LINGUIST List 7.552

Sat Apr 13 1996

Disc: Ungrammatical sentences

Editor for this issue: Ljuba Veselinova <lveselinemunix.emich.edu>


Directory

  1. Richard DeArmond, Re: 7.536, Disc: Ungrammatical sentences
  2. "David M. W. Powers", Re: 7.536, Disc: Ungrammatical sentences
  3. "Karen S. Chung", Re: Happy scissors (fwd)

Message 1: Re: 7.536, Disc: Ungrammatical sentences

Date: Fri, 12 Apr 1996 14:10:00 PDT
From: Richard DeArmond <dearmondsfu.ca>
Subject: Re: 7.536, Disc: Ungrammatical sentences

>Subject: 7.536, Disc: Ungrammatical sentences
>

>From: shaumyanminerva.cis.yale.edu (Sebastian Shaumyan)

>
>
>As to the essence of Chung's claim, I completely agree. The issues of
>what is commonly called "semantics" do not belong in grammar rules.I
>wish only comment on the terminology. The opposition of the terms
>`grammar/syntax' versus `semantics' may raise a problem. One may think
>that `grammar/syntax' does not deal with meaning, while semantics
>does. I would disagree with this understanding of the opposition
>`grammar/syntax' versus `semantic'. Both grammar/syntax and semantics
>deal with meaning but with very different kinds of meaning. And this
>is crucial. There are two kinds of meaning: 1) grammatical meaning
>(including the syntactic meaning) and 2) non-grammatical, or content
>meaning. Grammatical meanings are represented by grammatical morphemes
>while content meanings are represented by content morphems.Let me
>illustrate the distinction between the two kinds of meaning by some
>examples.
>
>Consider the Russian sentence:
>
>DEVOCHK-A CHITA-ET KNIG-U
>girl reads book
>
>`The girl reads a book'
>
>In the transliteration of the Russian sentence the hyphens are used to
>present in a simplified form an analysis of the words into grammatical
>and content morphemes. The first word consists of the content morpheme
>DEVOCHK `girl' and the grammatical morpheme A. The grammatical
>morpheme A denotes a bundle of grammatical meanings: 1) nominative
>case and simultaneously subject, 2) feminine gender, 3) singular. The
>second word consists of the content morpheme CHITA `read' and the
>grammatical morpheme ET. The grammatical morpheme ET denotes: 1) the
>present tense, 2) 3-d person 3) singular. The third word consists of
>the content morpheme KNIG `book' and the grammatical morpheme U. The
>grammatical morpheme U denotes: 1) accusative and simultaneously
>direct object, 2) feminine gender, 3) singular. Comparing the content
>meanings with the grammatical meanings of these words, we discover a
>profound difference between the both kinds of meaning. The content
>meanings are more concrete, more narrow, more specific while the
>grammatical meanings are more abstract, wider, more general. The
>content meaning DEVOCHK occurs in a dozen or at most in a couple of
>dozen of different words, while the grammatical meaning A occurs in
>thousands of words. Any language has only a closed limited set of
>grammatical meanings, while the number of content meanings is huge and
>unlimited, in principle.

Shaumyan's interesting repsonse to Chung should be updated a bit. It
is necessary to carefully distinguish between grammatical features
(Chomsky's 'phi'-features) and meaning. Grammatical features occur in
two classes: the first class includes meaning that is referential in
the Fregean sense, as opposed to menaing that has sense. All lexical
items have sense, but no direct reference. Derivational morphemes in
many if not all cases have sense. I would go so far to define
derivational morphemes as those that denote sense but no direct
reference. This distinction is not an easy one, especially for those
who not familiar with the distinction--and it is hard to expain this
distinction in an undergraduate linguistics class. This point would be
a good place for semanticists to join in.

The second class includes phi-features that have no reference. These
include Case and agreement. It could be claimed that agreement
features have reference through their antecedent features. This cannot
be the case. If both had direct reference, there is always the
possibility that each could end up having different references--this I
am certain never happens. Second, we have the problem of gammatical
gender which is under discussion elsewhere in LList. There is often a
connection with meaning for human and some animate objects, but in
most cases grammatical gender has no referential property. It is
determined by the inherent and unpredictable class of the noun.

>
>While in inflectional languages, like Russian or Latin, grammatical
>meanings are represented by sounds or sound sequences, in
>non-inflectional languages like English the grammatical meanings may
>be represented by word order. Thus, in the English translation of the
>Russian sentence, THE GIRL READS A BOOK, we must analyze GIRL into two
>morphemes GIRL-SUBJECT and BOOK into BOOK-DIRECT_OBJECT, but her these
>morphemes are determined by the word order.

Again I would consider Case or grammatical relation not to be a
morpheme but a feature (Case) or a structural definition (grammatical
relation). i would define a morpheme as a phonetic string that has a
grammatical function. This definition is no doubt controversial. This
definitial excludes suprasegmentals. Whatever we call suprasegmentals,
they must denote at least one grammatical feature.

>
>The grammatical morphemes belong in grammar, while content morphemes
>belong in the lexicon. If this assumption is correct, then we must
>replace the opposition GRAMMAR/SYNTAX VERSUS SEMANTICS by the
>opposition GRAMMAR/SYNTAX VERSUS LEXICON.

The term 'grammar' has acquired two meaning in linguistics. In one
sense (Chomskyan) it refers to the rules and forms of one's linguistic
competence. In the narrower sense it refers to the non-lexical part of
the Grammar (in the wider sense). This ambiguity is unfortunate.

>
>Traditionally, semantics has been a branch of linguistics which has
>been concerned with lexical meanings. Traditionally, many linguists
>have opposed semantics to syntax because they have not recognized that
>syntactic functions are a special kind of meanings. Every morpheme, no
>matter whether lexical or grammatical, is a bilateral linguistic unit
>consisting of sign and meaning. Language does not have meaningless
>morphemes. Meaning is any kind of information denoted by signs. A
>syntactic function is a meaning because it is a certain information
>denoted by a sign.

At the risk of stirring up another hornet's nest, I would say that a
morpheme (as I defined it above) may have no meaning, just a function.
Meaning is a function. I cite such alleged morphemes as Latinate
prefixes (de, re, in, pre, pro, ex, ad, and so forth), and the suffix
Latinate suffix -id (horr+id, horr+ible, horr+ify, ab+hor,
ab+horr+ent), Anglo Saxon prefixes (be-, under-, for-) in 'behead,
understand, forgive'. These prefixes have no meaning but they have a
function--to create a new lexical stem from a simple one or from a
non-lexical root). I would also cite the 't' in such Latinate words
as" de+stroy, de+struc+t+ion, de+struc+t+ive; re+ceive, re+cei+t,
re+cep+t+ion, re+cip+ient. -t- is not added to Latinate roots that end
in a sonorant or 'g': re+bell+ion; opin+ion; domin+ate, domin+ion;
leg+ion; re+lig+ion; reg+ion.

>
>Now, if we agree that both grammatical and content morphemes have
>meaning, one may ask: What is the place of semantics in linguistics?
>This is a good question: if semantics by definition must be concerned
>with meanings, and if we claim that not only lexical, but also
>grammatical morphemes have meaning, then we must admit that semantics
>must cover both the lexicon and grammar.
>
OK.

>I would agree with this claim. To be consistent, we must extend the
>notion of semantics to cover both the lexicon and grammar. But this in
>no way will undermine the fundamental distinction between the lexicon
>and grammar.
>
I shouldn't hope so.

>I propose the following structure of linguistics:
>
>
> LINGUISTICS
> /\
> / \
> / \
> PHONOLOLGY SEMANTICS
> /\
> / \
> / \
> GRAMMAR LEXICON
>
>
>These scheme is motivated by the following considerations:
>
>Every morpheme is a bilateral entity consisting of sign and meaning. A
>sign consists of phonemes, which are unilateral entities. So the first
>division of linguistics is into two areas of study: 1) phonology--the
>study of the unilateral, that is, meaningless entities, that is
>phonemes, 2) the study of meaningful entities, thar is, linguistic
>units proper. The latter study in its turn divides into the theory of
>grammar, or grammatical semantics and the theory of the lexicon, or
>lexical semantics. Syntax is not a separate level of natural language
>but an intrinsic part of grammatical semantics which is concerned with
>functional relations between the components of the sentence, the
>simple components being words and the compound components being word
>groups of different degrees of complexity.
>

Perhaps syntax is not a distinct level, but deals with configuring the
elements of grammar (both hierarchically and linearally). Another
controversy?

>
>My scheme in no way compromises the fundamental distinction between
>grammar and the lexicon. We must distinguish strictly what belongs in
>grammar and what in the lexicon. The distinction between content
>meanings and grammatical meanings is clear and not negotiable. Let me
>give an example of the confusion of content and grammatical meanings.
>
see above.

>Consider Fillmore's Case Grammar. In his "THE CASE FOR CASE" (In
>UNIVERSALS IN LINGUISTIC THEORY.Eds. Emmon Bach and Robert T. Harms.
>1968) has introduced a case system consisting of six cases: Agentive
>(A), Instrumental (I), Dative (D), Factitive (F), Locative (L),
>Objective (O). On page 25 of the book, Fillmore gives a list of
>sentences illustrating his case system. Here is a couple of examples.
>Comparing two sentences JOHN OPENED THE DOOR and THE KEY OPENED THE
>DOOR, Filmore assigns Agentive to JOHN in the first sentence and
>Instrumental to KEY in the second sentence. It is clear that from a
>syntactic point of view, both JOHN and KEY are subjects. They differ
>only with respect to their lexical meaning. Other sentences in
>Fillmore's list are treated in the same spirit. Fillmore also
>introduces the concepts of subjectivalization and objectivalization.
>But doing so Fillmore draws from different sets. As result, his case
>system becomes heterogeneous, being a mixture of grammatical and
>lexical concepts.

(Just as an aside, I would consider 'John opened the door' and 'the
key opened the door' to be in the active voice, but 'thefront door
opens with this key' to be in the middle voice. Has this been
discussed in the literature?)

Fillmore's "Cases" correspond with what have been called participant
roles, semantic roles, thematic roles, or more currently, theta
roles. His term differs from the more traditional usage of case
referring the morphological affixes marking grammatical relations. Of
course, today we have 'Case' referring to similar but abstract
grammatical relations. (Not without controversy.)

>
>My analysis is no way meant to compromise a very interesting work of
>Filmore. I wish only to present an example of a system consisting of a
>mixture of lexical and syntactic concepts.
>

Most of my comments are definitional, except for the distinction
between sense and reference. I believe that this is what we need to
make some of the disctinctions that Shaumyan is referring to.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 7.536, Disc: Ungrammatical sentences

Date: Sat, 13 Apr 1996 15:14:35 +0930
From: "David M. W. Powers" <powersist.flinders.edu.au>
Subject: Re: 7.536, Disc: Ungrammatical sentences

BAD MSG:
Sebastian Shaumyan comments in LINGUIST 7-536
> 
> Karen Chung reported in LINGUIST 7-506:
> 
> >	Since we distinguish the categories of 'grammar/syntax' and
> >'semantics' in our study of language, we should at least be consistent in
> >our usage of the terms. Any sentence that follows the rules of grammar is
> >'grammatical'. Semantic issues do not belong in grammar rules. They should
> >be treated separately in the study of semantics. I believe this is one
> >major area where syntax studies have gone awry.
> 
> As to the essence of Chung's claim, I completely agree. The issues of
> what is commonly called "semantics" do not belong in grammar rules.I
> wish only comment on the terminology. The opposition of the terms
> `grammar/syntax' versus `semantics' may raise a problem. One may think
> that `grammar/syntax' does not deal with meaning, while semantics
> does. I would disagree with this understanding of the opposition
> `grammar/syntax' versus `semantic'. Both grammar/syntax and semantics
> deal with meaning but with very different kinds of meaning ...
> 
> I propose the following structure of linguistics:
> 
> 
> LINGUISTICS
> /\
> / \
> / \
> PHONOLOLGY SEMANTICS
> /\
> / \
> / \
> GRAMMAR LEXICON
> 

The distinction between syntax and semantics is anything but clear,
and Shaumyan's notion that semantics subsumes both grammar and lexicon
is very appealing. This is particularly true if one takes a
diachronic view of the functional morphology, including the
development of the grammatical closed-class morphs from the more
purely semantic open-class morphs (e.g. be-hind and be-fore as members
of closed classes of such as prepositions, conjunctions and adverb,
but which evidently developed out of open class words; or the
development of full verbs into modals, auxiliaries and the like,
cf. keep: "He keeps missing" - Jim Entwisle drew this construction to
my attention).

Unfortunately, the distinction between grammatical and lexical morphs
is no more clear cut than is the distinction between syntactic and
semantic (un)acceptability. Many of the examples used in linguistic
texts and papers reflect this problem. Both the starred and the
unstarred examples seem awkward and unacceptable out of context, as
presented, but may be quite natural in context. This is particularly
true when it comes to determinations which hang on the
subcategorization of particular words, or the preferred near synonym
for a particular context. As an Australian who does not speak any
American dialect, it is sometimes difficult for me to to know whether
the frequent reversal of starring patterns which I would place on the
examples of an American text derive from dialectic differences or
carelessness on the part of the authors. I tend to suspect it is a
bit of both, but sometimes my objection can be definitively sheeted
home to dialect or idiom.

As a generalization, there is not a neat divide between grammatical
and lexical, but rather a continuum. My thesis is that the smaller
and more closed a substitutable class is the more it will tend to be a
syntactic class whose membership is functional, in the sense that
semantic reference is constrained to the domain of language, and the
smaller the class the less tendency there is to admit broader
spatio-temporal/ontological distinctions (cf. proximity, number and
gender marked articles versus articles exhibiting a purely
definite/indefinite distinctio. Pure functionality is rare or
impossible (it would imply a synchronically pointless distinction).
Conversely, the larger and more open, the class, the more
unconstrained the semantics of the individual members, semantic
reference being open to the full range of human experience.

There is a logic to this. Open classes are less constrained in their
membership and thus the expected range of reference of both the class
and its morphemes should be broader. As these classes are open, they
should be expected to grow bigger. Closed classes cannot grow bigger
and must remain small (such growth as occurs would be expected to take
place between generations, as phrases are lexicalized and
abbreviated).

The larger classes can however exhibit apparent subcategorization, but
it must be noted that strict subcategorization contradicts the notion
of openness. It is always possible to invent a new open class word,
and it is similarly easy to gain acceptance for new usages of an
existing open class word. Coinage of new words is actually rarer than
the pressganging of existing words, by analogy or extension, into new
contexts. Cutting across this dubious phenomenon of subcategorization
is the whole issue of entrenchment, which seems to be more closely
tied to referent than the lexeme.

When apparent synonyms show restriction of some form, my sense is that
this is more related to idiom than to grammar, and to the dialect than
to the language.

I am reluctant to point the finger at particular authors by quoting
examples from their texts, and I am sure that all of us have
frequently been brought up short by an unexpected * (or lack of it).
Indeed the surveys which take place on this list illustrate the
variation in what people regard as being grammatically acceptable.
Once we leave the close class morphs and peg our examples on
subcategorization, we are asking for this kind of trouble.

But to make the point, here is a class of examples, again brought to
my attention by Jim Entwisle: verbs like "shrugged", "winked",
"blinked", "smiled", "waved", "frowned" would normally be categorized
as intransitive, but allowed a prepositional phrase as an indirect
object ("waved to me") or noun phrase as a direct object ("waved his
hand") or perhaps both ("waved his hand at me"). However, all can be
used in a bitransitive frame like the following (substituting for
"gave"):

	He gave me an answer.

Some may seem a bit streched out of context, but it doesn't take a lot
to build up a context in which any one of them would be totally
acceptable. Virtually any action verb could be used. What about
"run", "jump" or "skip" you ask? OK, let's see what bitransitive
frames or contexts we can come up with:

	He ran me an errand.
	He skipped me a stone. 

Note that in elaborated contexts "answer" could be used - e.g. as the
function of the skipped stone! But with "jump" I want to illustrate
the dialectic point. I would tend to judge:

	* He jumped rope

That is an Americanism (and perhaps "jump rope" should be read as a
single lexeme) which is unacceptable in Australian or British English!
I could quite imagine that in some dialect the following was
acceptable (and children do come up with such things, which can thus
relatively easily enter the language):

	He jumped me a rope

Similarly
	
	He jumped me a stone
	He jumped me a tiddley-wink

Note that jump can quite acceptably take a direct object even in
British/Australian English:

	He jumped ship
	He jumped the puddle
	He jumped the burglar

As another dialectic example unacceptable to me/Australians:

	* He wrote me

but the following is acceptable:

	He wrote me a letter

Our response to all this is to decline to reject sentences on the
basis of subcategorization, and Entwisle's parser reflects this
attitude. Any other approach would seem invite an unacceptable number
of false rejections. The treatment of novel usages should not depend
on whether the word form is novel or not. This leads to quite a
different view of the distinction between syntax and semantics, as
subcategorization is no longer grammatical, but is purely semantic or
lexical in nature, and is thus subject to more variation than would be
expected of syntactic morphs.

David Powers
 
powersacm.org	 http://www.cs.flinders.edu.au/people/DMWPowers.html
Associate Professor David Powers		David.Powersflinders.edu.au
 ACM SIGART Editor; ACL SIGNLL President	Facsimile: 
+61-8-201-3626
Department of Computer Science			UniOffice: +61-8-201-3663
The Flinders University of South Australia	Secretary: +61-8-201-2662
GPO Box 2100, Adelaide	South Australia 5001	HomePhone: +61-8-357-4220
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: Re: Happy scissors (fwd)

Date: Wed, 10 Apr 1996 10:53:41 +0800
From: "Karen S. Chung" <karchungccms.ntu.edu.tw>
Subject: Re: Happy scissors (fwd)

Two forwarded responses (with permission from the writer) re the 'happy
scissors' question: 

	From: Glenn Bingham, Rowan College of New Jersey
	BINGHAMelan.rowan.edu

(1)
Karen,

Although your ideas about this subject may seem a bit green to some,
they have been sleeping furiously in my mind, waiting to emerge on the
linguistics community. Your ideas warmed my heart because, for a
couple of decades, I haven't been buying that standard ungrammatical
line dished out by many of the standard bearers of the discipline.

[Note: Analyze the grammaticality of this reply at your own risk. The
author is not responsible for any mental stress caused by it.]

(2)

I see grammaticality judgments as useful in eliminating the set of
utterances that are not among the possibilities for use in a
particular--or all of--language. In this light, whether semantic or
exclusively syntactic considerations are responsible for elimination
is not important on the broad scope, but only in the narrower scope of
deciding how to divide the study into convenient components. It
appears that not too much can be eliminated. Coming to grips with
that realization offers linguistics a hearty challenge.

If anyone is not convinced, ask them to judge the grammaticality of:

 "Yesterday, while surfing the net, I gophered over to Notre Dame."

As anyone lodged in a 1965 time capsule can plainly understand, this
outrageous combination of semi-English words and usage is simply
ungrammatical and useless for any English communication.

Glenn

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