LINGUIST List 4.293

Thu 22 Apr 1993

Disc: Markedness and Exception

Editor for this issue: <>

Directory

1. Bill Croft, Number markedness revisited
2. Anthony Rodrigues Aristar, Markedness and Exceptions

 The summary of markedness in number contains a number of
inaccuracies (to my knowledge, at any rate) and can be presented in a
slightly wider typological perspective.
 The phenomenon is better described as "zero-marking" because the
typological pattern of markedness involves many other cross-linguistic
phenomena than simply whether or not a grammatical category is
expressed by an overt morpheme or not. These phenomena tend to
correlate, but a full understanding of exceptions requires the
examination of these other phenomena, so my remarks here will
necessarily be incomplete.
 The Russian data does not have to do directly with the grammatical
category of number. They have to do with the form of nouns after
numerals, in which (in standard Russian) a noun takes the nominative
singular after 1, the genitive singular after 2-4, and the genitive plural
after 5-10. This pattern is repeated after numerals that end in 1, e.g. 21,
31, 101, etc. and likewise for numerals ending in 2-9. However, it is
not found in other syntactic environments. While this pattern is
interesting in its own right (see for example the Greenberg articles on
number in his recent anthology "On Language", ed. K. Denning and S.
Kemmer, Stanford), it isn't a part of grammatical number. (In fact, in
many languages, the noun form after the numeral has no number
marking at all.) Russian has suffixes indicating both number and case
which are obligatory in all syntactic contexts. There is in fact an
anomalous pattern there, namely that for certain declension classes the
genitive plural is zero-marked (see below).
 It should be mentioned that some languages overtly mark both
singular and plural (e.g. Latvian), though languages that zero-mark the
singular are much more common. (Languages that zero-mark both are
normally said simply to lack the category of number.) So the
problematic case in typological marking theory are languages that
zero-mark the plural but overtly mark the singular.
 There seems to be a fairly well-motivated pattern that runs contrary
to this rule, which is found largely among Afroasiatic and Nilo-
Saharan languages of North Africa, and also in Celtic (Orin Gensler,
are you reading this?). In these languages, many nouns have a
'collective' form which seems to be a plural, and an overtly-marked
singular which is called the 'singulative'. These nouns denote entities
which are likely to occur together naturally (e.g. herd of sheep---note
the zero-marked plural in English), which Anna Wierzbicka argues is
cental to the understanding of the assignment of countability in
languages ("Oats and wheat", in her "Semantics of Grammar", John
Benjamins). It is not surprising that such nouns would have zero-
marked plurals or collectives. Typical collectives include animals,
insects, fruits that occur in bunches, people, etc. Of course, this
likelihood is a gradient phenomenon, and languages in this area differ
as to how many nouns have collective forms; Kanuri has only one, the
noun for 'man'. And it is rather irregular in some other languages, so
that the semantic motivation is no longer transparent (assuming it was
there in the first place). Finally, some of these languages also form a
plural off of the singulative with an overt marker called the 'plurative'.
This has suggested to some that the collective-singulative relation is
actually derivational, not inflectional.
 So much for the motivated exceptions. Now for the real exceptions.
The Russian zero-marked genitive plural is one; another is Old French
and Old Provencal, in which the nominative singular ends in -s and the
plural in zero. However, this anomalous situation corrected itself by
the modern versions of the latter languages, and is in the process of
doing so for some Slavic languages other than Russian (see Greenberg,
"Some methods of dynamic comparson", in the aforementioned
anthology). Icelandic is similar to Old French in the nominative, but
hasn't shown any tendency to "correct itself" that I know of. Matthew
Dryer has pointed out to me that Imonda, a Papuan language, has a
zero-marked plural and an overtly-marked singular. The Sinhala case
may be related to the evolution of determiners, which are sometimes
the sole carriers of number marking in noun phrases, to noun markers.
The Kiowa language is by far the most bizarre number marking system
I have ever seen. A more complete though highly simplified
description of Kiowa is as follows. Kiowa has a singular, dual, or
plural. Depending on the noun class, any one (or two) of singular, dual
or plural is marked with -ga, the other(s) being zero-marked. But there
is no semantic rhyme or reason behind the classification of the nouns.
Nevertheless, these genuine exceptions are quite rare in the run of
languages.
 The typological evidence for Link vs. Ojeda is thus rather mixed. It
seems that some noun types are more inclined to "unmarked" plurals
than "unmarked" singulars, which suggests that a more complex story
must be told for their semantic representation.

Bill Croft

Bill Croft's message (immediately above) is, I think, a useful corrective
to some of the assertions about markedness which have been made here on the
list, and elsewhere. While it's always difficult for anyone who works in
typology to assess the data they use in their work adequately--by the
very nature of the field researchers cannot know well most of the languages
they are examining--it's nevertheless true that "exceptional" cases have
to be defined with considerable care. Thus, as Bill has pointed out with
regard to the Russian data, the zero-marking found in the plural after
certain numerals is primarily an issue related to the requirements
of numerals, and secondarily an issue of case-marking. It is only
tangentially related to number-marking as such.

There's another group of cases, however, which I'm disturbed to hear
called "exceptions" to marking patterns. This group comprises those
languages where exceptional patterns have been produced from
diachronically earlier, unexceptional, patterns by sound-change. The
Russian genitive plural is one of these, as is the Old Provencal/Old
French nominative pattern. Underlying the categorizing of these as
exceptional is the assumption that universals are defined exclusively
on the basis of synchronic pattern, i.e. that the patterns themselves in
some sense instantiate the universals. Now, it's obvious that universals
are largely only visible from synchronic patterns. But it is a considerable
theoretical leap to assert as a corollary of this empirical reality that the
patterns are direct representations of universals. After all, universal
patterns are not like other synchronic patterns: they cannot be discerned
from the point of view of speakers of the language containing the pattern,
for they have no access to the patterns of other languages. Thus universals
must be of a fundamentally different nature from other parts of a
synchronic grammar.

I'd like to suggest that there are two possible reasons for the existence
of "universal" synchronic patterns. One is that the patterns themselves
are universal. The other is that the synchronic patterns we see are
simply side-effects of what may be called universal processes, i.e. are
the result of the process by which categories, through time, initially
come into existence. If the first reason is valid, then indeed Russian
and Old French are exceptions. If the second reason is valid, then they
are not, for what happens to a pattern after the process which caused it
to arise is complete is essentially irrelevant. After all, the pattern,
in this view, is not itself the universal.

Which of these views is valid is a matter for research. But it does seem to
me that the possibility that synchronic universal pattern is essentially
a side-effect of language change should be investigated. It does, after
all, seem to offer us more of an understanding as to why so many of
the universals we have proposed seem to have exceptions.