classifiers and plural marking

[Alan R. King]

I would like to second (or "sixth") the praise expressed for Edith
Moravcsik's effort to systematize the range of possibilities, and add a few
annotations.


--- 1 ---

>      - number marking is restricted
>        to a specifiable subset of
>        plural-referent nouns:            Basque, Hungarian, Welsh
>                                          Aymara?, Hawaiian, Yukaghir,
>                                          Mari (Cheremis)
>
>      - number marking is optional
>        in all contexts:                  Orok (Tungusic), most dialects
>                                          of Oroqen other than Western
>                                          (Tungusic) /QU to Linsay Whaley:
>                                          what about with numerals?/

Okay for a start, but this linear articulation misses the complexity of the
(sub-)possibilities for number-marking distribution that Edith herself
mentions a little higher up in her message, namely:

>b/ the language has plural markers but it does not always
>           use them on semantically plural nouns:
>           i/ instead, it uses them in only a specificable
>              subset of such cases; or
>           ii/ it uses them optionally (i.e., in a subset of the
>               instances involving semantically plural nouns which,
>               however, does not seem specifiable).

Her i and ii involve two theoretically independent qualifications on a
language's range of uses of number marking.  To make this explicit at the
risk of repetition: a language may plural-mark NPs with plural reference
SOMETIMES in two significantly different senses of "SOMETIMES".  The
"SOMETIMES" may involve:

(i) a (syntactically?) specifiable subset of cases, i.e. according to
grammatical rules;
(ii) true optionality, i.e. depending on factors other than syntactic rule
(*which* other factors is a question that we haven't said much about).

In other words, we may talk about *restrictions* on the use of plural
markers, or again about *flexibility* or *optionality* in their use.

If a language can be found exhibiting simultaneously both aspects of
non-universal plural marking, i.e. both restrictions on and flexibility in
plural marking, the two types of "sometimes" might in principle interact or
combine in different ways.  Thus there may be a language where in certain
contexts (restriction!) plural marking is OBLIGATORY and in other contexts
OPTIONAL (flexibility!); or again we may find one in which in certain
contexts it is OBLIGATORILY ABSENT  (restriction) and others in which it is
OPTIONAL (flexibility).  Not the same thing.  Finally, there obviously
might be a language in which there are three sets of contexts in which
marking is, respectively, OBLIGATORY, OPTIONAL and OBLIGATORILY ABSENT.

Now both of these kinds of "sometimes" can give rise to separate scales of
contrasting degrees of use between languages.  Between two languages with
retrictions on plural marking, one of the languages may have more
widespread occurrences of marking than the other.  And between two
languages with optionality of plural marking, one language may tend to make
greater use of it than the other.

Finally, there can obviously be qualitative differences as well as
quantitative ones, within these categories.  So two languages placing
restrictions on plural marking may very well not place the same set of
restrictions.  And in two languages where plural marking is optional
speakers may follow different (non-grammatical) criteria to decide when to
plural-mark explicitly.

These nuances, which I believe are quite essential for a meaningful
investigation of the issue with any depth, get rather lost in the range of
types that I quoted at the beginning of this message.  Besides the type
described by Edith as "number marking is optional in all contexts" (which
is one possibility), there should at the very least be another one: "number
marking is optional in SOME contexts".  But the only other type listed that
comes anywhere close was: "number marking is restricted to a specifiable
subset of plural-referent nouns", but that's not the same thing because
this says nothing about the possibility of optionality.

Secondly, I think some of the languages listed in Edith's illustration of
the scale are misplaced, partly or entirely because of the coarseness of
the scale just referred to.  Since some of the possibly misplaced languages
are ones brought into the discussion by me, perhaps I was insufficiently
precise or explicit in the way I introduced them.  (On the other hand,
maybe the others are also misplaced, unknown to me because I don't know
those languages!)  Let's talk about the ones I do know, though.

When I mentioned Aymara (also mentioning that I have hardly any knowledge
of it and didn't have my sources on hand - I still don't, but I can see I'm
going to have to dig them out!), I did not mean to suggest putting it in
the "number marking is restricted..." type (where Edith has it) but rather
in the "number marking is optional..." one.

Also, when I mentioned Hawaiian, I think I talked not about restrictions
but about optionality, or at least about claims of optionality by an
eminent Hawaiian specialist, Emily Hawkins.  I'm not sure about
*restrictions* on Hawaiian plural marking, and if there are any, how these
interact with  optionality.  I have studied Hawaiian (out of books and from
written texts only) to some extent, but this probably requires somebody
with more knowledge, and besides I just haven't studied this question in
such detail.  (I suppose I could try.)

On the other hand, Hungarian, Basque and Welsh are languages that DO belong
in the "number marking is restricted..." type, just where they are listed
above, a point that I argued to justify the first time round.

So it may be that Hungarian, Basque and Welsh should not be in the same
type with Aymara and Hawaiian, because the difference in their requirements
of plural marking opposes *restriction* in the first group and
*optionality* in the second.  Even if that is not ultimately found to be
true for Aymara and Hawaiian, given that the information I/we so far have
is insufficient, I think the theoretical point that such a conceptual
differentiation should be made remains potentially valid, pending the
search for languages to exemplify each of the resulting types or points on
a scale.  But I'm also saying that the range of variation probably doesn't
fit on a one-dimensional scale.

Of the other languages mentioned in Edith's scale (I didn't reproduce the
whole thing above), I believe that Germanic, Romance and Fijian (all
representing extremes on the scale) have been placed correctly: as far as I
know, in Germanic and Romance languages nominal plural marking is
obligatory (unless spoken French is a candidate to be an exception?), in
Fijian non-existent.  And as I've already said, Hungarian, Basque and Welsh
are also correctly placed.


--- 2 ---

The second generalization put forth by Edith, in her conclusion, is:

>      b/ If a language has (non-generic) nouns with plural
>referents that are not marked for number in any context other
>than numerals, it also has such nouns unmarked for number when they
>occur with numerals. (I.e., there is no language where plural-referent
>nouns are plural-marked when occurring with numerals but not
>plural-marked in other contexts.)

If the basis for this hypothesis follows from anything claimed in the
discussion so far, I must have missed it.  I am also a little surprised by
David Gil's quickness to take sides in the absence of much empirical
evidence at this stage:

>Be that as it may, Edith's generalization b/ is almost certainly valid at
>least statistically; and of course it makes obvious functional sense -- to
>the extent that languages try to avoid redundancy.

In any case, I'll try to join in the hunt for counter-examples.  My bet
(since we're betting!) is that we'll find some.


--- 3 ---

To return to Edith's typological scale, Gontzal Aldai's question, whose
point of departure is a Basque phenomenon that found its way into our last
bout in the discussion, focuses on one particular way in which (at least)
one language *restricts* nominal number marking:

>Can somebody say whether for any of the languages where "number marking is
>restricted to a specifiable subset of plural-referent nouns" (other
>than Basque): Hungarian, Welsh, Aymara?, Hawaiian, Yukaghir, Mari
>(Cheremis), there is a different translation for:
>
>	(1) three musketeers
>	(2) the three musketeers
>
>If so, does the equivalent of (2) show number marking?

I can only say that I have not yet encountered, or noticed, another
language that does this in a way comparable to Basque (but then most of the
time I wasn't looking for one).  We might guess quite plausibly that, as
Gontzal previously suggested, the *diachronic* explanation for the Basque
system involves the assumed demonstrative origin of the morpheme (the
default determiner) which in Basque usually accompanies the plural marker
per se.  If (as I was claiming last week) this default determiner is not
synchronically, in general, a marker of definiteness, the usage here
focused on by Gontzal would still make a nice illustration of the
phenomenon in Grammaticalization theory that Paul Hopper has referred to as
*persistence*, in which "later constraints on structure or meaning can only
be understood in the light of earlier meanings" (Hopper & Traugott,
Grammaticalization, p. 90).

I don't know how useful counter-examples are in a case like this, but
here's one from Welsh, a language already in our "corpus":

tri dyn 'three men'
three man

y tri dyn 'the three men'
the three man

Definiteness in this Welsh noun phrase is expressed, unspectacularly, by
the use of the definite article.  Welsh has no indefinite article;
indefinite NPs simply lack a determiner, e.g.

dyn 'a man'
y dyn 'the man'

dynion 'men'
y dynion 'the men'

The Welsh definite article "y" is invariable for number.

Now, suppose we had a language like Welsh but in which the definite article
had distinct singular and plural forms.  We would then find, unless there
were other surprises, that what Gontzal Aldai asks for happens: 'three
musketeers' would (as in Welsh) have no grammatical number marking, but
'the three musketeers' would have number marking (in 'the').

David's Hebrew example in reply to Gontzal is very nice, although it
unfortunately only covers a restricted subset of cases:

>(1) (a) 9esrim iS
>        twenty man
>    (b) 9esrim anaSim
>        twenty men
>
>(2) (a)*9esrim haiS
>        twenty DEF-man
>    (b) 9esrim haanaSim
>        twenty DEF-men

This is also a nice example of what I was saying about different ways in
which languages can contrast.  In Hebrew, then, plural marking is mostly
OBLIGATORY, even with numerals, but the obligatoriness does not apply (it
is reduced to OPTIONALITY) in a (pretty small) specifiable set of contexts:
namely, when the numeral is one like "9esrim" AND the noun phrase is
definite (with the definite article ha-) - oh yes, AND when the register is
non-colloquial!  Still, it's an example.

Alan