31 32 An Aeent Iiierarchy in Some Coast Salisli Languages - -An Aeent Ilierarchr in Some Languages Eloise Jelinek University of ,\rizona, Tucson Richard A. Demers Eloise Jelinek Dear Colleagues, Richard A. Demers The following paper is an account of :;aps (non-occurring University of Arizona, Tucson sentences) in transitive and passive sentences in certain Coast 0. Introduction Saliah languages. The account makes reference to an agent hier- 1. The A~entIIierarchy and Pri-ativity archy, a different form of which appears in the languages we con- 2. sidered (Lummi, ~lallam., , and Squamish). 3. Squamish The question immediately arises, then, as to how pervasive this 4. Lushootseed phenomenon may be in Salisll in general. Since :iootka apparently 5. Conclusions has a similar feature ("Inverse person narking in Nootkan". Whistler

(1980), ns.),there is a possibility that the agency hierarchy is an 0. Introduction. Studies of -any of the Coast Salish lanppa~eshave areal feature. Since we haven't yet been able to get a hold of revealed certain classic problems in the analysis of active (transitive)

Whistler's paper, we don't yet know how similar the Nootkan features sentences and their passive counterparts.' These problems stem iron the may be to the Salish features. If the Interior Salish languages have fact that there are caps in the transitive and passive verb paradips,

similar agency properties, the source of the areal feature on the and suppletion between tkem. A question which has arisen is: can a sen-

Coast may be Salistl. tence be termed "passive" if it has no transitive counterpart? In

We would therefore be interested in hearing from you at the Squamish. for example, transitive sentences with a second person pro-

Salish Conference regarding the following facts concerning languages nominal patient (the msn helys you) are excluded, and a passive sentence in which you have expertise: (pu are helped by the nan> ~ustbe employed. Across Innmeyes, nassive

1) Are there any gaps in the transitive and passive sentence sentences have been described as constructions that make it possible to

paradigms? If so, what are they? state that some patient has been affected by some action, without

2) Are there any restrictions on the order of full noun phrases? identifying the acent of thot action. Yet in Lushootseed, for exarple,

That is, can inanimate nouns serve as subjects of sentences it is possible to specify a full noun agent (i.e., not a prononinal

with animate or human objects? (e.g. The rock hit the man) scent) only by employ in^ a passive construction, if the natient is also

Thank you. Eloise Jelinek a full noun. These problems in the analysis of a variety of Coast

Richard Ijemers 33 2) Lummi passive sentences Salish lansages h~vebeen pointed out by Hess (1973), Thompson and a-1 ' 'You are b-1 'I am know All Thompson (1971). Hukari (1976a), and Kuipers (1967) among others. known by me' by you'

the Coast Salish languages apparently exclude certain transitive and a-2 'It is known b-2 * 'It is known by me' by you' passive sentence types, but it is of interest that the particular sen- a-3 ' 'The man is b-3 'The man is tence types vary with each language. What is im~ortanthere is that the known by me' known by you' paradigm gaps are not lexically determined; that is. they do not reflect verb classes or specific verbs that are excluded from either transitive c-1 +i-t-q-san/ 'I/you are d-1 xbi-t-q-sari/ yo you are 6xW known' (by sxW a ca known by or passive sentences, as is frequently the case in other language someone swoy9qa9 the man'

families. The excluded sentence tynes may be characterized syntac- c-2 $i-t-q 'It is known' d-2 +i-t-rl *Itie know (by someone) a ca svay9qa3 by the tically, with reference to the argument types thnt can occur with the man'

transitive or intransitive verb of the sentence.2 Consider the follow- , c-3 zEi-t-q ca 'The man is d-3 Sci-t-q ca 'The boy is swa y'qa3 known' (by swipq6'aZ a known by ing (main clause) transitive and passive paradips from Lummi: I someone) ca sway-qa' the man'

1) Lummi transitive sentences In (11, -t is the transitivizing suffix. In (2), -9 is an intransi-

a-1 xsi-t-oqas-san 'I know you' b-1 $i-t-oq--sxV #YOU know tivizing suffix; sentences with -t and -3 have a passive interpretation. me ' In cia-1) and lb-l), -*is an undifferentiated pronominal suffix for a-2 $1-t-san 'I know it' b-2 +i-t-sxW 'You know it' both first and second person sinplnr accusative. In (la) -sari marks

a-3 #i-t-san ca I1 know the b-3 $i-t-sxW ca 'YOU know first percon nominiltlve, and -& (lh) marks second person nominative. sway7qa3 man' evay3qa9 the man' Both nominative and accusative third person prononinal are not overtly

c-1 ' 'ile/she knous d-1 ' 'The man perked; the 5 suffix in (lc) and (ld; will be accounted for below. The you/me ' knows me/ you' underlined fipaces in (1)and (2) show the excluded sentence tgpes in the

c-2 $1-t-6 'He/she knows d-2 'The man transitive and passive paradigms in Lunmi. Note thnt the speaker can it' know8 him' choose between an active sentence or its passive counterpart only %hen c-3 $i-t-s ca 'He/she knows d-3 $1-t-s ca svav7qa3 ca sway"qa7 the man' suiqqo9a~ the transitive sentence has third prrfion arflurento. There is a question. 'The man knows the boy' then, as to whether ths contrast between active versus passive sentences

is relevant in sentences containinr first and second Derson armnents.

