LINGUIST List 7.716

Mon May 20 1996

Disc: Ungrammaticality

Editor for this issue: Ljuba Veselinova <>


  1. benji wald, ungrammaticality

Message 1: ungrammaticality

Date: Sat, 18 May 1996 01:52:00 PDT
From: benji wald <IBENAWJMVS.OAC.UCLA.EDU>
Subject: ungrammaticality

I was delighted that Prof. Shaumjan went into as much detail as
he did to reply to my implied question about his grammatical
theory, or rather the diagram of it that he presented. I have a
few criticisms to make about his theory to the extent that I now
understand it, but first I want to ackkowledge that his reply was
sufficient for me to agree that his model does not threaten a
reply of "linguistic wars".

Inviting correction, I would characterise his theory as
consisting of two sets of components; one set is a Saussurian
language-specific set and the other, about which he did no go
into any detail, is a UG language-universal set. Most
significant, the Saussurian language- specific set, manifest in
his "relevance principle", includes the semantic (= meaning)
oppositions relevant to the particular language. These are what
he calls (but does not clearly distinguish other than by
assertion in his response to Powers) the lexical and grammatical
meanings of the language. In this way his model is clearly
different from the generative semantics model that he referred
to, since that model has only one semantic component, a UG one,
otherwise all other components map the UG semantic "structure"
onto surface syntactic strings by purely formal (or formalistic)
rules. Thus, the generative semantics model was even more of a
departure from Saussurian language-specific semantics (or
"meaning", I'll get back to that in a moment) than the
interpretive semantics model (ancestor of current "generative"
models, which generated our discussion of "ungrammaticality",
among others things).

S did not explain his concept of "meaning" in any way that was
clear to me, though I'm sure he was using it in a technical
sense, since he even coopts words like "reference"
(cf. "morpheme") from the philosophical/linguistic koine to give
them a significance convenient to his theoretical purposes. My
assumption is that "meaning" is eventually taken care of
somewhere in his UG component. This assumption is based on his
use of the term in the following passage:

"The English word WASH has different *meanings* in the context of
expressions WASH ONE'S HANDS and WASH THE LINEN." (*underlining*

This is in the context of explaining that English WASH
corresponds to two distinct Russian (or Swahili, for that matter)

That is not a language-specific use of the term "meaning",
obviously. Everything that S says makes it absolutely clear that
English WASH has ONE language-specific "lexical meaning". And
yet he spoke about different (= MORE THAN ONE) "meanings" of
English WASH (obviously not lexical meaning, then). What kind of
"meaning"? Must be UG.

In any case, one question it suggested to me is: if some language
turns out to have 15 more specific verbs where English uses WASH,
depending maybe on what's washed or what actions are involved in
washing (maybe even English has 15 such verbs), does that mean
that in the UG, English lexical item WASH has (at least) 15
different "meanings"?

I'm not sure I like this UG concept of meaning, or that it's
different from lots of other models that are already out there.

Meaning is always a problem for linguistic theories. In the
following passage my impression is that S overstates the case,
probably simplifying in haste:

"no matter how different, two distinct meanings belong in the
same class of meanings if they do not correlate with two distinct
phonic segments (or markers based on phonic segments)"

I'm sure S realises that things are not that simple (even if it's
a good heuristic to begin with). But if he insists, I ask: Does
that mean that there is no such thing as homonymys in a language?
The usual English example of a homonym is "bat", as in "baseball
bat" vs. "bat", the animal. Homonymy is to be distinguished from
polysemy, of course. There is no reason to question that
regardless of how many verbs some other language has to use,
English "wash" has a single coherent lexical "meaning", so, if
you want, you can say it is polysemous, but its various
correspondents in other languages do not make it a bunch of
homomyms. In contrast, who would argue, or even want to argue,
that the two "meanings" of "bat" are *irrelevant* to how "bat"
should be treated in the LANGUAGE-SPECIFIC lexicon of English?
Well, S would want to argue it, unless he makes more precise his
notion of "meaning" and "relevant (= language-specific)

This is a standard problem in Saussurian semantic theory, for
which various (partial) solutions have been offered, none of
which make the problem disappear. (The ultimate desperation of
the dominant forms of early structuralism emanating from
Saussurian ideas was to develop structuralism independent of
meaning. or claim they did -- even trying to banish the study of
meaning from linguistics -- but here it is again, and this time
we have to face it -- for the sake of UG?)

Last point on this. Maybe there are behavioral consequences to
the distinction between homonymy and polysemy. English speakers
recognise in different contexts that WASH is the "same", but many
might not notice that "bat" and "bat" are exactly the same
(phonologically) until it's pointed out to them. This is
dubious, perhaps. But Kindergarten children seem to be
delighted, as if they'd learned something new, when the teacher
points out that "to", "two" and "too" are different "words",
priming them for spelling. I'm sure what they learned was that
it hadn't occurred to them before that these words are pronounced
the same, not that they thought they were "the same word". I'm
just trying to be helpful, I don't insist that this is relevant
to every theoretically based uh theory.

