LINGUIST List 9.1726

Sun Dec 6 1998

Sum: Animacy and Definiteness II

Editor for this issue: Andrew Carnie <>


  1. Andrew Carnie, Animacy II

Message 1: Animacy II

Date: Sat, 5 Dec 1998 22:47:02 -0700 (MST)
From: Andrew Carnie <carnieU.Arizona.EDU>
Subject: Animacy II

Dear Colleagues,

A few weeks ago I posted a query ( LINGUIST 9.1462 ) about the link between 
definiteness and animacy, and received many interesting and informative 
messages. After posting a summary ( LINGUIST 9.1653 ), I received even
more interesting an informative messages. Here they are:

From: Sean Witty <>

Not sure if this is what you're looking for, but Korean expresses overt 
definition in Subject position and uses animation in the assignment of 

Witty's general rule of Gender assignment (all languages):

1. All nouns have gender of one sort or another, irregardless of 
biological sex.

2. Nouns referring to humans, or non-human nouns (including French 
occupational nouns) with definite Sex, are categorized +Animate, and 
receive Initial Masculine Gender.

3. Other nouns receive are categorized -Animate, and receive Initial 
Neuter Gender.

4. Gender by association. Nouns receiving Initial Neuter Gender may be 
reclassified as either Masculine or Feminine (or whatever, in the case 
of some African languages), depending on declension. Thus, the Latin 
'tabula' is feminine because it follows the first declension, in which 
all +Animate nouns are also Feminine (to include, curiously, four that 
typically refer to males).

5. In some languages, i.e., Anglo-Saxon, non-adult nouns may be 
assigned Neuter, or some other gender, to designate them as being 
+Animate, -Adult. In German, for example, indefinite children are 
assigned Neuter Gender, while only male children retain Masculine Gender 
(I'm not sure why, but I think it is for non-linguistic reasons, unless 
it is to designate that the child is male).

6. Among +Adult nouns, female nouns are reassigned Feminine Gender, 
while all other nouns retain the Initial Masculine. 

7. At all levels of distinction beyond the first, there may exist other 
gender categories to classify these distinctions. For each of these 
additional classifications, there may also be another variant applied to 
the Initial Neuter nouns. Thus, it is possible for a language to have 
in excess of 17 gender classifications.

Given these principles, which are by no means complete, a Korean adult 
noun will receive Initial Masculine Gender, unless the noun is of 
definite female sex. Thus, a word like 'hakseng', student, carries 
Masculine Gender when indefinite, and Feminine Gender when known to be 
female (possibly 'yohakseng', but this is rare usage, pronominal 
inflection is more likely; 'kunun', he and that person; 'kunyonun', 

Irrespective of this Definite/Indefinite relationship, a definite noun 
takes either the suffix '-i' or '-ga', depending on the final sound. An 
indefinite noun takes either the suffix '-un' or '-nun', also dependent 
on the final sound. Thus, 'kugo' (vowel assimilation of '-ga'), 
'kunyoga', 'kunun', and 'kunyonun' are all possible.

Hope this helps.


From: bingfu <>

Hi, Andrew,
	I just wonder whether are you aware of
Bernard Comrie's following works.
	1977: 	Subject and direct object in Uralic languages: A functional
	explanation of case-marking system. Etudes Finno-Ougriennes 
	12 (Budapest: Akademiai Kiado) 5-17

	1987: 	Definite and Animate Direct Objects: A natural class.
	Linguistica Silesiana 3: 13-21. 

Where he extensively discussed the phenomena related to your issue
and provide a quite convincing explanation.

	Bingfu Lu


From: "Oesten Dahl" <>

It strikes me that you and many of the people who have responded to your
query have been talking quite happily about "marking" without realizing what
it presupposes. The problem is that whereas you may claim that definiteness
is marked in many languages, you'll have to look hard for something that you
could truly label an "animacy marker". What you do find is that animate and
inanimate nouns are treated differently by various grammatical rules. For
as Martin Haspelmath pointed out, only animate datives can be raised by
possessor raising.
The closest you get to marking of animacy is semantically-based gender, but
it is seldom so clean-cut that you can say there is an animacy marker, or
inanimacy marker for that matter. You might want to have a look at the
following paper:

Oesten Dahl and Kari Fraurud. 1996. Animacy in Grammar and Discourse. In
Thorstein Fretheim & Jeanette K. Gundel (eds.), Reference and Referent
Accessibility, 47-64. Pragmatics and beyond 38. Amsterdam: Benjamins.

 Best regards,
Oesten Dahl


From: bwald <bwaldHUMnet.UCLA.EDU>

Andrew Carnie's inquiry into the relationship between animacy and
definiteness escaped my attention until I read his summary. Otherwise, I
would have discussed the Bantu situation, inlcuding Swahili. Fortunately,
Kieran Snyder gave him much useful information about Bantu, esp Swahili. A
few of K's points, however, seem to be incomplete or inaccurate, as I
understand them. EG

>Now, in the case of Bantu .... there's
>also object-marking to consider, but really the word order variation
>where it exists (say, in Swahili) boils down again to hearer-givenness,
>and the object-marking is completely determined by animacy and

Swahili is representative of a limited area of Bantu, and is relatively
extreme in requiring object-marking of an animate over an inanimate,
regardless of the information status of the animate and a possible
concurrent inanimate object. The more general situation in Bantu is
complicated and much more varied, but can be traced to a situation (still
existent for many Bantu languages) in which "hearer-givenness" (I usually
call it "contextual topicality", a slightly different concept) determines
object-marking. EG, a preverbal NP object is contextually topical by the
nature of its syntactic position and requires object marking, regardless of
the animacy or information status of a concurrent postverbal object. That
is, if A and B are both objects, the object marker MUST refer to A in the
structure: A SM-TM-*OM*-V (B). It cannot refer to B.

As K notes, that is not the case for Swahili, where, if B is animate, esp
human, and A is not, the *OM* can only refer to B (and that's only the
beginning of the conditions on Swahili OM reference, as can be gleaned from
my following comments).

One of the conditions which K gives for Swahili is inaccurate -- as I
understand K's statement.

>2) if [two objects] are equal in animacy value but not in hearer-givenness,
>object-mark the hearer-given one, whichever it is

That is not the case. In the context just given, thematic role restricts
OM reference. EG, if I say "I called some doctors for my mother", ONLY "my
mother" can receive OM reference -- because it is the "extended" object
(object of a verb extension). Similarly, if I say "I called my mother for
some doctors", ONLY "some doctors" can receive OM reference. The
hearer-givenness status *does NOT matter*. In the case of a first or
second person object, these are inherently more topical than a third person
object (they are always hearer-given). If I say "I called some doctors for
you", ONLY "you" can receive OM reference (as an extended object). If I
say "I called you for some doctors", I CANNOT make "some doctors" an
extended object at all. I have to use a prepositional paraphrase along the
lines of "I called you for the sake of some doctors". In that case, only
"you" is a GRAMMATICAL object at all, and thus can be (indeed must be)
referred to by an OM. "some doctors" is NOT an extended object, but the
object of the PREPOSITIONAL phrase "for the sake of". Such major
grammatical recasting is not obvious from K's statement quoted above.

K's next statement is consequently misleading:

>3) if they are equal in both animacy and hearer-givenness, object mark
>the recipient (ie indirect) object

No. Hearer-givenness plays no grammatical role in object marking, only
animacy and role. Well, in view of what I said above, my last statement is
not quite accurate. Hearer-givenness only plays a role in that when an
inherently more topical object (a first or second person) is put in a
direct object role and an inherently less topical "object" (third person)
is put in an indirect object role, the less topical "object" CANNOT be
expressed as an extended object at all, but must be expressed a
prepositional object. With regard to the verb, then, it is not a

For those who are unsure of what I mean by opposing "grammatical" and
prepositional object, it is a general Bantu grammatical distinction which
can be illustrated by the following Swahili examples:

ni-li-ku-it-*i*-a tabibu (I-call-OM=you-call-APPLIED-FV doctor)
 "I called a/the doctor for you"

