-
Lingua 32,239-253.0 North-Holland Publishing Company 1973
THE ERGATIVE: VARIATIONS ON A THEME
Bernard COMRIE Kings sbllege, Cambridge, Grtx t Britain
Recciwd Not.ember 1972
0.
A number of languages are referred to as ergative languages. In
the present paper I want to attempt a characterization of the
concept erga- tive construction within the general framework of
transformational syntax. For those who are not familiar with this
concept at all, the following brief description may serve as an
introduction to the prob- lem.
In the most typical European inflected languages, there is one
case (nominative) that expresses, basically, the subject of the
sentence, and another case (the accusative) that expresses the
direct object. Thus in Latin, where puer is boy (nom.), and pueZlam
girl (act.), we have:
(1) Puer venit. The boy comes.
(2) Puer puellam amat. The boy loves the girl.
In English the same distinctions may be observed, though it is
only within the pronominal system that nominative versus accusative
is marked (e.g. nominative he, she; accusative him, her). Basque
(examples from de Kijk 1966: l-2) is an ergative language, and the
expression of subject and direct object is different from that in
Latin or English:
(3) Miren etorri da. Mary came.
(4) Yon-ek Patxi jo du. John hit Bill.
In (3), the subject (Miren) has no overt inflection; in (4), the
subject has the inflection -ek, whereas the direct object (Patxi)
has no inflec-
-
240 8. Comrie, The ergative: variations on a theme
tion. In other words, the subject of (3) and the direct object
of (4) have the same inflection, while the subject of (4) has a
distinct ending. Quite generally, in Basque the subject of an
intransitive verb and the direct object of a transitive verb stand
m the same case (usually, though rather misleadingly, called the
nominative), while there is a special case (the ergative) for the
subject of a transitive verb. The nelevant syntactic constructions
(irrespective of the order of co:isiituents) are:
(5) Subject - Verb (6) Subject - Verb - Non-direct object (7)
Subject - Verb - Direct object
In (5) we have a verb with no object, in (6) a verb with a
non-direct object; in Basque the subject would stand in the
nominative. In (7) we have a verb with a direct object; in F;lsque
the subject would stand illI the ergative. I shall refer to (7) as
the ergative configuration, though this wili only be realized with
an ergative case in ergative languages. I shall assume (cp. Comrie
197 la: 45-46) that subject and direct object noun phrases can be
referred to together as non-oblique, whereas non- direct object
noun phrases are oblique.
I have defined ergativity in essentially syntactic terms. This
differs from the approach of, e.g., Fillmore (1968: 53--SS), who
would assign case rather on the basis of the semantic role of noun
phrases in the sentence. Below, I hope to show that in tne
instances I treat in detail ergativity is essentially a syntactic
phenomenon. From the data given by Fillmore, and his references to
work by Sapir, it would seem that in the languages he discusses
(they are American-Indian languages) seman- tic factors do
intervene, at least in addition to syntactic factors. The fact
th:lt in some language semantic factors may play a role in deter-
mining the case of some particular noun phrase does not necessarily
mean that syntactic criteria are irrelevant, or that they may not
be sufficient in some other language. In Finnish, for instance, the
direct object will stand in the nominative (plural) or genitive
(singular) where the whole of the object is affected, and in the
partitive if only part of the object is affected (Aaltio 1969:
154); however, the syntactic char- acterization of direct object is
still necessary in order to know to which noun phrase this
distinction may be applied (i.e. noun phrases in general are not
distinguished as to whether or not they are partitive). In Latin,
on the other hand, the direct object goes automatically into the
accusative case, irrespective of its semantics.
A similar example involving the ergative is found in the
Vejnakhian
-
B. Comne, The ergative: variations on a theme 241
languages (North-East Caucasian) (M&aninov, 1967: 78-82).
Chechen is a typical ergative language, and we havt: for
instance:
(8) So w&h. I go.