We will return to this question in section 5, below. 35 36

In this article we will survey transitive and passive verb pnra- patient arqment of a transitive verb end the subject of an intransitive

dims in sone Salish lan~uages,and offer an explanation for the verb receive the same case marking (absolutive), while the agent areu-

paradiepi gaps and the si~ppletionhetween them in terms of a gramm~tical ment of a transitive verb receives a different case marking (ergative),

feature shared by these lanpages: the atent hierarchy. An agent as shown in Fiwe 1: hierarchy in a 1anp;uage defines the rank of various elements that may serve as agents in transitive sentences; these rules require that the -- Transitive Agent Patient Lrent Patient, / element of kighest rank in the agent hierarchy in a sentence be the ,/' Intransitive Subject / S::hject /" subject of that sentence. in a transitive sentence, the agent is sub- I_- /..- . ,,,' ject; in a passive sentence, the single non-oblique armrnent is subject. Figure 1 Any tr3nsitive or passive sentence where the subject is not the element Some 1anp;uages do not nark case, but have ergative marking on verbs; of hiphest rank in the sentence is excluded. We will show that the this is the siti~ationin some of the Salish lnn$uages discussed here. apent hierarchy accounts also for another often-noted attribute shared In order to discuss this ercative marking on verbs, we will need first by those languages: the fact that in transitive sentences with a single to nake reference to the notion of an er~ativesplit. Dixon (1979) full notin ar~~nent,that full noun is always internreted as patient. emohasizes that lansin~estermed erpative iysniably show sone Finally, i.e will show the corJection between the ncent hierarchy and nominative/accusative features as well as erczitive/absolutive features. another rrannatical feature conron to these lanma~es: er~ativity. either in case-marking or in verbal inflectior.. For example, in Dyirbal,

1. The !.Tent 3ierarchy and Ermtivity. We ray define er~ativity pronoun6 show n~minative/nccusative rnarkinr, *'L.ile nouns show erptive/ hriefly as follows: Transitive verbs sre two-place predicates: t\ey absolutive mrkin~;thin kind of erytive srllt is t:-picnl of er~ative have an apent and a patient =;-u~ent. Intransitive verbs (a class which lnnwnges. Silverstein (1976) proposed a f-r.ctionn1 (or semantic)

includes passives) are one-olnce pre,:icntes; they have a single non- explanation of such erFative splits; he shci-ec: -l:?t it is possible to oblique ~rr~nent,the subject. In noninntive/accusative case markine specify certain semantic features hv neans cf which various terns that

systnn:;, '.he scent zr,-u-~?r.t cf R transitive verb and t!:e subject ar~ncnt may function 8s verbal arrunentn nnv he ranked ?;cording to the likeli- of nn intransitive verb rccel-ie tte sri-e cane -nrking (nominative), while hood of their servin~as arents in transitire :lauses. In conversation, th~?:stlent arvment of a transitive verb receives a different case the speaker nnd hearer are respectively flrst snc second nerson. They markin,: (accusative). In er~ative/absolutive case narkinp; systems, the are animate beings who see themsrlves as rossessed of a will and at 37 that certain constructions in Lumi and Lushootseed have ergative nroper- least to some extent in control over their actions; therefore, the ties. In all these languages, an er~ativesplit with respect to person probability of first or second person appearing as agent in a sentence appears. We will look at the Lumi case first. is very high. In discourse, reference is made to third persons, beings,

and objects some of which are inanimate and non-volitional, and there- 2. Lummi. The paradigms piven in (1) and (2) above shov that in Lunmi fore not likely to be a~entsin a transitive sentence. We may thus there are first and ~econdperson nominative and accusative pronominal recognize a 'potentinlity of agency' scale vhich is represented dia- elements. Nouns have only one non-oblique case; there is no nominative/ panmatically as follows: (Dixon 1979:85-86) accusative marking on nouns, as shown in (Id-3). When an intransitive

3) Demonstratives predicate has neither a flrst nor second person narker (nom.), it is Human Animate Inanimate 1st person 2nd person 3rd person Proper iY-'interpreted as having a third person pronominal subject; vhen a transi- pronoun nronoun pronouns nouns Common nouns .- tive verb has neither first nor second person narkers (nm. and acc.), likelihood of functioning as transitive agent it ifi interpreted as havinp; third person pronominal agent and patient the hierarchy piven here roufihly indicates the overall ". . . arwrnents. The following example sentences vill shov that the 2 suffix 'apency potential' of any p;iven NP. it is plainly most . . . that appears in (LC) and (Id) is not a third person pronominal enclitic: economical to 'mark' a participant when it is in the unaccus- 4) 8-1 sway?qa?-san 'I an a man' b-1 ye9-san *I wg to-ed role. . . . a nuqber of lanma~eshave 'split1 case a-2 svay7qa?-sxv 'You are a b-2 p7-sxW 'You so* na-king exactly on t5.i~principle: an 'ercative' case is used man'

with t.Tfi from the right-hand end up to ssq:e point in the middle a-3 svay?qa? 'He is a man' h-3 ye9 'He roes' of the heirarchy and an 'accusative' case from that point on, c-1 $i-t-san 'I know it'

over to the extreme left of the hierarchy. . . ." c-2 &i-t-swX @YOU know it'

Dixon observes that first and second persons are frequently of equiv-a- c-3 $i-t-s 'lie knows it'

lent ra~kin the hierarchy; as we will show, this is the case in d-1 $i-t-san ca .way9-qa7 'I know the can'

Salish. d-2 Xgi-t-s~Wca sway9"a9 'You know the man' 5:cntivity has bsen reported previously in Coast Sdish Innpagee. d-3 $i-t-s ca sway9qa9 'He knows the man' Kuirers ilO67) notes ernativity in Squ-lmish. liukari (19?Q)points out The -5 suffix that appears in (lc) and (Id) above does not appear in ergatire characteristics in Ralkomelem. and Gerdts (1980) documents (4 a-3) and 4b-3); therefore, the -2 is not a third person pronominal ergativity and ergative splits in that lanpo~e. We will shov here 39 40 3 enclitic. As (Id), (4c-3) and (4d-3) show, -5 appears on transitive 5) &i-t-a-san ca sway?qa? 'Do I know the man?*

verbs with a third person pronominal or full noun agent. Another possi- $i-t-a-sxw ca sway?"a? 'Do you know the man?'

bility would be that the -2 is a transitivizing suffix; however, the -5 $i-t-a-Y ca sway?qa? 'Do we know the man?'

that is common to all the sentences in (4c) and (4d) is the transitiviz- $i-t-6- a ca svayqqa? 'Does he/she know the man?'

Hukari (1976b) desi~atesthe -5 in Halkomelem a third ing suffix. In sum, the paradips in (1) show that Lunni has the following person transitive subject marker. This is correct; but since the restrictions on transitive sentence types: crucial feature of ergative systems is that they distinpuish between 6) Transitive sentences with a first or second persod transitive subjects (agents) and intransitive subjects, it seems prefer- pronominal patient are excluded, (cases (lc-1) a.d able to recognize the -2 as an er~ativemarker. Only third person forms (Id-1) ), unless first or second person is also the in L~inmi show ergativity. agent. (cases (la-1) and (lb-1) ) The er~ativesuffix on the verb nwks sentences with third person

7) Transitive sentences with a noninal acent and R third ?ronominal and full noun azents as the atypical constructionc: sen- person pronominal patient are excluded. [case (:d-?? ) trnces where an argument of low rank is agent. First and second persons,

who are norc likely to be agents, show nominative/accusative contrasts A consequence of (6) is :hat first and sccond yrononinal =xcvr.ents

in the pronominal elements. The er~ativesylit found in Lunmi is tberc- nay not bc outranked, and are eauivalent in rank. Their efluivalent

fore consistent with the ranking of nr@mcnts as potential agents rank as potential patients is connistent with the fact that first and

according to the went hierarchy. second person patients are marked by the same ac.hi(;uous accusative

The present analysis in which -s- is viewed as an er~ativ~marker suffix, as shown in (la-1) and (lb-1).

(and not n pronoun) is supnorted by the fact that the first nerson A consequence of (7) is that third person pronominal arplnents sincular nnd plural pronominal enclitics, (-E) and (-r),and the outrank full noun arplments. A verb with a transitivizins suffix alone second ;,rr:;on pronominal enclitic (-& 5.ave distributional pro?ert.isn (as in (lc-2) 1, means m,(("He/she knars it"). When a sin~lefull

that are distinct from those of the proposed erjintive marker. The noun arpment appears with n transitive verb, as in (lc-31, the full

?ror~o;~innlrnclitics always appear to tie ri~btof the yes/no question noun arpnent is inttqreted as patient, since 3 outranks N and 3 t?.ere- mor:>heme --9-, whereas the n.uestion marker anpears to the richt of the fore nust be agent. The a~enthierarchy, then, accoiints for the fact ercntive marker. These distributional facts are illustrated in (5): that sin~lefull nouns are always absolutive: the subject of an in-

transitive verb or the object of a transitive verb. %is interpretation 42

of transitive sentences with a single nominal argument is apparently We conclude that Lummi has the follow in^ apnt hierarchy:

shared by dl the Coast Salish languages. The only instance in which a 10) Lummi Agent Hierarchy: 1 and 2) 3) ti

non-ohlique noun in Lummi can represent an agent is when the patient is This ranking has the fiyntactic correlate that the arxunent of highest

also represented by a noun; that is, a full noun argument may outrank rank must be subject of the sentence. Firfit and sec@r,d persons are of

only another full noun argument. Put another way, transitive sentences equal rank. If neither of these corresponds to the agent referred to

where a noun represents the agent are excluded, unless the patient is in a sentence, then some third person prononinal is more likely to be

also represented by a noun. acent than some full noun, which may be an inaqimate, nonvolitional

The restriction on passive sentences in Lumi is: 'object'. The sentencefi where first or second person is not the agent

8) Passive sentences with a first or second person pronominal are the marked cases; indicated either by the ergative parkine on the

aeent (by,by you) are excluded. verb in transitive sentences, or by a passive construction when the patient outranks the ncent. As the examples in (2c) show, passive sentences with an unspecified

third person pronominal agent are permitted. If only one first or 3. Squamish. Data riven in Kuipers (1967) shows that Squamish exhibits second person prononinal arpment appears in a sentence, it must be the hierarchical rankine of afents, but that thin rankinp: differs somewhat

subject, since these armments outrank third person pronominal and full from that of Lummi. Table 1 shows that Squamish, unlike Lummi, does not

noun arements. Koninal a~cntsin passive sentences pay be marked exclude transitive sentences with a first permn ~atient. ?he corre-

ohliquely by prepositional phrases with the preposition a, as shown in sponding translations of Table 1 are riven in %ble 2. 3ut Squamish

the examples in (2d). It is possible to identify first and second (like Lummi) has tho follouing restriction on transitive sentences: person agents emphatically in Lumi passive sentences by use of oblique 11) Transitive sentences with a second person prono~inal 4 noun phrases that refer to fir~tor second person: patient are excluded, unless the a~entis first rerson

9) a. $i-t-? ca sway9qa? a tiva ?as 'The man is known pronominal. know-passive the man by DET myself by me myself. '

b. $i-t-q ca snayoqa? a tiva nakw 'The man is known Kuipers (1967:89) ob.?erves thnt "For the evression of 'he, they help(s) know-passive the man by Dm yourself by you yourself.' thee, you, Squamish has recourse to the passive forms. . . ." The But these noun phrases that refer to first and second person are not passive ~entencesthnt fill in the caps in the trnnsitive paradigm

pronouns. They differ from pronouns in that they occur with deter- given in Table 1 are in braces. From Tnhle 1 we see that the first miners and may occur in the focus position in the sentence, whereas person accusative suffix is -2, which is distinct from the second person pronouns may not occur in this position. 43 44 accusative suffix -5.Squamish -2 (passive) is copate with Lummi person, aside from the condition that first and second persons are of

-3 ; Squamish -&may be in part cognate with lumi -% (me/you: equivalent rank in sentencefi that contain only first and second person accusative). The fact that Squamish has separate suffixes (-2) and arguments.

(-L&) for first and second person accusative can be correlated with Kuipers ~ivesno pasaive paradigms and no information on what the fact that first and second person are of equivalent rank only in passive sentence typrs are excluded. But his outstandinp study of sentences that contain only first adsecond person arwments. Squami~hincludes no examples of first and second person pronominal

forms appearing as oblique arpments in passive sentences. We venture With regard to sentences with third person pronominal and nominal arp- to predict th.at Squamish excludes passive sentences with oblique first ments in Squamish, Kuipers states the following: and second person pronominal a~ents,and that (8) above applies. 12) ". . . a combination of a 3rd pers. trans. verb with a k!e conclude that Squ~mishhas an a~enthierarchy with the follow in^ noun in the absolutive case (Vtlia), when produced in two conditions: isolation, represents VtO and not VtS. For example, 13) Squamish Agent Iliernrchy: /nL&'a*m"t-as tabFqo mati?/ is understood as 'he bit the a. First and second person are of equal rank in dog' and not as *I the dog bit him', even though /n&'m?tas/ sentences that contain only first and second by itself means 'he bit him' and S and 0 are both in the nerson arpunents; otherwise, first person shares absolutive case. In isolntion, 'the dog bit him' is ren- the rank of N: dered by a passive construction V R - /nn-E'a'm9t-m P 1' h. 2) 3) N t-ta_sqoma'i9/, lit. 'he was bitten by the do^' (cf. the

sentence V S/ niLEbJ m7 -t-m ta4sq0ma'r/ 'the dog was The er~:ntivr: r,plit in Squarish is the sane aR in 1,urai: transitive P bitten')." (p. 172) sentences with a third person pronominal or full noun aqent have an

errrative suffix on the verb (sse Table 1). This passage shows that Squamish, like I,ummi, ranks third person pro- nominal arguments over full noun ar~l-ents,and that (7) above applies. 4. Lushootseed. Hess (1973) draws attention to the pann in the trnnsi-

Squamish, like Lummi, treats (non-oblique) nouns as nhsolutive tive and passive nnrndims in several Salish lnnnlnps. Accordinn to except under specified conditions. We have shown that in Lummi, only the data ~ivenby Hcss (nnd Hukari 1?7bb), Lirshootseed shows no nouns may outrank nouns. In Squamish, nouns may outrank nounfi and first hiernrchical ranking among pronominal wqments in transitive sentences. person as well, since first person does not share the rank of second There are two pieces of evldcnce to this effect: 1) there are senarate first and second person accusative suffixes as in Sauamish; 2) there are as natient and therefore the full noun cay be interpreted afi ergative.

no restrictions on the argument type which may be n~entin transitive :io full noun may be interpreted as er~ativewhile another full noun is

sentences with first and second person patients -- the went may be any absolutive in a transitive sentence. There are no transitive sentences pronoun or a full noun. The followinp, examples are taken from Hess where one full noun outranks another: full nouns are assifled the

(197390-93): abso1r;te lowest rank in the hierarchy, unless the presence of an accusa-

11 tive pronoun provides for an ergative interpretation. 14) "u$axwa-t-5. ti ZaEas IThe boy clubbed me.' clubbed- the boy ?ull noun aRents may be specified in Lushootseed only by employing transitive-me a passive sentence. In these intransitive sentences, the sin~lenon- 3 * 15) 7u$axW a-t-sid ti Eah 'The boy clubbed you.' clubbed- the boy oblicue ar~mentis absolutive. and the agent is specified in an transitive-you oblique noun phrase: 6

Recall that Lumni excludes sentences of the t.ypes represented in (14) 11 15; "uxad-t-b a ta stubg ta &has and (15); and Squamifih excludes the type shown in (15). Lilshootseed push-trans-passive by the fat man the hov lacks the er~ativerarkin~ that appears on the verh in third person 'The fat ran pushed the hoy.' transitive sentences in Lumi and Squmish. Lushootseed excludes the Lilsto3tseed -2 is cognate with Lummi 3 and Sauamish -m. Another way

follow in^ sentence type: to s:ate the restriction ~ivenin (16) is:

16) Transitive sentences with two (con-oblique) full noun ,:,: . Transitive fientences without a pronominal argument

ar-ments nre excluded. (ayent or n~tient)are excluded.

This remarkable feature of the erammar of Lushootseed is accounted for ?e izllowing sentence has a sin~lefull noun armnent (fiess 1973:W):

9 here with reference to the n~enthierarchy. Lushootseed permits tran- .?. * -5, "u:axwa-t ti basas sitive sentences to have only one full noun arcument, and that noun is 'He clubbed the boy' interpreted as absoiutive -- unless there is an accusative pronoun in 9.o z':solutive interpretation of this sinple f.111 nolin armnent depends the clause. The exclusion of transitive sentences with two full noun l~por.:wo features that we have seen to he cornon to these lnnr~a~es:

,umnents follows from the er~otivesplit in Lushootseed; pronouns show lj *..e third person pronominal a~enthas no overt mark in^; anl 2) third nominative/accusative markinp, vhile full nouns show no case nark in^ but Eers?:. rronominal outranks n full noun arorent. nerefore. in most often receive an ahsolutive interpretation. Nouns are interpreted - LuzkoooL.seed, as in Lumni and Squmlsh, (7) ahove ayplies. as absol~itiveexcept in sentences where a pronominal element is mnrked 4 7 48 corresponds to the Lummi one except that Squmish does not exclude cer- From other data riven by Hess and Hukari, we surmise that tain transitive sentences with a first person patient. The Lushootseed Lushootseed shares the restriction stated above in (8) on passive sen- case presents a still further reduction of the Lummi one since no transi- tences in the other Salish languages surveyed here. tive sentences with first and second person patients are excluded. ?,,is, Lushootseed requires transitive sentences to have some pro- Furthermore, transitive sentences with two noun ar~o~entsare excluded. nominal argument, a~entor patient, and requires passive sentences to While Lumi excludes transitive sentences with 1 and 2 person have unspecified or oblique full noun aflents. The excluded sentence patients and full noun a~ents,Lushootseed does not; and these are the types show that thr agent hierarchy in Lushootseed differs more from only transitive sentence types where full nouns m2y be aEents in the Lunmi case thrn does Squamish. Since there is less complementarity Lushootseed. Though the a~enthierarchy takes quite different forms in between transitive and passive sentence t,ypes in Lushootseed, it will Lummi and Lushootseed, the crucial feature that the two lanpages share be necessary to state a senarate acent hierarchy for the two sentence is that there are restrictions on transitive sentences with full noun types: apents, since accordina to the oflent hierarchy, full nouns are the 19) Lushootseed Aeent IIiernrchy ar~umenttype least likely to be agent, and most likely to be patient. a. passive sentences: 1 2 \, 3 N According to data riven in Thorpson and Tho~pson(1971) Clallam b. Transitive sentences: 3 ) ti correanonds to Lumni in excluded sentence types and in the agent hier- ri ) tr archy. According to data civen in Hess (1973) and Gerdts (1980)

5. Conclusions. 'ie hve been concerncd here to demonstrate the rres- H~lkonelemcorresponds to iiouaminh in the features we are concerned vith ence or .In apent 'ierarc?:; in certain Coast Snlinh lanmages, to relate here, and has the same a,;cnt hierarch.?. 'i'nt.le i. ~unmrizes

this accnt hierar:::! to certain ermtive phenoncnn, and to show how these features in tlir 1ani:uages surveyed. Zach ;nnri.qrn places 51111

certain transitlve and passive sentence types excluded in these noun areuTr?nts at tho hottom of the arent hierarchy. All of t::e lan-

1anWn:;es can be understood in terms of this hierarchy. Table flapes treat full nouns os absolutive in transitlve scntencrs except

3 d!snla~s the excluded ,~ontencetypes for each langua~e. Excluded under narrowly specified conditions, as follows: In Lushootseed, a

sentence tyr,es in eack Innp:i~eare st;irred and underlined. :Iote that (non-obli?ue) noun nay be cr~ative(that is, the nrent nr~mentof a

in Lunmi, wherever transitive sentences are excluded, passive ones Ne transitive sentence) if and only if the patient arpncnt is a first or

permitted, and vice versa, except in the 3 (and N) forms where both second person accusative pronominal. In Squ-mish nnd Halkomelem, a

sentence types are peraitted. It can be seen that the Squamish case noun may be erpxtive if and only if the patient arfllment is firfit 49

person accusative or a noun. In Lummi and Clallam, a noun may be The data reviewed here suggest that Proto Coast Salish may have had

ergative if and only if the patient is also a noun. ergativity, with a split between first and second person pronominds vs.

In his pioneering article on agent in Coast Salish, Hess called third person pronominals/nouns, as in the languages surveyed here except attention to the widespread presence of a patient interpretation of Lushootseed. There is evidence in favor of reconstructing the exclusion

full nouns in these languages and suggested that this feature, along of passive sentences with first nnd second person pronominal agents. with the paradigm gaps, held implications for the historical development There is evidence in support of an al:ent hierarchy in Proto Coast Salish,

of Salish. but the exact nature of this hierarchy cannot be reconstructed on the

bnsis of the data s~lrveyedhere. It is temptinl: to speculate that the

a~enthiernrchy in Proto Coast Salish resembled that seen in Lunmi and

Clallnn, since it is easier to interpret the a~enthierarchies in the

other lanpages as rodl~ctionsof the Lurmi-Clallm system than it is to

account for the other arent hier,uchies independently. Fione of the

lanflages surveyed here excludes transitive sentences with two full noun

arflments except Lushootseed; however, these sentences are extremely

rwe in the other lanruazes and it has been suggested that they we an

innov~tionthat has occlurred after contact. The nost conservative agent

hierarchy, then, rni~htrese~ble the Lummi-Clallan system with the added

exclusion of N over N in transitive sentences, as in us hoot seed.'

Historical reconstructions in the Northwest Coast area are always

made more problematical because of the presence of numerous features

that seem to be wed rather than ~enetic. There is evidence for an

agent hierarchy in Nootka (Whistler 19f50). Confirmntion of the extent

to which diffusion is reflponsible for the presence of these features in

Salish, and from which direction the features spread, awaits further 8 research. 52

The system of "inverse" marking described by Hockett (1966) for 20) a. hes-0-san "I sneeze" 9' Algonquian shows a close parallel with the Salish agent hierarchy. b. Ebn-9-san "I shiver"

"Inverse" marking appears on the verb in Algonquian when the patient The intransitivizer -f? in Haikonelem and -m in Squamish have parallel

outranks the agent. However, Sallsh differs from Algonquian in that middle voice functions.

Salish lacks obviative forms. e ' i ' While the contrast between active and passlve constructions may

not be relevant in Algonquian, this contrast apparently is present in

Salish. In the introduction to this paper, the question was raised

as to whether a sentence may be termed passive if it has no transitive

counterpart. We tentatively identify Lummi sentences with -t (transiti-

vizer) and -9 (intransitivizer) as ~assivesfor the follow in^ reasons:

there are sentence types in each of the languages surveyed here where

both transitive and passive constructions are permitted. For example,

Lummi permits the speaker to choose between a trnnsitive sentence with

two full noun arguments or its passive counterpart. Lushootseed

speakers may choose between transitive sentences with a third person

agent and a first or fiecond person patient and their passive analogues.

Where a choice between transitive and passive constructions is allowed,

there is a semantic contrast between the transitive and psssive forms

that corresponds to the contrast seen between these constructions

across languages.

Further support for the view that the intransitivizer -?marks a voice

contrast in Lummi is that there are derived intransitives with -2 (and with-

out a transitivizer) that seem to mark a middle voice, or less control on

the part of the subject of the sentence. 10 TABLE 1

The variation in types of apnt hierarchy that we have identified Transitive Verb Paradigm (Finite) in these Salish languaees has a more general theoretical consequence (after Kuipers 1967, p. 90) as well. Silverstein (1976) makes the point that his "hierarchy of

(semantic) features," and the resultant scale of potentiality of agency leaves open to the grammar of a particular lanpage the question of where along the scale a lan~agewill opt for an ergative split; but, by the nature of the scale, all terms above and below the split must natch in ergativity, That is, if third person is nominative/accusative, then second rjerson must be also, and if second person is nominative/ - -- us you them accusative, then first person must be as well. Some 'middle' terms are I z-n,z,al u-at-umi-ap E-n_d*a*%-at(-~it) said to show nominative/accusative and ergative/absolutive features;

The Salish er~ative that is, in some l;~npuaees, there is an overlan. you 1 E-ap>'alu-at-UNA ~-ap-~'atz-at(-ucit) splits confirm to this convention, as seen in many other languages they 1 n~'a'~-at-um~-a~~it'{~-a?d~va'_u-at-m(-pit)}ngE'alyat-as-Eit (Dixon 1979). Iiowever, the ranking of terms in an agent hierarchy, as shown by the transitive and passive paradir.s and the suppletion between The root verb is --- "help". To the left of the verb the then, does not. 'vie conclude thnt lanpza~eszay differ in the agent following morphemes are present: - indicates that the action of the verb is manifest; 2,first person singular subject; t- first person hierarchy, that is, in the ranking of terns as agents, and not neces- - ylural eubject; -2-,second person sinvlar subject; -9-second per- sarily differ in the point on the ~otentialityof aEency scale at which son plural subject; E- is a deictic which usually appears in sentences the errative split appears in a lankpage, as shown in Cable 3. We have a with third person subjects; third person subject is not marked. To the ri~htof the verb the following morphemes are present: -&-. a alco demonstrated that the a~enthierarchy na:r show non-scalar irrew- transitivizing suffix; -5- is the first person object; -as- is the lnrities in ?.articular lanppn~esthat may perhaps be attributed to ereative marker; -e-marks plurality for third person subjects: -a- the historical change in the Innwage family. ir; first pernon pli~relo1)ject; -I&- is the second ner:;on sinwlar object; -2-indicates thnt the second person is rlural; -m-- is an intrnnsitivizer that renders the sentence passive. The forms in braces are sunpletive forms; thc parenttlesized (-zit) is ontional. 55

TABLE 2

me thee him I I Lummi Squamish Luehootseed I help you (sg) 1 I help him Trans. Pass. Trans. Pass. Trans. Pass. we help you (sg) 1 we help him we I I thou / you (SKI help me 1 you (sg) help him ! you I you (pl) help me i you (pl) help him iI he he helps me 1 (you (bg) are belled he helps someone hy someone (SKI} ;

they help me / iyou (s~)are helped ! ha helps some people by some people I1

us you (pl) them I 1 , I / I help you (pl) I heln them I .,ie I we help you (pl) / we hell] them I I ! j you (s~)help them I you I you (pl) help us I you (~1)help them

I 1 he j he helps us I you (pl) are Llelped I he helps them t I \ \y someone} they they help us iyou (pl) are helped ' they help them ' by some people] 58

57 Footnotes

TABLE 4 'we would like to acknowledge the contributions of Mr.

Aloysius Charles and Mrs. Louise George, native speakers of

Excluded Trans. Avnt the Lummi and Lushootseed languages, respectively. Susan Steele Language Sentence Types Hierarchy Er~ativeSplit made some important suggestions for an earlier version of this

3Vl;NVl paper, and we thank her now for her assistance. We would also Lummi, 12/3N like to thank Ken Hale, Bernard Comrie, Larry Thompson. and

the anonymous reviewers of this journal for their comments and

Squamish, 3 V 2; N V 2 a) 1 & 2 of same 1 2 / 3 N suggestions. We would also like to thank the Office of the Halkomelem NV3 rank in 1 & 2 domain Vice-President for Research at the University of ARizona for b) 2) 3) N I its generous support.

I I a) Passive I 2~eassume a nounlverb distinction in Lummi, but will not sentences: 12>3~ present evidence on this question since it is outside the scope b) Transitive of this article. The agent hierarchy proposed here in no way ;e;t;nces: depends on such a distinction in any case. We refer the reader N) IT to the section on Lummi in Steele et a1 (in press).

3~nembedded sentences an -5 siffix optionally appears on intransitive verbs with third person subjects. A distinction

in either case marking or verbal inflection between main and

subordinate clauses is another type of ergative split found in

ergative languages. See Gerdts (1980) who shows this same kind

of split in Halkomelern and Dixon (1979) who discusses clause

splits elsewhere.

4~heglosses given in (9a) and (9b) employ the emphatic rather than the reflexive use of the English word -self. 59 60 lowe should mention that the Coast Salish languages share 5~quamishhas nouns that may be used to specify emphatic 1 another grammatical feature, not discussed here, vhich also oblique first and second person arguments, as in Lummi. See ,I demonstrates the emphasis placed on agency in these languages. Kuipers (1967: 143-144). Each language shows contrasts in the derivational/inflectional his sentence is from Mrs. Louise George and is from morphology of the verb, viz. in the transitivizing suffixes Demers' 1973 Skagit Fieldnotes. that mark differences in degree of control on the part of the 'Kinkade (19 ) proposes that the appearance of two full noun phrases in Coast Salish dialects may have been due to agent. This feature of the grammar has been described by

the influence of European languages. Hess (1973) and Thompson (1976). here are transiti~e/~assiveparadigm gaps in the Salish language family outside of Coast Salish; this suggests that

Proto Salish may have had an agent hierarchy. According to

Newman (1969) Bella Coola "kx-ct may be interpreted as he sees you sg, they see you sg, or you are seen: kx-tap, as he sees * you pl, they see you pl, or you pl are seen." It is suggestivd

that in Shuswap (Kuipers (1974)), Thompson (Thompson and

Thompson, undated ms), and Kalispel (Speck (1980)) there is

homonymy between active and passive forms with a first person

plural agent. The syntax of these forms is, however, in some cases distinct.

9~enHale (personal communication) reports the presence of an agent hierarchy in the Tanoan languages. Witherspoon (1980) has proposed a connection between the exclusion of certain transitive sentence types in Navajo and a hierarchy of control

in the Navajo world view. 61 REFERENCES 62 Dixon, R.M.W. 1979. Ergativity. Language 55:59-138. ------, and M. Terry Thompson. 1971. Clallam: A Preview. Gerdts, Donna. 1980. Some ergative phenonena in Halkomelem. Studies in American Indian Languages, ed. Jesse Sawyer.

Paper presented at the 15th International Conference on University of California Publications, no. 65. Berkeley: . , B.C.. Canada. University of California Press.

Hess, Thorn. 1973. Agent in a Coast Salish language. IJAL 39:89-94. ------, and M. Terry Thompson. undated. Thompson. ms.

Hockett, Charles. 1966. Whistler. Ken. 1980. Inverse person marking in Nootkan. ms. Hukari, Thomas. 1976a. Person in a Coast Salish Language. Witherspoon, Gary. 1980. Language in culture and culture in IJAL 42: 305-318. language. JIAL 46:l-13. ------1976b. Transitivity in Halkomelem. Paper presented at the XI International Conference on Salishan languages. . Kinkade, Dale. 1973. Salish evidence against the universality of "noun" and "verb"

Kuipers, Aert. 1967. The . Janua linguarum, series practica, no. 73. The Hague: Mouton. ------1974. The Shuswap Lancuage. The Hague: Mouton.

Newman, Stanley.1969. 9ella Coola Paradigms. IJAL 35:299-306. Silverstein, Michael. 1976. Hierarchy of features and ergativity.

Grammatical categories in Australian languages. Canberra: Australian Institute for Aboriginal Studies.

Speck, Brenda. 1980. An adition of Father Post's Kalispel grammar, University of Montana Occasional Papers in Linguistics. no. 1. Thompson, Laurence. 1976. The control system: a major category in the grammar of Salishan languages. The Victoria1 Conference

on Northwestern Languages, Heritage Record No. 4. Victoria:

British Columbia Provincial Museum.