Next, his example to indicate his concept of "causative" in

"A real causative is one with a causative morphonology as in the

Historically, of course, S is right about causative formation in
(pre-)English (resulting in umlaut and vowel change). However,
what about productivity in current English? Maybe that doesn't
matter to S's theory. In any case, with regard to the pairs
"sit/set", his analysis will lead to contradiction with other
principles in his theory. Thus, note:

"SIT/SET/SETTLE/SEAT the baby in her playpen"

They all have different "meanings" -- or don't they? Which is
the "causative" of "sit"? By the "phonic counterpart" principle,
SIT is the causative of SIT, not SET. And, indeed, this is a
more productive type in current English than umlaut, not that I
accept (without further argument) that in any sense the phonic
resemblence of "sit" and "set" is any more than a coincidence to
current speakers of English, "syllable" not "morpheme" in S's
terms. (Just as I suggested above for the two words spelled
"bat", as do most linguists.)

By the way, after SIT, maybe SEAT, still not SET, is the next
candidate for "causative" of SIT. SET is still waiting in/on

"fall/fell" is also suspect, in that "everything" falls, but you
can't "fell" everything that you can make fall = DROP/KNOCK
OVER/etc. S is building too much history into the language, and
his principles for doing so are not at all clear to me. Cf.

The cup fell off the table becauyse she dropped/??felled it.

S would certainly not claim that "fall" is a different "meaning"
depending on whether a tree or a cup "fell" (though he could in
UG because in some languages total separation from the ground
during fall is distinguished from falling *over* the way a tree

Making "drop" the causative of "fall" also has problems, of
course. especially for a Saussurian theory.

Two more quick comments relating to S's responses to other List
discussants (out of many more comments I hold in reserve)

In response to deArmond, S wrote:

"...we must distinguish two kinds of reference: lexical and
grammatical. Accordingly, we must distinguish two kinds of
meaning: lexical and grammatical. Yes, case and agreement, as
well as gender, have meaning. The meaning of grammatical is

In the larger context I found nothing that would make this
passage any less of an assertion without adequate explanation.
It does not explain the difference between lexical and
grammatical meaning, except to suggest that grammatical meaning
is MORE "abstract". That's just a matter of degree -- if that's
the case. If it's not, what is?

Finally, S's response to David Power's question troubled me most
- for many reasons, not least of which was that Power's comment
was most relevant to the context of a discussion of
"ungrammaticality" in which S inserted his theory, claiming it
had some relevance (which no doubt it does). S wrote:

"Under a clearly defined subject matter of linguistics, the
distinction between grammatical and lexical meanings does not
present serious problems except for special cases at the
interface of grammar and lexicon. These special cases have a
marginal significance; they in no way undermine the fundamental
opposition between grammatical and lexical meanings."

This and everything connected with it was nothing but assertion
without explanation, as far as I could see, with a hint that
assertion might even go so far as to dismiss certain problems as
"trivial" or "marginal" (though in the context of "marginal" in
this passage, I assume that S is making the assertion on the
basis of some empirical evidence, not just on the basis of what
he considers "important".)

As S acknowledges elsewhere, examples are important. So to
concretise a problem that occurs for many Indo-European (among
other) languages, we have the following which seems to fall on
the cusp between S's (and many others') grammatical and lexical
systems. There is a system of case, a grammatical system. There
is also a system of adpositions, or is there? I'm sure Saussure
would think there is. Is that grammatical or lexical? Why? Then
there is an interaction between them (some theories say that some
adpositions "govern" certain cases, some one case, some several).
Is this marginal?

My own immediate reaction is that insisting on a
lexical/grammatical dichotomy creates a false problem. We don't
need it. Whatever is going on between cases and adpositions,
that's the problem. Invoking lexical/grammatical oppositions has
nothing to do with it. Some systems may be more abstract than
others, and that may correlate sometimes with what is
traditionally recognised (and unchanged in S's theory) as
lexical/grammatical -- and in some cases the reverse may hold.
So what? So nothing, unless you insist on making a
lexical/grammatical distinction instead of analysing whatever you
are trying to analyse and just follow where it leads you. Of
course I may be wrong (but that's not for me to say; it's for me
to learn.)

In case you can't tell, I'm not interested in grammatical models
for their own sake, but for what they offer to tell me about
concrete problems encountered in trying to figure out what's
going on in languages. The phenomenon of "ungrammaticality" is
one. As S said about Columbus, getting an advantage from an
incorrect theory (to the benefit of some, not all, with Native
American slavery, annihilation, etc.), the fact that generative
grammar put this problem in relief does not mean that it is a
problem specific to generative grammar. No doubt S, like me, and
like the generativists themselves, has different opinions about
different examples, how they should be accounted for, and even
distinctions in terms of how significant or trivial one or
another particular type of problem is. I also sympathise with
the bind that he is in, at once trying to explain his theory to
those unfamiliar with it and at the same time trying to explain
how it is relevant to the motivating concept of the "limits of
grammar", which is manifest in such problems as
"ungrammaticality". I say this because I do not see that he has
made the connection. The problem for him (or for me, but I'm
sure there are many others like me) is this. Before I invest the
effort and energy to understand "yet another theory of grammar",
I want to know in some succinct and concrete form: what's the
payoff? What recognisable problems will it solve where other
theories have failed (or must fail)? What's the insight? --
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue