) are normative
where physical ones (water boils at 100 degrees C) are not.

This it seems to me is the crucial problem of linguistics
and this is what must answered if we are to accept that
linguistics is a natural science.

My brief remarks in Vol-3-371 were intended to suggest that,
even if we do not know enough about these matters, we can,
nevertheless make defend ourselves. That is, our claims
about "normal" utterances are not, I maintain, to be interpreted
as referring to utterances which SHOULD be used (that's normative)
but rather about utterances which ARE used UNDER CERTAIN
CIRCUMSTANCES. The problem is that we cannot precisely
define these circumstances, but I think we make a reasonable
case for this anyway.

Thus, whereas a phycisist can say precisely under what
conditions of atmospheric pressure and such water will
boil at an "abnormal" temperature, we cannot say with the
same precision which utterances will constitute "jokes"
or "speech errors". However, the fact that we have
discovered all sorts of constraints even on speech errors,
and that people have discovered ways of making subjects
make more than the usual quota of errors (by making them
read while performing some distracting task), both of
these things suggest that we can in principle specify
more precisely what a "speech error" is.

The situation with jokes and other deliberate manipulation
of language is analogous, though perhaps even harder. However,
I think I could safely maintain that the jokes that people
are capable of producing resp. understanding themselves are
subject to constraints which only linguists can explain.
Thus, one would probably want to say that a speaker of
with some high degree of probability to be unlikely to
produce a jocular utterance with one of these. Whereas,
of course, having been exposed to them, I can make fun
of Indian English, for example, by say 'Oh Ghod', when
an Indian speakers says, as millions of them do, 'ghost'
(with a breath 'gh').

The reason I put the relative clause in the previou para
in caps is that this is the crucial point. A linguistics
which wants to be a natural science and make predictions
about human behavior must always make these kinds of qualification.

I think this point is really Chomsky's (although it makes
especially good sense if you accept my interpretation of
Chomsky as holding that a grammar is a transducer which
takes in what it hears as input, but I digress):

 Linguistics cannot predict the behavior of people
 based on such classifications as English speaker
 vs. French speaker. Rather, we must know what
 the speaker has been exposed to in the way of
 "primary linguistic data" and other data, too.

Of course, Chomsky then goes to still make the competence (grammar)
vs. performance (behavior) distinction, and as soon as he does that,
Itkonen is home free. Linguistics then IS normative.

Rather (as I argue in another paper), we must regard "competence"
as a linguist's idealized, oversimplified idea of the system
that generates linguistic behavior. There is nothing wrong with
idealizations, of course, but we must take the theory of
behavior ("performance", if you must) as primary in order
to get around Itkonen's objection.

Thus, to summarize, to answer the claim that linguistics is
normative, we must

 (a) seek to define the conditions under which "jokes", "errors",
 etc., are possible and the constraints on these,

 (b) we cannot make statements about English, only about speakers
 who have never been exposed to any language other than English
 (or have to such-and-such a specified extent, and so on)

 (c) we must take behavior ("performance") as what linguistics
 studies, reducing grammar ("competence") to the status of
 a useful but ultimately misleading idealization.

Then and only then can we answer people like Itkonen, who, of course,
will probably reply, and will be right, that the kind of linguistics
THEY were describing as normative was the traditional linguistics of
Dionysius, Bloomfield, and Chomsky, which indeed does look normative
(in the relevant sense).

P.S. This, if correct, also has a bearing on the raging innateness
issue. I think that if you accept the view I am advocating (or
even just point (b), which is Chomsky's), then linguistics becomes
the study of individual people, not of languages, and it becomes
essential to assume innateness in the way Chomsky did. Of course,
this is not to say that any PARTICULAR feature of language is
ASSUMED to be innate. That, obviously, is, or should be, a factual
question (I always try to use 'factual' for the overused 'empirical').
What IS assumed is that, having accepted this view, one is committed
to being more interested in what is innate than anything else. For
the simple reason that THAT is what is shared by all the objects of
study (i.e., individual persons).
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue