28
Medial Clauses and Interpropositional Relations in Panare Medial Clauses and Interpropositional Relations in Panare Medial Clauses and Interpropositional Relations in Panare Medial Clauses and Interpropositional Relations in Panare Thomas E. Payne Universidad Católica de Táchira, and the University of Oregon May 15, 2006 1. Introduction 1. Introduction 1. Introduction 1. Introduction 1 Since the mid-sixties there have been many studies dealing with `clause-chaining' languages (McCarthy 1965, Healy 1966, Hetzron 1969, Hetzron 1977, Longacre 1972, Olson 1973, Thurman 1975, Gerdel and Slocum 1976, Haiman 1980 inter alia). The paradigm examples of clause chaining languages occur in the highlands of New Guinea, both Irian Jaya and Papua New Guinea (Elson 1964), though clause chaining is a well-recognized phenomenon in Australia (Dixon 1972, Austin 1981, Austin 1987) and the Americas (Longacre 1985). In descriptions of such languages there is normally a distinction drawn between `final' and `non-final' clauses. These terms are based on the fact that in clause chaining languages as identified in these previous studies, there are significant morphosyntactic differences between the sequentially final clause in a clause chain and those that precede it. The near-homophony between the term `final' and the traditional term `finite' is a happy coincidence in that final clauses are the most finite clause type in a clause chaining language. That is to say, final clauses take a broader range of verbal inflections than do non-final clauses, and final clauses do not take special subordinating/switch-reference morphology typical of non-final clauses. Nevertheless, the final/non-final distinction does not translate directly into the traditional finite/non-finite distinction in that non-final clauses also typically allow some verbal inflection. Hence in the traditional sense non-final clauses would be considered `somewhat' finite. Longacre (1985:264, and footnote 6) hypothesizes that, though it is a logical possibility for languages to have clause chains in which the functionally final (read `most finite') clause actually occurs chain-initially, no clear examples of such languages have been documented to date. More recently the term MEDIAL MEDIAL MEDIAL MEDIAL has begun to replace the term non-final in descriptions of clause chaining structures (Haiman 1987). 2 This term reflects the fact that this clause type occurs sentence-internally, i.e. in the `middle' of a clause chain. As defined by Longacre 1985:263, a medial clause is one which: 1. has a reduced range of tense-aspect possibilities in comparison to final clauses 2. usually specifies `subject' reference in terms of (i.e. as same as or different from) the subject of the final clause 3. usually directly expresses temporal relations such as `overlap' and `succession' with respect to other clauses in the sequence. 3

Medial clauses and interpropositional relations in Panare

  • Upload
    sil

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Medial Clauses and Interpropositional Relations in PanareMedial Clauses and Interpropositional Relations in PanareMedial Clauses and Interpropositional Relations in PanareMedial Clauses and Interpropositional Relations in Panare

Thomas E. Payne

Universidad Católica de Táchira, and the University of Oregon

May 15, 2006

1. Introduction1. Introduction1. Introduction1. Introduction1111

Since the mid-sixties there have been many studies dealing with `clause-chaining' languages (McCarthy 1965, Healy 1966, Hetzron 1969, Hetzron 1977, Longacre 1972, Olson 1973, Thurman 1975, Gerdel and Slocum 1976, Haiman 1980 inter alia). The paradigm examples of clause chaining languages occur in the highlands of New Guinea, both Irian Jaya and Papua New Guinea (Elson 1964), though clause chaining is a well-recognized phenomenon in Australia (Dixon 1972, Austin 1981, Austin 1987) and the Americas (Longacre 1985). In descriptions of such languages there is normally a distinction drawn between `final' and `non-final' clauses. These terms are based on the fact that in clause chaining languages as identified in these previous studies, there are significant morphosyntactic differences between the sequentially final clause in a clause chain and those that precede it. The near-homophony between the term `final' and the traditional term `finite' is a happy coincidence in that final clauses are the most finite clause type in a clause chaining language. That is to say, final clauses take a broader range of verbal inflections than do non-final clauses, and final clauses do not take special subordinating/switch-reference morphology typical of non-final clauses. Nevertheless, the final/non-final distinction does not translate directly into the traditional finite/non-finite distinction in that non-final clauses also typically allow some verbal inflection. Hence in the traditional sense non-final clauses would be considered `somewhat' finite. Longacre (1985:264, and footnote 6) hypothesizes that, though it is a logical possibility for languages to have clause chains in which the functionally final (read `most finite') clause actually occurs chain-initially, no clear examples of such languages have been documented to date.

More recently the term MEDIALMEDIALMEDIALMEDIAL has begun to replace the term non-final in descriptions of clause chaining structures (Haiman 1987).2 This term reflects the fact that this clause type occurs sentence-internally, i.e. in the `middle' of a clause chain. As defined by Longacre 1985:263, a medial clause is one which:

1. has a reduced range of tense-aspect possibilities in comparison to final clauses

2. usually specifies `subject' reference in terms of (i.e. as same as or different from) the subject of the final clause

3. usually directly expresses temporal relations such as `overlap' and `succession' with respect to other clauses in the sequence.3

T. Payne, Medial-clauses 2

These characteristics will henceforth be referred to as the `Longacre definition' of medial clauses.

A clause chaining language is a language that employs sequences of medial clauses completed by a final clause as a major discourse structuring device. Longacre (op cit.) describes such languages as `radically different' in surface structure from more familiar `co-ranking' languages, i.e. languages that tend to employ sequences of equally finite clauses in discourse.

In this paper, I will attempt to a) refine the notion of medial clause in terms of prototype theory, b) show that medial clauses are probably more common than has been previously suggested, and c) show how and, what is more important, why medial clauses (at least in Panare) express interpropositional relations that transcend the temporal relations usually associated with this clause type. Although Panare is not a classic clause chaining language in the tradition of highland New Guinea, still the notion of medial clause is viable and useful in Panare. Insofar as Panare does exhibit clause chains, these consist of an initial finite clause and a series of (three or fewer) medial clauses. This fact confirms the observation that the term `clause chaining language' does not describe a distinct language type, but rather that languages may exhibit clause chaining to a greater or lesser degree.

The organization of the presentation will be as follows. First I define the notions EVENTEVENTEVENTEVENT (section 2.1) and DEPENDENT CLAUSEDEPENDENT CLAUSEDEPENDENT CLAUSEDEPENDENT CLAUSE (section 2.2), drawing from recent literature in the functional analysis of discourse and cognition. These definitions are essential to the characterization in section 3 of prototypical medial clauses as `dependent clauses that encode events'. Then in section 4 I will identify and exemplify medial clauses in Panare, distinguishing them from other types of dependent clauses such as adverbial, complement and relative clauses. For purposes of this paper I describe the many interpropositional relations expressed in medial clauses, in addition to temporal relations. This range of interpropositional relations is motivated by (i.e. understandable in terms of) a cognitive model based on a spatial metaphor. The notion of cognitive models, and the particular model to be explicated in this paper, will be discussed in section 2.3 below.

2. Some terminology2. Some terminology2. Some terminology2. Some terminology

2.1. Events and non events2.1. Events and non events2.1. Events and non events2.1. Events and non events

The simplest way to conceive of an event is as an appropriate answer to the question `what happened?' (Chafe 1976, Hopper and Thompson 1984:726). Non-stative realis assertions are typically the only speech act types that satisfy this requirement, e.g.:

T. Payne, Medial-clauses 3

3

Speaker A: Speaker B: Speech act type: What happened? I hit him. NON-STATIVE REALIS ASSERTION ??He knew the answer. STATIVE REALIS ASSERTION ??I was angry. STATIVE REALIS ASSERTION ??I didn't hit him. IRREALIS (NEGATIVE) ASSERTION ??I will hit him. IRREALIS (FUTURE) ASSERTION ??In order to hit him. IRREALIS (PURPOSE) ASSERTION ??If I hit him. IRREALIS (CONDITIONAL) ??Because I hit him. PRESUPPOSITION ??While I hit him. PRESUPPOSITION ??After I hit him. PRESUPPOSITION

Many of the clauses marked by ?? above may in fact be appropriate responses to the question `what happened?' given a somewhat enriched context. For example, if it is obvious to all interlocutors that Johnny hit Billy, Johnny might understand the question `what happened?' as a request for an explanation of why he hit Billy. In which case `he knew the answer' or `I was angry' might be appropriate responses. It should be clear, however, that these responses assume a richer conceptual scene than that which is embodied by the question alone.

As with most functional distinctions, the terms event and non-event describe extremes of a continuum. Any given clause embodies a concept that is perceived as closer to one extreme or the other. If the distinction is at all relevant to communication, we would expect that languages should have grammaticalized means of grounding concepts (i.e. situations to be discussed) along this continuum. In fact, there is considerable evidence that languages have well-instantiated event and non-event cognitive models that underlie many morphosyntactic distinctions. For example, Hopper and Thompson's pioneering work on transitivity (1980) demonstrated that many aspectual and modal distinctions (their `components of transitivity') can be understood in a unified way in terms of the event vs. non-event continuum.4 In particular, IMPERFECTIVEIMPERFECTIVEIMPERFECTIVEIMPERFECTIVE aspect characterizes clauses that embody concepts close to the non-event end of the continuum, whereas PASTPASTPASTPAST----PERFECTIVE ASPECTPERFECTIVE ASPECTPERFECTIVE ASPECTPERFECTIVE ASPECT characterizes clauses that embody concepts close to the event end of the continuum (Hopper 1979).

2222.2 Grounding and the dependent/independent distinction.2 Grounding and the dependent/independent distinction.2 Grounding and the dependent/independent distinction.2 Grounding and the dependent/independent distinction

Some verbal morphological operations serve to GROUNDGROUNDGROUNDGROUND the situation described by the verb. These deictic operations establish the time, location, participant identity or actuality of a situation according to some reference point. For temporal operations the reference point is normally `now', i.e. the time of utterance. For spatial operations the reference point is normally `here', i.e. the place of utterance. Participant reference is usually grounded according to `person', i.e. speaker, hearer or non-speech act participant. Some modal and evidential operations ground a situation with respect to `actuality'. That is, these operations express, perhaps among other things, the speaker's

T. Payne, Medial-clauses 4

perception of how likely it is that the situation described by the verb did or will actually take place (Chafe and Nichols 1985).

In most multiclause constructions in Panare, there is one INDEPENDENTINDEPENDENTINDEPENDENTINDEPENDENT clause (i.e. a clause that is fully grounded in time and participant reference by its verb internal morphology), and one or more clauses that depend for their temporal and sometimes spatial grounding and/or the reference of their primary participant on the independent clause.5 The term `finite' could substitute for the term `independent' in this sense. However, the contrasting term `non-finite' would not correspond to `dependent'. Dependent clauses can be grounded. It's just that their grounding is specified by the morphology of some other clause.

2.3 Cognitive Models2.3 Cognitive Models2.3 Cognitive Models2.3 Cognitive Models

The term COGNICOGNICOGNICOGNITIVE MODELTIVE MODELTIVE MODELTIVE MODEL derives from Lakoff 1977, and is elaborated in Lakoff and Johnson 1984 and Lakoff 1987. It is analogous to the `frames' of Goffman 1974, Minsky 1975 and 1977, and Fillmore 1982. A cognitive model is a unified cognitive pattern of relationships that serves as the basis for conceptualizing various phenomena. For example, the cognitive model of WAR serves within my culture as the basis for understanding and reasoning about such diverse phenomena as argumentation and sexuality. This conclusion is derived partially from observation of such common locutions as the following (examples adapted from Lakoff and Johnson 1984):

ARGUMENT IS WAR: Your claims are indefensible. He demolished my argument. I attacked every weak point in his argument. She came at him with both barrels.

SEXUALITY IS WAR He is known for his many conquests. She devastated me. I couldn't resist his advances. She fought for him, but his mistress won out.

Some cognitive models may be universal while others are specific to individual communities. The cognitive model to be explicated in this paper will be termed the JOURNEY model. It consists minimally of a trajector (Langacker 1987) and a vector. The vector consists of a source, a present location and a goal. At any given point in a journey the trajector (the moving entity) is located at a present location. The source is the previous location and the goal is a location further along in the anticipated journey. This fundamentally spatial model serves as the basis in many languages for conceptualizing the temporal relations of anteriority and posteriority and the logical relations of cause/reason and purpose/result as well....6

In this paper I will show that this model provides a motivation for the use of certain medial clause suffixes in Panare to express a variety of interpropositional relations. The Panare data are not conclusive with respect to whether the temporal relations or the spatial relations are fundamental. Nevertheless, the data are consistent

T. Payne, Medial-clauses 5

5

with the model insofar as temporal anteriority corresponds to logical cause or reason, whereas temporal posteriority corresponds to logical result or purpose. The model would be invalidated if any of these correspondences were the other way around.

3. Towards a universal characteriz3. Towards a universal characteriz3. Towards a universal characteriz3. Towards a universal characterization of medial clausesation of medial clausesation of medial clausesation of medial clauses

With the definitions in sections 2.1 and 2.3 in mind, it is possible to characterize prototypical medial clauses universally as `dependent clauses that encode events.' If a language has a set of dependent clauses (as defined above) that typically encode events in discourse, such clauses can be termed `medial clauses'. Finite clauses and participial clauses with zero-pronominalized subjects in English come close to this definition though neither are prototypical medial clauses. Here I will briefly discuss finite clauses with zero-pronominalized (`gapped') subject only.7 First, the tense-aspect possibilities of gapped finite clauses are limited by the tense/aspect of the `main' clause:

(1) a. He sits down and starts to speak. b. *He sits down and started to speak. c. He sat down and started to speak. d. *He sat down and starts to speak.8

However, the verb in the second clause of these examples still inflects as finite verbs do. It's just that the particular inflection is limited to that which is chosen for the preceding verb. Consequently there is no independent means of determining the direction of dependency. One could as easily claim that the tense/aspect of the first clause is restricted to that of the second. Prototypical medial clauses, on the other hand, employ distinct verbal morphology that does not directly ground the event expressed by the medial verb. It merely points to the other verb for temporal, spatial and/or modal grounding, depending on the system of the language.

Despite this departure from the prototypical pattern for medial clauses, gapped finite clauses in English essentially possess all the characteristics of medial clauses listed by Longacre (1985:264): a) they are different in structure from other finite clauses (i.e. they allow a zero-pronominalized argument, normally strictly disallowed in finite clauses in English). b) They employ a system of same/switch reference that is controlled by the subject of the `main' clause: zero codes continuing reference and a free noun phrase or stressed pronoun encodes switch reference. c) They encode temporal relations, in this case strict sequentiality. Also, like prototypical medial clauses, in longer chains the temporal sequence proceeds from one clause to the next, rather than all clauses relating directly to the `main' clause. Of course in proposing these characteristics, Longacre had in mind the canonical medial clauses of New Guinea languages. Nevertheless, given a strict reading of his definitions it is not obvious how English finite clauses are to be excluded in a non-arbitrary way (e.g. by specifying that switch reference must be morphological, or that final clauses must actually come at the end of the chain). From this point of view we can see that the notion of medial clause (if not the language-specific definitions) is perhaps less unusual or exotic than previously assumed.

T. Payne, Medial-clauses 6

In addition to distinguishing medial clauses from independent clauses, it is important to distinguish medial clauses from temporal adverbial clauses. We will again use English examples to illustrate this distinction. As a `co-ranking' language English encodes simple sequences of events as sequences of finite clauses that are grammatically on a par. That is, neither is dependent on the other for specification of tense/aspect or participant reference, e.g. 1a and 1c above. This is distinct from adverbial clauses that also code temporal succession:

(2) He sat down before he started to speak.

The difference in function between co-ranking (ex. 1) and subordinating (ex. 2) construction types is twofold: a) in 2 there is some previous expectation that the two events `should have' occurred in the other order, and b) in 2 the second clause encodes a presupposition rather than an assertion (Searle 1970). This is one of the main differences between adverbial clauses and medial clauses: adverbial temporal clauses are presuppositions, i.e. they are presented as unchallengable or taken for granted, whereas medial clauses are assertions (Givón 1987). This universal characterization of adverbial clauses is based purely on their function. In English (and a lot of other language) this functional category is encoded in the grammar, i.e. adverbial clauses are introduced by `subordinating conjunctions' such as `before', `because' etc.

In Panare, medial clauses function more like the second clause of 1a and 1c than like temporal adverbial clauses. There is a free-standing clause connector, pake, that specifies sequentiality when the proposition is presupposed. Clauses with pake are more properly termed temporal adverbial clauses. These will not be discussed in this paper. Adverbial clauses in general in panare are distinguished grammatically from medial clauses in that adverbial clauses do not code reference relationships (continuing or switch reference) whereas medial clauses do.

4. Medial clauses in Panare4. Medial clauses in Panare4. Medial clauses in Panare4. Medial clauses in Panare

In the following sections I will show that three Panare suffixes mark prototypical medial clauses. These suffixes are -séjpe, -sé'ñape and -ñépe. Two other suffixes, -tááñe and -npan, mark clauses that would not be medial clauses under a strict definition of event couched in terms of temporal sequentiality. Nevertheless, under the characterization of an event as a non-stative realis assertion, -tááñe and -npan clauses also fall into the category of medial clauses. Finally clauses marked by a temporal suffix, -jpómën,9 almost certainly should not be considered medial clauses given our universal characterization. Nevertheless, -jpómën clauses remain formally distinct from temporal adverbial clauses and also exhibit the Longacre characteristics of medial clauses. The formal respect in which medial clauses differ from adverbial clauses is that adverbial clauses do not specify continuing or switch reference, whereas medial clauses do. Clauses with the suffix ----jpomën encode continuing subject, and so by the formal criterion should be considered medial clauses. These data illustrate that medial clauses and adverbial clauses delimit a continuum of possible morphosyntactic structures.

T. Payne, Medial-clauses 7

7

4.1 Subsequent action, high continuity: 4.1 Subsequent action, high continuity: 4.1 Subsequent action, high continuity: 4.1 Subsequent action, high continuity: -séjpe, -se'ñape

The suffixes -séjpe and -sé'ñape both signify that the situation expressed by the verb they are attached to closely follows in temporal sequence the action expressed in the previous clause. Clauses headed by verbs marked with these suffixes have the approximate discourse-pragmatic sense of independent clauses introduced by `then', `and then', `and so', etc. in English. These suffixes contrast in that -séjpe normally signifies that the primary actor is the same for both clauses, whereas -sé'ñape typically signifies that the absolutive argument of the first clause is the PATIENT of the second clause.10 In texts, however, there are exceptions to both of these generalizations. These suffixes essentially express a high degree of discourse continuity (Givón 1983, Scancarelli 1989) between the two clauses. This continuity typically coincides with referential identity between one argument of each of the clauses -- -séjpe signals actor/source/inception-oriented continuity, whereas -sé'ñape signals patient/result/termination-oriented continuity. This assertion is consistent with the fact that in séjpe clauses the AGENT is a surface subject, whereas in sé'ñape clauses the AGENT is demoted to an oblique case (see examples 15b, 17, 19 and 21b). For this reason (among others) I classify -sé'ñape as a marker of a passive clause (T. Payne 1990).

Examples 3 through 13 illustrate séjpe constructions:11

(3) a. Y-apooñe-n këj i'yaka-e y-uw-ëj-pu'ma-sa', 3-arrive-NONSPEC:I AN:PROX 3:family-POSS 3-I-DTRNS-hit-PPART b. y-o'koma-séjpeséjpeséjpeséjpe. 3-raise-HC:SEQ1 `Some counterparts of the fallen one arrive, and then raise him.'12 (PST.08.11.1)

(4) a. T-onkon-se e'ñapa i'yaka-e w-ës-uka-sa' IRR-insert-HAB Indian 3:family-POSS I-DETRANS-kill-PPART

ejchÿrëkun wasanto'-ya t-yo-n, sun straight-LOC 3-eye-POSS

`The person's counterparts place the dead one with his face towards the sun,' (MCW.76.65)

b. y-inapontë-séjpeséjpeséjpeséjpe ano ke, iye ke, to' ke. 3-cover-HC:SEQ1 earth INST tree INST rock INST `Then they cover him with earth, sticks and rocks.' (MCW.76.66)

(5) Y-an-ñe yu, y-ipima-séjpeséjpeséjpeséjpe tamaipe. 3-take-NONSPEC:T 1SG 3-pay-HC:SEQ1 later I'll take it, and pay later. (MCW.40.23.24)

T. Payne, Medial-clauses 8

(6) sïj y-onó'ma-'me, a-wë'-séjpeséjpeséjpeséjpe chankë y-ïmunko-pëj. INAN:VIS TRNS-ingest-IMPER:MVT 2-come-HC:SEQ1 towardsDC 3-hurt-COND13 `Take this and come back if it hurts.' (JP)

These examples illustrate that séjpe clauses depend on the preceding clause for inflectional grounding. In 3 the temporal grounding of the first clause is understood as past, and the temporal grounding of the séjpe clause is the same. In 4 the grounding of both clauses is habitual or gnomic. In 5 the grounding is future, and in 6 both clauses are understood as imperative.

In the text corpus, all situations expressed by a -séjpe clause have the same primary actor as the situation expressed by the previous clause. A switch in actor constitutes a level of discontinuity that is normally too disruptive for -séjpe to be appropriate. However, this is not a grammaticalized constraint, as examples can be elicited in which the main actor of the -séjpe clause can be either the same or different from the main actor of the preceding clause.14 For example:

(7) n-ïkëjtën-yaj kën y-úya arakon náápaka----séjpeséjpeséjpeséjpe (*yúya) 3-shout-PPERF AN:INVIS 1SG-DAT monkey encounter-HC:SEQ1 ``He shouted to me and then (I/he) encountered monkeys.')

(8) n-ïkëjtÅn-yaj kën kën-úya arakon náápaka----séjpeséjpeséjpeséjpe 3-shout-PPERF AN:INVIS AN:INVIS-DAT monkey encounter-HC:SEQ1 `Hei shouted to himj and then (either one) encountered monkeys.'

I speculate that -SÉJPESÉJPESÉJPESÉJPE is sanctioned in these examples, even though the primary actor is not necessarily the same as the primary actor of the previous clause, because there is no major disruption in continuity. There is no new participant introduced in the -SÉJPESÉJPESÉJPESÉJPE clauses. The dative participant and primary actor of the first clause are the only possible referents for the primary actor of the -SÉJPESÉJPESÉJPESÉJPE clause.

Although the core function of -séjpe is to mark a clause that occurs in close temporal sequence with the preceding clause, often a PURPOSEPURPOSEPURPOSEPURPOSE relationship holds between the clauses. Examples 9 and 10 illustrate two situations in which it is clear that the séjpe clause encodes the purpose for the preceding clause:

(9) Ya'ra y-án-ñe e'ñapa, y-u'-séjpeséjpeséjpeséjpe tato uya turtle TRNS-take-NONSPEC:T Indian 3-give-HC:SEQ1 non:Indian DAT `The people get mud turtles, (and then give/in order to give) them to the non-Indians.' (MCW.40.21)

(10) Tï-kïti-yaj t-yiño-n y-amaanë-séjpeséjpeséjpeséjpe ake yipí-j-pë. 1-cut-PPERF1 3-tongue-POSS 3-fix/make-HC:SEQ1 snake remedy-POSS-for `I cut off his tongue (and then prepared/in order to prepare) snake skin.' (MKW.62.32)

The difference between using -séjpe rather than the adverbial purpose clause marker -tópe is that a séjpe clause asserts that the situation described actually occurred,

T. Payne, Medial-clauses 9

9

whereas tópe clauses do not. This is consistent with the observation that séjpe clauses encode events whereas adverbial clauses normally do not. Nevertheless, bilingual native speakers will frequently translate a SÉJPESÉJPESÉJPESÉJPE clause with a Spanish purpose adverbial clause introduced by PARAPARAPARAPARA.

The remaining examples of séjpe in this section illustrate that séjpe clauses can be used independently of a preceding finite clause. Nevertheless, it is still clear that they are dependent in a pragmatic sense. For example:

(11) Y-ïïwa-n yëëwÿ-pë-to paken, y-uwarïpútë-séjpeséjpeséjpeséjpe. 3-son-POSS nose-puncture-for first 3-initiate:w/loincloth-HC:SEQ1 `After the boy's nose-puncturing, they initiate him.' (MCW.77.3)15

In this sentence, the main element of the first clause is not an independent clause. Rather it is a nominalization, much like the English translation. Hence the séjpe clause is the best candidate for `main' clause status in this construction. This example also illustrates that the sequential relationship encoded by a séjpe construction can be redundantly marked. The first clause contains the adverbial clause connector paken `first' that explicitly indicates that the first situation (nose-puncturing) precedes the situation described in the séjpe clause.

The suffix -séjpe can also be used outside of multi-clause constructions. For example, 12 illustrates -séjpe in a question:

(12) ëjkëi a-tamu-ta-séjpeséjpeséjpeséjpe? how 2-husband-INCHO-HC:SEQ1 `How then are you going to get a husband?' (MCW.34.8)

The context for this sentence is an exhortation by a mother to her daughter concerning marriage customs. The daughter refuses to take the bitter medicine (ingested through the nose) that will bring her success in finding a decent husband. There is, then, a latent proposition that is logically prior to this example, i.e. `Since you won't take the medicine, how then are you going to get a husband?'.

Even when a sequence of clauses is broken up by a significant interruption, such as a conversational exchange, -séjpe can still mark the second clause. For example:

(13) a. ñe'-ñe nkë ñij ty-ëkï chiwa. 3:bring-NONSPEC:T also there 3-CL:pet goat `He brings his goat from there.' (PST.13.18)

b. Audience: Tityasa chiwa? one goat `One goat?'

c. Speaker: Aye, tityasa. yes one `Yes, one.' (PST.13.19)

T. Payne, Medial-clauses 10

d. Y-ankë-séjpeséjpeséjpeséjpe y-uwëë'ya-po këj. 3-take-HC:SEQ1 3-place-LOC AN:PROX `Then he takes it to his place.' (PST.13.20.)

Now the second high continuity, sequential suffix, -sé'ñape, will be discussed. Examples 14 through 17 illustrate se'ñape constructions in what I consider to be their core function:

(14) a. Y-akomamë-n, y-u'-ñe oj-chichen maraka í'yan úya 3-darken-NONSPEC:I 3-give-NONSPEC:T party-owner maraca shaman DAT `When it gets dark the host of the party gives a maraca to the shaman,' (MCW.78.46a)

b. y-ëwa----se'ñapese'ñapese'ñapese'ñape . . . 3-play-HC:SEQ2 and it is played . . .' (MCW78.46b)

In this example clause b refers to an event in which the PATIENT is coreferential with the O argument of clause a, and which closely follows temporally and conceptually the event expressed in clause a. Example 15 is an additional example of O=PATIENT coreference:16

(15) a. Nï-kïti-yaj kën kawë-no. 3-cut-PPERF1 ANIM:INVISIB high-LOC `He picked it from up high, (PST.08.5)

b. Y-an----se'ñapese'ñapese'ñapese'ñape, inken úya tyei-yawo. 3-collect-HC:SEQ2 child DAT container-inside and it is collected by a kid in a container.' (PST.08.6)

Examples 16 through 19 illustrate that the antecedent for the PATIENT of the se'ñape clause can also be an intransitive S argument:

(16) a. Ta'meñe y-uwë'mukú'më-n e'ñapa, tyaj, ugly 3-approach-NON:SPEC:I people listen `Out of curiosity, many people approach , listen, (MKW.61.8)17

b. Y-ïkurajtë----se'ñapise'ñapise'ñapise'ñapi-nkë ayïñïpën pë. 3-heal-HC:SEQ2-also sickness OBL and are healed of sicknesses.' (MKW.61.9)

In this example, the PATIENT of the se'ñape clause is the intransitive subject of the previous clause. 17 is another straightforward example of S=PATIENT coreference:

(17) Y-owopataka-n wëiki ij-chawo y-uwë'----sé'ñapesé'ñapesé'ñapesé'ñape yu-uya 3-come:out-NONSPEC:I deer bush-LOC 3-kill-HC:SEQ2 1SG-DAT `The deer came out of the woods and was killed by me.' (JP)

T. Payne, Medial-clauses 11

11

Example 18 illustrates -sé'ñape in construction with -ñépe, the low continuity, sequential operator (section 4.2) to indicate S=PATIENT coreference:

(18) . . .ë-sa-ñépe y-ë'më'ka-sé'ñapesé'ñapesé'ñapesé'ñape yu-uya NEU-grow-LC:SEQ 3-cook-HC:SEQ2 1SG-DAT `. . . to grow, to be cooked by me.' (PS.97.57)

Since the controller of the same reference marking can be either an O or an S argument of the main clause, the coreference function of -sé'ñape can be summarized as indicating ABSOLUTIVE=PATIENT coreference.

Many of the above examples of -sé'ñape clauses illustrate that these clauses, in addition to expressing the temporal relation of succession, also express the logical relation of result. For example, in 17 the deer was shot as a result of having come out of the woods. The suffix -ñépe would be unacceptable in this circumstance, even though the actors are different in the two clauses, as it would tend to express that the deer came out of the woods and then the speaker shot something else or the deer came out of the woods in order to be shot (see section 4.2 below).

The following extended example illustrates how -sé'ñape indicates a resultative sense. The conversational exchange leading up to the sé'ñape clause in this excerpt is given in English translation:

(19) A: What (kind of) fruit? B: I dunno. A: mmmmm B: Big. A: Big, mmmmm. B: Larger fruit than around here. He put it on his bike . . . A: yes. B: Y-an-se'ñapese'ñapese'ñapese'ñape. 3-take-HC:SEQ2 `And it was taken.' (PST.12.16)

According to the function of -séjpe as indicating, among other things, same primary actor, -séjpe should be sanctioned in this environment as well, since the boy who puts the fruit on the bike is the same one who takes the fruit. However, the high number of mentions of fruit in the preceding context makes it clear that the fruit, and what happens to it, is the primary object of discussion rather than the boy and what he does. I take this as evidence for the claim that SÉ'ÑAPESÉ'ÑAPESÉ'ÑAPESÉ'ÑAPE expresss result/patient/termination oriented continuity.

The following examples illustrate some atypical uses of -sé'ñape. Example 20 illustrates that a sé'ñape clause can depend on a clause in another conversational turn. In this case coreference is O=PATIENT:

(20) a. ¿N-an-yaj? 3-take-PPERF1

T. Payne, Medial-clauses 12

`He took them?' (PST.04.13) b. Y-an----se'ñapese'ñapese'ñapese'ñape tonkwanan-úya 3-take-HC:SEQ2 others-DAT `So the other one takes them.' (PST.04.14)

It is worth noting at this point that -sé'ñape is clearly related to the PATIENT nominalizer -sé'ña. A few examples of the nominalizer -sé'ña are provided here:

(21) Ejke y-amenkë----se'ñase'ñase'ñase'ña kure atawën. NEG:EXIST 3-write-PART2 much all `There isn't anything to be written down.' (MCW.31.18 )

(22) Mo-ka n-aj y-apanawa-se'ñase'ñase'ñase'ña? EXIST-QP 3-AUX 3-rub:on-PART2 `Do you have something to rub on it?' (JP)

(23) Y-ërëëka-se'ñase'ñase'ñase'ña mën aro. 3-winnow-PART2 INAN:INVIS rice `The rice is to be winnowed.' (JP)

The meaning of -sé'ña as a nominalizer can be summarized as `one destined to be VERBed'.18 This correspondence between -sé'ña and -sé'ñape is analogous to the morphological correspondences between -ñe as an AGENT nominalizer and -ñépe (section 4.2), and perhaps the correspondence between -se as an `immediate' past participle and -séjpe. It is also parallel to the non-overtly coded correspondence between the `visible' past participle suffix -sa' and the perfect aspect/passive voice operator -sa' (T. Payne 1990).

The following example of a clause chain will close the discussion of -séjpe and -sé'ñape. This excerpt illustrates the interacting use of non-specific aspect (clause 24a), -séjpe (24b and d) and -sé'ñape (24c to express a narrative chain of events:

(24) a. y-an-ñe ñaj 3-take-NONSPEC:T s.w. `He takes it all at once.' (PST.14.25)

b. y-ïn-séjpiséjpiséjpiséjpi-nkë ñankë mënkën yaka. 3-put-HC:SEQ1-also returning1 same inside `Then he puts it inside the same thing.' (PST.14.26)

c. y-ankë-sé'ñapisé'ñapisé'ñapisé'ñapi-nkë mën, mën y-án-sa. 3-take-HC:SEQ2-also INAN:INVIS INAN:INVIS TRNS-OI:take-PERF `And it is also taken, it is taken.' (PST.14.27)

d. y-utë-séjpiséjpiséjpiséjpi-nkë. 3-go-HC:SEQ1-also `Then he goes also.' (PST.14.28)

T. Payne, Medial-clauses 13

13

Clause 24a initiates the event chain. Clause 24b takes -séjpe because the actor is the same as that of 24a. In 24c -sé'ñape indicates that the PATIENT is the same as the absolutive of the previous clause. Finally, in 24d -séjpe is again sanctioned because the actor is the same as that of the previous clause. Except for the fact that the finite clause occurs in chain-initial position, this excerpt is a near paradigm case of a clause chain as defined by Longacre 1985:264.

4.2 Subsequent action, low continuity: 4.2 Subsequent action, low continuity: 4.2 Subsequent action, low continuity: 4.2 Subsequent action, low continuity: -ñépe

The suffix -ñépe is probably the most difficult of the medial clause suffixes to characterize in terms of one unified function. One reason for this difficulty is that ñépe clauses span the functional scale between medial clauses and object complements. In this respect -ñépe can be compared to the complementizer/adverbial clause introducer to in English.19 However, -ñépe differs functionally in significant respects from English to. In particular, -ñépe is far more specific in semantic content, e.g. -ñépe clauses can encode events in discourse, and they always refer to situations that FOLLOW the situation described by the previous clause. -ñépe also normally indicates a switch in primary actor from that of the previous clause. Nevertheless, subsequent action + different actor is never a sufficient condition for the use of -ñépe, as there are many instances of subsequent action + different actor where fully independent clauses are used. For -ñépe to be appropriate there always must be some additional component of semantic continuity. It is the exact nature of this `additional component of continuity' that is difficult to characterize in one simple functional statement. Furthermore there are motivated exceptions to the generalization that a ñépe clause must have a different actor from the previous clause (see examples 32 and 33). In this section I will illustrate several uses of the -ñépe suffix and will attempt to show that MOTIONMOTIONMOTIONMOTION and PURPOSEPURPOSEPURPOSEPURPOSE are significant components to the semantic relations encoded by a -ñépe clause. The following example illustrates -ñépe in a clause that expresses an event 25b that immediately follows the event expressed by the previous clause 25a. The two clauses also have different primary actors:

(25) a. tu'ñen y-éñ-ñe añ-iñí-'ka y-intá-n-yen. medicine TRNS-ingest-NONSPEC:T NEG-ingest-NEG 3-mouth-POSS-LOC `They take medicine, but not through the mouth.'

b. Yuwachon t-inchawo y-a'kama----ñépeñépeñépeñépe inkenpëj kúya old IRR-think 3-teach-LC:SEQ youth DAT `Then the wise old men teach the younger ones.' (MCW.76.92-93)

This example represents what I consider to be the `basic' function of the -ñépe suffix. There are two events, taking medicine and being taught by the elders, that any person who aspires to become a healer must undergo in strict sequential order. In 25a the initiands are the primary actors, whereas in 25b the elders are the actors. Since there is a switch in actor, the sequential-high continuity suffix -séjpe is inappropriate. Also, since the PATIENT of the second clause is not coreferential with the absolutive argument of the first clause, -sé'ñape is similarly inappropriate. However, since the events are closely related conceptually, verbal suffixes used for independent statements are also inappropriate for 25b.

T. Payne, Medial-clauses 14

This function of -ñépe suggests that it contrasts with -séjpe in that -ñépe indicates switch reference, whereas -séjpe indicates continuing reference. In the vast majority of cases, this is indeed the case. However, it is not a grammaticalized rule. As noted above, it is possible to use -séjpe in situations of switch reference (examples 7 and 8). Below (examples 32b and 33b) we will see that -ñépe can be used in situations of continuing reference. The central respect in which these two dependent verbal suffixes differ functionally, then, is that -ñépe indicates a higher level of DISCONTINUITYDISCONTINUITYDISCONTINUITYDISCONTINUITY than does -séjpe. Switch reference is a major component of discontinuity, but it is not the sole factor that distinguishes appropriate contexts for the use of these suffixes (see Clancy 1980, T. Payne 1985:ch. 5, Carlson 1987 and Scancarelli 1989 for similar observations in other languages).

Even the more general notion of discontinuity does not exhaust the functional differences between -séjpe and -ñépe. Clauses marked with -séjpe are pragmatically nearly independent in that there are no special semantic constraints on the use of -séjpe other than close sequentiality and high continuity. In contrast, ñépe clauses must have some other conceptual link to the previous clause. In 25 above the link is the ceremonial progression of events leading up to becoming a healer. Often, however, the link has to do with purpose and/or movement towards a goal. Examples 26 and 27 below illustrate the use of -ñépe where both movement and purpose are expressed:

PURPOSE & MOVEMENT (26) y-apo'ma-ñe e'ñapa i'yakae kën akamo y-ápï-ñépeñépeñépeñépe

3-command-NONSPEC:T person comrade AN:INVIS wasp TRNS-test-LC:SEQ `The people sent a companion / to test the wasps.' (MCO.40.19) /.and he tested the wasps.'

(27) añimonton y-á'kapa-ñe ijka pú'ma-ñépeñépeñépeñépe 2PL:PRO TRNS-OI:ask-NONSPEC:T porcupine OI:kill-LC:SEQ `I'll summon you all / to kill the porcupine.' (MCW.75.11) / and then you'll kill the porcupine.'

In these examples the adverbial clause purpose suffix -tópe is inappropriate in that -tópe does not express movement, and carries a more immediate sense of purpose. Native speakers have difficulty accepting -tópe with adverbial clauses related to situations involving movement (e.g. `send' and `summon'). This is because -tópe seems to imply that the movement event itself accomplishes the purpose, rather than simply being the means by which the actor arrives at the place where the purpose is accomplished. So, for example, if -tópe occurs on the dependent verb in 27, one speaker's reaction is that the person who would utter such a sentence must have used some device to call on the hearer, knowing that the use of that device would kill the porcupine.

Use of the suffix -ñépe on what appear to be complements of the verb -ipi `want/need/invite' not only implies switch reference between the primary actor of the dependent clause and that of the matrix clause, but also implies that the actor of the dependent clause must move in order to accomplish the action:

T. Payne, Medial-clauses 15

15

(28) ay-i'-ñe këj kën wachi y-ú'-ñepeñepeñepeñepe 2-want-NONSPEC:T AN:PROX AN:INVIS anotto TRNS-give-LC:SEQ `He needs you to (go) give (him) anotto.'

(29) í-pi-mpëj amën wachi y-ú'-ñepeñepeñepeñepe o-ya 1-want-IMPERF 2SG anotto TRNS-give-LC:SEQ 2-DAT `You need me to (go) give you anotto.'

(30) a-y-ipi-mpëj chu a-saan-ñépeñépeñépeñépe 2-TRNS-want-IMPERF 1SG NEU-ascend-LC:SEQ `I want/need/invite you to (come) go up.'

(31) ay-i'-ñe yu yuwï' mááni-ñépeñépeñépeñépe 2-want-NONSPEC:T 1SG house repair-LC:SEQ `I want you to (come) repair my house.'

The next set of examples will illustrate situations where ñépe constructions are used even when the primary actor of both clauses is the same. In these cases the notions of movement and displaced purpose seem to override the tendency for this suffix to encode switch reference:

(32) a. y-uw-ïña-n nkë apoj, y-uw-ïña-n nkë wïnkïj, 3-DT-fall-NONSPEC:I also man 3-DT-fall-NONSPEC:I also woman `The man dives in and the woman dives in too, (MCO.71.14)

b. a-s-ëmënka----ñépeñépeñépeñépe tïna y-ijpi-n-pana. NEU-DT-reach-LC:SEQ water 3-lip-POSS-towards to/and get to the (other) river bank.' (MCO.71.15)

(Two versions of this story use the same form here).

If -séjpe were used on the verb in 32b, it would mean `and then they arrived on the other side.' Their swimming may or may not have been how they got to the other side, i.e. they swam for a while, and then they went to the other side.

The second example of -ñépe used in a continuing reference context is from a personal experience narrative about hunting monkeys:

(33) a. Ti-yaj yïm tonkwana-pákë', go-PPERF1 father other-by:way:of `My father went by another route,

b. i-naapaka----ñépeñépeñépeñépe arakon tonkwanan ij-cha. 3-encounter-LC:SEQ black:monkey other hill-LOC and found monkeys in the other hills. (MKO.50.8)

In this text the father was really looking for beehives and only happened to encounter monkeys. Hence purpose is not a component of the meaning of the ñépe

T. Payne, Medial-clauses 16

clause in this case. If -séjpe were used in 33b it would not necessarily express that the father found the monkeys at the end of the movement described by the previous clause. It would simply state that he went by a different route and at some later time encountered more monkeys. If the independent transitive verb suffix -ñe is used in 33b it would indicate that the monkeys were encountered at a later time or by a different person. The suffix -ñépe is appropriate in this context because of the spatial discontinuity between the situations described by the two clauses, even though the primary actor of both situations is the same.

The following two examples illustrate the fact that the exact semantic relationship between the two clauses in a -ñépe construction is largely a matter of inference. Native speakers have no problem whatsoever understanding 34 as a case of counter-expectation, and 35 as a case of purpose, though there is nothing, other than the pragmatics of the situations, that expresss these specific relations. It is simply nonsensical to try to force a purpose reading on 34 and a counter-expectation reading on 35. The only semantic elements that these two constructions have in common are sequentiality and discontinuity of participant reference:

COUNTER-EXPECTATION: (34) t-arïti-yaj chu kapucha e'-ñépeñépeñépeñépe o'mó-'ka

1-load-PAS:IMM 1SG shotgun be-LC:SEQ explode-NEG `I loaded the shotgun, but it didn't go off.' (MKW.62.15)

NEGATIVE PURPOSE: (35) Y-uputuurë-ñe aripéñe e'----ñépeñépeñépeñépe usuká-'ka

3-roast-NONSPEC:T well be-LC:SEQ rot-NEG `They roast it well so that it won't spoil. (MCW.77.10)

4.3 Simultaneity (`overlap'): 4.3 Simultaneity (`overlap'): 4.3 Simultaneity (`overlap'): 4.3 Simultaneity (`overlap'): -tááñe, -npan

Clauses that contain either of the suffixes -tááñe and -npan fulfill the language-specific definition of medial clauses in Panare. However, under a Labov and Waletzsky (1967) definition of event, such clauses do not fulfill the universal characterization of medial clauses as 'dependent clauses that encode events'. In this section I will present the data on these suffixes, and will show that the definition of event as an 'active realis assertion' does correctly identify these clauses as medial clauses.

Both -TTTTÁÁÑEÑEÑEÑE and -NPANNPANNPANNPAN indicate that the action of the verb they attach to is simultaneous with that of the preceding clause. The two suffixes contrast in that -tááñe indicates a switch in primary actor (XR, example 36b) whereas -npan indicates continuing reference of primary actor (CR, example 37b). As with the other dependent clause suffixes that do not require coreference between the primary actors of the two clauses, -tááñe requires that the AGENT occurs in the dative case (see section 4.2). Clauses with -npan, on the other hand, never express their primary actors overtly, as the identity of this argument is made explicit via the related independent clause. As with all non--past-perfective clauses, -npan and -tááñe clauses both reference the absolutive argument when the absolutive is not expressed as a preverbal nominal:

T. Payne, Medial-clauses 17

17

(36) a. i-naapaka-ñépe arakon tonkwanan ij-cha, 3-encounter-LC:SEQ black:monkey other hill-LOC `. . . and (he) found monkeys in the other hills, (MKO.50.8)

b. ana w-utë-tááñetááñetááñetááñe ana chima-tawën tësan-kën EXC 1-go-XR:SIM EXC road-on straight-only while we went by the straight path.' (MKO.50.9)

(37) a. Anan-ti-yaj wanëta-po pakëmÿn-ya, EXC-go-PPERF honey-LOC yesterday-LOC `We went for honey yesterday,

ana-ot-arááma-npannpannpannpan arakon-pë EXC-DTRNS-hunt-CR:SIM black:monkey-OBL and hunted black monkeys.' (MKO.50.5)

The following two examples of -npan occur in succession in a text about a burial ceremony:

(38) a. kën i'yan maraka y-ÿwa-ñe AN:INVIS shaman maraca TRNS-OI:play-NONSPEC:T `The shaman plays the maraca

b. y-u-waaren-npannpannpannpan ko'mampe sïpawï 3-INTR-sing-CR:SIM late at:night and sings late at night.' (MCW.78.47)

In 38 the temporal relationship between the events described by the two clauses is coextensive, i.e. the shaman's playing of the maraca occurs over the same period of time as his singing. The following three elicited examples provide additional illustrations of these suffixes:

(39) w-oromaipún-yaj chu y-amenkë-tááñetááñetááñetááñe Tose-úya 1IP-speak-PPERF 1SG 3NP-write-XR:SIM Tose-DAT `I spoke with Toséi and/while hei wrote it down.'

(40) w-oromaipún-yaj chu Tose-úya y-amenkë-tááñetááñetááñetááñe 1IP-speak-PPERF 1SG Tose-DAT 3NP-write-XR:SIM `I spoke with Toséi and/while hej (another one) wrote it down.'

(41) w-oromaipún-yaj chu Tose-úya y-amenkë-npannpannpannpan 1-speak-PPERF 1SG Tose-DAT 3NP-write-CR:SIM `I spoke with Tosé and I wrote it down.'

According to Labov and Waletzky (1967), an event/situation is not on the main event line if its position can be changed with respect to a clearly event line clause without destroying the continuity of the narrative. Though tááñe and npan clauses always follow the associated finite clause, the events these dependent clauses encode

T. Payne, Medial-clauses 18

are presented as occurring at the same time as that encoded in the finite clause. This would be enough, according to Labov and Waletzky, to eliminate the possibility that these clauses encode foregrounded events. However, according to the characterization of an event as a non-stative realis assertion, simultaneous events are not precluded. This, I suggest, is how tááñe and npan clauses function -- they indicate parallel event lines. The question arises, then, how do we know that these are any different from adverbial clauses indicating simultaneity (e.g. `while' clauses in English)? There are at least two responses to this question: 1) Panare has another simultaneous adverbial clause suffix, -mpëj. This suffix clearly marks adverbial clauses because a) it does not itself indicate continuing or switch reference, b) it indicates states or conditions that form the backdrop for events, and c) clauses formed with -mpëj are highly nominal in character (a property shared to a lesser extent by medial clauses). And 2) tááñe and npan clauses do refer to events in every other sense of the word, i.e. realis, active assertions (though they can refer to states at times).

In summary, I have illustrated several clause types that can be insightfully described as medial clauses according to the universal characterization of medial clauses as dependent clauses that encode events. In the following section I will discuss a suffix that marks clauses that are neither prototypical medial clauses nor prototypical adverbial clauses. This suffix illustrates the scalar nature of the medial clause/adverbial clause distinction.

4.4 Previous action, continuing reference: 4.4 Previous action, continuing reference: 4.4 Previous action, continuing reference: 4.4 Previous action, continuing reference: ----jpómën

The suffix -jpómën indicates that the action described by the verb it is attached to occurred prior to the action described by a preceding or following independent clause. This suffix meets all of the characteristics of medial clauses listed by Longacre (1985:264). It is structurally distinct from finite clauses, it requires that its subject be coreferential with the subject of an associated finite clause, and it is centrally concerned with temporal relations. Interestingly, however, it does not as clearly as the other suffixes discussed so far refer to events in discourse. This is manifested by the fact that a verb suffixed with -jpómën does not normally constitute the only expression of a given situation. Either the -jpómën clause expresses a non-event, as in 42 below, or it recapitulates a foregrounded event expressed in a previous clause. This is the case in examples 43 and 44:

(42) asoonwa okëtí-jpómënjpómënjpómënjpómën w-ï-të-séjpa yu a-wë'-yaka three sleep-CR:ANTER 1-INTR-go-FUT 1SG 2-place-to `In three days I will go to your house.' (Lit: `After sleeping three, I will go to your house.') (MCW.74.9.)

(43) a. y-anompa-ñe kën yï-pïj ayawa-ke 3-paint-NONSPEC:T AN:INVIS 3-wife black:resin-INST `He paints his wife with black resin. (MCW.79.21) (First expression of painting)

T. Payne, Medial-clauses 19

19

b. y-achima-séjpe y-ï'nanpa-jpómënjpómënjpómënjpómën 3-dance-CR:SEQ 3-paint-CR:ANTER Then they dance with each other, after/because he paints her.' (MCW.79.22)

(44) a. y-ankë-sé'ñapi-nkë mën, mën y-án-sa 3-take-HC:SEQ2-also INAN INAN 3-OI:take-PASS `It is then taken and carried off, (PST.14.27) (First instance of stealing)

b. yu-të-séjpi-nkë mën y-ámï-jpómënjpómënjpómënjpómën tikon 3-go-HC:SEQ1-also INAN TRNS-OI:pick:up-CR:ANTER child and then he goes also, after/because he had picked it up.' (PST.14.28-29)

In 44b the event of stealing pears is expressed in clause a. In clause b the next event in the event chain is expressed, and then the act of stealing is recapitulated. Even though a different verb is used, it is clearly the same event that is referred to in 44a.

As indicated in the glosses, -jpómën clauses often express a reason (`because') relation as well temporal anteriority. This is understandable in terms of the fact that reasons logically precede actions. Again, the morphosyntax of Panare provides evidence that logical anteriority is a metaphorical extension of temporal anteriority.

It is interesting to note that the suffix -JPÓJPÓJPÓJPÓMËNMËNMËNMËN might be analyzed as ----JPOJPOJPOJPO `locative' plus MËNMËNMËNMËN `inanimate specifier'. My arguments for not doing this are 1) the locative postposition ----JPOJPOJPOJPO never occurs independently on verb stems in our corpus. There are some examples of other postpositions occurring on bare verb stems, but the functions of such structures are not well-understood at this point. 2) The temporal meaning of -JPÓMËNJPÓMËNJPÓMËNJPÓMËN (`after') does not follow from the meaning of ----JPOJPOJPOJPO plus MËNMËNMËNMËN. This is not to say that semantic nuances never vary depending on contexts. But it is to say that if ----JPOMËNJPOMËNJPOMËNJPOMËN is analyzable into ----JPOJPOJPOJPO plus MËNMËNMËNMËN it has developed a very specific and non-predictable meaning just in those cases where it appears on a bare verb stem.

It spite of these problems for the synchronic analysis of -jpómën into ----JPOJPOJPOJPO plus mën, this analysis unites -jpómën clauses to nominalizations in the following ways: 1) for a noun to function as a predicate in Panare it must be followed by a specifier, either këj `animate proximal', nëj `animate distal' or mën `inanimate', or an auxiliary verb if temporal grounding is required. 2) The form ----JPOJPOJPOJPO, being a locative postposition, is a member of the set of morphemes associated with noun phrases. To hedge a bit, as mentioned above some of the more prototypical medial clause suffixes also contain syllables that resemble nominalizing or noun morphology. Nevertheless, the fact that -JPÓMËNJPÓMËNJPÓMËNJPÓMËN contains what appears to be a relic of a specifier, MËNMËNMËNMËN, obviously a relatively recent addition to predicate nominal structures (i.e. because they are still free forms, and appear last in the predicate phrase), indicates that -JPÓMËNJPÓMËNJPÓMËNJPÓMËN clauses are closer to their nominal roots than are the other medial verb suffixes. From this point of view it is understandable that JPÓMËNJPÓMËNJPÓMËNJPÓMËN clauses might not embody events -- the more a structure embodies prototypical events, the less likely it is to exhibit morphosyntactic characteristics of nouns (Hopper and Thompson 1984).

T. Payne, Medial-clauses 20

5. Discussion5. Discussion5. Discussion5. Discussion

To summarize what has been presented so far: there is a clause type in Panare that is grammatically dependent yet functions to code events in discourse. Clauses of this type have been termed medial clauses. Consistent with earlier treatments of non-final or medial clauses in other languages, medial clauses in Panare are a) sensitive to reference relationships and b) centrally concerned with temporal relations such as overlap and succession. The fresh facts brought to light in this study are that medial clauses do not just reflect reference relations and temporal sequencing. Rather, the same/switch reference distinction which at first blush appears completely inviolable, can be overridden. Furthermore, non-temporal interpropositional relations often overlay the basic temporal relations expressed by medial clauses. In fact, in some discourse situations where coreference relationships and temporal sequencing of events would allow more than one type of medial clause (e.g. a sé'ñape or a ñépe clause), other factors render only one sort acceptable.

Table 1 summarizes the raw data presented so far. In the rest of this section I will attempt to show how these properties make sense in terms of a cognitive model within which spatial, temporal and logical relations are unified.

Table 1: Medial clause suffixes in Panare (HC = high continuity, LC = low continuity) (Most prototypical medial clauses are listed first)

Operator Temporal Reference Other relation relation relations -séjpe Succession (HC) Actor=Actor purpose -sé'ñape Succession (HC) Absolutive=Patient result -ñépe Succession (LC) Actor=/=Actor movement/purpose -npan Overlap Actor=Actor none -tááñe Overlap Actor=/=Actor none -jpómën Anteriority Actor=Actor reason

In the following discussion I assume a view of semantics coming to be known generally as `cognitive grammar' or `cognitive model theory'. This view stems from a tradition of cognitive and discourse based approaches to linguistic explanation culminating in such works as Langacker 1987 and Lakoff 1987. The terminology I adopt comes largely from Lakoff 1977 and 1987, and Lakoff and Johnson 1984. Under this approach cognition is structured in terms of `cognitive models', i.e. holistic images composed of various schematic elements, including participants, props, events, states, etc. and the relationships among them. Such models are conventionalized and generalized images derived from common human interactions with the world. For example, people commonly experience situations known as journeys. Such situations can be described schematically as consisting of a trajector (i.e. the person that moves),

T. Payne, Medial-clauses 21

21

a source, a current location and a goal. This model, call it the JOURNEY model, is undoubtedly well-installed in the cognitive structure of all communities, since travelling from place to place is such a pervasive human activity. Other cognitive models may be less universal.

Once a cognitive model is established, it can serve as the analogical basis for conceptualizing various other phenomena. I have already mentioned (section 2.3) Lakoff and Johnson's observations that the model termed WAR is used to conceptualize argumentation and sexuality in my culture. The JOURNEY model is particularly well suited to serve as the conceptual model for various experiences involving progression from one situation to another, including temporal succession and logical relationships such as causation and purpose. Any given event is situated in a context that includes prior and subsequent events or situations. This is analogous to any given location of a journey being related to previous and subsequent locations. Typically, prior situations can be construed as causes, or reasons for subsequent situations. Anticipated situations can similarly be construed as purposes towards which current events are directed. These metaphorical correspondences are illustrated in table 2.

Table 2: Metaphorical connections between spatial, temporal and logical relations as expressed by medial clauses in Panare.

SPATIAL DOMAIN: Origin --> Here --> Destination metaphorical | | | extension | | | | | | TEMPORAL DOMAIN: Anterior --> Now --> Subsequent metaphorical | | | extension | | | | | | LOGICAL DOMAIN: Reason -->Current Event --> Purpose

The reason for doing something is logically prior to the act of doing it, hence it is understandable that morphology associated with temporal anteriority should metaphorically be extended to encode reason. Similarly, the purpose for doing something is logically subsequent to actually doing it -- we do X in order to bring about a subsequent situation Y, where Y is the purpose for X. Therefore, it is understandable that morphology associated with temporal posteriority should be metaphorically extended to encode purpose. Counterexamples to this hypothesis are well-defined: languages in which markers of temporal anteriority are extended to encode purpose (rather than reason) and/or where markers of subsequent action are extended to encode reason.

T. Payne, Medial-clauses 22

We would hypothesize that the order of development of these functions would be first the temporal functions and then the logical functions. Our basis for this hypothesis stems from universal principles and from evidence within Panare. Universally languages tend to have more spatial and temporal connectors than logical connectors. For those languages that have documented histories, the development of tense/aspect and clause combining morphology nearly universally follows the path of space --> time --> logic (Anderson 1973, Diehl 1975, Jessen 1975, Bennett 1975, Traugott 1978). Hence I hypothesize that the logical interpropositional relations encoded by medial clauses in Panare are extensions of the basic temporal meanings, rather than reflexes of an older stage in which the medial clause suffixes were adverbial clause markers indicating logical relations. The evidence for this directionality from within Panare is that the temporal relations are more inviolable than the logical or even the referential relations. For example, what examples 34 and 35 have in common is that in both sentences the medial clause expresses an event that temporally follows the event expressed in the previous clause. The logical relationships, however, are entirely different. Interpretation of the logical relationship between medial and finite clause is partially determined by the context, whereas the temporal relationships are very explicitly expressed by the morphology.

The cognitive model designated as the JOURNEY thus motivates the correspondence between cause/reason and temporal anteriority, as well as between purpose/result and temporal posteriority. However, there is at least one other function that enters into a full specification of the use of the medial clause suffixes in Panare. This is the function of coding continuity vs. discontinuity in discourse. There are three medial clause suffixes that encode sequential posteriority. These are distinguished by whether they code discontinuity or continuity. For the high continuity suffixes, ----séjpe and -sé'ñape (see table 1), there is a distinction between actor oriented continuity and patient-oriented continuity.

We can see in these suffixes at least one respect in which coreference relations are motivated by similar cognitive patterns as are the interpropositional relations. The purpose and the result of an action are both logically subsequent to that action. The purpose, however, is instigated by the actor, insofar as the event is consciously and volitionally conceived as being directed towards the purpose. The result may or may not be foreseen by the actor. In any case, the result of an action is registered with the PATIENT. Since the purpose for an action is the responsibility of the actor, it is understandable that the suffix that indicates continuity of actor should also express events that are purposes for previous events. Similarly, since the result of an action resides with the PATIENT, it is understandable that the patient-oriented suffix, -sé'ñape, should be associated with the interpropositional relation of result. For this reason I categorize -séjpe as indicating actor/inception/source oriented continuity whereas -sé'ñape indicates patient/termination/result oriented continuity.

The suffix -ñépe, on the other hand, normally indicates discontinuity of actor participants. It matters not whether the patient of the ñépe clause is coreferential with an argument in the previous clause. The extended interpropositional relation expressed by a ñépe clause is that the event expressed occurs at a different place than the event

T. Payne, Medial-clauses 23

23

expressed by the previous clause. Hence participant discontinuity and spatial discontinuity are both expressions of the general discontinuity function of this suffix.

6. Conclusion6. Conclusion6. Conclusion6. Conclusion

In spite of the fact that Panare possesses a functionally and structurally distinct category of clauses that can insightfully be analyzed as medial clauses, it does not employ extensive clause chaining structures as a basic discourse strategy. Chains involving more than one medial clause are rare, and chains of four or more medial clauses are non-existent in the corpus. These data make it clear that the distinction between `clause-chaining' and `co-ranking' languages is not so much one of kind as one of degree. That is to say, it is possible for languages to employ clause chaining to a greater or lesser extent as a means of expressing event sequences in discourse.

This study 1) highlights the importance of the notions of EVENTEVENTEVENTEVENT and NONNONNONNON----EVENTEVENTEVENTEVENT in discourse and sentence level grammatical analysis, 2) provides a unified cognitive/functional motivation for the structural categories of dependent and medial clause, 3) suggests that the notion `clause chaining language' is not an isolated type, but rather languages may exhibit clause chaining to a greater or lesser degree, and 4) illustrates how a cognitive model motivates semantic correspondences between temporal relations and logical relations between clauses in discourse.

T. Payne, Medial-clauses 24

NOTES

1. Research for this paper was conducted under the auspices of the school of Anthropology and the Philological Institute "Andrés Bello" of the Central University of Venezuela and the Catholic University of Táchira, with funding from U.S. National Science Foundation grants BNS-8609304 and BNS-8617854, National Endowment for the Humanities grant RV-20870-87, and the Oregon Foundation. Supplementary support was provided by Petroleos de Venezuela in the form of a video tape used for elicitation. Many thanks to our colleagues Andrés Romero Figueroa, Omar Gonzales Ñañez, Paola Bentivoglio, María Teresa Rojas and many others for help and encouragement with the research in Venezuela; to Monseñor Medardo Luzardo, Father José Pulido and colleagues, Father José del Rey Fajardo, Paul and Ginny Witte, and Raúl Prieto for logistical support in Caicara del Orinoco, and to Marie-Claude Mattei-Muller, Bob Dooley, Spike Gildea, Doris Payne and Jana Price for helpful comments on earlier drafts of this paper. The latter three colleagues are also to be thanked for providing some of the Panare data from their own field notes. Finally, I'd like to thank the Panare people for the privilege of learning their beautiful and intricate language. In particular, I thank Pragedes Salas of Caicara (originally of Caño Amarillo West, see Henley 1982), Miguel Castillo of Santa Fé (also originally of Caño Amarillo West), Manuel Castro of Camana, Tosé of Maniapure, his son Tëëna (José Gregorio), and the families of all of these men.

2. Haiman (1987) uses the term `medial verb' to describe the morphologically distinct verb type that heads non-final clauses. Longacre (1985), however, seems to use the terms medial clause and non-final clause interchangeably.

3. I follow Dixon 1979 in using the terms A, S and O to refer in a pretheoretical way to transitive subject, intransitive subject and transitive object respectively:

......................................................................................... A Subject of transitive (most AGENT-like argument)

Subject

...........................................................................................................S Subject of intransitive (only argument)

.........................................................................................................................................................................S

Absolutive

............................................................................................O Object of transitive (non-AGENT-like argument)

Within this framework, the SUBJECT category consists of the set of A together with S, while the absolutive category consists of the set of S together with O.

4. Hopper and Thompson define event strictly in terms of the discourse distinction between `main line' and `collateral' information in narrative texts, based on Labov and Waletzky 1967. The approach that views functional distinctions in terms of `cognitive models' is due to Lakoff 1977 (also Lakoff and Johnson 1982, Lakoff 1987 inter alia). Givón since 1981 has used the concept of `functional domains' to capture essentially the same insight. DeLancey 1987 deals at some length with the merits and demerits of the discourse/functional vs. semantic/cognitive motivations for morphosyntactic regularity, and comes down on the side of semantic/cognitive motivation largely because of its enhanced explanatory power. However, the discourse/functional characterization has the advantage of being more empirical. As difficult as it is to distinguish main-line from collateral information in transcribed texts, so it is even more difficult to observe and analyze cognitive schema directly.

T. Payne, Medial-clauses 25

25

The point of view taken here is that the observable and measurable behavior of

persons engaged in discourse constitutes data for drawing conclusions by induction regarding the structure of cognitive models. Cognitive models undoubtedly shape the structure of discourse, but perhaps discourse shapes the structure of cognitive models as well. In any case, discourse is what can be observed, while cognitive models are extrapolations from that.

5. By `fully grounded' I mean `sufficiently grounded so as to be understood as a complete independent utterance' in Panare. This does not necessarily mean that the time, location and participant reference of the situation will always be completely explicit. Since participant reference is not a consistent verbal morphological category (it is neutralized in many clause types), this cannot enter into the definition of independent clause. Furthermore, although central participants not referenced on the verb are usually identified via pronouns or full noun phrases, in texts even these free referring devices are often omitted. Hence, in order to avoid the questionable situation of equating the presence of zero pronominal reference with dependent clause status, I do not include full participant reference (either in verbal morphology or extra-verbal referring devices) as one of the defining characteristics of independent clauses in Panare.

6. See Traugott 1978 for discussion of the correspondence between spatial and temporal

morphology. See DeLancey 1987 and 1989 and references therein for discussions of the cognitive

relationships among agentivity, causality and temporal relations.

7. Participial clauses in English often code events, but it is far from clear whether this is their

`normal' or unmarked function. The following are examples of each type:

EVENT: She walked out, locking the door behind her.

NON-EVENT: She walked out, knowing the answer.

A finite clause in the second position of either of these chains would practically require an eventive reading.

For example, even non-active verbs in gapped finite clauses take on an eventive, in this case inchoative,

sense:

She walked out and knew the answer.

My intuition is that in this sentence knowing the answer is to be understood as the eventive process of

coming to know the answer, i.e. upon walking out the subject suddenly becomes aware of the answer. Though it might be argued that participial clauses more closely approximate the Longacre definition

of medial clauses (i.e. special morphology, subject co-reference and temporal semantics) than do finite

clauses with zero-pronominalized subjects, the latter are closer to the prototype according to the

definition suggested in this paper.

8 Kiparsky 1968 notes that examples such as 1d are possible if the non-initial clause is understood as historical present.

9 This suffix is pronounced -POMËNPOMËNPOMËNPOMËN by some of our consultants.

10. At several points in this paper I use the term `primary actor' to refer to the most AGENT-like argument of either a transitive or intransitive verb. The term `subject' is not

T. Payne, Medial-clauses 26

appropriate since the primary actor is sometimes demoted to an oblique case role. The term AGENT will be reserved for primary actors of transitive clauses. This term also must be understood as denoting a prototype. Any given AGENT, in this sense, may be more or less agentive. It is less obvious that semantic AGENTs (i.e. highly potent, dynamic initiators of events) constitute the prototype for the category of unitary arguments of intransitive verbs. Hence the term primary actor, rather than AGENT, is chosen to subsume unique arguments of intransitive clauses and the most AGENT-like arguments of transitive clauses. From an autonomous syntactic point of view such as Relational Grammar, one might substitute a term such as `initial 1' for primary actor. However, this term is inadequate on other grounds. Since the focus of this paper is interclausal semantic relations and not intraclausal grammatical relations I hesitate to burden the discussion with a lengthy digression into my bases for choosing one terminology over another. In another paper (T. Payne 1990) I discuss at more length issues of subjecthood and grammatical relations in Panare.

11. Most of the examples in this paper are from texts collected by myself and Doris Payne between October 1986 and April 1987 in Caicara de Orinoco. These data are indexed according to the speaker, whether the text is oral or written and a numerical index to the specific text and sentence number. For example, example 4 is followed by the index MCW.76.65. MC refers to the speaker, Miguel Castillo. W indicates that the text was written. O would indicate that the text was orally composed. 76 is the text number, and 65 is the sentence number within that text. Texts with the index PST.... are `pear stories'. These are oral retellings of a short film produced by the department of linguistics at the University of California at Berkeley. The film was designed to elicit a variety of grammatical and discourse structures (see Chafe 1980). Unfortunately the names of the narrators of the pear stories were not recorded. Examples that have no index are from my personal field notes. Examples with the index JP were graciously provided by Jana Price from her field notes, consisting of transcribed texts and conversational material.

Abbreviations used in the examples in this paper are the following: 1 First person 2 Second person 3 Third person AN:INVIS Animate, invisible (pronoun) AN:PROX Animate, proximal (specifier) COMP Complementizer COND Conditional CR:ANTER Continuing reference, anterior CR:SIM Continuing reference, simultaneous DAT Dative case DC Deictic center DERG De-ergative DETRANS/DT Detransitivizer EXC Exclusive (`first person' exclusive) FUT Future HAB Habitual HC:SEQ1 High continuity, Sequential, actor oriented HC:SEQ2 High continuity, sequential, patient oriented I Intransitive INP Intransitive non-past perfective IMPER Imperative IMPERF Imperfective aspect INAN Inanimate INAN:VIS Inanimate, visible (demonstrative pronoun) INCHO Inchoative INST Instrumental

T. Payne, Medial-clauses 27

27

IRR Irrealis LC:SEQ Low continuity, sequential LOC Locative NEU Person-neutral inflectional prefix NONSPEC:I Non-specific aspect, Intraansitive NONSPEC:T Non-specific aspect, transitive OBL Oblique postposition PASS Passive PERF:INVIS Perfect aspect, subject not visible PL Plural POSS Possessed noun phrase PPART Past participle PRO Pronoun PURP Purpose adverbial clause marker TP Transitive past-perfective TRNS Transitive XR:SIM Switch reference, simultaneous

12. The form i'yakae is based on the root yaka, meaning roughly `other'. It is the basis for many expressions dealing with perceived identity relationships between persons. The central concept that these forms embody seems to be something like `member of the same settlement group as X'. In this example, the term i'yakae refers to three boys who are of the same race and general age as the boy who is the main character of the story to that point. We have glossed this use as `counterpart', as there is no necessary assertion that they are family members in the English sense, though i'yakae also applies to a person's brothers, sisters and extended family members.

The stem meaning `to fall over' is grammatically a detransitivization of the verb PU'MAPU'MAPU'MAPU'MA 'mortally wound by hitting'. Its semantics are idiomatic. The non detransitivized past participle YUPÚ'MASA'YUPÚ'MASA'YUPÚ'MASA'YUPÚ'MASA' would be the proper expression of the notion of `the mortally wounded/dead one'.

13 Some imperatives that involve movement from the location at which the command is given are marked with the suffix -'MEMEMEME in the Western Panare (Colorado) area, and the particle MËNMËNMËNMËN in all other areas. Other movement imperatives employ the suffix -TATATATA in all areas.

14. By `preceding' here I mean `logically preceding'. This includes temporal precedence. As mentioned below, there are cases where séjpe clauses can be used apart from any formally instantiated independent proposition. However, even in these cases (ex. 12), there is some latent proposition that logically precedes the séjpe clause and upon which the séjpe clause depends for its referential and temporal grounding.

15. This example is difficult to translate for at least two reasons. First, the word ëëwëpëto is not a

verb. It is a noun that refers to the celebration of nose puncturing. The word for nose, ëëwë, is clearly incorporated into this term. Second, the verb warupútë means `to initiate with loincloth.' It refers literally to the ceremonial act of putting on a loincloth for the first time. For both men and women, this constitutes the

initiation into adulthood (see Henley 1982:141-150). Though it is unusual for a mono-morphemic verb root to

have more than two syllables, I have yet to distinguish any synchronic derivational morphology within this

root. The noun that refers to the initiation ceremony itself is kátya y-Ån-to, loincloth TRNS-put:on-PURP, `for putting on the loincloth.'

16. Since I analyze -sé'ñape as a kind of passive marker (see T. Payne 1990) the PATIENT in clauses 14b and 15b is an S argument, rather than an O. In these examples I will use the notation X=Y to indicate coreference between arguments of two clauses in sequence. The

T. Payne, Medial-clauses 28

terms X and Y can be instantiated by semantico-syntactic role labels, e.g. S, A and O, absolutive, etc. or by semantic roles such as PATIENT. X refers to an argument of the first clause, and Y refers to an argument of the second clause. The notation X=/=Y indicates non-coreference.

17 The expression TA'MEÑETA'MEÑETA'MEÑETA'MEÑE, deriving from the root TA'META'META'META'ME 'undesirable, ugly', has the connotation here of many people gathering with a skeptical attitude, out of curiousity. The form TYAJTYAJTYAJTYAJ is the bare root of the verb meaning 'listen'. The use of a bare verb root as a dependent verb form is grammatically non-standard. That is, in elicitation, consultants will change this form to a form inflected with -SÉJPESÉJPESÉJPESÉJPE, -MPËJMPËJMPËJMPËJ, -ÑEÑEÑEÑE or some other appropriate operator. Nevertheless, the use of bare roots is relatively common in discourse.

18. Many thanks to Marie-Claude Mattei-Muller for this insight.

19. As an illustration of the functional variability of to as a dependent clause introducer in English,

consider the following examples:

COMPLEMENTIZER: He wants John to wash the dishes.

This is a complement because the dependent clause is clearly the direct object of the main clause, cf. `He

wants X', X = `John wash the dishes' (`He wants that John wash the dishes' plus subject to object raising).

ADVERBIAL (PURPOSE) CLAUSE INTRODUCER: He washed the car to impress his girlfriend.

cf. `He washed the car in order to impress his girlfriend'.

INDETERMINATE He invited John to come to his birthday party.

This is not a prototypical complement, as evidenced by the fact that John, not the dependent clause, is the underlying PATIENT of the main clause, i.e. this is not a case of `raising' to direct object, cf. `He invited X',

where X = `John'; X cannot be the proposition `John come to his birthday party' (*He invited that John come

to his birthday party). However, neither is this a purpose adverbial clause, i.e. it does not mean `??He invited

John in order to come to his birthday party.' This sort of indeterminacy is also apparent with the suffix

-ÑÉPEÑÉPEÑÉPEÑÉPE in Panare.