2 Ergativity in Nez Perce

Nez Perce is a Sahaptian language spoken in Idaho, Washington and Oregon, USA.Footnote ^[5] The language is highly endangered; recent estimates count no more than 30 native speakers, all above the age of 65 (Harold Crook, p.c.). The data in this paper come from fieldwork on the Nez Perce Reservation in Lapwai, Idaho. Data are presented in the practical orthography used by the language program of the Nez Perce Tribe. A table of correspondences to IPA is given in the appendix.

Nez Perce has a nominative–accusative system of verb agreement alongside a case system that varies between nominative–accusative (for 1st and 2nd person) and tripartite ergative (for 3rd person).Footnote ^[6] The basic pattern is exemplified for 1st person in (4) and 3rd person in (5). As (5) shows, intransitive subjects, transitive subjects and transitive objects are all marked distinctly in the 3rd person.

The verbal agreement system directly distinguishes 3rd from non-3rd person and plural from non-plural number. Non-plural number and 1st and 2nd person – henceforth, ‘local person’ – are not marked on the verb overtly. The overt markers consist of the five prefixes listed in (6), along with the portmanteau suffixes listed in the rightmost column of (7).Footnote ^[7]

Plural verb agreement occurs only for animate arguments (Deal Reference Deal2015b). In the imperfective, habitual and imperative, the plurality of an animate subject is marked as part of a portmanteau suffix, as in (7). The basic form of the aspect/mood suffix is used if the subject is singular and/or inanimate. In other aspect/mood categories, the plurality of an animate subject is marked by the plural subject prefix pe; this prefix is simply absent for singular and/or inanimate subjects. Full paradigms for verbal agreement are given in Deal (Reference Dealto appear).

Case is marked by suffixes which attach to nouns and, optionally, to numerals, quantifiers, demonstratives and attributive adjectives. The table in (8) lists the core structural cases and their common allomorphs. Note that ergative and genitive are marked in the same way, as is often true in ergative languages.Footnote ^[8]

With the exception of relative clauses, noun modifiers are reliably prenominal.Footnote ^[9] Case marking on prenominal modifiers plays an important role in Section 4; it is exemplified in (9)–(11).

Nominative, ergative, genitive and accusative forms of personal pronouns are given in the tables in (12). Anticipating the conclusion of Section 4, I leave the cells corresponding to ergative local pronouns blank.

Note that the plural 2nd and 3rd person pronouns are identical in all contexts except as a transitive subject. In this environment, the 2nd person subject is nominative (’imé), whereas the 3rd person subject is ergative (’iméem).Footnote ^[10]

These examples can be diagnosed as unambiguously transitive thanks to the presence of the applicative verbal suffix.Footnote ^[11]

At the clausal level, the order of major constituents is quite flexible, and pronominal subjects and objects of all persons are often omitted.Footnote ^[12] Omitted arguments are indicated by pro in Nez Perce examples, with the gloss line reflecting the person and number features conveyed by the speaker’s translation; for ease of reading, I follow a convention of placing pros in SVO order. The person and number of a missing argument are frequently recoverable from the verbal inflection.

Recall, however, that verbal inflection includes no special markers for 1st or 2nd person. Clauses containing a 1st or 2nd person argument are typically disambiguated by the use of full pronouns or by the presence of a clitic from the table in (16).Footnote ^[13]

These clitics most commonly appear in immediate preverbal position, and may double the full pronouns.

Unlike the full pronouns, the clitics may not be coordinated or host focus suffixes such as -cim ‘only’ or -k’u ‘also’. They also differ from full pronouns in that they do not mark case distinctions. They may occur with subjects, as in (13) and (17), as well as objects, as in (18) and (19). No parallel set of clitics exists for 3rd or (non-inclusive-plural) 1st person arguments.

The ergative character of Nez Perce is confined to its case system. The language does not show syntactic ergativity in A’ extraction.Footnote ^[14] Verbal morphology remains constant across declaratives, wh-questions and relative clauses; there is no special clause type for extraction of an ergative.

The language also does not show an ergative split conditioned by clausal properties such as tense or aspect. Examples (23) show that 3rd person transitive subjects are ergative-marked across the language’s three tenses: present, recent past and remote past. These examples also show that ergative marking appears in the imperfective aspect.

Examples (24) show that 3rd person transitive subjects remain ergative in other aspectual categories, such as perfective and habitual.

Likewise, negation and clausal embedding have no effect on the appearance of the ergative case. The language’s ergative split is strictly on the basis of person.

With this background, we turn in the next section to proposed explanations for person-based split ergativity. We return to the Nez Perce facts in Section 4.

3 Two approaches to person-based split ergativity

Theoretical approaches to person-based split ergativity may be divided into two groups depending on the type of explanatory mechanism posited. On the morphological approach, the relevant mechanisms are active at the syntax–phonology interface, regulating the realization or exponence of case features assigned in syntax. On the syntactic approach, the relevant mechanisms are active in the syntax itself. In this section I briefly introduce the two styles of analysis before presenting arguments from Legate (Reference Legate2014) in favor of the morphological approach.

The central insight of morphological approaches is that patterns like (1)/(2) constitute an instance of syncretism. Subjects of all persons are assigned an ergative case feature in ergative languages. Person splits result when, for local persons, nominative and ergative share a morphological form. Applied to Nez Perce, this leads to a view of the pronominal system as partially depicted in (25): ergative forms of local person pronouns exist, but are identical to nominative counterparts.

The partial syncretism between nominative and ergative has been attributed to a variety of sources.Footnote ^[17]

The core of the syntactic analysis, by contrast, is that what you see is what you get: the absence of an ergative case form for 1st and 2nd person subjects is due to the failure of syntactic ergative case assignment.Footnote ^[19] It is not simply that ergative versions of the local pronouns are realized in a special way in languages with a person-based split; rather, in such languages, the relevant syntactic objects do not exist. A number of potential causes for their non-existence have been explored.

Legate (Reference Legate2014) discusses several key points where the syntactic and morphological approaches differ in their predictions, two of which are of central interest here.Footnote ^[22] The first concerns modifiers of the subject. If subjects of all persons have the same syntax and the same case features, as the morphological approach proposes, they should show the same pattern of case on appositive modifiers. If the modifier of a 3rd person subject is marked with ergative, the modifier of a 1st or 2nd person subject should be marked with ergative as well. This follows on the morphological approach because the syncretic realization of ergative and nominative holds only for the local person pronouns themselves, not for other material that may modify them. On the syntactic approach, by contrast, the entire subject lacks an ergative feature when 1st or 2nd person.Footnote ^[23] Therefore, both the pronoun and its modifiers should lack ergative case.

Legate discusses four languages with person-based split ergativity where data are available on modification. In all of these languages – Dyirbal, Udi, Kham, and Maraṭhi – modifiers of local person subjects show ergative, just like modifiers of 3rd person subjects. This supports the morphological approach. The pattern is illustrated below with data from Maraṭhi. Examples (28)–(30), from Dhongde & Wali (Reference Dhongde and Wali2009), show the basic pattern of person-based split ergativity. This pattern holds in Maraṭhi only in the perfective, and so all Maraṭhi examples given here use this aspect.Footnote ^[24]

Pronominal subjects in Maraṭhi may be modified by adjectives, in which case the adjective follows the pronoun and is case-marked. In the crucial examples, (31) and (32), we see that modifiers of local person subjects take the ergative case.Footnote ^[25]

This pattern provides evidence of a purely morphological basis for person-based split ergativity in Maraṭhi. Parallel facts, as noted above, hold in Dyirbal, Udi and Kham (Legate Reference Legate2014: pp. 188, 191, 193). The results of this diagnostic are summarized in (33).

A second diagnostic discussed by Legate concerns coordination. The morphological approach predicts that it should be possible to coordinate local and non-local subjects without altering the case pattern for each individual coordinate. Thus, a coordination like ‘he and I’, serving as transitive subject, should show ergative case on the 3rd person conjunct, but nominative case on the 1st person conjunct. The well-formedness of such coordinations follows because all transitive subjects are the same in structural properties and in abstract case features; they differ only in their morphological realization.

On the syntactic approach, predictions for coordination differ according to the precise mechanism implicated in the absence of [erg] on local person subjects, and the way in which this mechanism interacts with coordination.Footnote ^[26] To articulate a first set of predictions, let us suppose that the overall syntactic behavior of the subject is decisive in determining case assignment, rather than the behavior of either individual coordinate. Two types of predictions are thus possible. First, on views that reference the subject’s person value (27i,iii-v), the expectation is that coordinations including local persons should lack all ergative case. We learn from agreement in many languages that a coordination including a 1st person is 1st person plural, and otherwise, a coordination including a 2nd person is 2nd person plural (Corbett Reference Corbett, Gazdar, Klein and Pullum1983, Reference Corbett2006; Dalrymple & Kaplan Reference Dalrymple and Kaplan2000).Footnote ^[27] Thus, any coordinated subject including a local person will act as a local person subject, resulting in the absence of [erg] (by whatever mechanism). Second, on the view that the syntactic category of the subject is the decisive factor (27ii), the expectation for case assignment in coordination depends on which coordinate determines the categorial behavior of the coordination overall.Footnote ^[28] The entire coordination could in principle accordingly behave either like a local person argument (a DP) or like a 3rd person argument (an NP), respectively lacking or showing ergative throughout the coordination.

A different set of predictions flows from the syntactic approach under the assumption that the mechanisms determining case assignment also apply to individual coordinates inside coordinated subjects. First, if case assignment to each coordinate is determined according to that coordinate’s person value (27i) or syntactic category (27ii), the expectation parallels that of the morphological approach: local and non-local subjects should coordinate without altering their case pattern. Second, if local person features inside coordinated subjects require licensing by a head that prevents [erg] at the clausal level (27iii), coordinations including local persons again should lack all ergative case. Third (and most distinctively), if local and non-local coordinates of subjects must occupy distinct positions in the clausal spine (27v), the expectation is that such coordinations should be simply ungrammatical. The coordinates impose contradictory requirements on the position the coordination must obtain. A potential further prediction is that coordinated intransitive subjects should reveal a similar restriction, if subjects (as Merchant Reference Merchant and Bunting2006 proposes) and coordinates thereof must generally occupy specialized person-based positions. I will show in Section 5 that this final set of predictions can also be made in a system where local and non-local (coordinates of) subjects must merely agree with distinct person-related heads, instead of occupying distinct positions.Footnote ^[29]

The overall set of predictions for coordinations is summarized in (34).

Legate (Reference Legate2014) discusses three languages with person-based split ergativity where data are available on coordination. In all of these languages – Udi, Maraṭhi and Siberian Yupik – local person subjects may be coordinated with 3rd person subjects, and the coordinates retain the case pattern they show as simplex transitive subjects. This is exemplified for Maraṭhi in (35).

This result is expected on all versions of the morphological approach. It is expected on the syntactic approach only if the person or category features of individual coordinates are decisive in determining case assignment. (Legate Reference Legate2014 does not discuss this second possibility.) As noted above, similar findings hold for Udi (Legate Reference Legate2014: 191) and Siberian Yupik (Legate Reference Legate2014: 196). In Udi, as in Maraṭhi, the morphological approach provides a unifying explanation for the behavior of modifiers and of coordinations.

On the basis of the Maraṭhi data reviewed in this section, together with parallel facts from Dyirbal, Udi, Kham and Siberian Yupik, Legate concludes that ‘split ergativity based on nominal type is a morphological, rather than syntactic, phenomenon’ (Reference Legate2014: 209). In the next section, I argue that this picture is incomplete.Footnote ^[30] Applied to Nez Perce, the same tests of modification and coordination reveal a syntactic basis for person-based split ergativity.

4 Person-based split ergativity in Nez Perce is syntactic

We begin with the modification diagnostic. Recall that in addition to marking case on the head noun, Nez Perce shows optional case concord between a noun and its prenominal modifiers. We see this concord in (36) in the 3rd person subject yú’snim ’iceyéeyenm ‘poor Coyote’. Note that this sentence describes part of a traditional story in which Coyote is the main character, and so presumably ’iceyéeye ‘Coyote’ here is used as a proper name.

On the morphological approach, we expect the case marking on the modifier yú’snim to remain constant when the subject is changed from a 3rd person name to a 1st or 2nd person pronoun. This, however, is not what we find. The switch to a local person subject brings the switch to a nominative form of the modifying adjective. The ergative form is no longer acceptable.

This result, which is notably different from the Dyirbal, Udi, Kham and Maraṭhi facts reviewed by Legate, is as expected on the syntactic approach.

A second type of modification test yields results consistent with only some morphological approaches, but all syntactic approaches. In addition to independent adjectives, Nez Perce allows pronouns to be modified by various suffixes. One of these is the suffix ciwáatx̂ ‘alone’, which is special among the suffixes in that it attaches between the pronoun and its case marker. This is shown for the 1st person plural pronoun in (39).Footnote ^[31]

Modification by ciwáatx̂ ‘alone’ is of special relevance for morphological approaches like Aldridge Reference Aldridge2007, which posits a zero realization for the ergative feature in the context of a local person feature, and Deal Reference Deal2010b, which posits an overt realization for the ergative feature only in the context of a 3rd person feature. If allomorphy may only be determined by linearly adjacent material, as Paster (Reference Paster2006) and Embick (Reference Embick2010) have argued, then the former view leads us to expect that the ordinary, non-zero exponent of ergative should reappear when ciwáatx̂ ‘alone’ intervenes linearly between the local person pronoun and the case marker; the latter view leads us to expect that ciwáatx̂ should interfere with ergative case marking on 3rd person pronouns. Neither expectation is borne out. Rather, third person pronouns modified by ciwáatx̂ continue to bear overt ergative case as transitive subjects or (as in these examples) as appositive modifiers thereof, (40). In contrast, local person pronouns modified by ciwáatx̂ continue to lack ergative case, (41).

The facts about ciwáatx̂ may be accounted for on morphological analyses like Keine & Müller (Reference Keine, Müller, Richards and Malchukov2008), Woolford (Reference Woolford, de Hoop and de Swart2008) and Legate (Reference Legate2014) if, for instance, the mechanisms that prevent spell-out of ergative on local person pronouns apply to all words containing such pronouns. They may also be accounted for straightforwardly on all versions of the syntactic analysis, where local person subjects and portions thereof are expected to systematically lack ergative.

Table (42) summarizes the predictions about modifiers and the findings for Nez Perce, by contrast to the findings in Maraṭhi, Dyirbal, Udi and Kham.

This provides a first indication that person-based split ergativity is not a uniform phenomenon across languages.

Additional evidence in this direction comes from coordinations, which are formed in Nez Perce with the coordinators kaa ‘and’ or ’íitq’o ‘or’. Case affixes may appear on each coordinate individually, or just on the final coordinate. (The latter option instantiates what Johannessen (Reference Johannessen1998) calls ‘unbalanced coordination’.)

Examples (43) and (44) show that coordinated subjects pose no inherent grammatical problem in Nez Perce, whether in a transitive clause or an intransitive one. Examples of this type are readily accepted as grammatical, and volunteered in translation from English. The same can be seen in a coordination of two local person pronouns as subject, (46).

To express the coordination of local and non-local subjects, however, speakers shift to an entirely different sentence type – a comitative, or so-called Plural Pronoun Construction (Schwarz Reference Schwarz, Barlow and Ferguson1988, Vassilieva & Larson Reference Vassilieva and Larson2005, i.a.). The non-local argument is encoded in a comitative phrase and the verb shows agreement with a plural subject. (The presence of a plural argument is also marked in (48) by the 2nd person plural clitic ’eetx.) Notably, there is no coordinator, and presumably no coordination of local and non-local arguments.

In translating from English into Nez Perce, speakers shift to this sentence type both when the clause is transitive, as in (47) and (48), and when it is intransitive, as in (49).

The Plural Pronoun Construction is equally available when all arguments are 3rd person (Rude Reference Rude1985: 101–103); this construction imposes no person restriction. Simple coordination of subject DPs, however, does appear to impose a restriction. Generally, judgments on sentences with local and non-local subject coordinates range from skepticism and a suggested correction to the Plural Pronoun Construction to outright rejection. Note that this holds across a range of case patterns in transitive clauses: both coordinates in the nominative, as in (50), ergative on the final coordinate, as in (51), and ergative on a non-final 3rd person coordinate, as in (52).

The restriction also holds in intransitive clauses, where both coordinates are strictly nominative.

The one systematic exception is instantiated by examples like (54) and (55): the local person pronoun appears in the final position, and ergative case is absent throughout the coordination. While not perfect, such examples are considerably better than (50)–(53). (See note 32 for further discussion of the status of these examples.)Footnote ^[32]

It seems to me most plausible that examples of this type are a calque from English, a language in which all Nez Perce speakers are fluent. The primary fact in support of this conclusion is that coordinations like these are characterized as ‘beginner’s speech’ or ‘for students’. Interestingly, judgments on these coordinations faithfully reproduce a fact of English coordinations that may be attributed to prescriptive factors: in subject position, nominative 1st person pronouns must occur in the final position. Compare (50) and (53), with 1st person first, with the minimally different (54) and (55), with 1st person last; only the latter are accepted. Just like in English, the order effect in Nez Perce holds only for subjects, and not for objects. Compare (56), where either order is acceptable for an object coordination.

I draw two conclusions from this overall set of judgments. First, the English pattern of 1st person last in subject coordinations – a restriction found to hold almost categorically for nominative pronouns in acceptability and corpus studies by Grano (Reference Grano2006) – has been adopted into Nez Perce, at least by the speakers consulted. The pattern is independent of ergative case, as it applies both to transitive subjects and to intransitive ones. Second, the structure of English subject coordinations has been borrowed into Nez Perce as the ‘beginner’s speech’ construction (54)/(55). The native pattern seems to be that local and non-local person subjects cannot be coordinated.

From this perspective, the most relevant judgments on subject coordinations of local and non-local persons are (50)–(53). These judgments make for a clear contrast with (43) and (44), where local and non-local persons are not present in the same coordination. Also to be contrasted with these examples are sentences where the coordination serves as an object. Here, as we saw in part in (56), coordinations of local and non-local persons show no special behavior: they are well-formed, and allow case on both coordinates or just the final one.

This data set overall indicates that some syntactic problem is encountered in the combination of three factors: coordination, subject, and local plus non-local person. The response to this problem involves switching to the Plural Pronoun Construction, an alternative mode of expression that does not involve a coordinated subject.

These facts are different from the Maraṭhi example (35) and, overall, not predicted by the morphological analysis. Unlike the modification facts, they are also unexpected on the majority of syntactic analyses. Among the syntactic proposals listed in (27), the ill-formedness of subject coordinations (50)–(53) is not predicted by any version except (v). The results follow on this view with two crucial assumptions: (i) the syntactic requirements imposed on subjects are visited not only on entire subject coordinations, but also on the coordinates thereof, and (ii) these requirements apply both in transitive and in intransitive clauses. The overall results are summarized in (58).

Taken together with the evidence from modification in languages like Maraṭhi, these results confirm the need to recognize two sources for person-based split ergativity, one syntactic and one morphological. Beyond this, they reveal evidence that the syntactic variety of person split involves distinct, incompatible requirements holding for local and non-local persons, whether as subjects or as coordinates thereof. The conflict between these requirements plays the lead role in ruling out coordinations of local and non-local persons as subjects in Nez Perce.

5 The syntax of the person split

The results of the coordination and modification diagnostics lead to two core conclusions about the syntax of person in Nez Perce. First, and most fundamentally, local and 3rd person subjects differ not just in their morphology, but also in their syntax. Distinct grammatical requirements are imposed both for local persons and for 3rd persons, and these requirements extend to the coordinates of a subject coordination. The conflict between these requirements explains why coordinations of local and non-local persons are not well-formed as subjects. This ill-formedness extends across both transitive clauses, where subjects may receive ergative, and intransitive clauses, where subjects are always nominative. This suggests that the requirements in question should not be stated directly in terms of case. Second, the person-based syntactic system nevertheless interacts with case assignment in the transitive clause. Unlike 3rd persons, local person subjects are not assigned an [erg] feature in syntax, regardless of transitivity. This explains why ergative case is never present on modifiers of local person subjects. This section sketches a syntactic analysis incorporating these conclusions.

I begin with the observation that a variety of languages, both ergative and non-ergative, provide evidence for dedicated person-related functional projections for subjects in the inflectional domain of the clause. In some instances the evidence involves movement or cliticization to these projections. This is the case in some Northern Italian dialects, for example, where local person subject clitics occur higher than negation, whereas 3rd person subject clitics occur below negation (Poletto Reference Poletto2000). The same goes for local person pronouns versus 3rd person pronouns in Hebrew sentences with the negator ’eyn (Shlonsky Reference Shlonsky, Bennis, Everaert and Reuland2000); Shlonsky explicitly argues that both positions are within the inflectional domain. In other instances the evidence comes from a split between two distinct loci for subject agreement, one for local person and one for non-local person. This is so, for instance, in Euchee, an isolate spoken in Oklahoma (Linn & Rosen Reference Linn, Rosen and Griffin2003), in Athabaskan languages such as Slave (Rice Reference Rice2000) and in Salish languages such as Lummi (Jelinek Reference Jelinek, Bobaljik and Phillips1993) and Halkomelem (Wiltschko Reference Wiltschko, Johns, Massam and Ndayiragije2006). Finally, in some Romance varieties, the choice of auxiliaries is sensitive to the person of the subject (Kayne Reference Kayne1993, D’Alessandro & Roberts Reference D’Alessandro and Roberts2010). This pattern, too, may be explained by reference to person-sensitive subject-related functional projections (Coon & Preminger Reference Coon, Preminger and Choi2012).

Bianchi (Reference Bianchi2006) and Merchant (Reference Merchant and Bunting2006) posit that the heads in question are agreement heads specialized for particular person values. Adapting Bianchi’s terminology slightly, I will refer to them as LocS, for local person subjects, and 3S, for 3rd person subjects.Footnote ^[33] A central idea in this domain is that person-based height differences among arguments, as we see in Northern Italian dialects, Hebrew and Salish, arise because LocS-P is higher than 3S-P. This will be depicted in the trees below.

Person-sensitive phrase structure in the inflectional domain paves the way for an analysis in the general tradition of those developed by Jelinek (Reference Jelinek, Bobaljik and Phillips1993), Nash (Reference Nash and Zribi-Hertz1997) and Merchant (Reference Merchant and Bunting2006). All subject arguments must enter into a syntactic relationship with the appropriate subject-related projection; so too, I assume, must coordinates within coordinated subjects. Now, this second assumption raises a technical challenge if the required relationship is a spec-head configuration with an appropriate licensing head (as, for instance, in Merchant Reference Merchant and Bunting2006). In a subject coordination of two 3rd persons, neither coordinate obtains a spec-head relationship with a head outside the coordinate structure, but the corresponding sentences are nevertheless well-formed (see (43)–(44)). One response to this challenge would be to state separate positional requirements for subjects and for their coordinates – specifiers of appropriate heads in the former case, and coordinates within such specifiers in the latter case. This is essentially a disjunctive positional licensing requirement. I propose what I take to be a simpler alternative: what is required of subjects and their coordinates is merely agreement with a licensing head, understood as a transitive relation. In a coordination of two 3rd persons, 3S agrees directly with the overall coordinated DP in person features. The individual coordinates agree with the overall containing DP (possibly via its head, &); this allows them to share features indirectly with 3S, satisfying the agreement requirement.Footnote ^[34] I return below to the consequences of this agreement relationship for case assignment.

A parallel syntactic situation obtains when local person arguments are coordinated. The overall coordination agrees with LocS directly; both individual coordinates agree with LocS indirectly. All subjects and coordinates thereof agree, directly or indirectly, with the appropriate person-related head, and the result is well-formed.

The situation is different when local and non-local persons are coordinated. As a local person DP, the overall subject coordination agrees with LocS; both coordinates thus agree indirectly with LocS, rather than 3S. The result does not conform to the requirement that all subjects and coordinates thereof agree with appropriate person-related heads, in view of the presence of the 3rd person coordinate.

This pattern sheds light on the status of person-related functional projections for objects, as proposed by Bianchi (Reference Bianchi2006) and Merchant (Reference Merchant and Bunting2006).Footnote ^[35] Recall that coordinations of local and non-local objects are perfectly grammatical in Nez Perce.

This behavior makes sense if objects agree with a functional projection capable of agreement with both local and non-local person DPs. The asymmetry between subject and object coordinations reflects a greater degree of person specialization in the domain of subject agreement projections, and a lesser degree in the domain of object agreement projections.

We can now address the central question of how person-sensitive phrase structure for subjects interacts with ergative case. In Deal (Reference Deal2010a, Reference Dealb), I argue that ergative behaves as a structural case in Nez Perce, rather than as an inherent case assigned by $v$ to its specifier. I propose therefore that [erg] is assigned by 3S in the transitive clause.Footnote ^[36] LocS assigns only [nom]. The modifier facts follow straightforwardly. In (62), the 3rd person subject receives an [erg] feature from 3S; ergative is realized both on the modifier (by case concord) and on the head noun. In (63), by contrast, there is no 3rd person subject and thus no agreement with the head 3S. The subject agrees instead with LocS, which assigns it a [nom] feature. There is no source for ergative case on the modifier of the subject.

Case in subject coordinations also follows straightforwardly. Given the agreement requirement imposed on subjects and coordinates thereof, the only well-formed coordinated subjects are those that include only local person coordinates or only 3rd person coordinates. The former agree with LocS and receive nominative, (64). The latter agree with 3S and receive ergative, (65). Agreement within the coordination has the result that case features are shared with each individual coordinate.Footnote ^[37]

The syntax of person in Nez Perce thus comes down to two major factors. First, local and non-local person subjects and coordinates thereof are required to agree with separate person-related heads. This has the result that coordinations mixing local and non-local persons are not acceptable as subjects. Second, the head responsible for assigning the ergative case feature is person-related. Local person subjects are never assigned an [erg] feature in the syntax.

This analysis points to several dimensions that may be subject to cross-linguistic variation. The first is the possibility of phrase-structural variation, as emphasized by Bianchi (Reference Bianchi2006); languages may differ in the extent to which they project articulated person-sensitive functional categories. A language lacking such projections presumably would not have the syntactic type of person-based ergative split, though it may still have the morphological type of split ergativity. The second is the possibility of variation in the cases assigned by person-sensitive heads. In various of the languages for which such heads have been posited, case is on a strict nominative–accusative basis. This suggests that 3S may assign [nom] in transitive clauses in some languages and [erg] in transitive clauses in others. If the same goes for LocS, then an ergative language with no person-based split, or a split of the purely morphological type, might differ from Nez Perce not in its hierarchical structure, but in the cases assigned by elements therein: both 3S and LocS assign [erg]. Evidence for such a language could come from person-sensitive word order, placement of agreement morphology, or auxiliary selection, coupled with the absence of person-based split ergativity or coordination/modification data suggesting a morphological basis for such a split.