(9) As begabo palas. 1 shake the carpet.
where as and SO are ergative and nominative of the first person
singular pronoun respectively. Bats (Batsbi, Tush) is a closely
related language, in fact the forms for the pronouns are identical
(as and so). In the ergative configuration, the subject must be in
the ergative case:
(10) As jopst axo. 1 slough the land.
In other configurations, either the ergative or the nominative
may be used for the subject:
(11) So w&e. (12) As woke.
1 fell. (11) and (12) are not synonymous. MeslSaninov (1967: 82)
explains the meaning difference as follows: (12), with the
ergative, implies that it was my own fault that I fell down,
whereas (1 1), with the nominative, carries no such implication.
Semantic factors, then, are relevant to the choice of case. But
these semantic factors are not sufficient to distin- guish
nominative from ergative. In the ergative configuraticn we must use
the ergative, whether or not the action is the fmlt of the subject,
whether or not we wish to imply this. Thus the syntactic
characteriza- tion of the ergative configuration is still required.
There is a real prob- lem in the offing here, and one to which
cannot offer a satisfactory solution at present, namely: what is
the rationale behind the tie-up that seems to be felt, and is
explicit for instance in Bats, between the erga- tive as syntactic
expression of a transitive subject, and semantic cate- gories like
agentive, or non-accidental? The purely semantic approach to the
ergative is equally unable to account for this syntactic-semantic
tie-up in a non-arbitrary way.
In discussing ergative languages, one might conclude that the
notions subject and direct object, while useful for outlining the
ergative con- struction informally ta those familiar with
nominative languages, do not play any role internally to the syntax
of ergative languages. Apart
I take this to be the Qtenor of De Rijk (1966), which also
attempts a (transformational) syntactic treatment of the ergative,
in Basque. As this paper has not been published, 1 do not feel at
liberty to discuss it in detail; the line of argument is radically
different from that presented here, and does not take account of
such variations of the ergative as are found in Chukchee, Georgian,
Punjabi, etc.
-
242 B. Comrie, The ergative: variations on a theme
from the basic point tnat there is no direct morphological
expression of subject and direct object, ?or some ergative
languages there is other evidence that might seem to point in this
same direction. Verbal agree- ment is one instance in point, in a
wide range of ergative languages, where the verb agrees with an
intransitive subject (i.e. where therti is no direct object), but
with the direct object where there is one - i.e. in either case
with the nominative noun phrase, and not with the erga- tive. Thus
in Punjabi (Shackle 1972: 73, 82) we have:
( 13) d (masc. sg. ) kal lawr gya (masc. sg. ). he yesterday
Lahore went (= He went to Lahore yesterday.)
but with transitive verbs:
( 14) on2 (.~?rsc. pl. ) ne api_la k& (??rczsc. sg. ) ka@ya
(r~asc. sg. ) si. they erg. part. own house built was (= They had
built their own h.ouse )
(15) d&tar (masc. sg. ) ne t! fylmaN (jtim, pl. ) vekhia
Cfem. pl.) hoggia. doctor erg. part. these films seen will-be (=
The doctor will have seen these films)
The data from many ergative languages, however, indicate that
sub- ject and direct object are releva-;lt categories in describing
the syntax of these languages. For instance, in Georgian the verb
requires affixes mark- ing the person and number of subject, quite
irrespective of whether the subject is in the nominative or the
ergative, and different affixes mark- ing the person and number of
the direct object (TschenkCli 1958: 1.348356). Walbiri, a language
of Central Australia, marks full noun phrases as an ergative
language, but duplicates them by clitic pronouns on a
subject-object basis (Perlmutter 197 1: 89). Basic (unmarked)
word-order is determined in many ergative languages on the basis of
subject and object, a.lthough the data here are not always
watertight, given the possibilitie; for free word-order in the
languages under con- sideration. Georgian has a basic word-order
S(ubject)-V(erb)-O(bject) (TschenkGli 1958: 1.12); Dargva has SOV
(Abdullaev 197 1: 27-29), as has Punjabi (Shackle 1972: 48).
