The Structure and Functioning of Clauses in Upper Kuskokwim Conversational Discourse

Abstract

Upper Kuskokwim (Athabaskan, Alaska) is a polysynthetic language with morphologically complex verbs involving pronominal affixes denoting clause arguments. One goal of this paper is to see how clauses in this kind of language are organized and operate in conversational discourse. This study is based on a dataset of transcribed conversations, arranged as sequences of elementary discourse units. The issues explored in this article include the structure of clauses, their functioning in discourse, the composition and expression of clause arguments and other participants, as well as an assessment of more and less typical clauses. I find that clauses are strongly aligned with elementary discourse units; that there is a preference for verb-centered, independent, and one-place clauses; and that lexically expressed arguments are rare. Overall, the clause is a viable notion for the description of Upper Kuskokwim conversational discourse. The specifics of clause structure and clause functioning in Upper Kuskokwim can be explained by a combination of general principles of discourse production and the typological features of the language.

1. Introduction

Upper Kuskokwim is a language of interior Alaska, belonging to the Athabaskan (Dene) language family (which is the core of the larger Na-Dene family). Upper Kuskokwim (henceforth: UK) is a strongly endangered language, still remembered by some in the village of Nikolai (Collins, 2004). UK is among the 11 or 12 Athabaskan languages of Alaska and was traditionally spoken in the upper drainage of the Kuskokwim River, in the middle of Alaska. UK is very close to other adjacent Athabaskan languages: Lower Tanana, Koyukon, and Holikachuk. Less close are other Athabaskan languages neighboring the UK territory: Deg Xinag (Ingalik) and Dena’ina.

Like other Athabaskan languages, UK is a polysynthetic language. This means that numerous grammatical meanings are conveyed by bound morphology inside the verb form—much more numerous than in an average human language (e.g., Fortescue et al., 2017). As a result, verb forms have complex morphology and often include many morphemes. In particular, verbs include pronominal morphemes referring to clause arguments. A typical UK clause consists of a verb form with pronominal affixes plus, occasionally, some adjuncts and some particles. In the context of the present special issue, it is interesting to look at a polysynthetic language, since clauses in such languages are quite different from the common expectations based on English. Studies such as Thompson (2019) or Laury et al. (2019) make us wonder how clauses function in conversations in a wide range of typologically diverse languages. This is a case study of a polysynthetic language. (See Nakayama, 2002 for a discussion of clauses in a different kind of polysynthetic language).

In this article, I address the following issues: the representation of clauses in UK conversational discourse; the functions of clauses; the internal structure of clauses; the number of clause participants; and the expression of clause participants. I also offer a discussion of the presented facts, putting them in a theoretical and typological context. I demonstrate that the notion of clause is quite adequate for the facts we observe in UK conversations, though one can legitimately differentiate between fully typical and less typical clauses. The functioning of clauses in UK conversation depends both on general cognitive and interactional parameters of discourse production and on the typological properties of the language.

2. Materials and Methods

One of the main issues explored in this article is the expression of clause participants. So, it is important to mention a few key notions from this domain. I follow the common distinction between two kinds of clause participants: arguments and adjuncts. Arguments are imposed by the verb’s lexical properties. This imposition is semantic and, in many languages, formal. As will be clear in the discussion of example (1) below, UK arguments are conveyed by particular formal slots in the verb’s morphological template. (In UK, some arguments are expressed outside of the verb form. There is also a difference between core and non-core arguments. These aspects of argument structure remain beyond the scope of this article (A. A. Kibrik, 2012b)). Adjuncts are those participants that are not predictable from the verb’s lexical properties.

(Core) arguments are aligned in different ways across various languages. UK follows the accusative alignment, in which the agent-like arguments of transitive verbs are encoded the same way as sole arguments of one-place verbs, irrespective of the semantic role of the latter. In contrast, patient-like arguments of transitive verbs are expressed differently. Following (A. E. Kibrik, 1997), I call these two kinds of generalized role functions Principal and Patientive, respectively; see (A. A. Kibrik, 2012b) and the discussion of example (1) below.

Being polysynthetic is very common for the native languages of North America (see Mithun (2017)). Polysynthesis basically means the quantitative complexity of a typical verb form in a language. In addition, Athabaskan languages, including UK, are characterized by what can be called qualitative complexity, with respect to both inflection and derivation (see A. A. Kibrik, 2005, 2012a). The Athabaskan verb is heavily prefixing. Inflectional and derivational prefixes are intermingled. The order of inflectional morphemes is typologically unusual. There is a lot of multiple exponence, that is, the expression of a single grammatical meaning by two or more morphemes in different parts of the verb form. Morphophonemics is complex, so that many underlying morphemes do not show up in phonology.

In order not to overload readers with too complex morphological and morphophonemic phenomena, here, I provide just one example of a relatively straightforward verb form involving two pronominal prefixes referring to the clause arguments and offer detailed glosses.1 (This verb form also appears in example (12) below).

(1)	tse-	hi-	m-	ts’i-	na-	n-	e-	yut
	out-	3Pl-	3:Pat-	1Pl:Prin-	Der-	Conj-	Perf-	chase
	‘we chased them out’

The glosses provided in (1) are less than perfect and do not render all the important features of the morphemes. To give a better idea of the morphological structure, the following explanation may be helpful. Eight morphemes in the verb form, from left to right, include the following:

• The derivational prefix tse- ‘out’.
• An inflectional marker of third-person plurality.
• An inflectional marker of third-person Patientive or, more precisely, third-person non-Principal, occurring when the Principal’s person is other than the third; together with the previous morpheme, this marker is conventionally used to encode a third-person plural Patientive.
• An inflectional first-person plural Principal marker.
• The derivational prefix ni-, of a rather general meaning, here, surfacing as na-.
• The inflectional prefix n(i)- that can be interpreted as a conjugation-type marker.
• An inflectional perfective prefix of high allomorphy, here, surfacing as e-.
• The root -yut ‘chase’; more precisely, a root allomorph appropriate for the given tense-aspect-modality context.

This list could be extended with some additional zero morphemes, but that is not relevant for my current goals. Also, it is worth mentioning that the UK verb form follows a morphological template, so that each morpheme occupies a particular slot. The number of slots in an Athabaskan verb, depending on one’s approach, may be anywhere between one and two dozen, cf., Kari (1989), Rice (1989, p. 425), A. A. Kibrik (2005), and Kari (2024, pp. 536–537); see Rice (2000) for an alternative theory of Athabaskan morphology. In particular, the Principal and Patientive markers in (1) appear in strictly defined morphological slots. The placement of a pronominal morpheme in a particular slot encodes the role of the corresponding referent in the situation (see A. A. Kibrik, 2012b). As is discussed in Section 4, pronominal morphemes in the verb are full-scale arguments, even though separate noun phrases may share the argumental status with them. See Rice and Saxon (2005) for a comparative account of Athabaskan verb forms and clauses.

Because of the quantitively and qualitatively complex nature of UK verb morphology, I refrain from providing even reduced glosses such as in (1) in the subsequent examples. Showing full-fledged glossing of Athabaskan verb forms makes sense only in those contexts when the internal morphology of the verbs is the specific focus of interest. In other contexts, such glosses would become a mere distraction from the questions under discussion; see the analysis in Hargus et al. (2019). Below, I provide maximally simplified word glosses, avoiding an in-depth analysis of the verb forms.

This study is based on my field research on UK, conducted since 1997. The data include natural and arranged conversations recorded between 1997 and 2010. Already in 1997, and more so in the 21st century, Upper Kuskokwim was almost completely replaced by English in everyday communication. Still, some samples of conversational discourse could be collected. First, in 1997, I arranged two conversational meetings (28 and 47 min long) at which proficient speakers agreed to talk with each other in their native language, even though this practice had been almost abandoned. In 2001, a classroom interview was organized at the Nikolai school, in which questions were asked to a proficient speaker of UK about her life experiences; this was also an approximation to a natural conversation in the case of a strongly endangered language. Finally, in 2010, I asked a latent speaker, born in the early 1960s, to record conversations of his parents who were among the very last couples that still used UK on a daily basis. This undertaking lasted for six days, and about four hours of natural conversations were recorded. Interestingly, these conversations could only be recorded with permanent radio sound in the background, which was a normal accompaniment to their routine life. The radio could not be turned off as that would make speakers feel unusual and therefore uneasy. All three kinds of conversations were transcribed and translated with the help of my consultants in Nikolai. The recordings and transcripts are stored in the Alaska Native Language Archive, and some are available online in the “Andrej A. Kibrik Upper Kuskokwim Collection” at www.uaf.edu/anla (accessed on 14 January 2025). The analysis below is based on a dataset of 50 min of transcribed conversations. Appendix A provides a list of the conversational sessions included in the dataset, as well as codes of all speakers participating in the analyzed conversations.

The transcripts of UK discourses are organized as sequences of elementary discourse units (EDUs) (see A. A. Kibrik, 2019b—a study specifically devoted to EDUs in Upper Kuskokwim). EDUs are fundamental building blocks of discourse production, identified on the basis of similar criteria in many languages. These criteria are prosodic in nature and include the presence of a primary (rhematic) accent, an integral tonal contour, tempo and loudness patterns, and boundary pauses. It may be relevant to mention at this point that there is substantial literature on Athabaskan intonation and, more generally, discourse prosody, including studies such as McDonough (2003), the collection Hargus and Rice (2005), Hargus and Abou (2008), Schwiertz (2009), Berez (2011), Lovick and Tuttle (2012), McDonough (2018), and Palakurthy (2019). Among these studies, the ones by Berez and Palakurthy are particularly close to the present study in their methodology of identifying intonation units, being quite close analogs of the EDUs posited here.

In the practice of discourse transcribing, the above-mentioned criteria are used to represent quantized local discourse structure as a sequence of EDUs with a sufficient level of confidence. Further details on EDUs and principles of discourse transcription as applied to UK are provided in (A. A. Kibrik, 2019b). The dataset explored in this study includes 650 EDUs. In the context of this article, many elements of discourse transcription, such as pauses, disfluencies, intonation, pitch accents, etc., are ignored as they are not directly relevant to the topic. Also, in the analysis below, some EDUs are excluded from consideration, specifically, those fully consisting of code switching to English, truncated EDUs, and regulatory EDUs without propositional content.

As is known from various studies, such as Chafe (1994), a prosodically identified EDU displays non-trivial content-related properties. Cognitively, an EDU manifests a focus of consciousness. Semantically, an EDU typically conveys an event/state. Grammatically, an EDU often coincides with a clause (see also Croft, 1995 or Thompson & Couper-Kuhlen, 2005, inter alia). Below, I will address the latter point in detail with respect to the UK data. Since EDUs are building blocks of spoken discourse, the EDU structure is a firm foundation and a good starting point for exploring how clauses are instantiated in discourse. In accordance with this, I will start the discussion of clauses from the question of how clauses are identified vis-à-vis EDUs: Section 3.1 of the main section of this article (Section 3) addresses the clausal structure of EDUs.

The identification of clauses per se is relatively problem-free in UK. Since inflected verbs are so distinct in UK grammar, it is easy to recognize a typical clause as a structure centered around a verb form (and often coinciding with a verb form). Consider an example from a conversation involving three turns and four consecutive EDUs, each consisting of a single clause.

(2) Home conversation A
AgN:		ˀidiyats’		miye	tonoˀeneł
		really		in.it	you.pour.it.back
‘pour it back in the pot’ (about a piece of meat)
		diyats’		nich’ikosr
		really		something.is.rendered
‘it is really rendered’ (it is so lean that it is hard to swallow)
JN:	FINE		ts’eˀ	ch’ik’oˀiłnech
			Ptcl	he.cooked.something
‘he really cooked good’
AgN:		dineje		mamaˀ	ghotiniłnagh
		moose		food.Poss	I.not.crave.it
‘I am tired of eating moose meat’

Among these four clauses, one is a one-place clause (the second EDU), while all the rest are two-place, transitive clauses. Such a condensation of transitive clauses is not typical in UK conversation (see below), but it is good for illustrative purposes at this point. Each of the verbs in the transitive clauses includes two arguments, a Principal and a Patientive. These arguments are expressed within verb morphology. In the fourth line, one of the arguments, the Patientive, is additionally conveyed by a noun phrase, literally meaning ‘food of moose’. The first EDU contains an adjunct, ‘in it’, which is a postpositional stem with a third-person pronominal prefix. The third line involves an English code-mixed adverb (instances of code mixing are conveyed with all caps in transcripts), followed by a particle integrating this adverb in the clause structure (see A. A. Kibrik, 2004 on details about this particle). The first two EDUs contain the particle ‘really’ (in two slightly different variations). This example gives an idea of how UK conversational discourse is organized. In Section 3, I discuss the key questions associated with the functioning of clauses in UK conversation.

3. Results

3.1. Structure of Elementary Discourse Units vis-à-vis Clauses

As was pointed out above, there is a cross-linguistic tendency: speakers shape their EDUs as syntactic units, specifically, clauses. This has been found for many typologically different languages (see Section 4), and this is observed in UK as well. Example (2) includes four clausal EDUs. Example (3) demonstrates two consecutive EDUs, none of which coincides with a clause: the first EDU is more than a clause, and the second one is less than a clause.

(3)	Coffee party, BP
	nitsitł’a = ye		silak’a	ˀełchut
	it.is.small = Rel		on.my.palm	you.put.it
	‘give me the small one’ (the small cinnamon roll)
	ˀede	ts’inh
	that.one	from
	‘from that one’ (from the plate)

The main predicate of the first EDU in (3) is the three-place word ‘give’ that is conventionally phrased in UK as ‘put something on someone’s palm’. The Patientive of that verb is expressed within the inflected verb form and, additionally, by a relativized stative verb, ‘be small’. Therefore, the first EDU involves two clauses and can be called superclausal.

The second EDU in (3) is prosodically independent from the prior structure, though it is semantically very closely connected to the material in the first EDU. It is an adjunct to the main clause of the first EDU indicating the locative source: from where the cinnamon roll should be taken by the addressee. The second EDU in (3) is subclausal. It is a retrospective specification of the material found in the base clause, ‘you give it to me’, and consists of a demonstrative and a postposition; these kinds of specifications are sometimes called increments. Such subclausal specifications, uttered as prosodically separate EDUs but semantically integrated with a preceding base clause, are highly common in the discourse of various languages. There is a wide variety of subclausal EDUs in UK; see (A. A. Kibrik, 2019b).

Returning to superclausal EDUs, these typically come to be for the following reasons: a matrix clause plus an embedded clause; a clause with an inserted element that is a verb form morphologically but functions more like a particle synchronically; something like a serial construction in which more than one verb is required to describe an event; and a clause participant that is a relativized verb (as in the first EDU of (3)). A rather extreme example of a superclausal EDU appears in (4).

(4)	Home conversation B, AgN
	nongw	hwt’ana	łeka	t’iyats’	yitsa	dilghwsr	dine	ne	ˀin
	uphill	people	dogs	really	they.bark	they.howl	they.say	he.said	it.is
‘they say dogs of those people uptown are barking and making noise, he said, indeed’

Example (4) contains a clause that has a predicate consisting of two verbs, ‘bark’ and ‘howl’, that, in conjunction, describe the dogs’ behavior. This clause is embedded within a quotative construction with the matrix verb ‘they say’. This structure is again embedded within a broader quotative construction with the matrix verb ‘he said’. Finally, the whole thing is embedded within a construction with an existential matrix verb that could be translated as something like ‘really’ or ‘indeed’. According to the most likely prosodic analysis, (4) is uttered as a single EDU. There is a less probable prosodic analysis of (4) in which an EDU boundary is posited after the particle t’iyats’. In this case, the part before the break serves as an anticipatory topic EDU, while the second part still remains an EDU with five verbs in a row. So, in any case, (4) presents a highly complex superclausal EDU. This is a very rare occurrence: most superclausal EDUs include just two verbs. Quite often, superclausal EDUs involve phonologically and semantically light verbs, such as the two final verbs in (4) that function nearly as particles.

A quantitative analysis of the dataset gives rise to the following generalizations. A total of 72% of all analyzed EDUs are clausal, and 6% are subclausal, that is, they contain less material than a clause. The remaining 22% of EDUs are superclausal, most typically consisting of a matrix clause and an embedded clause. Cross-linguistically, this is a very high percentage of superclausal EDUs; see Section 4 for a discussion.

3.2. Functions of Clauses

In this subsection, I address the functions of clauses in UK conversation, that is, the difference between independent and dependent clauses, with the latter class falling into the types of complement clauses, adverbial clauses, and relative clauses (see Shopen, 2007). This subsection is based on the results of an earlier study (A. A. Kibrik & Markus, 2009) that looked specifically into the functions of UK clauses and used a different but intersecting dataset of UK discourse. This dataset included 750 clauses. The general trend discovered in that study was that the share of independent clauses was very high: 86.1%. This number includes concatenated clauses that, under a certain syntactic analysis, can be considered clause coordination (see A. A. Kibrik, 2004). Example (5) demonstrates one of the strategies of clause coordination with the help of a comitative postposition, ił, which, in this case, is attached to a clause uttered by way of English code mixing.

(5)	Coffee party, PE
	SHE TOOK SAW		ił
			with
	‘she took the saw and’
	miˀił	ch’ididit’oz = e		ˀił	hwts’its’aneyo
	with.it	something.is.sawed = Rel		with	she.left
	‘she left with the saw’

(The comitative postposition ˀił occurs three times in (5), each time in a different function. The first occurrence, as already mentioned, marks a non-final coordinate clause. The second occurrence, miˀił ’with it’, is a part of a lexical item, ‘saw’, which literally translates as ‘an object with which something is sawed’. The third occurrence is just a marker of a comitative adjunct).

Among the 13.9% dependent clauses, the most frequent type is complement clauses (9.8%), with the dominant (7.5%) subtype being clauses in a quotative construction with a matrix verb of speech or thought (see (4)). Less frequent (3.7%) are adverbial clauses. Example (6) shows three dependent clauses: an adverbial temporal clause in the first EDU and double clause embedding in the second EDU, including in a quotative matrix clause.

(6)	Interview at school, LP
	nezeyonh	daˀ
	you.grow.up	when
	‘when you grow up’
	nigugaˀena	ch’ihinˀesh	ts’idihtoneł	siłne	ts’ihwnagh
	your.children	they.steal.something	they.will.become	she told me	instead
	‘your children will be stealing, she told me instead’

(Example (6) is part of a story about the speaker’s grandmother who refused to allow the girl to have a pet puppy).

Finally, the rarest type (0.4%) of dependent clauses are relative clauses. An example of these appeared in (3) above. Note that forms with a relativizing enclitic, such as in (3), are the only kind of non-finite verb forms in UK. There are no converbal forms of any kind.

Overall, UK demonstrates a strong dispreference for using dependent clauses in conversational discourse, even though the grammar possesses devices to construct complex sentences with subordination. This dispreference is overridden primarily in cases where a verb lexically requires an embedded complement; more often, this happens with matrix verbs of speech or thought.

3.3. Internal Structure of Clauses

In this subsection, I return to the main dataset explored in this article. Most typically, a UK clause is built around a verb form functioning as a predicate. The Athabaskan verb is so morphologically specialized that identifying it as a nucleus of a clause is straightforward. A sufficient number of typical verb-centered clauses was demonstrated above (see examples (2)–(6)). A total of 80% of all clauses in the dataset are verb-centered.

Atypical clauses, other than verb-centered ones, account for 20% of the dataset. Three salient kinds of atypical clauses missing a verb predicate are nominal (6%), adjectival (5%), and negative-existential (4%). These three kinds of clauses are illustrated in the following three examples, respectively.

(7)	Coffee party, BP
	mega	ghelheˀ
	bread	maybe
	‘bread, maybe’ (a response to the question ‘What do you call this?’)

(8)	Home conversation A, AgN
	OVEN	goyajit = heye	ˀidiyats’	RICH	k’int’a
		small.meager = Rel	really		it.seems
	‘even that little piece [of meat] in the oven is really rich, it seems’

Two comments are in order regarding example (8). First, UK has very few true adjectives (one of them is goya ‘small’, appearing in this very example) and sometimes uses code mixing, inserting English adjectives. In (8), the adjective RICH appears as a clause predicate. Second, the final word k’int’a morphologically is a verb; it functions as a matrix verb, with all of the preceding material in (8) serving as its complement.

(9)	Home conversation B, AdN
	denje	kwl	ts’uk’wda
	money	not.exist	yet
	‘there is no money yet’

The predicate of the clause in example (9) is the negative-existential predicative word kwl ‘none, absent, does not exist’.

The remaining 5% of the non-verb-centered clauses are paraclausal structures, most of them verificational, such as in (10), and a few interjections.

(10)	Coffee party, BP
	ghwlaˀ
	unknown
	‘I don’t know’

The word ghwlaˀ in (10) may etymologically be related to one of the existential verb roots, but currently, it functions as an unanalyzable particle.

Potentially, non-verb-centered clauses may also include vocatives, onomatopoeias, and other holophrases, but such examples were not encountered in the dataset.

3.4. Number of Clause Participants and Clause Arguments

Clauses are linguistic manifestations of events or states. Accordingly, event/state participants are conveyed in clauses by linguistic expressions, called clause participants. In the dataset, the number of clause participants varies normally from zero to three and, very rarely, more than three. Since clauses with zero to three participants were sufficiently illustrated above, here, consider a rare example of a clause with four participants.

(11)	Coffee party, PE
	yada	ˀeko	dinaˀił	k’oy’nalˀesh?
	what	for	with.us	he.is.hiding.something
	‘what for is he hiding from us?’

Two major kinds of participants include arguments and adjuncts. The clause in (11) has two adjunct participants, expressed by separate phrases: ‘what for’ and ‘with us’. As for the arguments, there are two of them, Principal and Patientive, expressed inside the verb ‘he is hiding’. As for the Principal, the verb in (11) involves a zero pronominal morpheme that clearly refers to the third person. An explanation is due for the Patientive: why is it that the one-place situation ‘hide (oneself)’ is conveyed by a two-argument structure? The UK transitive verb ‘hide’ involves two participants, a Principal and a Patientive. The intransitive variation of this verb, illustrated in (11), uses the indefinite pronominal prefix y’- as the clause’s Patientive. Literally, the intransitive ‘hide (oneself)’ means ‘hide something’. Since the Patientive slot in the verb is filled by the indefinite pronominal prefix, it is faithful to the language to say that this verb involves a Patientive participant.

Table 1. Number of clause arguments: illustrating examples.

characterizes the clauses contained in the above-presented examples in accordance with the number of arguments.

Shaded in Table 1 are those clauses that, in addition to arguments, have an extra adjunct participant. Accordingly, the clauses in the first and the third EDUs of example (2) have three participants, and the clause in the second EDU of example (5) has two participants. Apparently, in the set of examples used in this article, there are not too many instances of adjunct participants. However, in the whole dataset, the breakdowns of clauses according to the number of all participants vs. arguments alone are somewhat different.

Table 2. Number of clause participants: shares in the dataset.

Zero	One	Two	Three or More
7%	49%	38%	6%

and

Table 3. Number of clause arguments: shares in the dataset.

Zero	One	Two	Three
7%	71%	21%	1%

show how the clauses in the whole dataset fare in terms of the number of all participants and arguments, respectively.

As Table 3 suggests, three-place verbs are very rare. Most typically, they are represented by transfer verbs such as ‘give’ or speech verbs; see examples (3) and (6), respectively. Two-place verbs, most of them transitive, are perfectly well equipped with the language’s fundamental grammatical resources but are quite rare in conversational discourse. Clearly preferred are clauses with just one argument. If we look at all participants (Table 2), general preferences remain the same, but the language is more tolerant of having two and even three participants.

3.5. Participant and Argument Expression

As people talk, they often mention referents that are already activated in their cognitive system; see the notion of activation cost proposed by Chafe (1994). Activated referents are expressed by reduced referential devices, such as free or bound pronouns or zero forms; this point was formulated in various ways on many occasions (see, e.g., Gundel et al. (1993)). Furthermore, activated referents are often sustained in the cognitive system for a certain time (A. A. Kibrik, 2011, p. 383). This leads to a high frequency of clauses in which all arguments are expressed by reduced referential devices. In a polysynthetic language such as UK, this means that a typical clause consists of a verb form alone, sometimes with some additional particles. For example, consider example (2) cited above: all arguments of three out of four clauses and one of the two arguments in the fourth clause are encoded exclusively by bound pronouns. Two similar examples are presented in (12) and (13).

(12)	Coffee party, PE
	tsehimts’inaneyut	k’int’a
	we.chased.them.out	it.seems
	‘it seems we have chased them out’

(13)	Coffee party, DE
	“nut	zisdo		dihighˀin	ˀine
	here	I.stay		because	it.is
	dohdisnaˀ“		yinezinh		ghelheˀ
	they.are.quiet		he.thinks.it		maybe
	‘“it is because of me being here they are quiet” maybe he thinks’

The glosses in both examples suggest for each verb form which and how many arguments are expressed by that verb form. For example, in (12), the first verb ‘we chased them out’ contains the explicit elements him- (third-person plural Patientive) and ts’i- (first-person plural Principal), while the first verb ‘I stay’ in (13) contains the first-person singular pronominal morpheme s-. In UK, the majority of activated arguments are expressed only within verb morphology; one exception concerns clauses with transfer verbs in which the recipient, even when activated, is expressed by a phrase outside of the verb (see the first EDU of (3)). Among the examples above, we have seen quite a few clauses consisting of an inflected verb form alone; these concern, for instance, the clause ‘we chased them out’ in (12) and the clause ‘they are quiet’ in (13).

Of course, there are instances when an argumental referent is mentioned with a lexical referential expression, particularly because it is not fully activated at the moment or for some other reason. For example, see (2), the fourth EDU, where the relatively heavy noun phrase ‘food of moose’ is used.

Adjuncts can be mentioned with reduced referential devices, as in the first EDU of (2) (‘in it’), but generally, they are more prone to full referential expression than arguments (see, e.g., ‘with the saw’ in (5), second EDU). These two examples demonstrate that UK adjuncts, either activated or not, are expressed outside of the verb form.

Table 4. Clause arguments and clause participants: type of expression.

	All Participants	Arguments
Verb-internal only	52%	58%
By means of a separate phrase2	48%	42%

shows the shares of morphological vs. verb-external expressions of clause arguments and participants in the dataset.

It is interesting to look specifically at the clauses with two arguments, that is, those headed by a transitive verb. How common are situations where both arguments are expressed lexically? Dominant are situations in which no or one argument is expressed lexically: 43% and 48%, accordingly. We have seen above a fair number of both of these kinds of transitive clauses. Only 9% of clauses with two-place verbs have two lexically expressed arguments; see example (14), in which both lexically expressed arguments represent English code mixing.

(14)	Home conversation B, JN
	ˀiłt’e	MAP	nełˀanh	THE PILOT
	always		he.sees.it
	‘that pilot always watches the map’

Prosodically, this material is pronounced as a single EDU. In fact, the word order in (14) is highly unusual: UK is a verb-final language, and postverbal noun phrases usually turn out to be retrospective specifications, that is, additional subclausal EDUs. If such an interpretation is chosen, even this example would not deviate from the general principle “no more than one lexically expressed argument in a clause”.

4. Discussion

Upper Kuskokwim, just like Athabaskan in general, is a very unusual language. The ordering of morphemes in a verb form is typologically exceptional (see, e.g., Rice, 2000; A. A. Kibrik, 2005). Equally striking is the generally prefixing character of the Athabaskan verb, with the background of the cross-linguistic preference for suffixation; for example, Dryer (2005) demonstrates that only about 6% of the world’s languages are strongly prefixing, and Athabaskan languages are the only strongly prefixing languages in the USA/Canada part of his sample. In many ways, Athabaskan languages, and UK in particular, are as far from “standard average European” as you can possibly get.

However, in some respects, Upper Kuskokwim behaves in a rather expected way, if we base our expectations on relatively familiar languages. The identification of clauses is quite straightforward. The functioning of clauses in discourse is also more “normal” than what one may expect in a typologically unusual language. In this section, I discuss the same five aspects as in Section 3, putting them in a broader theoretical context and assessing the UK system from the point of view of its typological (a)typicality.

The structure of elementary discourse units vis-à-vis clauses: The UK local discourse structure, as in other known languages, is a sequence of EDUs, and prosodically identified EDUs demonstrate a high degree of coincidence with clauses.

Table 5. Share of clausal EDUs in various languages in various studies (after A. A. Kibrik, 2019b).

Language, Study	Percentage of Clausal EDUs
English (Chafe, 1994)	60%
Mandarin (Iwasaki & Tao, 1993)	39.8%
Sasak (Wouk, 2008)	51.7%
Japanese (Matsumoto, 2000)	68%
Russian (A. A. Kibrik & Podlesskaya, 2009)	68.6%
Upper Kuskokwim (A. A. Kibrik & Markus, 2009)	70.8%

shows the share of clausal EDUs (or intonation units) in various languages, according to a number of studies. To be sure, the methodologies of counts and the discourse types explored differed across these studies. However, it is remarkable that in most instances, the share of clausal EDUs is between one half and two thirds. (See the discussion on this generalization in A. A. Kibrik, 2019b, p. 145).

Note that Table 5 includes a percentage for Upper Kuskokwim, taken from a different study that was based on a different dataset compared to the present study; that dataset included mostly monologic discourse. We can infer from this that the observation about the high number of clausal EDUs in Upper Kuskokwim is quite robust. Remember that in the present study, the share of clausal EDUs in conversation was found to be 72%. Relying on the evidence in Table 5, I suppose that the share of clausal EDUs is correlated with the degree of a language’s (i) morphological complexity and (ii) grammatically marked distinction of inflected verbs from other kinds of words. UK fares high for both of these parameters, and this may be the reason why we see a larger share of clausal EDUs than in the other languages listed in Table 5. To phrase this hypothesis somewhat differently, perhaps languages overtly marking verbs as dedicated predicative elements more consistently align clauses with EDUs. To verify this hypothesis, one needs to perform a thorough analysis of many different languages, including other polysynthetic languages.

It must be noted, however, that Lovick and Tuttle (2012) found that in Dena’ina, another Alaskan Athabaskan language, only 46% of the discourse units coincide with clauses. I suppose this may be due to their different methodology: they define discourse units in terms of “pause groups”, which is an idea quite remote from what I call EDU in this study.

Of course, as in any language, in UK conversational discourse, there are EDUs not coinciding with clauses. An important difference of Upper Kuskokwim compared to better studied languages is a relatively high number of superclausal EDUs. In a different dataset, mostly based on monologic UK discourse, the share of superclausal EDUs was found to be 14% (see A. A. Kibrik, 2019b, p. 155). If one compares this with the results for Russian, a relatively well-explored language, the UK number is already more than twice as high as in Russian (A. A. Kibrik et al., 2020, p. 43). However, in the conversational dataset underlying the present study, the share of superclausal EDUs turned out to be even higher: 22%. As I reasoned in (A. A. Kibrik, 2019b, p. 155), the relatively high frequency of superclausal EDUs may possibly be due to the polysynthetic character of the language: as information is packed more tightly in verb morphology, it may be easier to collapse more than one verbal predicate within the bounds of a single EDU.

The functions of clauses: Much emphasis is laid in linguistic theory on complex sentences, see, e.g., Shopen (2007). The evidence from UK conversations suggests that speakers of the language have a strong preference for using independent clauses. Mithun (1984) and Jacobsen (1992) demonstrated similar phenomena in other polysynthetic languages. The main exception from this general trend in UK is the use of quotative clauses, embedded in matrix clauses with a verb of speech or thought. This exception is quite natural as these kinds of complement clauses are lexically predetermined by the properties of matrix verbs. More discourse-oriented kinds of dependent clauses are rare: there are few adverbial and relative clauses. Also, UK has very few devices for creating non-finite clauses. These observations offer a contribution to the ongoing discussion of the role of clause subordination from a cross-linguistic and functional perspective (see Laury and Suzuki (2011)).

The internal structure of clauses: Typical clauses headed by a verb form account for a large majority of all clauses in the UK conversational dataset. Typical clauses may consist of a verb form plus some particle or may include a noun phrase or another separate phrase. Atypical UK clauses have a predicate other than an inflected verb form. These kinds of clauses are, in the first place, clauses with nominal, adjectival, and negative-existential non-verbal predicates; cf., Haspelmath (2025) with an alternative classification of non-verb-centered clauses.

The number of clause participants and clause arguments: UK conversational discourse prefers clauses with one or two participants. As for argumental structure, there is a strong preference for clauses with just one argument. This observation is in line with the findings of other languages. In fact, the numbers we have seen for UK conversational discourse are very close to the counts of Thompson and Hopper (2001) who found that in English conversation, the shares of one-place and two-places clauses are 73% and 27%, respectively.

I believe that the limitations on the typical number of participants stem from the basic cognitive limitations of the human attentional system. Attention is a limited resource and a selective process, and a person can only attend so much at a given time. When one produces discourse, attention to entities, or referents, is represented as a mention of those entities (A. A. Kibrik, 2011, p. 376). When we describe events or states, we only pick a few participant entities that can be included in those events or states. As we have seen, the limitation on the typical number of arguments, that is lexically predetermined participants, is even more severe. This must result from the process of event/state conceptualization in the semantics of verbal lexemes: solidified representations of events/states are even more restrictive in the number of concurrently attended entities.

Participant and argument expression: It has been pointed out before, for a number of languages, that lexical expressions denoting arguments are not common (Ono & Thompson, 1997; Ewing, 2019; and Mithun, 1987). We have seen that UK also follows this pattern. Clause arguments in UK discourse are often expressed only by verb morphology, such as pronominal affixes; this is typical for polysynthetic languages (see, e.g., Chafe (1994, pp. 148ff)). Pronominal affixes are full-fledged manifestations of clause arguments. As is typical for polysynthetic languages, when an argument is expressed by a full lexical phrase, it is expressed in verb morphology as well. In other words, pronominal elements on the verb are present irrespective of whether there is a coreferential full noun phrase. (For some exceptions to this general principle in UK, see (A. A. Kibrik, 2011, p. 198). Also see (Rice, 2003) on the discussion of this phenomenon in Athabaskan languages). In (A. A. Kibrik, 2011, p. 96), I used the term “tenacious” to describe this property of pronouns in a language such as UK. In this connection, a question arises: if there is a full lexical phrase and a tenacious pronominal morpheme on the verb, both concerning the same referent, which form has the property of argumenthood? Consider example (15).

(15)	Athabaskan lunch, MK
	łats	k’onwhdineł
	dirt	you.guys.are.moving.it
	‘you guys are moving dirt’

The verb in (15) definitely contains morphological elements expressing both the Principal (nwh-, second-person plural) and the Patientive ‘dirt’. If there were no noun phrase in the clause, the verb would still be a clause with the meaning ‘You guys are moving it’. So, the Patientive pronominal morpheme, even though with zero expression in this case, is definitely within the verb.

Therefore, the Patientive is referred to both by that pronominal morpheme and the noun phrase. As I argue in detail in (A. A. Kibrik, 2019a), there is no need to insist that either the noun phrase or the pronominal morpheme is the sole manifestation of the argument. We can allow for the idea of double manifestation in these kinds of instances, thus avoiding an artificial and forced problem of choice. So, when there are two manifestations of a referent within a clause, the argument is expressed by both in conjunction.

Returning to the question of third-person morphological zero in a modified version of (15), from which the Patientive noun phrase is excluded, I can testify that speakers of UK consistently explain and translate these kinds of structures as involving a clear third-person referent. That is, the idea of ‘it’ is definitely expressed in the verb, even though we cannot posit any explicit form that marks it. This situation is different from what Ono and Thompson 1997 describe for “syntactic” zeros in Japanese, suggesting that this kind of zero reference is vague in some sense. There is clearly a need for a broader theory of putative zeros, including morphological and syntactic ones. A recent monograph by Givón (2017), “The story of zero”, only addresses the functioning of “syntactic” zeros.

As was pointed out above, there are not too many two-argument clauses in the conversational dataset. If we look at argument expression specifically in this modest number of examples, we see that clauses with two lexically expressed arguments are very rare. Usually, just one argument is manifested lexically, or no argument at all. This should not be surprising, as similar facts have been reported for other languages. Chafe (1994) used the notion “one new idea constraint”. This constraint suggests that the human cognitive system activates from a previously inactive condition no more than one item at a time. So, UK discourse follows this strategy that is likely to be a general feature of human cognition.

To sum up, our observations definitely support the clause as a basic unit of conversational discourse in UK. Just a priori assumptions about typical clauses, based on written language and the tradition of non-empirical theorizing about language, are not supported by the evidence from UK conversations. Such assumptions might make one expect to see many clauses such as The farmer kills the duckling that were mocked many times by discourse-minded linguists as very improbable examples of real language use. We see instead that conversation is full of sequences such as He got mad, he kicked it, it died. In a polysynthetic language such as UK, this imaginary sequence would consist of three clauses, each involving just one word, an inflected verb.

A polysynthetic language with a morphological verb template involving slots for pronominal arguments indeed favors a clause-based analysis even more strongly than languages like English that only require syntactic slots around the verb. In fact, morphology appears as a more robust and more nuclear part of grammar compared to syntax, which is more fluid and more prone to various discourse-based factors.

5. Conclusions

Relying on the analysis presented above, we can now summarize what constitutes typical and atypical clauses in Upper Kuskokwim conversational discourse.

Very typical is a clause with a verbal predicate. This kind of clause has one argument, expressed solely within verb morphology by a pronominal morpheme. Apart from the verb form, a typical clause involves a particle and/or an adjunct.

Moderately typical are clauses with a verbal predicate involving two arguments. The arguments are expressed by verb morphology. One argument may or may not be additionally represented by a lexical noun phrase. Again, there is often a particle and/or an adjunct.

Atypical are clauses with predicates other than verbs, such as nouns, adjectives, or negative-existential words; transitive clauses with two lexically expressed arguments; and clauses with three arguments.

Completely absent from the dataset are clauses with more than three arguments.

Furthermore, clauses most often coincide with elementary discourse units, established on prosodic grounds. However, superclausal EDUs are more common in UK than in some other languages, and this may be related to the polysynthetic nature of the language. Clauses are usually syntactically independent and finite, even though the UK grammar has some devices to express clause subordination and the non-finite status.

Overall, we can conclude that the profile of UK with respect to clause structure and clause functioning in discourse is derived partly from the general cognitive and interactional parameters of discourse production and partly from the basic grammatical features of the language. Hopefully future research will help to better understand the relative contributions of these two kinds of factors.