amerindia n 16, 1991









Split ergativity in Shuswap Salish




Simon Fraser University




0 Introduction


In this paper we wish to present a set of observations that raise interesting questions about the status of the pronominal system in Shuswap[1]. There is a form /-s/ in Shuswap that occurs in several common syntactic constructions. It behaves as a third person marker in the grammatical system, but also appears cliticized to /-w-/ and has developed a broader function triggered by aspectual and discourse factors. This development is shared only with the closely related Thompson language (Thompson 1976) although Tillamook may have separately developed a similar system (Newman 1980). We provide an analysis in which Shuswap is shown to have a split ergative morphological case marking system. It is shown that this system is triggered by distinctions in person, aspect and clause type.


Shuswap /-s/ shows up as a 3rd person pronominal agreement marker in Wh-question, focus, and relative clause constructions.


(1) swέtỷ k-čntέs[2]

swέtỷ k čw-n-t--έs

who irr-punch-fc-tr-3abs-3erg

Whom did he punch?


(2) swέtq k-čntέms[3]

swέtq k-čw-n-t--έm-w-s

who irr-punch-fc-tr-3abs-pass-incompl-3nom

Who punched him?


In (1), it is the argument corresponding to the Patient that is questioned and the verbal morphology is identical to declaratives (3). In (2), it is the Agent argument that is questioned. We assume that the 3rd person marker appears as /-s,-έs/ or /-s/ depending on the phonological environment. The notional verb is grammatically intransitive, having been turned into a passive and is then extended with /-w-s/. It is similar, but not identical to the independent clause (4) which is an impersonal passive contruction.

(3) čntέs



He punched him.


(4) čntέm



He was punched.


The form /čntέm/ can also occur in Wh-constructions.


(5) swέtq k-čntέm

swέtq k-čw-n-t--έm

who irr-punch-fc-tr-3abs-pass

Who was punched?


Independent impersonal passives can never be formed with /-s/.


(6) *čntέms


Focus constructions with a pronominal argument in preverbal position exhibit a similar pattern. These constructions are always emphatic.


(7) nwi/s mε/ čntέs

nwi/s mε/ čw-n-t--έs

3emph exp punch-fc-tr-3abs-3erg

He's the one whom he punched.


(8) έčw/ lu/čnčέ[č]ms

έčw/ lu/ čw-n-sέ[č]m-s

Iemph deic punch-fc-tr-1acc-3erg

I'm the one whom he punched.


(9) nwi/s čntέms

nwi/s čw-n-t--έm-w-s

3emph punch-fc-tr-3abs-pass-incompl-3nom

He's the one who punched him.


(10) έčw/ čntέms

έčw/ čw-n-t--έm-w-s

Iemph punch-fc-tr-3abs-pass-incompl-3nom

I'm the one who punched him.


In (7-8) the preverbal emphatic pronoun corresponds to the Patient. The verbal is identical to the corresponding non-focus clauses in (11):


(11) čntέs He punched him.

čnč[č]ms He punched me.


In (9-10) the preverbal emphatic pronoun corresponds to the Agent. The predicate has been made passive and is then extended with /-w-s/. Non-focus clauses are shown in (12).


(12) čntέs He punched him.

čntέ[t]n I punched him.


A third construction that displays the same pattern is one that has properties resembling relative clauses.


(13) člx*mstέ[t]n F-sqέlmx l-m-čntέs

c-lx*-m-st--n F-sqέlmx l-m-čw-n-t--έs

hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-3abs-3erg

I know the man whom he punched.


(14) člx*mstέ[t]n F-sqέlmx l-m-čntέx

č-lx*-m-st--n F-sqέlmx l-m-čw-n-t--έx

hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-3abs-2nom

I know the man whom you punched.


(15) člx*mstέ[t]n F-sqέlmx l-m-čntέms

č-lx*-m-st--n F-sqέlmx l-m-čw-n--tέm-w-s

hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-3abs-pass-incompl-3nom

I know the man who punched him.


(16) clx*mstέ[t]n F-sqέlmx l-m-čnčs

c-lx*-m-st--n F-sqέlmx l-m-čn-n-t-ss

hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-2acc-3erg

I know the man who punched you.


The relative clauses (13-14) are headed by a nominal that corresponds to the Patient internal to the clause. All of the persons are marked on the predicate of the relative. In (15-16) the relative clause is headed by a nominal that corresponds to the internal Agent. In (15) the predicate is made intransitive by passive /-έm/ and the /-w-s/ form is added and in (16) there is a pronominal agreement suffix on the predicate corresponding to the nominal head.


These constructions can also occur without nominal heads.

(17) člx*mstέ[t]n l-m-čntέs

c-lx*-m-st--n l-m-čw-n-t--έs

hab-know-middle-caus-3abs-1nom deic-compl-punch-fc-tr-3abs-3erg

I know whom he punched.


(18) člx*mstέ[t]n l-m-čntέms

č-lx*-m-st--n l-m-čw-n-t--έm-w-s

hab-know-middle-caus-3abs-1nom deic-compl-punch-fc-tr-3abs-pass-incompl-3nom

I know who punched him.


The pattern of person marking in relative clauses parallels that of Wh-questions and focus constructions.


Jacaltec (Craig 1976) a Mayan language has a similar construction in which extracted subjects[4] behave differently than objects. Transitive clause structure is shown in (19).


(19) xil naj ix

saw cl/he cl/her

He saw her.


In (20-22) third person pronominal objects are extracted in question, cleft, and relative clause constructions and there is no mark on the verb.


(20) mac xil naj

Who saw cl/he

Who did he see?


(21) ha' ix xil naj

cleft cl/she saw cl/he

It's her that he saw.


(22) wohtaj ix xil naj

I know cl/her saw cl/he

I know the woman that he saw.


In (23-25), the reference to subjects requires additional marking on the verb /x'il-ni/.


(23) mac x'il-ni ix

Who saw-suffix cl/her

Who saw her?


(24) ha' naj x'il-ni ix

cleft cl/he saw-suffix cl/her

It's he who saw her.


(25) wohtaj naj x'il-ni ix

I know cl/he saw-suffix cl/her

I know the man who saw her.


Before turning to the syntactic properties of the Shuswap constructions we provide an overview of the pronominal system.


1.0 The Pronominal System


Person is referenced by sets of affixes and clitics that occur on the predicate. We first discuss first and second person.


Intransitive clauses are referenced for person by pronominal clitics :


(26) '/ε[/]k-kn


I go/went.

(27) '/εk-k


You go.


(28) '/εk-kt

go-1pl incl nom

We (including you) go.

(29) '/εk-kux

go-1pl exl nom

We (but not you) go.

(30) '/εk-kp

go-2pl nom

You (pl) go.


The clitic paradigm is shown in the following chart.



Singular Plural

1st -kn -kt (incl)

-kux (excl)

2nd -k -kp


Transitive clauses are referenced for person by pronominal suffixes. The predicate is suffixed with /-t/ and the pronominal subject follows the object marker.


(31) čnčέ[č]mx


You punch me.


(32) čnčn


I punch you.


The pronominal subjects are shown in the following chart.[5]


Singular Plural

1st -(έ)n

2nd -(έ)x -(έ)p


The pronominal objects are given in the following chart.


Singular Plural

1st -s(έ)m- -(έ)l-

2nd -s()- -()lm-


First and second person are marked according to a nominative/accusative system. In intransitive clauses first and second person is supplied by clitics. In transitive clauses the first and second person markers are suffixes.

The nominative clitic paradigm behaves differently from the suffixal paradigms. Evidence that these are different paradigms comes from the question suffix /-n-/ :

(33) m-'/εk-n-k

Did you go?


(34) čnčn


Did he hit you?


The question suffix /-n/ intervenes between the predicate stem and the subject clitic in intransitive clauses whereas it is suffixed after the pronominal objects and subjects in transitive clauses.


1.1 Third Person Marking


In this section we establish an ergative/absolutive system for third person. The behaviour of third person pronominals in intransitive clauses is shown in the following examples :


(35) '/εk-

He goes.


(36) wik-m-

He is looking.


(37) wik-t--m

He was seen.

Clauses (35-37) are intransitive and the third person marker is /-/. The following clauses are transitive.


(38) wik-t--s

He saw it


(39) wik-t--s F-sk'lέp

see-tr-3abs-3erg deic-coyote
He saw the coyote.

(40) wik-t--s F-sqέlmx F-sk'lέp

see-tr-3abs-3erg deic-man deic-coyote
The man saw the coyote.


Third person behaves according to an ergative/absolutive system. Subjects of transitive clauses in (38-40) are marked with /-έs,-s/, the ergative. Third person objects are marked with /--/, the absolutive, parallel to subjects of intransitive clauses (35-37).


Shuswap then has a split ergative system for person. First and second person are nominative/accusative and third person is ergative/absolutive. The common types of split systems recognized are on the basis of person marking, aspect and clausal type. Having established the presence of a split system based on person, we now turn to aspect.


1.2 Aspect


Split ergative systems frequently make a distinction based on aspect. Often in the completive aspect third person behaves as an ergative/absolutive whereas in the incompletive aspect third person behaves as a nominative/accusative. Clauses (41-43) are marked with the completive prefix /m-/.


(41) m-'/ε[/]k-kn


I went.


(42) m-'/εk-k


You went.


(43) m-'/εk-


He went.


The completives (41-43) are all independent clauses. First and second person are nominative and third person is absolutive. Incompletive constructions have the particle /w/εx/ in predicate initial position and the notional verb is a dependent clause.


(44) w/εx F-'/ε[/]k-w-n

exist deic-go-incompl-1nom

I am going.


(45) w/εx F-'/ε[/]k-w-x[uxw]

exist deic-go-incompl-2nom

You are going.


(46) w/εx F-'/ε[/]k-w-n

exist deic-go-incompl-3nom

He is going.


Notice that the pronominal markers are attached to /-w-/ in (44-46) parallel to being attached to /-k-/ in (41-43).[6] The incompletive pronominal paradigm is given in the following chart.


Singular Plural

1st -w-n -w-t

2nd -w-x -w-p

3rd -w-s -w-s


The behaviour of third person would follow from an account of split ergativity based on aspect. In (47) the third person of the completive clause is absolutive whereas in (48) the third person subject of the incompletive clause is nominative.


(47) m-'/εk-


He went.


(48) w/εx F-'/εk-w-s

exist deic-go-incompl-3nom

He is going.


Incompletives have a strong active interpretation as shown by the derived nominals in (49-50). In (49) the predicate /k'lms/ is focussing on the action of 'making' and the Patient argument is an oblique. In (50) it is the action that is questioned by the form /kεnm/.


(49) w/εx F-k'lms t-misx '

exist deic-make-middle-incompl-3nom deic-basket

He is making a basket./He's a basket-maker.


(50) /εx k-kέnm-ux

exist irr-do-w-2subj

What are you doing?


Constructions formed with the particle /w/εx/ do not force incompletive aspectual agreement as can be seen by the following pair.


(51) w/εx F-pp-m-

exist det-hunt-middle-3abs

There's a hunter around.


(52) w/εx F-pp-m-w-s

exist det-hunt-middle-incompl-3nom

He's hunting.


Third person splits along aspectual lines with completives using an ergative/absolutive system whereas the incompletives employ a nominative/accusative system. Incompletives also exhibit behaviour paralleling subordinate clauses. We now go on to look at the status of these clauses.



1.3 Clausal Type


Split ergative phenomena often appear on the basis of clausal type. Third person marking behaves as an ergative/absolutive in independent clauses and as a nominative/accusative in subordinate clauses. Independent clauses are shown in (53-54).


(53) '/εk-

He goes.


(54) wik-t--s F-sk'lέp

see-tr-3abs-3erg deic-coyote
He saw the coyote.

Third person marking in independent clauses is ergative/absolutive. The subject of the intransitive clause (53) is /-/, the absolutive, parallel to the object of the transitive clause (54). On the other hand the third person subject of (54) is /-έs/, the ergative.

Subordinate clauses are underlined in (55-57).


(55) / kxčn t-spqpέq ε qwnέnux

/ kx čn t-spqpέq ε qwnέn-w-x

expect give-2acc-1nom deic-berries if want-incompl-2nom

I'll give you some berries if you want.


(56) č-ppqw-stn yFέy lu/ w/έxws q'wiylxws

čppqwst--n yFέy lu/ w/έx-w-s q'wiylx-w-s

hab-watch-caus-3abs-1nom that one part exist-incompl-3nom dance-incompl-3nom

I was watching him when he was dancing.


(57) sq'lέlnmstn lu/ yFέy l-sčnms

sq'lέlnm-st--n lu/ yFέy l-sčn-m-w-s

listen-caus-3abs-1nom part that one deic-sing-middle-incompl-3nom

I was listening to him when he was singing.


The subordinate clauses are marked in the same way as incompletives, taking a member of the clitic paradigm suffixed to /-w-/parallel to clause (58).


(58) w/εx F-k'l-m-s t-misx

exist deic-make-middle-w-3nom deic-basket

He is making a basket./He's a basket-maker.


The split in person marking systems can be seen in contrasting dependent clauses (59-60) which take ergative/absolutive marking with (61) which takes nominative/accusative marking.


(59) člpmstέtn F-sqέlmx l-m-qwčέč

c-lp-m-st--έ[t]n F-sqέlmx l-m-qwčέč-

hab-know-middle-caus-3abs-1nom deic-man deic-compl-leave-3abs

I know the man who left.


(60) člpmstέtn F-sqέlmx l-m-čntέs

c-lp-m-st--έ-[t]-n F-sqέlmx


hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-3abs-3erg

I know the man whom he punched.


(61) člpmstέtn F-sqέlmx l-m-čntέms

č-lp-m-st--έ[t]-n F-sqέlmx l-m-čw-n-t--έm-s

hab-know-middle-caus-3abs-1nom deic-man deic-compl-punch-fc-tr-3abs-pass-3nom.

I know the man who punched him.


A split in clausal type has been reported for Halkomelem with pronominals behaving as ergative/absolutives in dependent clauses and as nominative/accusatives in subordinate clauses (Gerdts 1988). The Shuswap facts however give credence to an analysis based on aspect.


The following chart for third person displays the distribution of morphological case marking :


Nom/Accus Erg/Abs


Third +

Non-Third +


Completive +
Incompletive +

Clause Type

Independent +
Dependent + +
Subordinate +


The chart shows a complementarity of ergative/absolutive and nominative/accusative systems for person and aspect. Clause types appear to split however, taking nominative/accusative for subordinates and agentive dependent clauses, and taking the ergative/absolutive system for non-agentive dependent clauses.


2.0 Syntactic Properties of Wh-Constructions


In this section we examine the syntactic properties of Wh-constructions. An analysis of focus constructions and relative clauses will follow from this treatment. Active and impersonal passive constructions are repeated in (62-63).


(62) swέtq k-čntέs

swέtq k-čw-n-t--έs

who irr-punch-fc-tr-3abs-3erg

Whom did he punch?


(63) swέtq k-čntέm

swέtq k-čw-n-t--έm

who irr-punch-fc-tr-3abs-pass

Who was punched?


Both of these constructions are Patient-centered.[7] Syntactically, Wh-constructions are formed by extracting the absolutive. The Wh-question form /swέtq/ occurs in pre-verbal focus. In (62) both the Agent and Patient are morphologically present whereas in (63), the Agent argument is suppressed and the clause has passive marking with /-έm/. The construction is an impersonal passive.


In (64) Patient-centered Wh-constructions are compared with independent clauses. The person markings for both sets are identical.


(64) swέtq k-čntέ[t]n Whom did I punch?

swέtq k-čntέx Whom did you punch?

swέtq k-čntέs Whom did he punch?

swέtq k-čntέm Who was punched?

čntέtn I punched him.

čntέx You punched him.

čntέs He punched him.

cntέm He was punched.


The Wh-construction in (65) is Agent-centered.


(65) swέtq k-čntέms

swέtq k-čw-n-t--έm-w-s

who irr-punch-fc-tr-3abs-pass-3nom

Who punched him?


Shuswap permits absolutives to be directly questioned as in (62-63). In (65) it is the ergative that is questioned. The construction is made passive and then the form /-w-s/ from the incompletive aspectual paradigm is suffixed to represent the Agent.


The situation is similar to the Coast Salish language Halkomelem (Gerdts 1988).


(66) nw ni lm-θ-am-/e.n/

2emph aux look-tr-2obj-1subj

It's you that I looked at.


(67) /.θ ni q'wqw-θ-m/-s

1emph aux club-tr-1obj-3subj

It's me who he clubbed.


(68) / ni q'wqw-t(*q'wqw-t-/.n/)

1emph aux club-tr

I'm the one who clubbed it.


(69) nl l slni/ ni q'wqw-t(*q'wqw-t-s)

3emph det woman aux club-tr

It's the woman who clubbed it.

Objects (66-67) and obliques have pronominal copies whereas subjects (68-69) do not. There is no shift in transitive marking.


In (70) Shuswap Agent-centered Wh-constructions are compared with independent clauses.


(70) swέtỷ čnčέ[č]ms Who punched me?

swέtỷ čnčs Who punched you?

swέtỷ čntέms Who punched him?


cnčέ[č]ms He punched me.

čnčs He punched you.

*čntέms/(čntέs) He punched him.


In Agent-centered Wh-constructions (70) with first and second person Patients it is clear that there is a pronominal copy of the Agent manifested as a pronominal agreement suffix on the verb. The person marking is exactly the same as in independent clauses. In the third person Agent-centered construction the form /k-čntέms/ is grammatically marked as passive and is morphologically the same as the independent clause /čntέm/. There is, in addition, the /-w-s/ ending. In Lummi, another Coast Salish language (Jelinek 1987) the Patient-centered paradigm (71) is in all relevant respects the same as in Shuswap.[8]


(71) c-x*či-t-n the one that I know


c-x*či-t-x the one that you know


c-x*či-t-s the one that he knows



In (72) the Agent-centered paradigm is given.



(72) c-x*či-t-oNs the one that knows you/me


c-x*či-t- the one that knows him



Jelinek assumes that because there is no ergative argument in the Agent-centered construction, there is no absolutive argument in the Patient-centered construction.


In (73) the corresponding Shuswap forms with Patient-centered interpretations are given.


(73) l-č-lx'-m-st--έ[t]n the one I know


l-č-lx'-m-st--έx the one you know


l-č-lx'-m-st--έs the one he knows



In (74) forms with Agent-centered interpretations are given.


(74) l-č-lx'-m-st-sέ[č]m-s the one who knows me


l-č-lx*-m-st-s-s the one who knows you


l-č-lx*-m-st--έm-w-s the one who knows him



In both the Patient-centered and Agent-centered paradigms, pronominal forms are fully represented. To summarize, Shuswap has the syntactic strategy of extracting absolutives in focus constructions and has innovated a mechanism based on aspect to extract ergatives. The Shuswap facts suggest that there are no gaps involved in the expression of focus.


3.0 Conclusion


Although there is strong evidence of a grammatical split in the Shuswap pronominal system, the appearance of /-s/ as a third person Agentive in dependent clauses still requires an explanation. It has been observed (Dixon 1979) that direct/inverse grammatical systems are often mistaken for split ergative ones. One promising line of investigation would be to view the Shuswap system as reflecting an animacy hierarchy, certainly well recognized in being involved in split systems (Silverstein 1976). This line has been followed for the Coast Salish language Lummi (Jelinek and Demers 1983).


In Shuswap there is evidence of an animacy hierarchy with inclusive first person plural pronominal subjects. The Shuswap form for 'we saw him' occurs as a passive and has the alternate interpretation 'he was seen'.


(75) wik-t--m


He was seen./We (incl) saw him.


Evidence for an animacy hierarchy suggests that Shuswap may have behaviour parallel to direct/inverse person marking systems well recognized for the Athabaskan language family. Whistler (1985) has looked at direct/inverse marking in Nootkan suggesting that this person marking system typologically may be more widespread in the area. Jelinek and Demers have explored this possiblility for the Salish language Lummi (1983).

There are some intriguing similarities in the Navaho paradigms in (76-77) (Jelinek 1987) and the Shuswap paradigms in (78-79).


(76) yiztal He kicked him.
yiztal-e the one who was kicked...

'ashkii yiztal He kicked the boy.

'ashkii yiztal-e the one who kicked the boy...


(77) biztal[9] He got kicked by him (approx).
biztal-e the one who kicked him...

'ashkii biztal He got kicked by the boy.

'ashkii biztal-e the one who got kicked by the boy...


(78) m-čntέs He punched him.

l-m-čntέm the one who was punched...

m-čntέs F-sqέlmx He punched the man.

l-m-čntέms F-sqέlmx the one who punched the man...


(79) m-čntέm He got punched.

l-m-čntέms the one who punched him...

m-čntέm F-sqέlmx The man got punched.

m-čntέm t-sqέlmx He got punched by the man.

l-m-čntέms F-sqέlmx the one who punched the man...

l-mcntέm t-sqέlmx he one who got punched by the man...


The two languages have several features in common: the predicate is fully marked for person, and there is an animacy hierarchy banning certain argument co-occurrences.


Finally it should be pointed out that the behaviour of /-s/ is not triggered strictly on grammatical grounds. It also occurs frequently in spatial and temporal deixis.


(80) tl/έn/ kx-č-n-s

from this expect give-2acc-1nom-3nom

I'll give you some from this.


(81) kεnm k-sčlpmst--έx pnhέ/nyFέy qwčέč-s

do irr-know-3abs-2nom when that one left-3nom

Do you know when that person left?


In (80) the predicate /kx-č-n/ is fully marked for its arguments and the /-w-s/ is triggered by spatial deixis as it is with the temporal deixis in (81).


Kuipers (1974) has noted the frequent occurrence of /-s/ in texts and that the texts often have forms with and without the suffix. This is suggestive that /-s/ may have diverse functions in discourse styles. These properties suggest that an analysis based on aspect has some credibility. Given that /-s/ has such strong active readings it may be functioning as a reference tracking device in discourse, somewhat akin to switch-reference systems. The Shuswap form /-s/ appears to be what Nichols and Woodbury (1985) call an emergent linguistic category, one with "tendencies in discourse which are almost, but not quite, rigid and grammaticalized".[10]





craig, Colette G.

1976 "Properties of Basic and Derived Subjects in Jacaltec" in Subject and Topic, ed. Charles N. Li. New York: Academic Press.


dixon, R. M. W.

1979. "Ergativity". Language 55:59-138.


gerdts, Donna B.

1988. Object and Absolutive, in Halkomelem Salish. New York : Garland Publishing Inc.


jelinek, Eloise

1987. "'Headless' Relatives and Pronominal Arguments: A Typological Perspective" in Native American Languages and Grammatical Typology: Papers from a Conference at the University of Chicago, eds. Paul D. Kroeber and Robert E. Moore. Bloomington: Indiana University Linguistics Club.


jelinek, Eloise and Richard A. Demers

1983. "The Agent Hierarchy and Voice in Some Coast Salish Languages". IJAL 49 :167-185.


kuipers, Aert

1974. The Shuswap Language. The Hague: Mouton.


newman, Stanley

1980. "Functional Changes in the Salish Pronominal System". IJAL n 46:155-167.


nichols, Johanna and Anthony C. Woodbury

1985. Grammar inside and outside the clause: Some approaches to theory from the field. Cambridge: University Press.


silverstein, M.

1976. "Hierarchy of Features and Ergativity" in Grammatical Categories in Australian Languages, ed. R.M.W. Dixon, pp. 112-71. Atlantic Highlands, N.J.: Humanities Press.


speas, Margaret

1990. Phrase Structure in Natural Language. Dordrecht: Kluwer.


thompson, Laurence C.

1976. "Salishan and the Northwest" in The Languages of Native America: Historical and Comparative Assessment, eds. L. Campbell and M. Mithun, pp. 692-765. Austin: University of Texas Press.


whistler, Ken

1985. "Focus, perspective, and inverse person marking in Nootkan", in Grammar inside and outside the clause: Some approaches to theory from the field, eds. Nichols, Johanna and Anthony C. Woodbury, pp. 227-265. Cambridge: Cambridge University Press.

[1] Shuswap is spoken on the Interior Plateau of British Columbia and is the northernmost member of the Interior Salish language family. It is a predicate-initial language with a system of pronominal person marking on the predicate. Shuswap is spoken with several minor dialectal differences. The data in this paper is representative of the Deadman's Creek/Kamloops area. We would like to thank Leslie Jules of Kamloops, Mona Jules of Chu Chua, Annie-May Jules, Sam Camille, and Basile Deneau of Skeetchestn who have helped us to understand their language. We would also like to thank Henry Davis, Ewa Czaykowska-Higgins, Mandy Jimmie, Dale Kinkade, Aert Kuipers, and Paul McFetridge who provided valuable comments on this paper. A version of this paper appeared in Working Papers for the XXV International Conference on Salish and Neighbouring Languages, Vancouver, B.C., 1990.

[2] The following abbreviations have been used: abs (absolutive), acc (accusative), caus (causative), cl (classifier), compl (completive), deic (deictic), emph (emphatic pronoun), erg (ergative), exp (expectational), fc (full control), hab (habitual), incompl (incompletive), irr (irrealis), inv (invisible), nom (nominative), pass (passive), qu (question), [...] (reduplication), tr (transitive).

[3] Evidence for positing the /-w-/ comes from the following:




He hit me.

The [] is expected to delete in unstressed environments. We therefore posit a /-w-/. This is supported by data from Thompson were the equivalent form is [-us]. Also in intransitive constructions that are not derived by passive the suffix triggers labialization.

w/εx F-λ'/έk-w-s

exist det-go-incompl-3nom

He's going.

[4] The Mayan facts are described in terms of subject and object extraction. Throughout this paper we remain neutral regarding the configurational status of subject and object. The facts can alternatively be accounted for by referring to morphological case or to thematic relations such as Agent and Patient.

[5] Shuswap does not permit 1st person plural ergatives. The equivalent of 'We punched him' would be expressed by a passive. An example is given in (74).

[6] Phonological developments often obscure this grammatical process.

[7] The terms Agent-centered and Patient-centered are used in Kuipers (1974).

[8] We assume that // corresponds to /3/ in Jelinek's orthography.

[9] The bi- prefix has been argued to be a topic marker. See Speas (1990).

[10] Nichols and Woodbury, pg.8.