LINGUIST List 14.3169

Wed Nov 19 2003

Sum: Grammatical Gender

Editor for this issue: Steve Moran <>


  1. Greville Corbett, Grammatical Gender

Message 1: Grammatical Gender

Date: 5 Nov 2003 20:16:55 -0000
From: Greville Corbett <>
Subject: Grammatical Gender

Editor's note: Due to problems during our server move this message has
been posted much later than our 48 hour target period. We apologize
for any inconvenience this may have caused the authors or our


Alex asks for an 'in-a-nutshell' answer (Linguist 14.3007) to the
question as to why languages can have apparently unmotivated instances
of gender assignment. In broad brush terms, here is an attempt.

Gender systems, whether sex-based or based on a human/animate
classification, always begin as semantically based. The instances
where we can see new genders being born, as in Daly languages, are
semantically based. And such systems can be stable over many
centuries, as Dravidian languages show.

Yet however simple the system of classification, there may be tricky
borderline cases: is the noun denoting an infant to be masculine or
feminine (or not yet within that classification)? Is an the noun
denoting an animal which talks (which occurs in various legends)
really to count as non-human? Is the noun denonting a god who takes
the form of an animal to be treated as non-human? Is an object with
immense religious significance to be treated linguistically as
inanimate? In many instances, such nouns are given the
.higher. classification; for instance, in Algonquian languages,
significant religious objects are treated as animate.

Again, we find numerous stable situations, in which the gender system
is predominantly semantic, but in which we have to specify that for
some nouns the gender is not absolutely straightforward.

In some languages, however, these nouns act as .Trojan horses., which
allow an incursion of much larger numbers of nouns into the apparently
.wrong. gender. So in several Bantu languages, the original gender for
nouns denoting humans is expanded to include large numbers of
non-human animates. In Konkani, most nouns for female humans have
followed the word for .girl. to become neuter.

Given sufficient mixing of this type, we can reach a situation where a
substantial part of the noun inventory is not assignable by semantic
rules. However, though we often read that .there is no principle for
gender assignment in language X., whenever the work has been done we
find that there are regularities covering the vast majority of
nouns. Subservient to semantic assignment rules we find formal
assignment rules. These may be phonological, as in Qafar, where the
final segment is a clear indicator of gender, or morphological, as in
Russian, where the inflectional class of a noun (which the speaker has
to .know. in any case) is an excellent predictor. And there are more
complex instances, where there is a substantial list of potential
predictors, giving rise to the interesting psycholinguistic question
as to which of these are real for some or all speakers.

That was 'in a nutshell'. The references for the examples I have
quoted can be found in Gender (Cambridge 1991). By the way, there will
be maps on Gender, and on all sorts of similarly interesting puzzles
in the World Atlas of Language Structures (eds Dryer, Haspelmath, Gil
and Comrie), to be published by OUP.

Greville Corbett
Surrey Morphology Group
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue