How Does A Language Acquire Gender Markers?

Universals of Human Language
Edited by Joseph H. Greenberg
Associate Editors:
Charles A. Ferguson & Edith A. Moravcsik
VOLUME3
Word Structure
Stanford University Press, Stanford, California

1978
• l
e
o <J
UT.
I,0622l
71/
e
-
G-
z-
0
o
T.t
Stanford University Press

Stanford, California
@1978 by the Board of Trustees of the
Leland Stanford Junior University
Printed in the United States of America
ISBN 0-8047-0968-8
LC 77-89179
Several of the papers in this volume were published originally in Working Papers on Language
Universals, published for limited distribution by the Language Universals Project at Stanford University
in sixteen numbers, 1970-76, as indicated in the opening footnotes to the individual papers. Most have
been modified to some extent for publication here.
How Does a Language
Acquire Gender Markers?
JOSEPH H. GREENBERG
ABSTRACT
The bulk of this paper is devoted to outlining a process by which,

in a number of languages, a definite article which agrees in gender
with the noun passes through a stage in which it combines the uses
of a definite and indefinite article and finally becomes a gender
marker on the noun. It is also shown that this process can occur
even when the definite article does not agree in gender, in which
case it ends up as a sign ofmere nominality. The role of the de-
monstrative from which the definite article usually arises in gen-
erating agreement phenomena is also discussed. The problem of
the origin of gender is then briefly reconsidered in the light of the
evidence adduced in the body of the paper.
Joseph H. Greenberg GENDER MARKERS 49
48
1. Introduction
By a noun gender system will be meant a system in which the

noun stems of a language are divided into a set of genders, this
CONTENTS distinction being based on the fact that the choice of a noun belong-
ing to a particular gender determines the choice among a set of
alternative "agreeing" forms in one or more other classes of mor-
l. Introduction • 49
phemes or words, e.g.: articles, demonstratives, adjectives,
unbound anaphoric pronouns, pronoun incorporated in a verb com-
2. Types of Noun Classification 50
plex, etc. Such systems are often called noun class systems.
However, it is not usual to apply this term when sex is among the
3. Overt and Covert Systems 53
bases of classification. Structurally, such systems do not differ
3.1 From covert to overt gender: Niger-Congo in any basic way from those in which sex does not figure, 'Gender'
languages with simultaneous prefix and suffix ultimately comes from LA TIN genus which simply means ''kind,'
marking . 54 and there is ample precedent for applying the term gender to all
3.2 The development of class markers outside of such systems. When sex figures among the bases of classification,
Niger-Congo 58 we will refer to sex gender. There is a further advantage in using
3.3 The cycle of the definite article: the initial stage 61 the term gender system rather than noun class system. There are
3.4 Stage II of the definite article. 62 at least two other systems of noun classification than gender sys-
3.5 The Stage Ill article. 69 tems numeral classifier systems and possessive classifier systems.
We are therefore free to use noun class system as the superordinate
4. The Spread of Grammatical Agreement 75 term which includes all systems of this type.
5. The Problem of Origins 78 The present study is · base d on a b roa db u t by no means exhaus-
. . d t ms The approach
tive sample of languages with noun gen er sys e · . .
Bibliography . 80
:is largely in terms of diachronic
: : process. Our ultimate aim is to
. th typical course of develop-
establish how such systems arise, eir 1 Whil
ment, and the factors inv olved in their ultimate extinci "? ,
· d here it is ope a
such an ambitious goal is clearly no t a tt aine '. . d' tion
thi, .: <fi t progress in this tirec ·
s study at least represents signitcanT . Th first
The content of this paper falls into four main sections •t t t: e study
. b • stions relevan o
o f t h ese (Sec. 2) poses certain a sic que t d,·finitional
. . l d' me fundamen a 1 e
O1 f nominal gender systems includi ng so1 Eh < body of
. h .stitutes t e main
ones. The second one (Sec. 3) whicl con h . languages
the paper develops a processual theory as to ow, int acquire gen-
whi h . c] h already have gender sys t ems, the nouns come·15r o
period or not.
d tat an ear1e
er markers, whether they were presen l" f Old markers which
If th e former is the case, we h ave a 'renewa.. Ott 'tion The exis- •
. d b honetic a r1 •
in some cases have been weakene 'y P . h cases usually
t f ·kers is in suc
ence of two historical layers o mar If however, such
evid, ent from the existence o: fd, Iou ·ble markers. Many gender
d'd 1 not exist previously, we have a pr istinek case, on the noun, and
sy t h ·ert mar ers
s ems, of course, do not ave ov forms outside of the
genders are distinguished only by agreeing . s are discussed
non the »maining top1c
un. For reasons of space e re
•
\,
50 Joseph H. Greenberg GENDER MARKERS 51
much more briefly, details being reserved for future treatment. Noun classification can exist without concord just as concord
The third section {Sec. 4) deals with the spread of syntactic agree- can exist without noun classes, though whether, synchronically
ment which is coincident with the process by which the nouns receive viewed, such a classification can be considered grammatically sig-
overt marking, but which also develops in its absence. The final nificant is another question. Of course any number of classifica-
section (Sec. 5) is devoted to some speculations regarding the rise tions of nouns, as of any other word class, can be carried out,
of gender and the relations of the conclusions of the present study which are syntactically irrelevant, though some may be relevant
to previous theorizing concerning gender origins. The important from some other point of view, e.g. the declensional classes of
topic of the classificational system as such in relation to its semantic LATIN. Others might be perfectly arbitrary, e.g. a classification
bases is only touched on incidentally and is also reserved for future of bases into those with at least one stop consonant. The following,
treatment. however, turns out to be a kind of classification which, while it
does not involve agreement, may be of potential diachronic rele-
2. Types of Noun Classification vance to the origin of noun classes. In MALAY and certain other
languages there is widespread and compulsory use of certain super-
Our preliminary definition of a noun gender system in the initial ordinate terms in lexicalized phrases parallel in formation to ENG-
paragraph is clearly in need of further specification. It will be LISH 'apple tree. 1 Thus, in MALAY one cannot say merely Djawi
repeated here for convenience. By a noun gender system will be for 'Java I but rather tanah Djawi where tanah means 'land' and
meant a system in which the noun stems of a language are divided occurs as an independent word, and so for a number of other super-
into a set of genders, the distinction being based on the fact that the ordinates. One might say that this constitutes a classification in
choice of a noun belonging to a particular gender determines the which each superordinate term defines a class and in which those
choice among a set of alternative "agreeing" forms in one or more nouns which do not have such a superordinate term constitute an
other classes of morphemes or words, e.g. articles, demonstra- additional class with a zero marker. In JAVANESE this type of
. function,
classification has absolutely no syntactic : an d'1 t rmg
' ht
tives, adjectives, unbound anaphoric pronouns, pronouns incorpor-
ated in a verb complex, etc. therefore be considered far-fetched to consider it grammatically
relevant.
A noun gender system may be regarded, then, as involving the
intersection of two basic factors, classification and agreement, the . . . l. ·xists in the Daly River
However just such a system in principle e
two being in a relation of mutual determination, the gender being subgroup of general AUSTRALIAN, spoken in the north weSt ~f tt~e
defined by the agreements and the agreements being determined by continent. In some of these languages it seems to have abso y
the genders. The mutual determination here defined as constitutive no syntactic function. Yet, for example, in Tryon's grammar o
of gender systems is not, however, a logical requirement. Agree- thi type of system 1s
MARANUNGKU, a language of this group, nus • t
ment is possible without classification. For example, in HUNGARIAN 70) A noun must be ong o
ca 11 ed a noun class system ( Tryon 19 · d db awa
ebben a kertben 'in this garden' there is agreement in case between . . t f nouns precese y _,
one of four classes. The first consis s o1 A,{El
eb-ben 'this and kert-ben 'garden,' both being in the locative. Yet t: <:h occur ind ependent.y.
itself a noun meaning 'meat' whicl can wa pat-
HUNGARIAN is not a gender language, since differences in agreement Th, ese nouns all designate edible
; :
animal ls or. insects, ,f
e··Ins "
always
patterns do not divide the noun stems of the language into classes, Patt , grasshopper. 1 Similarly, class. .
t
two
g
consists o nou
tools weapo
ns and wooden
In HEBREW yeled tov 'good boy,' versus hayyeled hattov 'the good P receded by yili ' stick' and compn srn ' db miya 'vege-
boy,' there is agreement in definiteness between yeled 'boy' and tov irn 1 are precede Ygetable foods.
I d I -
plements, while class three nouns f
goo@, but although HEBREW does have sex gender, it i s not shown table food' and precedes nouns indicating kind s",'f +eterogeneous
in this construction because all noun bases behave similarly in regard Th, 'ker and is ug
e fourth class which has a zero mar be seen from
to it. It would be useful, then, to distinguish the wider notion of con- is t As far as can
noted as being by far the larges · t tic relevance,
cord from agreement, the latter being a subtype in which the choice Tryon's account this classification has no syn ac up there are
of alternative concord elements depends on the class to which the B ' D 1 River gro '
st em of the governing item belongs, whether marked by an affix or Owever, in other subgroups of the )alY ., ·1early related to
lang ,f classification c! t
not. uages which have a system o c b • s of agreemen
that of MARANUNGKU in which it forms the as1
phenomena, For example, in MARENGAR (Tryon 1974:122) the a classification is supposed to be exhaustive and to divide the uni-
same four classes are found, but we find, for example, agreement verse under consideration into mutually disjoint sets. The latter
with adjectives. Thus, corresponding to class two in MA RANUNGKU requirement is probably not fulfilled in any of the languages which
are words prefixed by yeri 'stick' = MARANUNGKU yili_. With the are usually considered to have gender systems. To begin with
adjective -kati 'good' we have yeri-kunt ikin' yeri-kati 'good boom- there are instances of individual stems which belong to more than
erang.' With nouns in the large class with a zero marker we find, one class, e.g. SPANISH hij-o 'son,' hij-a 'daughter. This is what
for example, watYan kati 'dog' in which neither item has a class is traditionally called 'motion, 1 and insofar as there is a systematic
marker. semantic relationship, it becomes one of the objective bases for
assigning meaning to classes. More drastically, two inflectionally
In this instance, for reasons discussed at the end of Sec. 3.5, different classes may share all of their stems. This is frequently
I believe we have to do with a nascent system. At the other end of the case with paired singular and plural classes. Such classes may
the development, diachronically, are languages like LOBI in the be said to be in equipollent relationship, and in a sense, to be con-
VOLTAIC subgroup of NIGER-CONGO which has paired singular sidered a superclass. Specialists in BANTU, for example, have
and plural suffix sets, historically related to those of VOLTAIC differed as to whether in such pairs they are dealing with one class
languages with functioning systems, but no longer having syntactic or two. Since viewed as a superclass it does have a distinct mem-
relevance. Viewed synchronically, LOBI simply seems to have a bership, we may say that this does not detract from the overall
very elaborate system of plural formation by affix replacements, system in classifying the noun stems.
falling into a number of different patterns,
Sometimes, however, the relationship between two inflectionally
Viewed synchronically, then, and no doubt tautologically, a noun different classes is not equipollent. This is often true for diminu-
classification system, if it is syntactically relevant, involves some tive, augmentative and piace gender as found, for example, in many
form of agreement. But what of numeral classifier systems? These BANTU languages. Most, if not all, stems found in such classes
classify nouns in accordance with the classifier they take, and the are also found in others. There may be a productive rule by which
classification is clearly a syntactically relevant one. A choice of a member of any other class becomes a number of this class by
a particular noun determines the choice of the classifier. There affix replacement or even by addition. In these cases we may say
are, however, some differences between such systems and those that we are dealing with a derivative gender. Wh ere, as is some-
. hi h' ique
usually called gender systems. One important difference is that in times the case, a gender does not have a single stem w ic is un '
a full-fledged numeral classifier language, there are considerable it will be called a minor gender. An example is the neuter gender
number of nouns which are non-countables and do not take classifiers, in NAMA HOTTENTOT. In view of these considerations, we will
. . .
rn ean b y a gender classification . hi .h there are at least two
whereas in gender languages, literally every noun has a gender. Sec- one in wuc
1 , la s I as men-
ondly, in numeral classifier languages, with a few exceptions, agree- c_ asses, including in the term 'class' also superc s '
ment is confined to the construction with quantifiers. In spite of honed above, but excluding minor genders in our count.
th ese differences and others that will be mentioned later it will be
shown in the final section that there is a possible diachr~nic connec- 3• Overt and Covert Systems
tion between numeral classifier and gender systems. What is said
I 'd b ut the existence of
here about
· numeral cla ssifi
"f" er sys:ems
t ·
applies mutatis· mutandis
· to n defining agreement, nothing was sai a o1 hi in a
a g d f, ·ly that members up
possessive systems, such as those found in some OCEANIC and en er marker on the noun itsel , mere . f However,
AMERIND languages · Ag reemen t is · confined
· · part; 1, 1, ·ti. ' agreeing torms.
to the possessive Vicuiar gender determined selecti on 1n ·ert class
st in itself has an ov
con ruction, Corresponding to non-countables in a classifier lan- such paradigm cases as BANTU, the noun ' .ti li-nee_hi-ki
age are inalienably possessed nouns which are not subject to clas- mark, : ·A SWAHILI phrase !lz'!"i""
1
er as, for example, in tle in INDO-EUROPEAN.
sification. ""ool large this,' and this is also generally "!{ maruu saev5rum
II 1
!Pir mentions the "relentless rhythm of LA . ts in the noun
One further aspect of the definition of gender system as stated v rG. 1, Wh e a marker ex1s
~ of those savage men.' ere t is not found, the
1 1
here is in need of further elucidation. Wh at is meant by the noun self, the system will be called overt; where diachronically,
stem of a language b eing
. d;livi
ad..d ·
.ec into a set of genders? Ideally, System will be called covert. Viewed once more
54 Joseph H. Greenberg GENDER MARKERS
55
there are two types of covert systems, those in which overt expres- and suffixes. However, the detailed grammar of Chantoux, Gontier
sion formerly existed, e. g, FRENCH, as compared to ITALIAN, and Prost (1968) shows that whereas the suffixes are a normal part
and those in which, as far as can be seen, it never existed. There of the noun, the prefixes are often omitted. When a prefix is present,
are many examples of this, for instance, the NORTHERN CAUCA- it corresponds fairly closely to the definite article in languages like
SIAN languages (except the northwestern group which has no gender) ENGLISH and FRENCH, e.g. niti-ba 'men,' ba niti-ba 'the men.'
In fact the authors of this grammar write the "prefixes" as separate
This is one more point of difference between gender systems and words. Furthermore, as we shall see, there can be no reasonable
other systems of noun classification. In numeral classifier systems doubt that the preposed article is recent and the suffix old.
the classifier goes syntactically and is sometimes fused with the
numeral and never seems to occur additionally with the noun itself. GURMA gives its name to the whole subgroup of VOLTAIC lan-
Similarly, in possessive systems the classifiers go with personal guages of which it is a member. An examination of the other lan-
possessive affixes and do not appear on the noun. They thus fail to guages of the subgroup shows that in regard to what corresponds
give rise to phenomena such as the "alliterative concord" of BANTU historically to the preposed article of GURMA, they fall into a
languages. number of typologically distinct groups. To begin with, there is
MOBA which shows no prefixes at all. It is thus similar to the
3.1 From covert to overt gender: Niger-Congo languages with great mass of VOLTAIC languages which have class suffixes only.
simultaneous prefix and suffix marking There is GURMA itself in which, as we have seen, the preposed
class marker functions as a definite article. Another language,
The NIGER-CONGO languages have an inherited system of noun GANGAM, has a preposed marker like GURMA, but unl ike GURMA,
genders involving the existence of overt class markers on the noun. it is not merely a definite article, It combines, roughly speaking ,
In some of the major subgroups, these are generally prefixed to the the uses of the ENGLISH indefinite article when it is [ +specific],
noun (as in the well known instance of BANTU), but in other branches that is it involves an existence assumption and can in general be
, . k I
they are normally suffixed. The markers themselves, whether pre- replaced by 'a certain,' e.g. 'I am looking for a certain noteboo •
fixed or suffixed, exhibit clear relationship in phonetic form and Statistically, it occurs with the noun in the vast majority of its
semantic function. The WEST-ATLANTIC, KWA and BENUE- textual occurrences. However, there are a minority of construc-
CONGO branches are basically prefixing, while the VOLTAIC and .
ti ons, especially . ones, e.g. as a ver b object in negative
generic
ADAMAWA-EASTERN branches are suffixing. The MANDE branch, sentences in which it does not occur, I shall call th is type of .
which is the most divergent genetically, shows no clear traces of 11 • ' will be indicated in
article'' a non-generic article, However, as . .
this system. An example of prefixing is SWAHILI ki-ti, pl. vi-ti l at ; : fabsence of this arti-
aer discussion, by no means all instances o. .st
'chair' and suffixing, MOSSI (VOLTAIC) bi-ga, pl. bi-se 'e+a. cle involve genericness, and the name is not an entirely appropr1ate
Alongside of this, however, there are a few languages in each major , . ber of languages in
one, Articles of this kind are found in a num . i·ttl at-
branch except, of course, MANDE which have prefixes and suffixes v · . s relative 1y .1 .e
arious parts of the world and, as it appear,
simultaneously. An example is AKASELE (VOLTAIC) ke-ji_-ke/ tention has been paid to this phenomenon,
n-Ji-m (< t.<m-ji-m) 'knife.' How is this diversity to be explained?
AKASELE and TOBOTE-
One of the suggestions has been that languages with both prefixes Finally, within the GURMA group are th e t between prefixed
and suffixes represent the original state of affairs and that subse- BASARI languages in which there are no contras s !ions to be
quently some languages lost their suffixes and some their prefixes and 1 w1"th a few . excePand the pair
non-prefixed forms. In genera.,
(e.g. Welmers 1973: 209), However, a closer look at languages d; s d suffixes,
1scussed, all nouns have both prefixes an ffi by themselves
with double affixing tells a quite different story. 'unctions basically in the way as prefixes or sut ixes
1n the NIGER-CONGO languages.
We will st art with GURMA, a VOLTAIC language which was often
four stages of a pro-
cited in th e literature as a typical example of a language with prefixes We thus have within the same subgroup, set of prefixed
ce 1 ' nt of a new
ss .eading ultimately to the developme! ti g suffixes, These
H owever, in most CAUCASIAN languages there are a few words Clas, : sly exist.n ·ti:le
which have class mark ers. Thi, s matter is
, discussed later at the s markers reinforcing the prev1out 3) n-generic artc1?»
stage ,Gt article, no
end of Sec. 3.5. 4, 'es are: 1) no marker, 2) defi n1e
) class prefix.
That this is the order of development is shown in a number of Similarly, in the KWA group which is prefixing, there are lan-
ways. To begin with, the suffixes are representative of a set found guages like KEBU and AVA TIME in which suffixes are found, In
through the entire VOLTAIC subgroup and reconstructible for the case of AVA TIME once more they function like the definite
PROTO-VOLTAIC, They are often irregular phonetically and some- article. In KEBU we have a language in which nouns are normally
times reduced to zero or tonal alternations. The prefixes, in con- both prefixed and suffixed and in which, on internal evidence, it
trast, are transparent and in fact identical with the verb subject appears that the suffixes are more recent, though the evidence is
pronouns which agree with the noun in class. Moreover, that a perhaps not decisive.
functional prefix detaches itself from a noun, gradually becomes a
definite article, and then disappears is like running the historical Among BENUE-CONGO languages, TIV which is closely related
camera backwards. Manessy, well-known as a VOLTAIC specialist, to BANTU has both prefixes and suffixes. The suffixes are clearly
in an article discussing the phenomenon of double affixes in VOL- recent. The suffixes essentially exhibit an advanced stage of the
TAIC languages, comes to the following conclusion. ' Thus Tamari non-generic article. Nouns can appear without the suffix mainly in
and Ngangam appear to mark an intermediate stage in the develop- one construction, the locative. As we should see from a more de-
ment of a process quite comparable to that by which the Latin tailed examination of the non-generic article, this is a typical con-
demonstrative became the French article, a process whose com- struction in which the form without the article tends to survive.
mencement can be discovered in Gurma and of which Akasele
illustrates the highest stage of development" (Manessy 1965: But if suffixes can be renewed by prefixes (e.g. GURMA) and
175). prefixes can be renewed by suffixes (e.g. DYOLA), why cannot
prefixes be renewed by additional prefixes or suffixes by suffixes?
Manessy here mentions, in addition to NGANGAM, TAMARI as Regarding the former of these possibilities, there are examples
illustrating the same stage of a non-generic article. However, in in NIGER-CONGO, but I do not know of any instances of the latter,
subsequent publications he correctly assigned TAMARI to a different
genetic subgroup of VOLTAIC. In almost all dialects of TEMNE, a language of the Southern
branch of WEST ATLANTIC (the other languages already mentioned
But if suffixes can be renewed by prefixes, why cannot prefixes belong to the Northern branch), the prefixed class marker has an
be renewed by suffixes? In fact this process is well attested in the additional marker, usually the same as the vowel of the prefix
languages of the WEST ATLANTIC subgroup of NIGER-CONGO which functions as a definite article. Wilson (1961:13) notes that
1
which are, as noted earlier, basically prefixing. The definite ar- in one dialect that of KONIKE, 11the nouns in all contexts are ~ bl
ticle stage is represented by DYOLA which has, in addition to the what 1ooks 1le h e definite form. " This dialect, then, has
. th. f
PPP?
r definite artic.e as
usual WEST ATLANTIC prefix system, a suffixed article which entered into the final stage in which e torme
duplicates the prefix, e.g. fu-nak 'day, ' fu-nak-af 'the day' (Sapir become part of the class marker on the noun.
1965). In certain other WEST ATLANTIC languages the original
fi being renewed by pre-
prefixes have been reduced to initial consonant alternations and, in The most important example of pret1xes eli, re, the well-
the case of WOLOF, even these have been reduced to the status of fi : fr so I el1ev,
1xes is that of the BANTU languages, 0 ' hould
archaic survivals in a few words. The system has been renewed, k . BANTU languages s
nown pre -prefix vowel found in many
however, by suffixed class markers which can in certain construc- be interpreted.
tions still be separated from the noun and which function as a def- .
fix functions as a typical
inite article. SERER-NON illustrates the stage of the non-generic In the majority of instances the pre-pre t bundant evidence
article. FULA which still has functional ini tial consonant alterna- no d "des the mos a
n-generic article and indee provi h" tage, Examples
t
tions illustrates the last sage. Th e class suffixes
· have become we h, ·t ' tics of thi s s U
. ave anywhere of the charac:er1s DE, LAMBA, ZUL'»
markers on virtually all nouns,and no contrasts exist between con-
structions with and without the suffix. WOLOF SERER-NON and
include LUGANDA, KIRUNDI, BEMBA,
XHOSA and SOUTH ERN KIKONGO. Thus
"9,,
"°, o tea only one
aien functions as
FULA are the three memb ers o f one subgroup of ' northern WEST
examp le or a BANTU 1anguage with a pre-""N,"" l eer gro up of
ATLANTIC. a d e.fi.nite
. article, namely DZA MBA · The sou
58 Joseph H. Greenb erg GENDER MARKERS
59
BANTU including such languages as HERE RO, OV AMBO and OCHIN- Another example of a pristine marker, in this instance without
DONGA illustrates the last stage in which the pre-prefixes have connotations of either gender or number, is provided by the South-
become part of the class marker on nouns in general. 2 western MANDE languages. We have here a suffixed -i which func-
tions as an ordinary definite article in KPELLE, but is-a non-generic
3.2 The development of class markers outside of Niger-Congo article in the remaining languages: MENDE, LOKO, GBANDI and
LOMA,
A priori, there is no reason to expect the process by which a
definite article becomes a marker on the noun to be confined to As a final African example, we may cite a main subgroup of lan-
NIGER-CONGO languages. Secondly, there is no reason why it guages of the Central group of Southern KHOISAN languages including
should only occur in gender systems which do not include sex as here the various dialects of HOTTENTOT and the closely related
a semantic basis for gender. Thirdly, there is evidently no neces- languages of the Naron Bushman. These languages have a system
sity for the process to be one of renewal confined to languages which of three genders, masculine, feminine and neuter. They all, with
already have overt markers, as is true with NIGER-CONGO lan- the possible exceptions to be noted, have a pristine non-generic
guages outside of the MENDE group. Fourthly, as we shall see, suffixed article as gender-number markers, However, the extinct
there is no reason even for this process to be confined to languages HOTTENTOT dialects of Griqualand East and West are known from
with gender classification at all, word lists in which the gender-number suffixes do not occur. In
languages with the non-generic article, it is usually employed in
All of these possibilities can be illustrated from languages in citation forms. Hence, in these dialects the affixes were probably
Africa. The languages of the EASTERN NILOTIC subbranch of either definite articles or not used at all (cf. Planert 1905).
the NILOTIC languages, which in turn belong to the EASTERN SU-
DANIC substock within CHARI-NILE, the major branch of NILO- The phenomenon under discussion is not confined to Africa, al-
SAHARAN, all have sex gender. These languages fall into two though it seems to be more frequently encountered there than in any
subgroups, one consisting of BARI and other languages very other major region. I will not, here, cite either the numerous cases
closely related to it. The other consists of a set of internally of languages with definite articles of the usual sort or languages
more divergent languages including MAASAI, LO TUKO and TESO. with gender classification in which the noun has an overt marker,
BARI has a sex gender classification without any marker on the but rather those instances in which the tell-tale and hitherto neg-
noun or a definite article. The gender elements appear in certain lected stage of the non-generic article is known to occur. A par-
constructions, e.g. in the demonstratives, pronominal possessives ti cuiarly
1 .
interesting . ARAMA. IC wh1cl,
case is h' h 1n a recorded history
s. d
and relatives. In the other group of languages, there are geneti- of almost 3000 years, has gone through the three stages mentione
cally related prefix markers for sex gender (TESO, in addition, h ere. An -a (<·ha) 'suffixed to nouns in . h
the masc
uline and feminine,
has a neuter) which are typical non-generic articles. As against • - ~ "d d with sex/num er
Singular and plural, which were already prov e
the NIGER-CONGO case, these are from all evidently pristine k ' TIC tern functions as a
marl:ers based on the inherited SEMI sys th tu 'y
markers, and none of the languages has reached the stage at which d f" . . . 1 language (nint cent r
eti nite article in the earliest inscriptiona) , d Nehemiah
the prefix has become a mere classificatory mark on the noun. B,C ) ·ti of Daniel an ene
• • and as late as the ARAMAIC por ons Christian
th 1
(generally dated second century B.C.). From e ear y Western
A further possibility is exemplified by the SOUTH ERN NILO TIC era · h tw dialect groups,
onward, we have to reckon witl 'o ti e in this mat-
(KALENJIN) languages which constitute another subbranch of and E • re conserva 'V
astern, in which the former is mo C BABYLONIAN
NILO TIC. The NANDI-KIPSIGIS and POKOT (SUK) groups have ter. The Eastern literary dialects (e,g, SYR!Ah 'western literary
genetically cognate suffixes for singular and plural without gender TALMUD; . ·ticl e while the
. IC) have a non-generic ar c . . Palestinian SYRIAC,
classification. In POKOT these function like a defini te article dialects still have a definite article (Christia" 4al eets of ARAMAIC
while in NANDI-KIPSIGIS they are in the stage of the non-generic TARGUM ONOELOS, etc.). The cont empora" ,,avanced one
article.
;till are distinct in this respect, but each group generic article
2 urther stage. WEST ARAMAIC now has a 0 "" ~arker. One
However, these languages do have the unarticulated form for Wh.il . . . genera noun . d
e in EASTERN ARAMAIC, it is a da new prefixe
the vocative of common nouns so that, strictly speaking, they are Eastern dialect, that of TUR-ABDIN, has develoP' ,, u s renewing
still in the stage of the non-generic article, albeit an advanced stage, art· ' • d and num
1cle, this time distinguished for gen er
h . ch started in prehistoric times at least 3000 years

the process wi There are also some instances of what appears to be the last
earlier. stages of a non-generic non-classifying article, e.g. the -s of
KLAMATH and other PLATEAU PENUTIAN languages of Oregon,
KHASI, an AUSTROASIATIC language of Assam, has a three which appears on most nouns and in most constructions. The "ab-
gender system; masculine, feminine and neuter with a prefixed non- solutive'' of UTO-AZ TECAN and other AMERIND languages pos-
generic article. sibly also belongs here. This requires further investigation.
A number of AUSTRONESIAN languages on New Britain and New The foregoing review is far from exhaustive. There are doubtless
Ireland possess non-generic articles. These include MENGEN other examples of languages in the stage of the non-generic article
(New Britain) with a suffixed article (Muller 1907) and TOLAI (New or instances of closely related languages which exhibit different
Britain), and the only dialectically differing LABEL (New Ireland} stages of the development just outlined.
with a prefixed article (Franklin 1962 and Peekel 1929).
3.3 The cycle of the definite article: the initial stage
Native Australia offers interesting examples. DYIRBAL in
northeastern Australia, a member of the PAMA-NYUNGAN group In the foregoing sections three stages in the process by which a
which comprises most of the continent, has a non-generic article definite article ultimately may become a noun marker were briefly
at an advanced stage associated with a classification of nouns into indicated. It remains to consider this process in greater detail
four genders: masculine, feminine and two inanimate genders, The three stages were: 1 ) d e f1n1 1
.. t e ar ti c.e, 2) non-generic article,
.
with animals divided between the masculine and feminine (Dixon : : b; uent discussions to
3) noun marker. It will be convenient in su,s eq .
1972). In the northwest of the continent there is a concentration of tid es respective1 Y•
call these Stage I, Stage II, and Stage Ill ar '
languages belonging to various other subgroups than PAMA-NYUN-
GAN, most of which have gender classification. An interesting ·n1"tiated by a significant change by means
Each of these stages is l ft
case is that of WORORA and the closely related UNGARINYIN. there are instances o rans1-
of which it can be defined. However' the borderline
Both languages have a system of four genders much like that of . hth t ·tain languages are on
tional phenomena sue a cer 1 "bl t see that a lan-
DYIRBAL, with class markers that are clearly related to each other b etween two stages. In some cas es it is poss1 e obile in other
historically and more remotely to those of DYIRBAL. WORORA, :hi rticular stage, w
guage is well advanced wit rn a pa d th tage recently.
however, has suffixed class markers on a large proportion of its . h th nly entere e s
instances it is clear t at 1 as O • d single continu-
nouns which have reached the final stage of complete amalgamation · to be viewe as a
Hence, the whole developmen t 15 • oints
and universal use while these are lacking altogether, as far as can
ous process marked by certain d, .ec1'sive turning p '
be made out from the rather sparse information on UNGARINYIN
(Love, n.d.). . . f the definite article is the demon-
Since the most comm on origin o: 1s and well-attested
strative a development of wac h . h there are numerous s stage zero.

Th
e
I do not know of any certain instances of the non-generic article th demonstrative a
examples we might speak o e b n studied in a num-
in AMERIND languages, except possibly GOAJIRO, an ARAWAKAN
historical development or defini te article h?",,,, rrom a purely
language as described by Holmer (1949). This is a sex gender lan-
ber of instances in considerable detail •. It evel ment as previously
guage in which gender marking has been renewed through a set of
deictic element which has come to identify an e : dditional function
suffixed non-generic articles. The development, if reliably reported, . s often an .
mentioned in discourse. Such a use1 B ut sometimes there is
must have been very rapid, since Celedon less than a century pre- deictic, ud this as its • b asi·c
O. f an element which is also a pure
viously, although describing GOAJIRO as a sex gender language,
a particular demonstrative which has assume e which points to lo-
gives no indication of the existence of the specific markers £ . • . : ost often on d person,
cited by Holmer. There are a number of AMERIND languages wi th unction. The source deictic is mc the first or secon
cation near the third persons rather thand ' urse deictic becomes
sex gend er with markers on the nouns, which indicate that they should hi h a lisco read
have gone through the sequence of stages postulated here, e.g. CHI- e·g. LATIN l le. The point at wIc lsory and has sp.
a d . -- . b mes compu th including
NOOK, a PENUTIAN language, and CHIQUITA, a MACRO-GE lan- efinite article is where it eco!a' ' in general, 1us ,Eh
guage of Bolivia. to the po int at which it means 'i dentifie ., knowledge, or as WI
t,: t ext, gener
YPIca Uy things known from con e »
63
'the sun' in non-scientific discourse, identified because it is the grosso_modo to the combined uses of a definite and indefinite article.
only member of its class. Such an article may, as with GERMAN Frequently, existing descriptions are simply not sufficient for us
der, be an unstressed variant of the demonstrative, which continues to decide. We must also include in our definition a maximal condi-
in its former use in stressed form. tion to distinguish Stage II from Stage III in which the former article
is a mere marker. There must be at least one construction in which
An interesting case in a gender language of a nascent article common nouns regularly appear in their non-articulated forms so
which is, so to speak, at a point between a zero stage demonstra- that all common nouns have two contrasting forms, one with and
tive and a Stage I definite article is BWAMU, a language which one without the article. This may involve minimal contrast as when
appears to be a genetic isolate within the VOLTAIC branch of the non-articulated form of the noun is used in generic sentences
NIGER-CONGO. In this language the original suffixes have been like 'I don't like meat ,' as against an articulated form in 'I don't
phonetically eroded to the point at which they no longer exhibit a like the meat, 1 (i.e. some particular specimen which has just been
clear indication of membership in a gender. Moreover, all vestiges mentioned or can be identified by non-verbal context). However,
of agreement have disappeared with the exception of two demonstra- in languages with Stage II articles the choice of articles is always
tive type elements, both of which precede the noun and indicate largely grammaticalized, being determined by the syntactic con-
previous mention. Unfortunately, our only major source for the struction and is thus redundant. For example, most of them have
language (Manessy 1960) is devoted to noun morphology and says generalized either the articulated or non-articulated as negative
little about syntax. Manessy calls one of the two elements a determ- object, and they do not express the semantic distinction which has
inative and the other a demonstrative pronoun. However, even the just been noted. This is a large step in the direction of a Stage III
first of these does not appear to be a definite article. He notes article. Languages with Stage II articles show varying degrees of
that its presence is ''facultative et relativement peu frequent dans advancement towards a situation in which the distinction between
~e discours" (Manessy 1960: 93). Moreover, he does not translate the two forms is completely redundant, being dependent on th e con-
it as an article. Yet it is only these two elements which indicate struction, e.g. negative objects are always articulated or always
the gender membership of the noun. The determinative is a mono- non-articulated the same for nominal predicates, etc. Wi th th e
syllabic element, basically similar to the obsolescent class affixes, .
a tt ainment '
of complete . . ti .on, the analogical tendency.
grammaticiza
We have, it would appear then, a case of renewal of gender by a . ersal. If the artic-
sets in for one or the other form to become un1V
demonstrative element which is not yet an article and whose next l
u.ated . h
form, which is the one which as usua lly spread to more
st age would presumably be comparable to that of a language like un marker. If not,
contexts predominates we have a Stage III no h d
GURMA which has a clear Stage I article. , , . th" is what has appene
the articulated form disappears. I believe is 3 h t
. f In t e pro O-
in Eastern BANTU languages without pre-pre ixes._ hi h 'n
3.4 Stage II of the definite article 1 St e II article w c 1
anguage the pre-prefix was probably a tag d • Stage Ill
' some move 1n
some languages remained in this stage, in . d d 1opments
St age II was in the earlier sections of this paper referred to a d .
n in some languages disappeare .
d Th e van e
es
eve o
f the same sub-
0
as the non-generic article. We shall see that this is not really an often took place in quite closely related languages
adequate designation. We may define Stage II as the stage in which group.
we have an article which includes, along with possibly other uses,
both defini te determinaticon an d non-efinite
.d., <fi ticulated form has become
specific uses. Speciiic, In languages with Stage II articles the ar c . al citation form
opposed
. here to gener· ·
1c, 1s th e use of such an article in contexts th, ·
e normal form of the noun. It is usua»
lly the 1 .ex1c
f such languages
in which a specifiac but id,.entified item is referred to that is and it
· heavily
·
ut uni predominates in text, Gammars o
s ; •
there is a pres iti, f '> ° r

· upposit.on o, reference. Thus, ENGLISH 'I am look-
ing for a hook' is ambiguous as between specific reference, i.e. also sociolinguistic
there is a certain ho 0k f h: h
which th . or w.ic I am looking, and a reading in ·There are in some instances, probablY ,uerts that the Tonga
ere is no such assumption I might mean any old book, for factors. Carter (1963) in an interesting article ,_ith and without
example, to prop up an unbalanced table leg. are ; ,. .: ·t, between noun: n-
quite conscious of the distinco n ,f the former as 'u
Pre-prefixes and consider the excessive us"! ,e variations in
Our definition is minimal In f t 1 dignify ge considera
11 · 1 d . • ac., anguages in Stage II gener- 11ed. 11 There are in this langua, tyl s .
a Y inc u e instances of non-referential use so that they correspond rear various s' .e
uency of these forms in texts 0
65
habitually list not the uses of the articulated form, but rather situa- existence of contrast between articulated and non a ti ul
- r c ated in
tions in which it is not used. As is evident from the earlier dis- common nouns as a part of the definition of a languag ·th S
--- . 'e w1 a Stage
cussion, no more than is the case with the ordinary definite article, II article. For some of the languages a minimal contrast is re-
or for that matter individual cases in case languages, is there a ported when the same noun is used both as a common and proper
single overall semantically based definition which will completely name. For example, in BEMBA amafupa or mafupa, depending on
delineate the respective uses of the articulated and non-articulated the usual rules for the use of the articulated and non-articulated
forms. As is usually done in the grammars, however, it is easier form, is a common noun meaning 'bones,' but as the name of a
to distinguish the relatively limited set of uses of the non-articulated river it takes the form Mafupa, without the pre-prefix. In some
forms and view the articulated forms as an unspecified remainder. languages an animal name, when used to designate a character in
a folktale, is in the unarticulated form, but when used as a common
As we may conclude from the earlier discussion, we cannot name, is articulated or non-articulated in accordance with the gen-
expect that all languages with Stage II articles will have the same eral rules for the use of the two forms in the language.
set of uses for the non-articulated forms. However, there is a
common core of functions which recur in languages of this type. Common nouns, when used in the vocative, commonly occur
without the article. In one language of Stage II, LOTUKO, both
Most of the functions of the non-articulated form can be placed articulated and non-articulated forms occur, but the latter is said
in one of two categories. Strangely enough, on a scale of degree to be more emphatic as well as more elegant.
of determination, these two categories are at exactly opposite ends.
In some instances the article does not occur, because the noun is In construction with demonstratives, there is some variability.
inherently determined (e.g. proper names) or because it is deter- The predominant tendency is to add the article to the noun redun-
mined by something else in the particular construction (e.g. a dantly, a tendency already visible in some instances in languages
demonstrative modifier). In these instances the noun which would with ordinary Stage I articles, e.g. ARABIC and CLASSICAL GREEK.
not usually have an article in Stage I does not acquire it in Stage II. Sometimes, when the article is not on the noun, it is added to the
It is because of these uses that the name 'non-generic article' is not demonstrative. There are some indications that the principle in-
really appropriate. At the other end of the determination spectrum volved here is that definiteness is a property of the noun phrase as
are the generic uses, as in negation and predication. These oppos- a whole. There are instances of word order varia1on · ti where, ' in
ing uses of the non-articulated form bear a certain resemblance to . th ther on a demon-
one order the marker is on the noun an d in e O '
the use of non-articulated forms in languages with both a definite AI 'th Stage II pre-
strative or other modifier. In TESO and MAAS • wi
and indefinite article. In such languages proper names do not usu- f' f 11 th noun However,
1x articles almost all noun modifiers o ow e · d
ally have an article, but the article may also be absent in generic th ' h' h · that case oes
ose form which precede take the article w ic rn h
uses, e.g. FRENCH je n'ai_pas d'eau 'I don't have water:' il est rd order s ow
not occur with the noun. Cases of alterna ti ve wo d
tail1egr 'he is a tailor.' The generic uses of the non-area.ea l . . · s marke once
c early that what is involved is that definiteness 1: A
forms in languages with Stage II articles, it may be noted, largely o 1 ·n these languages.
n Y and always initially in the noun phrase 1 , with its alternative
coincide with the totality of two sets of categories enumerated by AASAI example is il-kuti_tunana 'few people,
Moravcsik in her discussion of determination on the one hand as
!!-tunana_k ti 'people few.'
inherently definite and on the other, as generic (Moravcsik 1969:
especially p. 72). . the demonstrative is the
The interrogative which corresponds to st for identifi-
spe 'f . . . , It age as reque: ·
In stances of automatic definiteness will be considered first.
ifying interrogative 'which? :s us it employment in
cat· • . . · h d from 1.s
These include proper nam es, voca'ti ves and noun modified
• · 1on in ENGLISH is to be distinguisne ,E?' illustrates
by d .e- th . d you wan . l
e sense of 'what kind of?' 'Which wine to itis not usually
monstratives and personal possessive pronouns. Of these proper th 'f' tional sense
1
names ' both personal an d place, ·' _" Potential ambiguity. In the identifica' II languages·
are the group most consistently
companied by an article on the noun in Sta
used in unarticulated form. No exception was found to the rule that
in languages with Stage II . th demonstrative
ar ti c.es,
1 h article
tl.e . . not found with
is .
Th tion like e a. tly
proper names. This was the reason for the. specification of the e personal possessive construe d' fi •teness redun an
Con, E, ·k .e: .n1
s ruction, shows a tendency to mar
67
on the noun, and once more, this is a tendency which is already that either negative objects always take the article, or they never
visible occasionally in languages with class I articles, cf. ITALIAN do, The latter is more usual.
la mia casa '(the) my house.' Kinship terms, as inalienables, how-
ever, show far less affinity for the definite article, a point that can Two BANTU languages from very different areas, SOUTHERN
also be illustrated from ITALIAN which has mio padre and not il KIKONGO and LUGANDA, show the following variation. A nega-
mio padre. tive object takes the article when it precedes the verb but does not
when it follows. A closer consideration of these two cases shows
When, as often, there are two constructions, one for alienables the following. The usual order is for the object to follow the verb.
and one for inalienables, it will usually only be the latter in which When it precedes, it is always definite in meaning, initial in the
the possessive pronoun is directly affixed to the noun. Often, also, sentence, and referred to by a pronominal object on the verb. In
the kinship term can only occur with possessives so that, e.g. other words, it is topicalized and the definiteness of topicalization
'John's mother' is always John his-mother or the like. In these takes precedence in the linguistic expression. An example from
cases the article is, so to speak, blocked from attachment to the Tucker and Mulira's grammar of LUGANDA is the following: tog-
nouns. Whatever the factor involved, kinship terms rival proper gyawo bitanda 'don't take away the bedsteads' in which bitanda
names in their resistance to the acquisition of a Stage II article. does not have the pre-prefix, but e-bitanda tobiggyawo 'the bed-
steads, don't take them away I in which we find e-bitanda with the
We turn now to the instances in which a Stage II article is not pre-prefix, and in addition, the incorporated object pronoun bi-
used with nouns taken in a generic sense. These may be classified in the verb (Ashton et al. 1954: 33).
as falling roughly into four main types: 1) negation, 2) predication,
3) adverbial and locative uses, 4) generic verb objects and depen- A second major. type of construe ti on 1n
· which the articulated
dent genitives in compounds. These last two items can be shown . h • l predicates in sentences
form usually does not occur is wit! nom1na
to be basically similar. k , b t not where the sentence
of the type 'the man is an iron wor er, u 'h . th hie£,
is equational so that the predicate is definite, e.g. • c t ·
In regard to negation, two observations can be made. The first ly and usually o no
Predicative adjectives are treated ana.logous
is that non-articulated forms are confined to the objects of verbs
have the articulated form.
in negative sentences and to subjects in negative existential sen-
tences. It does not ever seem to occur with definite subjects in . 1 d r which may b e subsumed
negative sentences, e.g. with 'girl' in the sentence 'the girl did not The third major type is adverbiai unue: . ilar to expres-
l ± : These are s1m1
ocative and temporal constructions, h, id,' 'on foot,'
pick the flowers. 1 With indefinite subjects we are evidently dealing 1
sions i ENGLISH with out the article such a°."! ,, as adverts
with instances in which an existential sense is involved. The some- 'th rds whicl tou
what strange sentence 'a girl did not pick the flowers I is either a ome,' and 'at night,' as well as wo h . here as in
.. ' b 1 There 1s '
contrastive and really predicative that is equivalent to 'it was not and prepositions like 'behind' and a ove, maticalization
oth er instances,
· a strong ten d ency towards gram d by prepos1- •
a girl who picked the flowers, 1 or equivalent to the negative exis- whj ·+ All nouns governe :
tential 'no girl picked the flowers,' 'there was not a girl who picked 1ct takes the following form. 4 their meaning is
1 ti Ons are in the non-articulate d f or m even w enh the noun has a
the flowers. Such sentences are generally expressed in these two sp : . appears w en . h
latter ways in languages, in which case they fall under predication ecific, but the articulated form re <Ere, even if the
.
q ua..ifier
1 . dent geni 'v
and existential negation, in both of which cases non-articulated such as an adjective or depenC ;ometimes even
rn · h" rt occurs 5 • ·th
forms are common in languages with Stage II articles. caning is generic. A rule of this SO +/ANIAN in which, wI
With a Stage I article as exemplified by RU constituting the
a f repositions, fol-
The second observation is that verb objects in negative sentences ew minor exceptions, all these P s articulated nouns .
can also be divided into instances with generic and with definite ob- large majority which are not themselve t the postposed article·
l
.,"d by the geni '
tive, take the noun '
·thou
ee noun has qualif,,
·r rs
jects, but very few languages ia Stage II make a consistent distinc-
tion. The languages which I have examined which are closest to
making th is distinction consistently are ZULU and XHOSA, two closely
related BANTU languag es. El sewh ere, the rule is· generalized
· so
, "">
.,"articulated form is used, howev ' ,
we have f_grading& 'in garden'
e garden the large.'
" "
gradinuacea_mare
Joseph H. Greenberg GENDER MARKERS
68 69
Since body part nouns are frequently the source of adverbs and Numerals appear more frequently in indefinite constr ·ti .
, . . · 'uc ons like
their corresponding prepositions, we often see a contrast between 'five houses than in definite constructions like 'the E:1ve h Ouses.'
body parts as nouns which will in Stage II languages have the artic- However, even the · former construction is not generc. H tence,
ulated forms in most constructions, as against their adverbial and the noun should in Stage II have the articulated form w'th 1 numera.s, 1
prepositional uses in which they are non-articulated. A gramm ar and this is generally the case, but there are occasional deviations.
of A TESO states that "many prepositions are derived from nouns MENDE, a language without noun-classes with a suffixed Stage II
by dropping the noun prefix' (Hilders and Lawrence 1957: 66). article -i, distinguishes indefinite from definite numeral construc-
tions along the lines to be expected in a language with a Stage I
The last major group is that of generic noun objects and nominal article, e.g. maha feleng a 'chief two,' as against maha feleisia
compounds based on a genitive construction. Again, we may cite 'the two chiefs.' In the latter expression, as is normal in MENDE,
examples from languages like ENGLISH in which we have phrases it is the following modifier which takes the definite form. An even
like 'take care' and even occasional object-incorporation as in 'baby- more striking example is ARAMAIC. In modern WESTERN ARA-
sit.' An example of a Stage II language in which this occurs exten- MAIC, an advanced Stage II dialect, the only productive use of nouns
sively is KHASI. In GUNWINGGU which is really a Stage III lan- in the non-articulated form which has survived is with numerals.
guage, nouns in classes Ill and IV, the two thing-classes, may be As noted earlier, the development of the article through the stages
incorporated in the verb as objects, in which case they do not have enumerated here was much more rapid in EASTERN than in WEST-
their class-prefixes. ERN ARAMAIC dialects.
Compounds based on the genitive construction, or genitive ex- 3.5 The Stage Ill article
pressions in the border of lexicalization, show similar characteris-
tics. The genitive in such cases is taken generically, just as is It was indicated that during the second stage there is a decreasing
the noun object of the verb. These expressions can be distinguished set of environments in which there is direct contrast between the
from genitive expressions in which the regens is non-generic. articulated and non-articulated form. In general it is the articu-
These are often systematically distinguished. An instance in point lated form which spreads until it becomes the normal form of the
is TAMARI, a VOLTAIC language with older noun class suffixes noun. In the absence of significant contrast, there is an analogical
which has a prefixed Stage II article. From fa nafa 'cow' and li tendency for one of the forms usually the articulated, to spread
t0 ' h · ally the mass
yini 'horn,' we have li nayini 'une corne de vache,' 'a cow-horn,' all the remaining environments so that, sync romc ' h. h
as against fa nafa k wa 11 y1n1 : : ] a corne de la vache' or 'une corne f
0 common nouns now only have a single form, usua lly the one wuc
.
de la. vache.' The reason ta £ occurs 1n . cow's horn without either a ' hi fc Wh en this happens ,
1s istorically the reflex of the articulated form. •
prefix or suffix is that ·1n th e f.1rst round of gender suffixation, the . • pure mar er
we are in Stage III in which the former artic 1 e is a ..
same phenomenon occ urre d so th at tlh e first . whi+] . +ith definiteness or
. member of a compound 1c no longer has any synchronic connection h t arbitrary.
did. not acquire th, e su ffix, as can be seen in VOLTAIC languages spec<f : {I dIII is somewa
cifi city. The line between Stages I] an ±. fe .s of the
which are suffixing o n.y.1 I n f act, b ecause this is productive con- Bow . . t the alternating torm
restricted and non-productive mus h than Stage
trast, one can say that languages like TAMARI were still in the noun b . dt O Stage III rat er
e for a language to be ass1gne h blems of
final stages of a Stage II su. fffixed article. when they began to develop II? W · h with t e pro
· e will not so much be concerned ere mewhat rough
a new prefixed article. assi . t ms of our so
gnment of borderline languages in er of the process,
and . . th first stages
ready definition, as in considering 1e i1ch are character-
A special case here is . th a t o fd;liminutive
· . and th ' . vivals w
b ase.,
.
d in many lang
constructions which are • e synchronically non-functiona 1 sur
. thi . . {uages, on a compound of the form 'child of x. 1 1sti
S c of Stage Ill languages.
d'mce. . s is highl y p r o d uciti ve and any noun in principle can have a
d, d, those
1 t O be consi tere '
. n;~utifve, t~e unarticulated form of the noun sometimes survives • There are two distinct classes of cases • e to the process
in s ormation even wh th d in wh;, ·h:h gave r1s
article at a late/ . . en e ependent genitive acquires the ch the original demonstratives wm1c' f gender markers
point in the development of Stage II languages. Were 1 lt . s a se O • 1
c..assifying so that the final resu 1 that the fina
on th, tso, so f
One further constructi +h e noun, and those in which this was no' er, a mere sign O
artic l t d on s ould be mentioned in which non- esult : 1 sty' mare. »
norn· is rather the existence of an 'emP' o uns·
u a e nouns sometimes appear, namely with numerals.
inality on the large majority of common
Where the original demonstrative was a classifying one, the work here is that place demonstratives in these languages often
essential outcome is, of course, that the nouns are now classified have two elements, one meaning 'place' and the other a deictic
by markers, either pristine or renewing an older system (i.e. double element found also with the regular gender demonstratives. For
affixation). There are, however, always tell-tale signs of the pro- example, in SWAHILI we have ha-pa 'here, this place,' parallel
cess by which the markers have come into existence. The first of to ha-wa 'these,' referring to nouns in class 2 (plural personal
these is that proper names, kinship terms, and frequently, bor- class) which have the prefix ~-· Syntactically it is plausible, as
rowed nouns which entered the language at a point in the develop- pointed out in Giv6n 1976, that such agreement should start with
ment in which there was no longer a synchronically relevant sentences in which place is topicalized. The final chief area of
alternation between articulated and non-articulated forms, all end survival of the non-articulated form is, as might be expected, in
up without having gender markers. Involved as they normally are fossilized nominal compounds and incorporated verb objects. We
in a system with gender agreement, they are assigned gender on .
might also add here that sometimes 1 e person, ' 'thi«g
wor d s 1k n, , I
a semantic basis. etc. survive in pronominal uses without the article (cf. FRENCH
personne, pas, point ; ITALIAN oSa 'what?').
This is the presumed explanation for the existence of the class
La in BANTU languages, a phenomenon which is in fact found Instances of class 3 noun markers where the original demon-
throughout NIGER-CONGO with the exception, of course, of MANDE strative article was not classifying are especially intriguing,
which has no noun classes. Class l.a typically contains personal because we do not have here the guide of gender-agreement . All
proper names and kinship terms, lacks the class affix, and has the we have is that virtually every noun in the language has a partic-
agreements of class l, the singular personal class which in BANTU ular marker which. . were, b rec ome a sign of nominality
has, as it
has the prefix "mu-· . . t '
as such. Such cases will illustra e in a
. .
graphic manner how the
.
tends the reach of in -
study of generalized diachronic process ex g, .}
The second consequence of this process has to do with locative d fr the interpretation o sue
ternal reconstruction. Our evi ence O • h the
and adverbial expressions in general. It will be recalled that body a marker as a former article rests on the fact that it slows
e there is gender classifica-
part terms and other nouns, e.g. 'earth' (='down') in Stage II lan- same characteristic survivals as Wher
guages, frequently contrast non-articulated forms in adverbial and tion.
prepositional uses with articulated forms in nominal uses. Such
adverbs and prepositions tend to survive because of the absence of w·th are exceptions HAUSA nouns
A case in point is HAUSA. 'i r A text count
productive alternations. Prepositions and adverbs, being a restricted d · · 1 · lmost always 1 ong.
en in a vowel. This vowe, 1s a . 1ike 99% final long vowels,
and fairly closed set, also often survive in the face of lexical re- of nouns would probably show something 1ng into two main
placements of the corresponding common noun. An examination of However, forms with s h or t vowe ls do exist, a ind adverbial expres-
er names an
adverbs and prepositions will generally uncover relict forms with- and apparently unrelated types, pro p : ith, mini mal con-
out class markers. For example, OVAMBO, a BANTU language : :h, re to begin w: ,
s1ons. To take the former, t ere a ' and proper name,
in Stage III, in which nouns only occur in pre-prefixed form, has ommon noun .,.
d I trasts between the same wor d as a c . but Gambo is a
• I
PO 'under' to be analyzed as po- 'on' and -si 'earth,' the latter ' .,. .,.
Th us, gamboo designates a c 1 h'ld born after twins,
h'ld There is a town
without the pre-prefix. The root ?-si is general BANTU root for proper name which may be given · to such a ca+i1 da4a1{{· e
'horses, ,
'earth » I b ·u t rn
· OVA MBO and related -
languages it has been replaced p; .z ze ; l 1l ength wi "- ]
-2awaaki contrasting in finai vowe! have short vowels
by edu. th e source of its . M roper names
name. any pi e: hich are cities. I In
. .
'th k, Zaariya, wr
without such a contrast, e.g. ban0o ·'< .e borrowed words,
·ad
. There are also consequences for the nominal member of loca- some instances the reason may be that they ar
(from FUL), ?Alhaji
,2
-,. '··' (f tr
om
tional and other pre pos1'ti ona 1 constructions.
. It was seen that the for example the personal names Bell ·tut another short vowel
noun governed by a · ti d , consti e e
. prepos1 on oes not usually take a Stage II ARABIC). Borrowed common nouns .A from NUPE and siisl
article. A characteristic result in Stage III is the reinterpretation . z +l,·' borrowec !I
category, e.g. ?angulu 'vu ture
of the prepo 'ti ·t
Sl on as i .self a gender marker since the noun has 'six pence' from ENGLISH,
none and the d l ' a whole series
' . eve opment of agreements with it parallel to that
ions there are ·th
of the earlier gender l
nd
. c asses, Hence, the rise of so-called place With regard to adverbial express h' h
have special forms Wl
ge ers as in BANTU languages, A further facilitating factor at of terms, many of them body parts w ic
73
short vowels when used with the preposition ?a. These forms are of Abraham's dictionary shows that many other compounds with the
called locatives in the grammar. They also exhibit tonal and other non-suffixed form as second member are reflecting an older direct
differences. Thus, a6 is 'eye,' but 2a 2fa6 'in the eye;' baa16 construction without a class marker of the re gens. There are also
'mouth,' but ?a baka 'in the mouth.' Examples of words that. adverbial uses of words like 'ground' for 'down. 1
not body parts are gut44 'fire,' but ?a_g t 'in the fire;' k's
'country,' but 2ak'as or just k'asa 'on the ground, below.' There Where the article was not a gender classifier, one additional
are also time expressions with the preposition~ e.g. daree phenomenon may be noted. Since in this case virtually all nouns
'night,' but da dare 'at night. I In addition words like jiya 'yester- receive the same mark, it becomes a sign of nominality as such
day,' (2a)a± 'last year,' yanza 'now,' etc. have short vowels and is used to derive verbal nouns. The -s of PLATEAU PENU-
without contrasting long vowel forms. Those numerals which end TIAN functions this way.
in vowels have them short, e.g. biya 'two, 1 except units, which,
as generally in languages, are treated as nouns, e.g. dar{[ 'hun- The entire account given here has been on the basis of the cate-
dred.' --- gory of determination. This seems natural enough, because we
are dealing with developments deriving from what is a definite
None of the adverbial expressions can have modifiers without article in its first stage. There is, however, another way of look-
being replaced by long vowel forms. At present these forms are ing at the process which at first blush seems very different, but
more and more being replaced by the corresponding long vowel is in fact ultimately related to the notion of determination, namely
forms as alternatives even without modifiers. 4 case.
In light of the above account, we can explain the long vowel forms If we consider the Stage II article which is the decisive turning
as the resultant of former article which has reached Stage III, while ·
point, we can roughly state the ma tt er as fo 11 o ws • The character- .
the short vowel forms are the original forms still surviving in some istic constructions without Stage II articles include the vocative
of their most typical uses. Note that there is no possibility of a and adverbial uses corresponding . t o th e 1 oca t·ve
i and instrumental .
unified account of HAUSA short vowel forms on a purely synchronic 'tive and accusative
of languages which have these cases. Th e gem d
base. HAUSA does have sex gender, but it appears that the demon- a re the cases in which. real contras t bi e tw ee ·n the articulated an ti
strative-article which is the presumed source of the long vowel . 1 1 th latter where nega ve
non-articulated form occurs, particu ar Y e . th ti-
forms, just as with ARAMAIC -a, did not vary for gender or num- and generic, sometimes incorporate oDj ecs,
d bi ·t are without e ar
llence
ber. b' t case is par exce '
cle which occurs in other uses, T h e suoyec . '. l ·dica-
. 1 and adjectiva. pre
the case which takes the article, Nomina b' t. articulated,
A somewhat similar case to that of HAUSA, namely TIV, a lan- ti on has to be considered separa t ely.1 Here the suj ec 1s
th h b.object
guage of the BENUE-CONGO branch of the NIGER-CONGO family, as wit t e ver -
but the predicate often is not, However, anent prop-
may be m en ti one d m
· passing.
· The old prefix system has here been
renewed by suffixes which, however, only appear in those classes
there is the possibility of contrast roug Y
.
hl between perm
t (non-articula
. ea)
1 e •
° Proper
erties (articulated) and non-permanen , d on determina-
which had consonantal affixes, Their presence in the other classes account ase
names do not enter here. For th em an
can be detected by tonal changes which parallel those of the conso- tion is primary.
nantal class. The earlier non-suffixed form of the noun called the
ted to definiteness•
Prepositional by Abraham (1940), survives after prepositions, e.g· d . clear 1 y re la
mger@m 'water ' s ng ': th The case hierarchy just sketc h e 15 fi ·t In some languages
---- {er 'in e water.' This form may not be b.
Suvjects, .
as favorite topics, ten °
d t be de n1 e.
used with a preposition if the noun has modifiers. It also survives
in diminu tives and augm tat ; hi
th entav es, whi ch are old compounds with . ti ve, the latter
e noun as second membe rs, in · mostly identical
• form. A perusa 1 5 The SLAVIC short and long form O f theti' adjecbelong here. The
a.a=.a.ca,a+Mesaa
. tle
in
.
h existi
. 1it na vowel quantity is very inadequate
ng aerature. I am indebted to W. Leben and Mohammed
. ·
containing a suffixed demonstrative
or re 1 av e,
become like a Stage
long form starts out as definite and tends tlo . RUSSIAN, it only
a.·ti, f example 1n
II
T airu for checki g ._ r.cle plus adjective, until, as or ly in certain uses,

1n some cases of final vowel quantity in HAUSA, s erves as a predicate adjective,
. : a .nd here on
they must be so. The accusative where the contrast is clearest 4. The Spread of Grammatical Agreement
is precisely the case which in some languages, eg. TURKISH,
PERSIAN, is the only one which formally distinguishes definite- From the account in foregoing sections we see that the demon-
ness by a marker. strative, as the normal source of Stage I definite article, plays the
role of initiator for the whole process described here. It also plays
All this bears on the origin of case markers. In INDO-EURO- a further role in that it constantly generates concordial phenomena,
PEAN it is striking that the two cases which can be reconstructed sometimes producing gender agreement where the demonstrative
with zero inflection are the vocative and the locative. The latter is classifying, and sometimes not. If we look at the impressive
has, in the singular, two variants: zero and -i. In the plural tables of concord series often found in grammars of BANTU lan-
it has long been noted that -si (with a variant -su) is really -s guages, we can see that for every one of them, there is abundant
'plural 1 + -i. In fact a form with just -s i s also found. In SAN- synchronic and diachronic evidence of the role of the demonstrative.
SKRIT the plural oblique forms for instrumental, dative, ablative A detailed treatment of this process is not within the scope of this
and locative have sandhi forms like that at word boundary, thus paper. We may note, however, a few principal considerations.
being in effect added to the pure base form. Note that the locative
1
and other oblique cases, being marked forms, do not usually re- The term used up to now has been simply 1 demonstrative, but
main in the zero form, but acquire prepositions or postpositions this requires some further specification. If we consider once more
as with HAUSA ?a, sa
TIV and, presumably, INDO-EUROPEAN our initial example of the GURMA group of VOLTAIC, we note that
i. This view also strengthens the case for those who equate the the new prefixes which arise match synchronically not the demon-
nominative singular -s with the demonstrative ?so. Accusatives, stratives of these languages, but the verb subject and object pronouns.
on the other hand, often derive from old prepositions marking an This does not exclude the diachronic possibility that they come from
indirect object. The case hierarchy just described is obviously earlier demonstratives which, on the one hand, developed into pro-
related to that discussed in much recent work in regard to sub- nouns and on the other, into articles. The present demonstrati ves
jectivization and topicalization, including relational grammar. also would have these as source with the addition of new deictic
elements. If we look at present day FRENCH, the article matches
.
h object . h from the primarily
The entire foregoing account of how noun classes acquire mark- the pronoun. Historically, botl come
ers is subject to two important reservations. One is that some- demonstrative ille which also had anaphoric uses in LATIN,
times only a few nouns within a class come to have markers, and
th forms is a shifting
this by a mechanism which is essentially different from that just In fact the synchronic boundary O f ese d d nstra-
t . ronoun an emo
described. In almost all such cases we have ''motion,'' that is, one. In most languages the demonstra:ive p .
es the third person
a minimal contrast of the same stem in different classes. This is ti ve adjective are identical. In many languag fcle is
t . e and often an ar l
common in NOR TH CAUCASIAN. The nouns here do not in general pronoun is identical with a demons t ra iv ' h t th article
have markers, but there are instances like AVAR v-as 'boy,' j -as d. . · t excluded t a e
1uentical with one or the other. It is no ti but come
1
girl.
1
I believe that these arise from substantivization of adjectives sh 1 . . • t 1 demonstra ve,
ouid arise from what is ultimately a he NEO-ARAMAIC
which do have concord markers. rn ore directly
· •
from a third persona.l p ronoun. In tfed e article has
of T ur-'Abdin, it was noted earli ·1 er that a new prenx :. while Siegel,
Another and more fundamental reservation is the following. While a · ' th d monstrative,
T1sen. N ldeke derived it from e e1 ctly for a pre-
the development of markers from the article is no doubt the usual in ·obably corre
a more recent treatment, argues pr d di onstrative are very
.
process, there is what at the moment is , to my knowledge , the fix e d pronoun. In any event the prono un an emt ·ative in the presen t
~mque case of the DALY RIVER language in Australia described si- <], t the demonsr ti
1milar. Hence, when talking about , Between demonstrav e
» : tline e
in Sec. 2. Here, a new system has been superimposed on an older Connection, I will not be drawing a str1c
Australian-wide formation of the usual type. In this new system, and third person pronoun.
as it appears, the source of the marker is a superordinate noun and . d heuristic fashion
in which concord subsequently develops in the adjective to produce We may distinguish in somewhat rough a~ tis that of a noun
a gender system in the sense defined here. three types of concordial phenomena. The F°' .g. adjective,
• phrase, · ·
with its immediate modifiers in the noun
with its immediate modifiers in the noun phrase, e.g. adjective, the basic structure is similar to that of SWAHILI:
demonstratives and numeral. The second is predicate agreement,
wa-kwe wa-zuri wa m-ti
that of a subject noun with predicate adjective or demonstrative.
"women beautiful those -of village"
The third is anaphoric use. We may distinguish under these two
types. Wh ere a deleted noun has no modifier, we have a pronom- In fact, in Tagliavini 's gramm ar of RUMANIAN, he notes the basic resem-
inal substitute. Wh ere the noun has modifiers, there may be a blance in such instances to langua ges like SWAHILI (Tagliavini 1923: 274).
substitute, e.g. 'one' in ENGLISH 'the good one.' Languages with
noun classes in such cases, however, usually attach the class Demonstratives, then, are constantly producing concord phenom-
marker to the noun modifier. This is probably one of the sources ena. However, specific demonstratives, as they become bleached
of modifier agreement, as soon as it is redundantly applied, when of deixis by anaphoric uses, are constantly being replaced by new
the noun is present also. Another kind of redundant anaphora is demonstratives usually formed from the older ones by the addition
intraclausal, but not within the noun phrase as when the verb com- of new deictic elements, by reduplication, etc. These in turn lose
pulsorily contains a pronominal mark of the subject or other noun their deictic force to be replaced by others. In relation to gender
phrase head. The origin here is no doubt in topicalized sentences. there are several possible courses of development. If the demon-
stratives themselves have gender classification, they will give rise
Much of this may be illustrated by the development in RUMANIAN to further phenomena of gender agreement. If they pass through
of two new so-called improper articles which now exist alongside the phases described earlier in the paper, they will end up in th e.
of the earlier suffixed definite articles. The first of these, eel, is noun, all the while producing concord from diachronically_ successive
originally a demonstrative from LA TIN ecce illum 'behold that,' layers of demonstratives. The exis . t ence o f such layers is shown
the last element of which by itself gave the RUMANIAN suffixed . h h . g concords for the same
in languages like BANTU in whic. t e varym . f th
article. Alongside of om-ul bun 'man-the good." one can have om-ul ff t fc s The persistence o1 1e
gender exhibit a number of di eren orm · t f
eel bun "man-the that good." With deletion of the noun, we get eel t t O gether in the same se o
genders provides the cement whic,h pus '
bun 'the good' = LA TIN bonum, eel'.' drepµ 'the just' = LA TIN justi, concords, forms which differ phonetically.
and with relative-like uses om-ul cel cu boi 'the man with cattle. 1
The second article is al in the masculine singular. It derives . d h t of the other concordial
The development of the article an t a If the former
from the preposition a and same demonstrative ille, found in the phenomena may not take place at the same P%,,y gender is shown
suffixed article, and in the first improper article al. Its basic use proceeds more rapidly, we have languages m w \ r·c pronoun also,
is possessive and can be seen from such examples s the following: with the articulated noun and usually with the a"%$'_..,, Examp l es
b n noun mod1ne. "·
ut not with non-articulated nouns or O h BRASS dialect, with
pom-ul bun al vecinu-lui of such languages include IJO, particularly t e . e feminine and
"apple-the good that-of neighbor-the" . ti, ish ma scui1n 1 '
suffixed Stage I articles which d listingu , ·tic in pronominal
e distinc .on
neuter in the singular and with t h e sam h ase TUNICA, a
pom-ul meu si al vecinu-lui t · the noun P r •
reference, but without agreemen 1n d . definite nouns
"apple-the my and that-of neighbor-the" G DLF AMERIND language, only shows h gen er in
,d • the verb, but no
: ·orporate 1n
and also has pronominal elements inc
ai nostri sosira
agreement in the noun phrase.
"the our (people) have:arrived'
·t will still produce
. fying one, 1 d n-
If the demonstrative is not a classi d. 1 if the same .emo
These parallel BANTU examples in that eel goes with the noun mod- Ph enomena which are on the sur.a f ce concor ia ' 11 h b 0 y the-
ifier, agreeing with the noun in gender and number, and al, as with 1 d hat-tov t e- d"
strative is at work, as in HEBREW hay-Y, aa the "concor
the BANTU genetive, agrees with the regens which is followed by '
g Ood.'' Where, however, . clas si i • ds we
there is I0 ,different perio '>
the rectum. In such a phrase as el, ,deictics o. fc s as in
ements involve varying forms O : ng some orm
will h, .: <larities amo thers e.·
femei-le ce-le frumoase a-le satu-lui ave at best suspicious sim» t whereas of e: ·» ·tion
ENg Eh,
NW'+LISH 'this ' 'that,' 'the,' 'they, ' 'them,
. Hibit no over ·tconnec ·
"woman-the those beautiful those-of village-the" 'sh1 ' o ·gin, will ex
e, though of demonstrative on i '
Joseph H. Greenberg GENDER MARKERS 79
78
Incidentally, we see from this why it is the noun par excellence human, male and female, tend to occur in demonstratives, third
which gives rise to classificational systems of syntactic relevance, person pronouns and interrogatives as a guide to identification.
It is not so much that the noun designates persisting entities as
against actions or temporary states of persistent entities. It is It is possible that gender systems in their initial stages are of
that nouns are continuing discourse subjects and are therefore in this type, The way in which gender arises needs not be the same as
constant need of referential devices of identification. As soon that by which the system can expand by the development of new
as we wish to talk about an action as such, we nominalize it; genders. A mechanism by which place gender might develop was
classification is a help in narrowing the range of possible identif- described in the course of this paper, namely, through tli.e rein-
ication. terpretation of a preposition or postposition as a class marker.
Such prepositions frequently derive from locational nouns. Minor
The theory of the special role of the demonstrative in the devel- genders such as diminutives, and less often augmentatives, probably
opment of agreement advanced here is not intended necessarily to arise by a similar mechanism involving an element meaning 'small'
encompass concord in case or number. It is plausible in regard or 'large. 1 Three parallel instances may be cited in which it ap-
to number and is in fact almost always an incidental by-product of pears that a neuter has just developed or is in the process of devel-
gender agreement where that occurs, since gender systems, except oping in a system with masculine and feminine. Boas, in his gramm ar
in a few instances in Australia, always intersect with the category of CHINOOK (1911: 602), notes that the neuter in CHINOOK can be
of number. used with any masculine or feminine stem to give an indefinite
meaning. He notes also that the neuter is close to being what in
5. The Problem of Origins the initial discussion was called a minor gender, He states that
. l • th neuter gender is so
''the number of words which appear only in e .
It is the thesis of this paper that, given a classifying demon- small that we may almost suspect tha h t th e neu ter was recently in -
11
strative and the constant tendency of demonstratives to generate definite and used to indicate both singular and plural. In ~-H: SI,
agreement, the result will be a gender system. Further, if the alongside of a masculine and femini ne, there is a neuter
.: diminutive. tis my
""""S
demonstrative goes through the stages outlined, the result will be does not distinguish number. Its meaning 1s ·1
t here which are exc u-
a marker on the noun as well as on the agreeing element. Given overall impression that there are no s ems th culine
. OTTENTOT also wi mas
the existence of classifying demonstratives, the whole process will sively neuter. A third case is H • ·d, 1ik
d r used for wor s l e
unfold with something close to inevitability. However, it has not an d feminine and a neuter, common gen e th KHASI, there
' hi d , • l , As w1 ,
been explained how classifying demonstratives arise in the first c ild' in distinction from 'boy' an gir · KHASI has
1 · ely neuter,
place. This remains for further investigation. There is one pos- appear to be no stems which are exc.us1v f euter singular,
: 'ENTOT has -i tor n
sible origin which is just that. That is, it is not asserted that in -!. as the neuter marker and HOTT .- A remarkable
any actually reconstructible case it occurred this way, nor that th . 1 f diminutive, ,
e comm on sound symbolic vowel tor : ;hon in Moi.' 3f's
15 5
there are no other mechanisms. parallel between CHINOOK and HOTTENTOT Th , f the genu s
s OT II e 1 O
tatement regarding NAMA HOTTENT ,' h'1 -h had nothing to
. Th'1s poss1'bl e source is
· numeral classifier systems. As was ~mmune was originally an indefinite article; ~resent time, It
pointed out in an earlier study (Greenberg 1972), the first construc- do with gender and it is still used this way at e line or feminine
tion to which the numeral classifier spreads is to the demonstrative, ca b h ther of mascu )
n e added to every substantive w e . 1ass' (1909: 48.
as has happened in CHINESE, TH AI and other instances. This gend. .f th ppropriate t
er to signify an instance o e a . f' •teness, I canno
~ives us Just what we need, a classifying demonstrative, In fact, WA, . f mm indefi n1 i
Y comm on gender should arise tro , ri te meaning o' l

in KIRIWINA in the Trobriand Islands, it has spread from the de- really say. Meinhof believes that "thi s inde""",tion was paid
monstrative to some adjectives; in standard THAI it occurs with ea,1 , 1;ttle or no =
s1.y led to the further result that .1 e came into ex1s-
adjectives in some c ons t rue ti ons so that true agreement phenomena to . . s commune
. sex in its use so that its use as genU''-
b_egrn to appear. In northern THA I languages the classifier functions tence. '
like an article and 1n JACALTE CMAYA as ananaphoricpronoun, to treat the
ons of space,
It may, however, not be necessary to go so far afield. The funda- It has not been possible here, for reash and the large and
mental bases of contrast, animate and inanimate, human and non- rel ti ted ere
a.onship between the views presen
80 Joseph H. Greenb erg GENDER MARKERS 81
frequently murky literature on the origin of gender. Suffice it to de la Grasserie, Raoul. 1896. De l'article. Memoires de la
point out that the notion of a special role for the demonstrative in Socit e Linguistique de Paris 9. 285-322, 281-394.
the development of gender is far from novel, being found after a
fashion even as early as Bopp, who found the origin of inflectional Dixon, Robert W. 1972. The Dyirbal language of North Queensland.
elements in pronoun. In most speculation about gender, it has Cambridge University Press.
been assumed that classification starts with the noun, and the prob-
lem is to discover how it spreads. If the thesis is valid that the
usual course of events is that it starts with the demonstrative and
only sometimes ends up in the noun, the statement of the problem
Fodor, Istvan. 1959.
8.1-41, 186-214.
The origin of gramm atical gender. Lingua
v
Franklin, Karl J. 1962. Tolai language course. Division of Ex-
should be reversed. There are some examples of earlier specula- tension Services, Territory of Papua and New Guinea.
tions which point in this direction, though none of them seem to be
very clear. The closest approach encountered was in the writings Givon, Talmy. 1976. Topic, pronoun and grammatical agreement.
of that frequently acute but generally neglected thinker Raoul de la Subject and topic, ed. by Charles Li, 149-188. New York, San
Grasserie. After I had arrived at the notion that the NAHUATL Francisco, London: Academic Press.
absolutive was probably an old determiner, I found in his essay on
the article the following statement: "In the Nahuatl language, the Greenb erg, J.H. 1972. Numeral classifiers and substantival
substantive ending tli is not a derivational suffix, but an indication numbers· Problems in the genesis of a linguistic type. Work-
of determination whose function has disappeared' (1896: 293). . · . 1 9 • 1 - 40 · Stanford University,
ing Papers on Language Universa..s
Hilders, J. and J. Lawrence. 1957. An introduction to the Ateso

language. Kampala.
BIBLIOGRAPHY
I nouns and associated
Holmer, Nils M. 1949. Goajiro (Arawak ) I:
Abraham, R.C. 1940. The principles of Tiv. London: Crown morphemes. IJAL 15.110-120.
Agents for the Colonies.
Studies in
Lov e, J • R..B • n d. • An outline of Worora gramm ar·
24
· A p Elkin 112-1 · Sydney:
Ashton, E.O. with E.M.K. Mulira, E.G.M. Ndawula and A.N. Australian linguistics, ed. by · • '
Tucker. 1954. A Luganda gramm ar. London, New York, Science House.
Toronto: Longmans, Green and Co.
. du nom en Bwamu (bobo-
Manessy, George. 1960. La morphologie ,
Boas, Franz. 1911. Chinook. Handbook of American Indian Lan- oul), dialecte de Bondoukuy. Universite Dakar·
guages, Part I. 561-677. Bureau of American Ethnology,
, . t suffixes dans les
Washington, D.C. 1965. Les substantifs a prefixes ' ,es 4. 170-181.
langues voltaiques. Journal of African Langua
Carter, Hazel. 1963. Coding, style and the initial vowel in north
Sprache. Berlin:
Rhodesian Tonga: a psycholinguistic study. African Language Meinh of, Carl. 1909. Lehrbuch der Nama
Stuclie s 4.1-42.
Georg Reiner.
. WPLU 1. 64-98.
Celedon, R. 1878. Gramatica, catecismo i vocabulario de la
Moravcsik, Edith A. 1969. Determination·
lengua goajira, ed. by E. Uricoechea. Paris. h An-
Mengen -Sprache·
Muller, Hermann. 1907. Gramm atik der
Chantoux, A ., A. Gontier, and A. Prost. 1968. Gramm aire
thropos 2. 80-99, 241-257. hnis
gourmantche. Dakar: Institut Fondamentale d'Afrique Noire. Wc:,rterverze1c
duge un d 0
6For historical reviews of the literature on gender, see partic- Peekel, G.P. 1929. Grammatische Grw ,,renen-Sprachen ?"·
· ft fur Einge
ularly Royen 1919 and Fodor 1959. der Label Sprache. Zeitsc.hr1
10-33.
82 Joseph H. Greenberg
Planert, W. 1905. Uber die Sprachen der Hottentoten und Bushm::in-

ner. Mittheilung des Seminars fur orientalische Sprachen. Abt. The Nature of Future Tenses
3, Berlin.
Ro.yen, Gerlach. 1919. Die nominalen Klassifikations-System in RUSSELL ULTAN

den Sprachen der Ertle. Medling bei Wien.
Sapir, J. David. 1965. A grammar of Diola-Fogny. Cambridge

University Press.
Siegel, Adolf. 1968. Laut- und Formenlehre des neuaramdischen

Dialekts des Tur Abdin. Hildesheim.
ABSTRACT
Tagliavini, Carlo. 1923. Grammatica della lingua Rumena.
Heidelberg: Julius Groos. The general observation that future tenses often differ from
other tenses more than present or past tenses do from one another
Tryon, D.T. 1970. An introduction to Maranangku, Northern was explored from two standpoints: structurally in terms of marked-
Australia. Australian National University. nes s and semantically in terms of atemporal functions of future
tenses categories associated with the latter, their etymologies,
_ _ _ _ • 1974. Daly River languages, Australia. Department and atemporal categories that also may be used to mark future time.
of Linguistics, Research School of Pacific Studies, Australian Markedness features investigated included: relative boundness of
National University. forms, temporal gradation, obligatoriness of occurrence, and neu-
tralization in several environments (subjunctive, gapping, negative,
Welmers, William E. 1973. African language structures. Berkeley subordinate temporal and sequential clauses, participles,and others).
and Los Angeles: University of California Press. Markedness features were also examined for correlations with a
propose db asic · t ense typology·· prospective (future may be. neutral-
.
·izecd wi"th presentt) vs. retrospective (past may be neutralized with
Wilson, W.A.A. 196l. Temne: an outline of the Temne language.
London: School of Oriental and African Studies. presen t) • I n genera,, 1 l·t was demonstrated that future tenses are
more marked than either past or present. Semantically, futu re
·
t enses s h ow a grea t er tendency to evolve from and develop into.
modal categories representative of varying degrees of uncertainty
. h is
w h ic · in· conor
f ·mity with the inherent uncertainty of ·futurity.
While future tenses often evolve from spatial categories, they
apparently do not develop into the latter.
J.
Reprinted from Working Papers on Language Universals 8,

August 1972, 55-100.

How Does A Language Acquire Gender Markers?

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

How Does A Language Acquire Gender Markers?

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

How Does A Language Acquire Gender Markers?

Uploaded by

Copyright:

Available Formats

Universals of Human Language

Edited by Joseph H. Greenberg

Stanford University Press, Stanford, California

Stanford University Press

The bulk of this paper is devoted to outlining a process by which,

By a noun gender system will be meant a system in which the

h . ch started in prehistoric times at least 3000 years

strative a development of wac h . h there are numerous s stage zero.

there is a pres iti, f '> ° r

T airu for checki g ._ r.cle plus adjective, until, as or ly in certain uses,

Y comm on gender should arise tro , ri te meaning o' l

Hilders, J. and J. Lawrence. 1957. An introduction to the Ateso

Planert, W. 1905. Uber die Sprachen der Hottentoten und Bushm::in-

Ro.yen, Gerlach. 1919. Die nominalen Klassifikations-System in RUSSELL ULTAN

Sapir, J. David. 1965. A grammar of Diola-Fogny. Cambridge

Siegel, Adolf. 1968. Laut- und Formenlehre des neuaramdischen

Reprinted from Working Papers on Language Universals 8,

You might also like