"you" is a GRAMMATICAL object. It is the extended object made possible by
the extension -i-, usually called "applied" or "applicative". The object
happens to be interpreted as benefactive in contexts like the one in the
example. "doctor" is also a GRAMMATICAL object, as indeed the "direct
object" (object of the root transitive verb) always is. The example CANNOT
mean "I called you for the doctor".

ni-li-ku-it-a kwa ajili ya tabibu (I-Past-OM=you-call-FV for sake of doctor)
 "I called you for (the sake of)/on behalf of a/the doctor"

ONLY "you" is a GRAMMATICAL object. There is NO extended object or verb
extension. "doctor" is the "object" only of the prepositional phrase "for
the sake of/on behalf of". Swahili balks at saying: ????ni-li-mw-it-i-a
wewe tabibu (???I-Past-OM=him-call-APPLIED-FV you doctor), or any other
word order permutation of the same.

K continues:

>In Kinyarwanda, which lacks the relevant word-order variation but
>which has object-marking, the above hierarchy of hearer-givenness and
>animacy also obtains.

K's comments for Swahili above are better suited to Nyarwanda, NOT for
Swahili (and various adjacent languages).

The way I figure it on the basis of extensive Bantu data, originally
object-marking in the context of concurrent objects (where one is an
extensional object and the other a direct object) was indeed based on
contextual topicality. Whichever object was more topical could receive
object-marking reference; even a third person could receive object marking
reference in spite of a concurrent first or second person object, as long
as it was more topical in context (more what the longer stretch of
discourse was about). That remains the case in some Bantu languages, e.g.,
Zulu, etc. Most often such object-marking reference would be the only
reference to the object in the clause. And almost invariably it would be a
DEFINITE object. Pragmatically, humans are inherently more topical than
inanimates. Thus, by the process of grammaticalisation in a certain area
of Bantu including Swahili, humans became grammatically more topical and,
regardless of their contextual topicality, require object-marking (as long
as they are referential -- even Swahili still varies on object-marking for
non-referential human objects, such as "virgin" in "marry a virgin", where
that object does not have a specific referent but only indicates a type.)

Stated simply, the connection between animacy (esp human vs. non-human) and
definiteness is the pragmatic connection of inherent topicality (which
includes the invariable hearer-givenness of first and second persons). As
far as I know, when inherent topicality grammaticalises, any grammatical
processes that applied to definiteness generalise to (indefinite) humans.

Thematic role (or "case" or whatever) is a distinct dimension in Bantu
historical change. It adjusted differently to grammaticalisation of
inherent topicality in different Bantu areas, and is too complex to discuss
here. I've written a cycle of papers on object-marking in Swahili and
various other Bantu languages (and plan to return to to it with additional
details at some point).
My most complete statement about the facts and origins of Swahili and most
other systems of Bantu object-marking was published (more recently than the
date suggests) as:
"East Coast Bantu and the Evolution of Constraints on Passivisation" in
Sprache und Geschichte in Afrika (SUGIA) 15 (1994) 211-315. -- Benji


From: Asya Pereltsvaig <>

I am very interested in your work on the relationship between
definiteness and animacy. I am working on the relationship between
definiteness/specificity/animacy and Case marking. In particular, in
many languages (both nominative-accusative and ergative) Case marking
depends on some notion of specificity of the object. However, some
languages go by definiteness or specificity (either overtly marked or
understood) while others go by animacy (and some go by both). I was also
looking for a way to unify definiteness and animacy in one phenomenon.
However, this seems to be intuitively wrong. This is why I am interested
in the solution proposed by Jelinek to this problem of animacy
patterning with definiteness in word order and Case marking phenomena. I
would be very interested in exchanging ideas on this subject.
Good luck with your work,

From: ana de la fuente <>

[Editor's note, I had to delete some of the lines of data from
this message because they were displayed in a character set that
LINGUIST is currently unable to support (support for such non-USASCII
symbols is coming soon). AC]

I'm a mexican student living in Paris (Paris III). I'm writing you
because I would like to contribute to your question about <animacy and
definiteness>. Unfortunately I just subscribed to THE LINGUIST LIST a
week ago, and I had only access to the summary of your question.

I'm working on the syntax of classical nahuatl, and last year one of the
things I explored was the order of constituents in the transitive phrase,
and the possibility of the existence of a VP in this language. I found
out that the definite/ indefinite reading of the NPs affects the order of
the constituents. With a definite reading of the arguments the order
relatively free, but with indefinite reading the order is fix. Another
characteristic that could interest you is that both definite and
indefinite noun can be marqued by animacy (which is overt only in plural
contexts). Here are some examples from my thesis:

[Relative order (VOS et VSO) in many mayan languages is determined by

These can also be observed in classical nahuatl. An indefinite object
occupies the first position to the right of the transitive verb, the
complement position [ [ [ VO ] ]S]

[Nahautl data deleted]

We can see that a definite noun is necessary preceded by a definite
particle. The object noun (definite or indefinite) can be animate or
inanimate. The difference in animacy is overt in plural forms with both
definite and indefinte object arguments.

[Nahautl data deleted]

These restriction on definiteness and animacy of plurals is also
reflected on the verb, by means of the obligatory agreement with the
arguments of the verb.

a. With definite agreement, there is preference for the animate object in
bitransitive verbs. The 3person agreement disappears" (4a), only the
plural agreement "stays" (4b):

[data deleted]

b. With indefinite pronominal argument agreement, which is overtly
marqued by the accusative form (inanimate) and a dative form (animate),
both objects appear simultaneously in the verb stem, with an animacy
hierarchy reflected in the order of the clitics:

[data deleted]

These only happens with the object of the verb, since the subject (the
agent) cannot have an indefinite reading in nahuatl.


From: Sarah G. Thomason <>

 I don't remember seeing your original LINGUIST query, but the
one response you got about Slavic could be expanded a bit, maybe
usefully. In the earliest Slavic texts (which are Old Church
Slavic), the incipient category animacy that your respondent
mentioned is itself linked very closely to definiteness: in the
particular noun class (masculines, but it's not the only noun
class where masculine nouns occur), in the earliest texts, only
*definite* grammatically animate nouns get the special animacy
marking. Usually. So in "I saw a man", "man" would not get the
animate marking; but in "I saw the man", "man" would be marked
for animacy. (The only grammatically animate nouns at that early
stage in the development of the category were free adult human
males; the semantics of the category has changed in all Slavic
languages, and the link to definiteness also disappeared long
ago -- now any accusative noun in the appropriate noun class(es), 
which vary from lg. to lg., get the animacy marking.)

 The Slavic literature on the definiteness-animacy link says -- and
maybe someone else who answered your query already pointed this out?
- that it makes good intuitive sense, at least, for animate nouns to
get the special accusative marking, because sound changes had merged
the nominative and accusative suffixes in the relevant noun class,
and the chances for confusion in a sentence like "dog bites man",
given that Slavic had & has free word order, are a lot greater than in
a sentence like "dog bites hamburger". (That doesn't explain the
definiteness link, but there are of course other lgs., e.g. Turkish,
where only definite objects get marked accusative; and read on.)

 The source of the animacy category in Slavic is a mystery, the
subject of long controversy. I gave in to temptation some years
back and published my own hunch, which is that it arose when
speakers of Uralic (specifically Finnic) languages shifted to Slavic
and brought along their own comparable category -- namely, definite
nouns with a special marking in the accusative. The grammatical
patterning is identical to the early Slavic one, EXCEPT that Uralic
lgs. have no noun-class system at all, so there's no link in Uralic
with animacy: only with definiteness. And it's not all of Finnic,
but only a couple of Finnic lgs.; that, together with the problematic
early age of the proposed Finnic-to-Slavic shift, is the main 
stumbling block for my hypothesis; one doesn't know how old the
feature is in Finnic.

 Still, the link with definiteness is indisputable even within
Slavic, and if it DOES come from Finnic substratum influence, it's a
case where definiteness eventually got reinterpreted as an animacy
distinction -- and the earliest Slavic texts show the beginning stages
of that set of changes.

 This probably doesn't help you much, but I thought I'd tell you 
about it just in case it helps you make a connection with something
in your own data.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue