1. Introduction
This paper examines two instances of person complementarity in Eastern Canadian Inuit (Eskimo-Aleut).Footnote 1 The first is an apparent Person Case Constraint (PCC) effect involving transitive agreement in the Labrador Inuttut dialect, illustrated in (1), whereby φ-marking can index a third-person internal argument but not a first- or second-person internal argument.
The second instance of person complementarity is an unexpected gap in posses- sor marking in South Baffin Inuktitut, whereby synthetic first and second person possessor forms are no longer allowed in oblique cases, as exemplified in (2).
While Johns and Kučerová (Reference Johns, Kučerová, Coon, Massam and Travis2017) have proposed that these patterns involve a PCC effect – the presence of one argument acting as an intervener for person agreement with another – it is argued herein that both instances of person complementarity arise from sources other than intervention, and that the properties of these phenomena are uncharacteristic of the PCC, both empirically and in terms of their theoretical explanation.
1.1 The Person Case Constraint
One area in which the nature and organization of person features have been the sub- ject of ongoing inquiry is the Person Case Constraint (PCC) (Bonet Reference Bonet1991, Reference Bonet1994; Perlmutter Reference Perlmutter1971). Building on earlier work in individual languages, Bonet (Reference Bonet1991) observes that an array of unrelated languages ban the co-occurrence of an indirect object clitic and a first or second person direct object clitic. She states the constraint as follows:
For instance, in French the combination of a first or second person dative clitic and an accusative third person clitic is grammatical, but the inverse where the third person clitic is dative is ungrammatical:Footnote 2
Bonet identifies two varieties of this phenomenon: a Strong PCC, as exemplified in French, in which the DO must be 3rd person, as well as a Weak PCC, found in other languages, where if only one of the two arguments is 3rd person, then it must be the DO. Subsequent authors, particularly Nevins (Reference Nevins2007), have analysed three additional patterns that include an Ultra-Strong PCC (which deviates from the Weak pattern by also blocking 2 > 1 combinations), a Super-Strong PCC (in which only 1/2 > 3 combinations are possible), and a Me-First PCC (in which the DO cannot be 1st person). Each pattern is schematized in Table 1.Footnote 3
Table 1: Varieties of the Person Case Constraint (adopting the terminology of Pancheva and Zubizarreta Reference Pancheva and Zubizarreta2017)
1.2 Person complementarity in Inuit
Johns and Kučerová (Reference Johns, Kučerová, Coon, Massam and Travis2017) propose that a PCC effect is present in Labrador Inuttut, albeit between ergative subjects and absolutive objects, instead of between indirect and direct objects.Footnote 4 This difference in the arguments involved is not inherently prob- lematic, as what is crucial in most accounts of the PCC is that both arguments be in the same agreement domain with respect to the relevant probes (see, e.g., Béjar and Rezac Reference Béjar and Rezac2009: 47). Furthermore, various accounts of φ-agreement in Inuit have proposed that the head (or heads) responsible for agreement with both ergative and absolutive arguments is at the periphery of the clause and thus the two arguments are arguably in the same agreement domain.Footnote 5
Johns and Kučerová observe that while φ-marking indexing a local (i.e., 1st or 2nd person) agent and a non-local (i.e., 3rd person) patient is grammatical in the Labrador Inuttut dialect, the opposite is not possible in the declarative/participial clause-type, repeated from (1).Footnote 6
Furthermore, this dialect also bans combinations of two local persons from being co-indexed on the verb in this clause-type, causing otherwise expected forms (ap- pearing in other dialects) to be ungrammatical, as shown in Table 2, adapted from Johns (Reference Johns2000: 120).Footnote 7
Table 2: Person complementarity in Labrador Inuttut (Johns Reference Johns2000: 120)
The person complementarity effect in Labrador Inuttut holds in the declarative/participial clause-type, regardless of the number marking of either argument. The pattern is schematized in Table 3. We can note that it matches the Strong PCC pattern from Table 1.Footnote 8
Table 3: Person complementarity in Labrador Inuttut (based on Johns Reference Johns, Burgess, Dziwirek and Gerdts1995, Reference Johns2000)
The equivalent forms are grammatical in closely related dialects, as illustrated below from South Baffin Inuktitut (see also Yuan Reference Yuan2018 for additional examples of these person combinations):Footnote 9
Similarly, such person combinations are grammatical in the North Baffin di- alect in the same clause-type. The following examples are from a speaker from the community of Igloolik (data collected by the author):
(8) Absence of person complementarity in North Baffin Inuktitut
a. apu-lauq-tagit
meet-dist.past-part.1sg.2sg
‘I met you.’
b. apu-lauq-tara
meet-dist.past-part.1sg.3sg
‘I met him/her.’
c. apu-lauq-tarma
meet-dist.past-part.2sg.1sg
‘You met me.’
d. apu-lauq-tait
meet-dist.past-part.2sg.3sg
‘You met him/her.’
e. apu-lauq-taanga
meet-dist.past-part.3sg.1sg
‘He/she met me.’
f. apu-lauq-taatit
meet-dist.past-part.3sg.2sg
‘He/she met you.’
g. apu-lauq-tanga
meet-dist.past-part.3sg.3sg
‘He/she met him/her.’
