Fluid Ergativity in Gujarati*

Peter Hook....and ....Bharati Modi
University of Michigan
.............MSUniversity of Baroda

....... The two case-marking patterns most commonly found among the world's languages are the nominative-accusative and the ergative-absolutive. In the first, case-marking (or agreement in the verb) groups the subjects of intransitives together with the subjects of transitives as a single morphosyntactic category in opposition to transitive direct objects. In some of its pronouns, English shows this pattern:

(1) I ate fish. .........Tran subj (Nom) ....'I' .... (English)
.... I was there. ......Intr subj (Nom) .....'I'
.... They left me. ....Tr dir obj (Acc) .... 'me'

In the second, intransitives subjects and transitive direct objects form a single morphosyntactic category in opposition to that of transitive subjects [meanings and word order as in (1)]:

(2) mye khyov gaaD ......Tran subj (Erg) ....'mye' .... (Kashmiri)
.... bi oosus tatyi ..........Intr subj (Abs) ..... 'bi'
.... tyimav troovus bi .... Tr dir obj (Abs) .... 'bi'

....... In addition to these two (broadly defined) case-marking patterns there is a third, that of active-stative languages. In some of these, intransitive subjects are treated as belonging to two groups: subjects of "active" intransitives (laugh, cry, run, jump, etc.) are grouped morphosyntactically with transitive subjects while subjects of "stative" intransitives (be, seem, belong, own, exceed, etc.) are grouped with transitive objects. In other languages there is no grouping by lexical identity of the predicate. Rather if the referent of the subject of an intransitive is felt to be actively engaged in the action, an ergative or agentive case is assigned. If not, then a nominative or absolutive (or, as in these examples from the North Californian Pomo language, an accusative) case is assigned:

(3) a. to: shinu:-cade............................... a'. ?a: shinu:-cade
....... me.Acc be.drunk-NrFut......................... I.Erg be.drunk-NrFut
....... 'I'll get drunk.' (a prediction) ................. 'I'm gonna get drunk.' (a promise)

.... b. mo:wal to: shu?uci?-khemna .............. b'. mo:wal ?a: shu?uci?-khemna
....... him.Acc me.Acc forget-Refl.Fut.............. him.Acc I.Erg forget-Refl.Fut
....... 'I'll forget him.' (a prediction) ................ 'I'll forget him.' (an intention)

......................................................(data from O'Connor 1985:231-2)

When semantic factors (rather than lexical ones) determine the choice of the case of the subject noun phrase, we speak of "fluid subject-marking".
....... Most fluid subject-marking languages are found in North America and in the Caucasus mountains of Eurasia. Fluid subject-marking in South Asia has been rarely noted nor much discussed (fn 1). In this paper we present evidence for it as a covert category operating in the morphosyntax of the NIA language Gujarati.
....... Like most Indo-Aryan languages spoken to the west of Bihar, Gujarati shows a split ergative pattern in its verb agreements: In the preterite or perfect tenses (ie, in any tense that is morphologically based on the OIA participle in -ta) the verb concords with the direct object (if there is one):

(4) mE kaam kar-i lidh-U
.... I.Erg work.Nsg do-CP took-Nsg
.... 'I did the job.'

Otherwise, it agrees with the subject:

(5) chokr-o jaldi doD-y-o paN chokr-i jaldi na doD-i
.... boy-Msg quick run-Pst-Msg but girl-Fsg quick not run-PstFsg
.... 'The boy ran quickly; but the girl did not.'

In other tenses it agrees with the subject, not the object:

(4') hU kaam kar-i la-ish
.... I.Nom work.Nsg do-CP take-Fut1sg
.... 'I will do the job.'

(5') chokr-o jaldi doD-she chokr-i jaldi nahI doD-e
.... boy-Msg quick run-Fut3sg girl-Fsg quick not run-Pr3sg
.... 'The boy will run quickly; the girl will not.'

This rule of concord has traditionally been considered as being purely a rule of syntax, its application (with a few lexical exceptions like laav 'bring') being governed solely by tense and predicate argument structure. However there is evidence for the presence of "fluid" or semantically based concord in certain Gujarati verb sequences. It is possible to give an active-stative analysis of this fluid concord.
.... I. In Gujarati the set of phasal verbs includes the morphologically related pair maND and maaND 'begin'. The intransitive maND is used with either intransitive or transitive infinitival complements:

(6) koi koi vaar bij-U bol-vaa maND-i ja-t-U........................ (Broker 1972:21)
.... some some time second-Nsg talk-Inf begin-CP go-PstHab-Nsg
.... 'Occasionally another would start talking (while one was halfway through).'