According to Lafltte (1967: 46), Basque is SVO; De Rijk (1966:2),
on the other hand, claims that the order of maJor constituents in
Basque is determined by informational strticture; in either case,
the morphological distinction nominativ+- ergative is not held to
be the basis of determining word-order.
-
B. Comrie, The crgative: variations ora a theme 243
1.
Chukchee (Chukothan, Luorawetlan) is spoken by some 11,000
people, mainly iz-$ the Chukothan peninsula (Eastern Siberia)
(Skorik 1968: 248). In Chukchee, intransitive verbs regularly take
the nomina- tive construction, as in
( 16) Tumg-at jegtel-g?e t. The friends escaped.
whereas transitive verbs require the ergative construction
(Skorik 1968: 267):
(17) Tumg-e na-ntawat-an kupre-n. The friends set the net.
MorphologicaJy, Cht: nominative singular ends in --12 or zero,
the nomi- native plural in -b, -ti, or --at, the ergative (singular
or plural) in -e or -te (for further &tails, not relevant here,
cp. Skorik (I 968: 253)).
One of the syntactic characteristics of Chukchee is the ergative
con- struction. Another is that known as incorporation, whereby a
constitu- ent of the sentence is embedded inside some other
constituent: incor- poration is an optional transformation. In (
17) above we have the verb rztawat with, as its direct object, the
noun phrase h-upre, the two con- stituents forming a verb phrase.
An alternative way of phrasing (17) would be to incorporate the
direct object into the verb itself, forming a new verb stem
kopra-ntawat to net-set. Note in particular the change in the
vocalism of kuprelkopra: Chukchee has vowel harrnony (Skorik
19&t : 250-25 l), so that when the direct object is
incorporated into the verb, forming one phonological word with it,
its vocalism must agree with that of the head constituent, namely
rztawat (cp. Comrie 197 1 b: 453-454). The syntactic transformation
of incorporation (in this particular case) may be illustrated as
follows:
(19) S
(18) S
NP-P
I I I I I I tumg ntawat kupre tumg kopra ntawat
-
So far, I have not indicated how, if at all, the transformation
of incorporation effects the case-marking of subject and direct
object. TO avoid further suspense, I shall now give the Chukchee
version of (17) with incorporation:
(20) Tumg-at kopra-ntawat-gvat. from which it can be seen that
the subject is in the nominative, not the ergative. In the
derivation of (20), we have a deep structure with the ergative
configuration, but a surface structure with the norninative con-
figuration. The case-marking in Chukchee is determined., then, not
by the deep structure configuration, but by the surface structure
con- figuration - in (19) kopra is not direct object of VP, which
has only the one immediate constituent V. There is a general Goint
to be noted here, one that will recur throughout this article:
where a sentence has more than one level of representation (as
generally in a transformation- ai anaiysis), the syntactic
relations may be different at the various levels. In particular, an
ergative configuration may be transformed into a nonergative
configuration, as in (18) and (19) above. These transfor- mational
possibilities present the variations on d theme of the title.
Chukchee has yet a third way of ,bhrasing sentence (17) (Skorik
1968: 2671, namely by putting the direct object into the
instrumental (morphologically, always the same as the ergative), as
in
(2 1) Tumg-at ena-ntawat-gpat kupre-te. with the subject again
in the nominative, not the ergative. (2 1) would appear to be
derived from the same deep structure configuration as (17) and (20)
(i.e. an ergative configuration), but like (20) at some stage in
derivation the ergative configuration is transformed into a
nominative configuration (in surface structure, kupre is nDt direct
object in (21), for then it would appear in the nominative; rather
it has been trans- formed into a non-direct object).
I was surprised and gratified to find, after I had worked with
the above data from Chukchee, that there is at least one other
language, quite unrelated to Chukchee, that allows the parallel
sentence patterns (17) and (21). D argva is a Dagestanian
(North-East Caucasian) language; like Chukchee, it is an ergative
language, though unlike Chukchee it does not have incorporation
(therefore no possibility of sentence type (20)). In the f 11 o
owing examples, the ergative-instrumental ending is -rti or 4, the
nominative is endingless:
-
B. Comrie, fie ergative: variations on a theme 245
(22) Nu Ztiz-li uEulra. (23) Nu-ni Zuz duEulra.
1 read the bdok (lit. 1 book read) [Abdullaev 197 1: 261). In
Dargva, factor in distinguishing nominative surface structure
configuration.
then, it appears that the decisive from ergative constructions
is the
2.
Georgian differs from the typical ergative languages in that it
uses the ergativc construction in certain tenses only while using
the nomina- tive construction in others. In the Present group of
tenses (present, future, imperfect, conditional, present and future
subjunctive) the sub- Lad
JLUL of the sentence stands in the rlominative (whether the verb
is transitive or not) and the direct object in the accusative
(which is morphologically identical to the dative). In the Aorist
group of tenses (aorist and optative) the subject of an
intransitive verb stands in the nominative, as does the direct
object of a transitive verb, while the subject of a transitive verb
stands in the ergative (which in Georgian is morphologically
distinct from all other cases and cases plus postposi- tions). An
account of the tense groups in Georgian is given by Tschen- kili
(1958: I. 64, 150, 500). The following examples illustrate the
vari- ous possibilities. For ease of reference, the following
morphological inflections should be noted: nominative -i, ergative
-m(a), accusative- dative -s:
(24) Student-i midis. +he &dent goes.
(25) Student-i cers ceril-s. The s&dint &ites a
letter.
(26) Siudent-i mivida. The s&dent went.
(27) Student-ma dacera ceril-i. ?he s&dent wrote ;he
lette:.
Note that although in the nominative construction the
morphological ending of the dire& object is the same as that of
the dative case, the different syntactic functions of the two noun
phrases are brought out in the ergative construction, where the
direct object appears in the nominative, but other datives remain.
e.g. (Tschenkkli 1958: I= 157):
-
246 B. Comrie, The ergative: variations on a theme
(28) Mascevlebel-i aCukebs bavSv+ cign-s. The teacher presents
to the child a book.
(29) Mascevlebel-ma aculka bavsv-s qigr-i. The teacher presented
to the child a book.
In Georgian, then, the rule that assigns ergativity is sensitive
not only to the syntactic configuration of subject and object noun
phrases, but also to a tense-group feature on the verb. In what
follows I shall take this tense feature for granted, and
concentrate on defining more accurately the syntactic
configuration(s) that require the ergative. In older stages of the
Georgian language? apparently, there *was no requirement of a
tense-group feature, and the ergative occurred with all tenses
(Cikobava 1967: 15).
The first set of verbs requiring us to go somewhat beyond the
charac- terization of the ergative environment outlined above are
those which, though generally taking a direct object, may (in the
same meaning) occur without any such object. One such verb is qera
write, used with an object in (25) and (27) and without an object
in
(30) Student-i cers. The student writes.
What happens if we put (30) into the aorist? Although the
surface structure of the sentence contains no direct object, the
subject must still stand in the ergative:
(31) Student-ma dacera. Thp student wrote.
Let us assume, as has generally been done in transformational
syntax, that sentences like (30) and (3 1) are derived from
underlying structures with a direct object, this direct object
being subsequently deleted. In deep structure, then, (30) and (31)
are ergative configuration, whereas in surfa.:e structure they are
non-ergative configurations. Georgian ap- pears tc be sensitive to
underlying syntactic configurations, as does, for ?tstance, Basque,
for similar constructions (Lafitte 1967: 188), and unlike Chukchee.
