Clause-level coordination and discourse continuity in Tohono O’odham

1 Introduction

The purpose of this work is to present a Minimalist analysis of coordination in Tohono O’odham, a Southern Uto-Aztecan language, with a specific emphasis on CP (i.e., clause) level coordination. We argue that the special ku- prefix, characteristic of CP coordination, is a switch reference complementizer. We argue that ku- is syntactically merged in C⁰ in Tohono O’odham and that it consistently marks switch-reference semantic interpretations throughout the Uto-Aztecan languages examined in this work. Furthermore, in Tohono O’odham, O’ob No’ok, and Comanche, it appears to mark obviation not only between subjects but also between eventualities occurring in a successive linear fashion.

Subsequently, we present the characteristics of category-sensitive coordination in Tohono O’odham within the general cross-linguistic typology of coordinate structures. We then go over a few examples highlighting exactly how c-selection of the conjuncts is realized by different coordinators in the language and highlight the word-order flexibility of Tohono O’odham clauses.

We then narrow our focus to CP coordination to illustrate how the special form of the auxiliary that surfaces in between two coordinate clauses can suffix onto C⁰ phonologically realized as ku-, which we follow Hale (1983) in analyzing as a marker of obviation deriving from the Uto-Aztecan switch reference system. We show that the auxiliary suffixing onto a functional element high in the left periphery is not unique to ku- and that ku- itself is not restricted to coordinated structures. To this end, we present data from information questions that license ku- in the left periphery, polar questions introduced by n-, subordinate clauses introduced by m-, and Tohono O’odham narratives where ku- can function as a marker of discourse discontinuity. We also offer a diachronic account of ku- as originating from a Proto Uto-Aztecan suffix *-ko ^∼ -ku and present cognates throughout the language family.

2 Tohono O’odham coordination strategies

Tohono O’odham is a Southern Uto-Aztecan language spoken in Southern Arizona and Northern Sonora, Mexico by the Tohono O’odham. It is mutually intelligible with Akimel O’odham and closely related to O’ob No’ok. Zepeda (1983) describes the system of coordination as sensitive to the syntactic categories being coordinated. Haspelmath (2004:7) notes that category-sensitive coordination is common cross-linguistically, being found in roughly half of the world’s languages. A further typological split exists in such languages depending on whether VP-coordination employs the same strategy as NP-coordination or clausal coordination (Haspelmath 2004:8). This difference can be represented as follows:

Tohono O’odham’s category-sensitive system for coordination is typologically closest to strategy (1a) since VPs in the imperfective aspect and NPs use the same coordination strategy, distinct from clausal coordination. However, we see in later examples that it makes a further distinction between imperfective and perfective VP coordination, making (1a) an over-simplistic generalization.

In instances of coordinating lower in the syntactic structure than the CP level, the grammar utilizes (k)c (Zepeda 1983:24). Below are examples for both NPs in (2) and (3) and imperfective VPs in (4). These examples also highlight the flexibility of Tohono O’odham word order. Except for the auxiliary invariably surfacing as the second constituent in the clause, nominal constituents in simple declarative sentences can appear in any order (Zepeda 1983:130). Thus, intransitive sentences may surface as Subject-Verb or Verb-Subject (Zepeda 1983:129):

The g determiner present in (2b) is always omitted in a sentence-initial NP (Zepeda 1983:13). However, it surfaces preceding any sentence-internal NPs. In instances of coordination, it surfaces only once, preceding the first NP in the linear order. Note that NP juxtaposition cannot be interpreted as coordination; the coordinator must be repeated between each NP in the conjunct:^[3]

In (4a), we see the phonologically determined alternation between kc and c, as the conjunction surfaces as kc if and only if the preceding word is vowel-final (Zepeda 1983:25). In (4a), the first verb in the linear order cicwi ‘play’ is vowel final; therefore, the conjunction surfaces as kc. In (4b), the ordering of the verbs within the coordinated VP has been changed to illustrate the other allomorph of the coordinator. Furthermore, the constituents within the clause have been rearranged to show the possibilities of Tohono O’odham word order:

A separate strategy is attested for coordinating VPs in the perfective aspect (Hale 1983:7). In these instances, the grammar employs (-)k, which may fuse onto a perfective verbal form and prevent truncation (Hale 1983: 7), as evident in the example below:

When k is not post-verbal, perfective truncation takes place,^[5] and k does not fuse onto the preceding element (Hale 1983:7):

These data indicate that Tohono O’odham does not quite fit the typological generalization made in (1). Similar tripartite divisions of coordination strategies are cross-linguistically attested (e.g., Somali in Haspelmath (2004:8)). However, in Tohono O’odham, aspectual considerations determine the appropriate strategy for VPs. Thus, we can amend (1) to account for these data:

We leave an account of the aspectually-determined alternation in VP coordination to future research and turn to the main focus of this work; let us consider some examples of coordination of finite clauses.

When coordinating separate clauses in the imperfective aspect with distinct subjects, the auxiliary of the second clause surfaces between the two clauses in phonologically reduced form, agreeing in person and number with the subject of the second clause (Zepeda 1983:26). In the following examples, we show how two separate clauses can be conjoined using this strategy. The (a) line in each example represents the sentences as they would appear independent of each other, and the (b) line represents their structure when coordinated. In this initial section, we have glossed all coordinators as exhibiting both properties of conjunctions and auxiliaries. As we proceed with our analysis, we can see that this is not an entirely accurate description:

Once again, if an NP that would otherwise appear in sentence-initial position is now sentence-internal, the g determiner must be realized:

Thus, we see a special form of the imperfective auxiliary that is employed specifically for CP coordination. The full paradigms of the imperfective auxiliary and its special coordinative forms are given in Tables 1 and 2.

As a descriptive generalization, initial-vowel deletion holds between the first and second-person of the auxiliary and CP conjunction (Zepeda 1983:26). The third-person auxiliary, however, does not follow this pattern, surfacing as k (e.g., (6)). Below, we analyze the emergence of k as a reflex of the Uto-Aztecan obviative switch reference prefix ku- that has undergone final-vowel deletion. For the purposes of this work, we adopt the definition of switch reference presented by Baker (2020:1):

SR is a “grammatical phenomenon in which one clause is morphologically marked to show whether or not one of its nominal arguments refers to the same entity or entities as a nominal argument of a structurally nearby clause.” SR contrasts between same subject (SS) (i.e., proximate) and different subject (DS) (i.e., obviative) marking.

To initiate this line of reasoning, we now present a variant of CP coordination that gives phonological realization to the ku- prefix in the first and second person. Consider the following example:

Thus, the paradigm for these forms can be understood as being composed of two diachronically distinct parts: the obviative marker ku- and the appropriate auxiliary agreeing in person and number with the subject of the second clause, as in Table 3.

Hale (1983:15) notes that the obviative prefix ku- and the perfective proximate (-)k illustrated in the coordinate constructions in (5–6) above have separate diachronic origins. We return to the diachrony of ku- in §3.2, presenting cross-linguistic evidence and a phylogenetic tree to support an obviative analysis of ku- throughout the language family. We analyze such obviation not only in terms of subject reference but also with respect to the identification of separate eventualities in a temporal sequence. Although we have seen above that the appropriate allomorph for VP coordination is aspectually determined, the same does not hold for CP coordination, as the strategy of perfective CP coordination is the same as that of imperfective CP coordination.

Before turning to specific instances of perfective CP subordination and coordination with ku- in (13–14), consider the following illustrative example highlighting the morphological changes between the imperfective (12a) and perfective aspects (12b):

The full paradigm for the perfective auxiliary is given in Table 4.

Glossed instances of perfective clause coordination are not attested in Hale (1983); however, the SR origins of the language’s strategies for coordination are highlighted in (13), as the same strategy for CP imperfective coordination is employed for SR perfective subordination, signaling a switch in topic (Hale 1983:9):

There is, however, a clear instance of perfective coordination in Scancarelli (1988), where the auxiliaries are inflected following the perfective paradigm, and lexical verbs undergo truncation. We may thus deduce that the ku- + AUX strategy remains the same for CP coordination with an auxiliary in the perfective aspect:

Hale (1983:9) notes that the first-person auxiliary nt in (13) is the residue of kunt following deletion. As is the case in the imperfective aspect, both the form with and the form without ku-are grammatical. Overall, licensing of ku- or of a null element bearing the same features in a higher projection of the subordinate clause in (13) is consistent with its suggested obviative nature, as the subjects of the two clauses are, in fact, distinct. The superordinate clause surfaces with the second-person singular auxiliary apt, and the reduced auxiliary of the subordinate clause is the first-person singular nt.

(13) effectively broadens the picture of possible analysis of Tohono O’odham coordination. In the following section, we explore the appearance of ku- in information questions, as presented by Zepeda (1983:55). Furthermore, we present further corpus data from a collection of Tohono O’odham stories published by the Summer Institute of Linguistics (SLI) in 1969 (here cited as Enos et al. (1969)) to highlight the pervasive use of ku- as a discourse marker in narratives.

3 A closer examination of obviative ku-

3.2 A diachrony and comparative typology of ku-

Following a comparative approach to diachronic lexical reconstruction, Hale (1983:16) suggests the Tohono O’odham obviative prefix ku- derives from a Proto-Uto-Aztecan subordinating suffix *-ku ^∼ -ko. Numerous synchronic cognates in the language family support this claim.

In O’dam, the particle cu is categorized by Willett (1991:219–220) as an “additive conjunction used only with clauses,” signaling that “the action in the preceding clause is […] necessary to enable the action in the clause introduced by cu.” Thus, Willett (1991:234) suggests that there is a common ground between cu and O’dam subordinators of purpose, cause, and reason, noting, however, that cu serves a broader purpose as a generalized CP conjunction between eventualities with a “higher degree of continuity” and can function as “something like an inflectable auxiliary” on which the subject enclitic is suffixed:

(20)

O’dam
Bai’-p	xɨ-minda-’	gu	ja’á	cu	jòtom	baiy-a’	gu	bav.
TWD-2SG	IMP-burn-FUT	DET	pot	so	fast	cook-FUT	DET	beans.
‘Fire up that pot so the beans will cook fast.’ (Willett 1991:220)

Hale (1983:16), on the other hand, suggests that O’dam cu indicates a “break in continuity” between eventualities, as it is not consistently obviative between subject referents. These contradictory analyses can be reconciled if we extend obviative phenomena beyond subject reference to include temporally or causally ordered eventualities that remain, however, distinct between the two clauses (see also (13) in Tohono O’odham). We leave the specifics of such a proposal for O’dam to future research; however, we hold to the tentative conclusion that ku- and its cognates switch between subjects as well as distinct eventualities as markers of narrative discontinuity.

Comrie (1983:19) notes the occurrence of a DS suffix ku- in Wixárika. Unlike other Uto-Aztecan languages, the productive usage of the Wixárika SR system is limited to temporal clauses:

(21)

Wixárika
ʔuuka	nua- ku ,	nee	ne- petɨa.
Girl	arrive-DS	[1SG]	1SG-leave
‘When the girl arrived, I left.’ (Comrie 1983:19)

In the much less closely related Ute, Givón (2011:102) notes the usage of a suffix -ku as a DS marker in “equi-object” verbal complements (22a) and a subordinator marker for “most adverbial clauses” (22b):

(22)

Ute

a.

Mama-chi	’áapa-chi	wʉ́ʉka-vaa- ku	may-kya.
Woman-S[BJ]	boy-O[BJ]	work-IRR-COMP	tell-ANT
‘The woman wants the boy to work.’ (Givón 2011:102)

b.

…kulkwi-kwa-pʉga-y- ku ,	’uwa-rugwa-pʉga-y-k…
fill-go-REM-O[BJ]-SUB	him/O[BJ]-give-REM-OBJ-it
‘… when it (the bowl) filled up, (he) gave it to him…’ (Givón 2011:102)

Another cognate in the Northern branch of the Uto-Aztecan language family is also present in Hopi, where it is realized as [-q] from an underlyingly vowel-final/-qö/^[6] (Jeanne 1978:67). Jeanne (1978:183) remarks that it is the “sole obviative ending for adjoined clauses in Hopi”:

(23)

Hopi
ʔi-pava j	paki- q	(pɨɨʔ)	pam k	qatɨptɨ.
My-brotherj	enter-OBV	(then)	hek	sat:down.
‘When my brotherj. came in, hek sat down.’ (Jeanne 1978:183)

Another reflex of Proto-Uto-Aztecan *-ku ^∼ -ko can be found in the Comanche -ku (Charney 1993:231–235). Interestingly, -ku surfaces on the verb of a subordinate clause if the subject of the verb of the subordinate clause is distinct from the subject of the verb of the main clause and if “the time of the verbs is simultaneous” (Charney 1993:231), as seen in the following example:

(24)

[ninse wihnu uvásaʔeeku huhnupehtu nunuraʔeYU]
nini-se	wihnu	u-pasa-ʔe- ku
we=EXCL-CNTR	and=then	it=OBJ-dry-RPT:ASP-ku
hunu-peHtu=nu-nuta-ʔe-yu=
creek-toward REDP-run-RPT:ASP-PROG:ASP
‘When it (mud) dried we ran for the creek.’ (Charney 1993:231)

In cases where the subjects of the two clauses are distinct and where the eventualities denoted by the verbs occur at different times, the embedded verb is suffixed by either -h/H/ka = or by -h/H/kaku. Crucially, kaku indicates a “longer duration for the subordinate verb” and combines -h/H/ka = and -ku (Charney 1993:232), demonstrating that the synchronic descendants of *-ku ^∼ -ko are not limited to switching between subject referents but also serve different functions within the event structure of the clause.

Hale (1983:14) further notes the possibility that Mayo -k ^∼ -ko and Tubar -ko may also be cognates to Tohono O’odham ku-. Whether or not this conclusion in Hale (1983) is ultimately correct, the fact remains that the morpheme in question is widespread throughout the language family and consistently switches subject reference or identifies separate eventualities within a temporally ordered sequence. A phylogenetic representation is given in Figure 1.

Figure 1:

Phylogenetic tree of the Uto-Aztecan switch-reference marker.

However, there remains a split between the Tepiman branch and other more distantly related languages. In the former, it is a prefix hosting the auxiliary, and in the latter, it is a verbal suffix. This difference is schematized below:

(25)

(26)

Thus, there must have been a reanalysis of the affixal nature of *-ku ^∼ -ko in the Tepiman branch of the UA language family. Originally a post-verbal suffix, as in the other Uto-Aztecan languages mentioned above (Hale 1983:15), it was transcribed as a pre-nominal prefix in the earliest written records of the Tepiman branch, such as the Arte de la Lengua Névome, a grammar of O’ob No’ok (i.e., Névome) written by Jesuit missionaries in the 17th century and published by Buckingham Smith in 1862 alongside the Doctrina Christiana y Confesionario, a confessional and short book of doctrine.^[7] In the Arte, it surfaces preceding the second-person marker p and the auxiliary -ta, which are cognates with the Tohono O’odham second-person singular perfective ’apt (Hale 1983:15):

(27)

Névome
Taso	hakida	ay	si-m’	himi	tani	an’ t’
Yesterday	at:this:time	hither	really-you	go	order	[PFV].1SG
igui, co
igui, OBV
‘p’ ta	ta	pima	ay	s’ imi	mu.
[PFV].2SG	ta	NEG	hither	really-go	DESI
‘Yesterday at this time, I ordered you to come here, but you didn’t want to come.’
(Hale 1983:16)

Interestingly, it is not uniformly the case that is consistently obviative with respect to subject reference. Consider the following example:

(28)

Névome
Dodaki	Padre	divia-na,	co	hunu	ni	maca-na.
OPT	priest	come-SBJV,	[OBV]	corn	me	give-SBJV
‘Would that the priest would come to give me corn.’ (Hale 1983:17)

In (28), the subjects of the two clauses are coreferential. However, this example fits in a broader understanding of ku- and its cognates as identifying separate eventualities within a temporally ordered sequence, as is characteristically the case for O’dam in (20).

As seen in §2, the auxiliary in O’odham consistently surfaces as the second constituent in the linear order. Thus, given a stage in the protolanguage where the entire protasis amounted to the first-place constituent, the structural profile of the protolanguage is as follows:

(29)

Hale (1983:18) suggests that the linear string in (28) could have been sufficiently ambiguous for the acquiring child to trigger a reanalysis of ku- as a prefix hosting the auxiliary. Such a reanalysis would be consistent with the distribution of other prefixes hosting the auxiliary in the left periphery, as we have already seen with the polar question complementizer in §3.1. As further evidence, also consider the complementizer m- introducing both adverbial relative clauses (30a) and CP complements (30b):

(30)

a.

Hegai ceoj mo cipkan ’o wuḍ ñ-we:nag.
Hegai	ceoj	m-o	cipkan	’o
DEM.3SG	boy	COMP-AUX.IPFV.3SG	work:IPFV.SG	AUX.IPFV.3SG
wuḍ	ñ-we:nag.
COP	POSS.1SG-brother.
‘That boy that is working is my brother.’ (Zepeda 1983:105)

b.

Hegai ’uwĭ ’o s-ma:c map ’a:pi mumku.
Hegai	’uwĭ	’o	s-ma:c	m-ap	’a:pi.
DEM.3SG	woman	AUX.IPFV.3SG	knowing	COMP-AUX.IPFV.2SG	2SG
mumku.
sick.
‘That woman knows (is knowing) that you are sick.’ (Zepeda 1983:108)

As we can see, m- fits the pattern described thus far, where the auxiliary inevitably raises to the left periphery to suffix onto an overtly realized complementizer. Thus, ku- in synchronic Tohono O’odham has been reanalyzed to follow the broader pattern of complementizers in the language.

4 Conclusions

This work explored the syntactic and semantic functions of the Tohono O’odham ku- prefix as a marker of CP coordination, switch reference, and discourse continuity, paying special attention to its diachronic origins in Proto-Uto-Aztecan and its synchronic cognates throughout the language family. It analyzed how ku- patterns with other complementizers in Tohono O’odham as a prefix high in the left periphery of the clause, although its diachronic origins have been traced in the literature to a Proto-Uto-Aztecan post-verbal switch reference suffix *-ku ^∼ -ko. In fact, in a number of Uto-Aztecan languages (e.g., Ute, Comanche, Tubar), it has remained a suffix. Furthermore, it appears to be the case in Tohono O’odham and throughout the language family that cognates of ku- not only switch between subject referents but also play an important role in switching between eventualities denoted in the clause and discourse with respect to their relative order of occurrence.

Thus, the novel import of this work is that switch reference systems in Uto-Aztecan languages should also be analyzed primarily considering the sequence of distinct eventualities in the discourse, even with coreferential subjects, and not exclusively as markers of subject discontinuity. Given these generalizations, further research would have to consider the distribution of temporal adverbs in clauses containing an SR morpheme and consequent temporal restriction on subject-coreferential clauses with ku- and its cognates. The pervasiveness itself of descendants of Proto-Uto-Aztecan *-ku ^∼ ko and the different ways in which they have been reanalyzed can shed light on different syntactic strategies to switch between subjects and eventualities in adjoined clauses. An implicature-based theory of SR can account for non-canonical DS SR and offer an intriguing path forward in understanding how the GCIs associated with this type of DS SR marking can become conventionalized to specific aspects of discourse.