(7) taaro e-ne prem thi shikhav-vaa maND-i jaay che
.... swimmer him-Dat love with teach-Inf begin-CP goes is
.... 'The swimmer begins affectionately teaching him.'.............. (Broker 1972:109)

No matter what the transitivity of its complement is, maND is indistinguishable in its grammar from that of other intransitive predicates in Gujarati and we will not discuss it further.
.... While maaND , the transitive form of the phasal, is often used with transitive complements (8), it can also be used with intransitive complements (9) and (10):

(8) tyAA-naa maNDaL saathe buddhidhan-e sambandh vadhaar-vaa maaND-y-o
.... there-Gen people with Buddhidhan-Erg relation.Msg increase-Inf begin-Pst-Msg
.... 'Buddhidhan began to cultivate a relation with the people there.'.......... (Tripathi 1988:53)

(9) chokro gai.kaal-e caal-vaa maaND-y-o
.... boy.Nom yesterday-Loc walk-Inf begin-Pst-Msg
.... 'The boy began to walk yesterday.'...................... (Cardona 1965:5.20.3)

(10) caare baaju bhaavo vadh-vaa maaND-y-aa ha-t-aa
.... four sides prices.Nom.Mpl grow-Inf begin-Pst-Mpl be-Pst-Mpl
.... 'Everywhere prices had begun to increase.'................... (Broker 1972:95)

.... In Gujarati, as in most of western Indo-Aryan, the occurrence of ergative case-marking on transitive subjects is entirely automatic: If the verb is transitive and the tense is preterite or perfect the subject gets the ergative case; otherwise, not. However, when used as a phasal verb, maaND is an exception. With phasal maaND used in the appropriate tense a subject entitled to get the ergative case may or may not actually get it. It is as if the general, categorical, morphology-based rules specifying the case of the subject are held in abeyance, perhaps by the occurrence of a lexically transitive phasal auxiliary with lexically intransitive main verbs (fn 2). The assignment of subject case now depends on a complex interplay of a number of syntatic and semantic variables. For example, if the infinitive dependent on transitive maaND is itself intransitive, the subject may get the ergative case, as in (11):

(11) ja-vaa-no samay thayo jaaN-i tem-Ne paachaa vaL-vaa maaND-y-U
.... go-Inf-Gen time became know-CP he-Erg back turn-Inf begin-Pst-Nsg
.... 'Realizing it was time to go, he began to turn around...'......... (Broker 1972:52)

Or it may not, as in (12) and (13):

(12) baapuji-naa bijaa mitro paN aav-vaa maaND-y-aa
.... Bapuji-Gen other friends too come-Inf begin-Pst-Mpl
.... 'Other friends of Bapuji began to come over, too.'................. (Chavada 1993:13)

(13) vaajAA vaag-vaa maaND-y-AA
.... instruments.Nom.Npl be.played-Inf begin-Pst-Npl
.... '...the instruments began to play...' ........................ (Broker 1972:52)

Sometimes even if the dependent infinitive is transitive the subject may occur in the nominative:

(14) maarU man bijU j druSy jo-vaa maaND-y-U
.... my mind.Nom other Emp scene see-Inf begin-Pst-Nsg
.... '...my mind began to behold a completely different scene...' ...... (Broker 1972:106)

Given these variations there are eight types of phasal constructions possible in Gujarati: (Intransitive vs. Transitive) x (Animate vs. Inanimate) x (Erg vs. Nom) = 2 x 2 x 2 = 8:

(15) a. IntrAnErg: ....e-Ne caal-vaa maaND-y-U....'She began to walk.'.... (SJoshi)

...... b. IntrAnNom: ..chokro caal-vaa maaND-y-o.... 'The boy began to walk.' .... (Cardona)
............................ kaantaa has-vaa maaND-i ....'Kanta began to laugh.'.... (SJoshi)

...... c. IntrInanErg: ....patthar-e naas-vaa maaND-y-U....'The stone began to flee.'.... (SJoshi)

...... d. IntrInanNom: ..gaaDU caal-vaa maaND-y-U.... 'The cart began to move.' .... (Tripathi)

...... e. TrAnErg: ....vaartaa mE shodh-vaa maaND-i....'I began to look for a story.'.. (SJoshi)
......................... story.Fsg I.Erg search-Inf began-Fsg

...... f. TrAnNom: ....loko copDi vAAc-vaa maaND-y-AA
........................... people.Nom book.Fsg read-Inf began-Pst-Npl
........................... 'People began to read the book.' ............ (Bharati Modi)

...... g. TrInanErg: ....e-ni kalpanaa-e duhkh-naa taakaa ne taakaa vaN-vaa maaND-y-aa
.......................... that-Gen thought-Erg pain-Gen stitches and stitches knit-Inf begin-Pst-Mpl
......................... 'The thought of that begin to knit stitch after stitch of pain.' (Meghani 1990:45)

...... h. TrInanNom: ..maarU man bijU j druSy jo-vaa maaND-y-U
........................... my mind.Nom other Emp scene see-Inf begin-Pst-Nsg
.......................... 'My mind began to behold a completely different scene...' .. (GBroker)

.... A small, random sampling of forty-one -vaa maaND-constructions in texts shows that they are not evenly distributed into these eight possible classes:

a. IntrAnErg:

9

e. TrAnErg:

11

b. IntrAnNom:

3

f. TrAnNom:

0

c. IntrInanErg:

1

g. TrInanErg:

2

d. IntrInanNom:

14

h. TrInanNom:

1

(16) a. IntrAnErg: .... ....9 ........e. TrAnErg: .. ....11

..... b. IntrAnNom: .... ..3.........f. TrAnNom: .... .0

..... c. IntrInanErg: .......1........g. TrInanErg: ......2

..... d. IntrInanNom: ...14........h. TrInanNom: .....1

.... It is possible to account for the asymmetries in (16) as the reflections of markedness relations holding among syntactic and semantic variables: Animacy of subjects is associated with transitivity and control over volitional acts. Inanimate subjects tend to go with intransitive predicates and be associated with non-volitional events. Thus, in ex. (11) the action expressed by the dependent infinitive, even if intransitive, is one which is under the conscious control of the subject tem, which gets the ergative:

(11) ja-vaa-no samay thayo jaaN-i tem-Ne paachaa vaL-vaa maaND-y-U
.... go-Inf-Gen time became know-CP he-Erg back turn-Inf begin-Pst-Nsg
.... 'Realizing it was time to go, he began to turn around...'........... (Broker 1972:52)

Such is not the case in exx. (12) and (14): Musical instruments cannot play of their own accord (12):

(12) vaajAA vaag-vaa maaND-y-AA
.... instruments.Nom.Npl be.played-Inf begin-Pst-Npl
.... '...the instruments began to play...' ............... (Broker 1972: 52)

and the mind's eye "sees" the images before it in (14) without being able to control their appearance:

(14) maarU man bijU j druSy jo-vaa maaND-y-U
.... my mind.Nom other Emp scene see-Inf begin-Pst-Nsg
.... '...my mind began to behold a completely different scene...' ...... (Broker 1972: 106)

This association of ergative versus nominative case with a variable semantic rather than a categorical factor allows manipulation of it by speakers (and authors) to create special effects:

(17) patthar-e naas-vaa maaND-y-U
.... stone-Erg flee-Inf begin-Pst-Nsg
.... 'The stone began to flee.'........... (Suresh Joshi)

The context of (17) is the stream of consciousness narration of a dream sequence in which events are presented as if occurring against the narrator's will; that is, with a will of their own.
.... While semantic variables such as control appear to have a major role in determining whether the subject of a phasal construction in Gujarati gets the ergative or the nominative, there are syntactic factors that also play an important part. Or where syntax and semantics are so completely aligned that variation in subject case is impossible. For example, the ergative case cannot occur with the noun phrase denoting the source of an experience in intransitive expressions of experience:

(18) a. e-ne taav aave che ..............................'She has a fever.'
........ her-Dat fever comes is

...... b. *e-ne taav-e aav-vaa maaND-y-U..........'She began to have a fever.'
........ her-Dat fever-Erg come-Inf begin-Pst-Nsg

...... c. e-ne taav aav-vaa maaND-y-o...............'She began to have a fever.'
........ her-Dat fever.Msg.Nom come-Inf begin-Pst-Msg

Contrast (18a-c) with (19) in which the experiential aspect of the situation is absent:

(19) taav-e aav-vaa maaND-y-U ............'Fever began to occur (in the town).'
...... fever-Erg come-Inf begin-Pst-Nsg

.... The subjects of morphologically marked passives also cannot occur with the ergative case:

(20) a. e saari kavayitri gaN-aa-y che
........ she good poetess consider-Pass-3sg is
........ 'She is considered a good poet.'

...... b. *e-Ne saari kavayitri gaN-aa-vaa maaND-y-U
........ she-Erg good poetess consider-Pass-Inf begin-Pst-Nsg
........ 'She began to be considered a good poet.'

...... c. e saari kavayitri gaN-aa-vaa maaND-i
........ she-Nom good poetess consider-Pass-Inf begin-PstFsg
........ 'She began to be considered a good poet.'

In its behaviour the Gujarati phasal verb maaND shows striking parallels with the Kashmiri phasal hye 'take; begin'(fn 3). For instance the blocking of the ergative case in experientials (21) and passives (22) is found there, too: [Kashmiri data from Hook and Koul (1987)]

(21) a. *kooryi hyot temyis khar-un .................. [compare (18)]
........ girl.Erg took him.Dat annoy-Inf.Def

........ b. kuur hyetsi-n temyis khar-iny ......'The girl began to irritate him.'
...... girl.Nom took.Fsg-3sg.Erg him.Dat annoy-Inf.Fsg

(22) a. *tamyi hyot jaan sheeyir maan-ini y-un ............ [compare (20)]
...... she.Erg took good poet consider-Inf.Abl come-Inf.Def

........ b. swa hyetsi-n jaan sheeyir maan-ini y-iny
...... she.Nom took.Fsg-3sg.Erg good poet consider-Inf.Abl come-Inf.Fsg
...... 'She began to be considered a good poet.'

As these parallels cannot stem from common inheritance (the lexical items in question have distinct OIA sources), we must be dealing with underlying universals of human language that are able to surface in peripheral situations when the operation of categorical morphosyntactic rules is suspended.

Endnotes

* This paper was first drafted in Baroda (a.k.a. Vadodara) during the summer of 1994. The sojourn in Gujarat of one of its co-authors was made possible by the American Institute of Indian Studies. The transcription used for Gujarati in this paper is the one usually found in the linguistics literature. In this system reduplicating a symbol denotes contrastive length. The voiceless palatal fricative is indicated with a digraph sh. T, Th, D, Dh, and N are retroflex stops; L, a retroflex lateral; and S, a retroflex fricative. Capitalization of vowel symbols indicates nasality. Abbreviations include the following: ..........Abl.........................ablative ....Loc.....................locative
..........Acc......................accusative ....M....................masculine
..........Caus................causative affix ....N.........................neuter
..........CP..........conjunctive participle ....NF...................non-finite
..........CTF.........counterfactual mood ....Nom................nominative
..........Dat...........................dative ....Obl......................oblique
..........Def..........................default ....pl..........................plural
..........Emp.............emphatic particle ....PP...............past participle
..........Erg.........................ergative ....Pst....................past tense
..........F...........................feminine ....sg.......................singular
..........Fut............................future ....1....................first person
..........Gen.........................genitive ....2.................second person
..........Inf.........................infinitive ....3....................third person
1. We have looked at the listed works by Bhayani, Cardona, Kothari, Masica, and Shah et al. Cardona (1965: 5.20.3) observes some of the grammatical patterns we discuss here but without relating them to the typological literature on subject-marking in active-stative languages. Hook and Koul (1987) and Hook, Koul, and Koul (1987) are studies of parallel phenomena in Kashmiri.
2. However, the concept of abeyance (reminiscent of ideas in David Stampe's "Natural Phonology") cannot be directly applied to explain the nominative case in examples like (14).
3. Note the presence in the finite verb of an ergative suffix (-n) which serves as a null element or as a kind of place-holder for the ergatively-cased subject that one normally finds with transitive finite forms like hyot 'took; began'. See Hook and Koul (1987) and Hook, Koul, and Koul (1987) for detailed discussion.

References

Primary sources:

Secondary sources:

Draft of 5.January.1997. Comments welcome. Send e-mail to: pehook@umich.edu.

Back to main index page.

Back to list of publications.

Back to top.