Within a transformational framework, this can readily be captured
by ordering the trar&ormation that distinguishes nomina- tive
from ergative noun phrases (ergativiration) before object-deletion,
in Georgian; in Chukchee, ergativizatior, would follow
incorporation. 2
2 For footnote, see next page.
-
B. Comrie. The ergative: variations on a theme 247
A further point worth noting in this context is that where a
normally transitive verb occurs with a sentential object, if the
verb is in the amist its subject will still be in the ergative.
eg.
(32) Student-i cers, rom mova. The student writes that he will
come.
(33) Student-ma dacera, ram mova. The student w;*ote that e
would come.
In Comrie (197 la: 105) I pointed out that for certain
languages, there is evidence that such sentential objects are noun
phrases in deep struc- ture, but not in surface btructure. 1 know
of no such evidence for
sting that the noun phrase node dominating ram m~va is pruned in
sentences like (32) and (33) - i.e. it is possible that even in
surface structure rorn mova is a noun phrase, where it was one m
deep structure. If any such evidence were forthcoming, this would
be anoth- er point in favour of an analysis of the type outlined,
since studer?f would be surface structure subject of (33), and also
in the ergative configuration in underlying structure (though not
in surface structure). In the absence of such evidence, sentences
like (33) do not tell either way. Incidentally, the noun phrase
nature of sentential objects to such verbs is not restricted to
indirect speech, compare (Tschenkeli 1958: 11.77):
(34) Mela-m tkva: me da mgel-ma aklem-i movkalito. The fox said:
1 and the wolf killed the &nel.
3.
Punj(abi, like Georgian, restricts the ergative construction to
certain tenses only (Shackle 1972: 82, 1X), and I sha 1 here be
concerned with the relevant tenses, i.e. those formed from the past
participle (preterite, perfect, pluperfect, future perfect). In
Punjabi, unlike Georgian, nomi- native (and accusative are not
distinguished overtly, i.e. in the nomina- tive constr;iction,
barring certain exceptions to be- discussed below,
2 In a theory of transformational-generative grammar extended by
the inclusion of non-local derMiona.l constraints (Lkoff 197(l),
the ergativization rule could operate at the same level in both
hwuage-types, but would refer to configurations at different levels
in the two cases. In the abench+ of more detailed analyses of other
aspects of the syntax of the languages involved, It is not de&a
whether this increased descriptive apparatus is required, in this
case.
-
248 B. Comrie, The ergative: variations on a theme
subject and direct are not morphologically distinct. In the
ergative con- struction, the transitive subject requires the
postposition pze (Shackle 1972: 82). In the nominative
construction, the past participle and auxi- liary verb, if any,
agree with the subject; in the ergative construction, they agree
with the direct object without postposition:
(35) meri ma kade c$ pindi e. my mother sometimes tea drinks My
mother sometimes drinks tea.
(36) tiv?uem. sg.) kal l&r gai (fem. sg. ). woman yesterday
Lahore went The woman went to Lahore yesterday.
(37) tivY ne kvirj ky t5 bg Cfem. pl. ) mez te rakkhig vern. yl.
). woman erg. part. so me boo ks table on put The woman put some
books on the table.
In general, the direct object in Punjabi takes no postposition?
where . 1 n71 as the hdirect object takes the postposition ni2
(Shackle I 7 I L : 47).
Sometimes, however, the direct object is constructed with nM
(only where there is not also an indirect object): this happens
when the direct object is a pronoun (obligatorily), and may happen
when it is a definite noun phrase (Shackle ! 972: 69-70). The
precise rules are not relevant, though the existence of the
phenomenon is:
(38) e nu &he na rakkho ji! it plzrt. here not put please
Please put it here.
How does this construction interrelate with the ergative? In the
relevant tenses, where the direct object has ni.?, the subject must
still stand in the ergative with ne, and the verb may not agree
with it. However, the verb may no longer agree with the direct
object, instead it stands in the impersonal form (morphologically,
third person singular masculine):
(39) tiv? ne dujia kytabgnti mez te rakkya (masc. sg. ). wormyi
erg. other books part. table on put The woman put the other books
on the table.
Similarly, where the object of an object-deleting verb is
deleted, the subject is in the ergative, the past particple
impersonal (Shackle 1972: 82-83):
(40) tivY ne vekhya (masc. sg.) :i? *Ivoman erg. seen 1 Qas Had
the woman seen?
-
f3. Corntie, The ergative: variations on a theme 249
The same applies with direct and indirect speech (cp. the
examples in Shackle (1972: 102)):
(41) tivr ne kya (ntasc. sg. ) si ky m5y panj5bi X ~82~~~2~1
erg. said ivas that 1 PkIl?;abi am The woman said: Im Punjabi.
(42) tivine kya si ky 6 pairjabi e. The woman said that she was
Punjabi.
Let us return to the rule that inserts ~t7i after certain direct
objects, comparing it with object-deletion. The effect of
n&insertion is to change what would otherwise be
morphologically a non-oblique noun phrase into an oblique noun
phrase, i.e. make it like a non-direct object, i.e. the
transformation brings about a change in the syntactic relatiors of
the sentence. Object-deletion also brings about a change in
syntactic relations, by deleting the direct object. What the two
transformations iaave in common is that they get rid of the direct
object, in one case transforming it into something else, in the
other case deleting it. Both are direct-object removing
transformations. In the underlying struc- ture of (39)-(40) there
is a direct object, and it is the presence of this underlying
direct object that triggers ergativization, whence the post-
position ne in these sentences. In derived structure, however, they
have no direct object. The verb agreement rule states that the verb
must agree wi;h the direct object noun phrase if the subject is
ergative, otherwise with the subject noun phrase; where there is no
such noun phrase (i.e. impersonal (subjectless) sentences in the
nominative con- struction, and sentences that have undergone
direct-object removing in the ergative construction), the verb goes
into the third person singular. In other words, as far as verbal
agreement goes it is the derived struc- ture that counts. The
transformations considered are ordered as fol- lows: (i)
Ergativization (ii) Direct-object removrng transformations (iii)
Verb agreement . In this way we can account for the apparent
discrepancy between erga- tivization and verb-object agreement.
One assumption, theoretical rather than empirical, that has
guided my thinking on the ergative is that in the various languages
that have
-
250 B. Comrie, The ergative: variations on a theme
the ergative construction, we are dealing with a unified
phenomenon. This does not, of course, mean that there are no
differences between ergative constructions in different languages
(cp. the tense-group re- quirement in Georgian and Punjabi,
influence of semantic factors in Bats, importance of surface
structure configuration in Chukchee, but not in Georgian or
Basque), but it does imply that the basic framework of the
construction should be the same in the various languages.
One view, well represented in the traditional literature on the
erga- tive, but which I slave not so far considered, is that the
ergative is in some way intimately connected with the passive
construction in lan- guages like English (cp., e.g., Schuchardt
1896; Uhlenbeck 1917; these and several other older essays on the
ergative have been collected, in Russian translation, in Bokarev
(1950)). There is a certain amount of circumstantial evidence in
favour of this view, the main points being as follows (not all of
them are true of all ergative languages): (i) In the ergative
construction, as in the passive, the (underlying) ob- ject of a
transitive verb is treated morphologically the same as the subject
of an intransitive verb, and differently from the (underlying)
subject of a transitive verb. In Punjabi, even this parallel is not
strictly correct, since t!le direct object in an ergative
construction may be trans- formed into a non-direct object with
nii, whereas tile subject (transitive or intransitive) never takes
nti, (ii) In many ergative languages. the verb agrees with an
intransitive subject and a transitive object in like manner, and
either does not agree, or shows different concord markers, with a
transitive subject; this is not true, for instance, of Georgian,
where there are specifically subject and object markers,
irrespective of nominative or ergative construction (Tschenkeli
1958: 1.348-356); it is true, for instance, of Basque (De Rijk
1966: 6); (iii) The inflection of the ergative is often the sane as
that of the instrumental (e.g. Chukchee, Dargva; not Georgian); ill
languages that have a passive, the agent noun phrase is often in
the same case or takes the same preposition/postposition as the
instrument; (iv) Ergative languages tend not to have a passive
distinct from active ergative; again, Georgian is an exception
(Tschenkbli 1958: 1.254), e.g.
(43) Ceril-i icereba student-is mier. The letter is written by
the student.
(44) Ceril-i daicera student-is mier. The letter was written by
the student.
-
B. C?mrie. The ergative: variations on a theme 251
(The subject of (44) is, of course, in the nominative; at no
stage of derivation is peril subject of an ergative configuration.)
There are also many non-ergative Lnguages that have no passive
construction.
The ergative might therefore be construed as a variant of the
passive, differing basically in that whereas the passive is
typically an optional construction, the ergativl= would carry out
the same structural change, but obligatorily. There is some
historical evidence that the ergative in modern Indo-Iranian
languages is in fact the reflex of an originally optional passive
transformation (cp. the arguments in Pirejko 1968; this historical
account is not universally accepted).
Nonetheless, and particularly if we want to retain the ergative
as a unified phenomenon, I think there are more cogent arguments
mili- tating against the obligatory passive as a synchronic
solution to the ergative. The solution will obviously not do for
languages like Georgian where there is a passive separate from the
ergativc, where moreover the verb requires the same concord marker
for transitive and intransitive subject, and a different marker for
the dir :t object. In general, any rule that requires reference to
the concepts of subject and direct object, as defined for
nominative ianguages, will become cumbersome if these same
definitions are not applicable to ergative languages too. A number
of such rules were discussed above. Historically, it may well be
true that in (some) ergative languages the ergative construction
has developed from a passive at an earlier stage of the language,
but this does not seem to be of synchronic relevance.
Sync;lronically, the rule assigning ergativity operates on
precisely the same kinds of structures as rules assigning
nominative and accusative in non-ergative languages. On this basis,
it is much easier to handle s of the more marginal uses of the
ergative that crop up in some e ive languages. For instance,
although most transitive verbs in Punjabi require the ergative
construc- tion, there are a few exceptions, like &a- bring,
bol- speak, myl- meet (Shackle 1972: 74), requiring the nominative
construction:
(45) tivY dnil lai makkhan (masc. sg.) lyai Cfenz. sg.). woman
them for butter brought The woman brought butter for them.
In Georgian, most intransitive verbs require the nominative
construc- tion, but there is one exception: iarn he went requires
its subject in the ergative, although it cannot take a direct
object, and although other verbs meaning to go take the nominative
construction in the Aorist tense-group (Tschenkeli 1958: 1.169). It
is interesting to compare the
-
252 B. Comrie, The ergative: variations on a theme
different developments of the ergative construction in the
related Kartvelian language Zan. In one form of Zan, Megrelian
(Mingrelian), the subject of an aorist verb invariably (transitive
or intransitive) stands in (what i:: etymologically) the ergatslve;
in the other form, Chan (Laz), only the transitive subject stands
in the ergative, but in all tenses (i.e. a typical ergative
language) Kiziria 1967: 73-74). In certain languages (Bats, cp. (1
1 )-( 12); some American-Indian languages) the choice of the
ergative is influenced by semantic factors, which combine with the
syntactic-constructional environment of the ergativization
rule.
On this analysis, the ergative may appear as a totally arbitrary
rule for a language to have - yet the construction turns up in a
number of genetically unrelated and geographically separated
languages. One pos- sible explanation would be historical - if we
accept that diachronically the ergative is a passive made
obligatory, then it is not too far-fetched for this change to have
occurred independently in different languages. There is a chance
that even synchronically the rule is not so unnatural as it might
appear, though this suggestion, in terms of a measure of closeness
of nodes (Comrie 197 1 a: 273-274), is ctill very tentative. Let us
define the closeness of nodes as: A node X is closer to a node Y
than to a node x if and only if the lowest node dominating both X
and Y is dominated by the lowest node dominating both X and Z.3
(46)
(47)
3 This is a natural relation to deli 23 on hierarchical
structure, e.g. the members of a given sub- group are more close
than members of the same group belonging to different sub-groups.
In Corntie (1971a: 270-273) the measure is used in connection with
Equi-W-deletion, It remains to be seen whether the measure does
have general linguistic validity.
-
B. Cornrie, The ergative: variations on a theme 253
Referring to configurations (46) and (47) (cp. (5) and (7) -
non-direct objects (oblique noun phrases) are not relevant)? we see
that in (47) the non-oblique NP closest to the verb is the direct
object. whereas in (46) the non-oblique NP nearest to the verb is
the subject. Perhaps, then, ergative languages are putting the
non-oblique NP nearest to the verb in the nominative, and other
non-oblique NPs (only the transitive subject comes into question)
into the! ergative.
Reference 3
Aaltio, M.-H., 1969. Finnish for foreigners. Helsinki: Otava.
Abdullaev, Z.G., 1971. OEerki po sintaksisu darginskogo jazyka.
Moscow: Nauka. Bokarev, E.A. (ed.), 1950. Ergativnaja konstrukcija
predlogenija. Moscow: Izd-vo inostrannoj
Iiteratury. Cikobava, A.S., 1967. Problema ergativnoj
konstrukcii v iberijsko-kavkazskix jazykax. In: V.M.
Zirmunskij et al. (eds.), Ergativnaja konstrukcija predloZenija
v jazykax razlicnyx tipov Leningrad : Nau ka.
Comrie, B., 197la. Aspects of sentence-complementation in
Russian. Cambridge: University of Cambridge, Ph.D.
dissertation.
Comrie, B., 1971 b. Some thoughts on defining the word in
phonology. Pap: Linguist. 4.3. De Rijk, R.P.G., 1966. Redefining
the ergative. Cambridge, Mass.: MIT, unpublished. Fillmore, C.J.,
1968. The case for case. In: E. Bach, R.T. Harms (eds.), Unrversals
in linguistic
theory. New York: Holt, Rinehart and Winston. Kiziria, A.I.,
1967. Zanskij jazyk. In: V.V. Vinogradov et al. (eds.), Jazyki
narodov SSSR, vol.
IV (E.A. Bokarev et al. (eds.)). Moscow: Nauka. Lafitte, P.,
1967. Grammaire basque. Bayonne: Editions des Amis du musee basque
et
I kaS. Lakoff, G., 1970. Global rules. Language 46.3.
MeSEaninov, II., 1967. Ergativnaja konstrukcija v jazykax razlicnyx
tipov. Leningrad: Nauka. Perlm:ltter, D.M., 1971. Deep and surface
structure constraints in syntax. New York: HoIt,
RiTehart and Winston. Pirejko, L.A., 1968. Osnovnye voprosy
krgativnosti na materiale indoiranskix jazykov. MOSCCW:
Nauha. Schuchardt, H., 1896. uber den passiven Charakter des
Transitivs in den kaukasischen
Sprachen. Sitzungsberichte der kaiserlichen Akademie der
Wissenschaften, philosophisch- historische Klasse, vol 133.
Shackle, C,, 1972. Punjabi. London: Teach Yourself Books.
Skorik, P.Ja., 1968. Cukotskij jazyk.. In: V.V. Vinogradov et al.
(eds.), Jazyki narodov SSSR,
~01. 5 (PJa. Skorik et al. (eds.)). Leningrad: Nauka.
TschenkCli, K., 1958. Einfuhrung in die georgische Sprache. 2 vol.
Zurich: Amirani VerIag. UhIenbeck, C.C., 1917. Het passieve
karakter van het verbum transitivum of van bet verbum
actionis in talen van Noord-Amerika. Verslagen en mededeelingen
der Akademie van Weten- schappen, Amsterdam, Afd. Letterkunde, V.
2.