An analogous effect of person complementarity is also observed in possessor marking in the South Baffin dialect of Inuktitut. In most Inuit dialects, possessed nouns bear obligatory marking for the person and number of possessors, as well as the case and number of the possessum, with their exponents often being fused. In South Baffin, however, formerly synthetic combinations of local person possessor marking and oblique cases have been replaced by periphrastic constructions, involving a free pronoun and the (otherwise) third person form of possessor marking on the noun. Third person possessors are unaffected. For instance, in the conservative form in (9a), the φ-features of the possessor, the number of the possessum, and ablative case are encoded in the portmanteau -nnit, while in the innovative form in (9b), the person of the possessor is instead indicated by a separate pronoun, with the exponent typically found for the third person occurring on the noun (cf. example 9c).Footnote 10
Once again, Johns and Kučerová propose the innovative construction in (9b) to be due to a PCC effect, noting that the effect arises with a local person (ergative) possessor and a third-person oblique possessum. We can note that the phenomenon resembles a subset of the transitive verbal agreement pattern above in Table 3 whereby 3 > 1/2 forms were illicit, as shown in Table 4.Footnote 11
Table 4: Person complementarity in South Baffin synthetic possessor marking (Compton Reference Compton2014)
As part of a larger line of argumentation that φ-markers in Inuit are clitics and not agreement, Johns and Kučerová interpret the default third person form arising in such constructions as a type of repair, and thus characteristic of clitics, following Rezac (Reference Rezac2008) (although see Preminger Reference Preminger2014 for arguments that such defaults are in fact characteristic of agreement).
Similarly, Johns and Kučerová also take the presence of PCC effects in both Labrador Inuttut and South Baffin Inuktitut to be a diagnostic of these markers being clitics, following Nevins (Reference Nevins2011) who argues that the PCC may hold only of clitics, and not of genuine agreement (i.e., the exponence of the AGREE relation on a functional head), although later work has called into question whether the PCC is a reliable diagnostic for differentiating clitics from agreement (e.g., Preminger Reference Preminger2019). We return to Nevins's approach to the PCC below in section 2.2.
Despite the textbook-like quality of the pattern of person complementarity in transitive verbs in the Labrador dialect, I propose that these data are not in fact instances of the PCC, insofar as it is not the combination of φ-features on the two arguments that yields ungrammaticality. Instead, I argue here that the effect is the result of a defective or missing probe that cannot value local person on absolutive arguments. This instance of person complementarity thus has a source distinct from other cases of the PCC, which involve (at least descriptively) the incompatibility of particular sets of features between two arguments (or their corresponding clitics).
Further, I argue that the person complementarity observed in possessor marking is also not a PCC effect, insofar as it is motivated by a desire to resolve ambiguities brought about by regular phonological change, and is not due to any particular combination of φ-features in the syntax.
1.3 Organization
The next section gives an overview of two influential syntactic analyses of the PCC. Section 3 provides arguments that the observed person complementarity is not consistent with the expected behaviour of the PCC. In particular, it is shown that person complementarity in φ-marking on verbs extends to intransitives and that it is alleviated by changing the clause type—both of which are unexpected of the PCC. The asymmetry in periphrastic possessor marking in South Baffin is shown to correlate with the loss of place distinctions in heterorganic consonant clusters, leading to a loss of contrast between first and second person in oblique forms, with the resulting ambiguity leading to innovative forms. In section 4, I present an alternative account of the observed transitive agreement facts, whereby the declarative/participial clause-type's agreement probes lack the ability to value local person on absolutive arguments.
2. Competing analyses of the PCC
The PCC has received a number of treatments, including morphophonological accounts (Bonet Reference Bonet1991, Reference Bonet1994, among others), as well as syntactically-oriented ap- proaches based on the mechanisms of Multiple Agree (Anagnostopoulou Reference Anagnostopoulou, Heggie and Ordóñez2005; Nevins Reference Nevins2007, Reference Nevins2011), the Person Licensing Condition (Béjar and Rezac Reference Béjar, Rezac, Pérez-Leroux and Roberge2003), Cylic Agree (Béjar and Rezac Reference Béjar and Rezac2009), and more recently the P(erson)-Constraint (Pancheva and Zubizarreta Reference Pancheva and Zubizarreta2017).Footnote 12 However, most syntactic approaches have in common that PCC effects arise from the interaction of probes and the relative markedness of φ-features of the two arguments (or their corresponding pronominal clitics).
In this section, I give a brief overview of two syntactic accounts of the PCC; those of Béjar and Rezac (Reference Béjar, Rezac, Pérez-Leroux and Roberge2003, Reference Béjar and Rezac2009) and Nevins (Reference Nevins2007, Reference Nevins2011).
2.1 Split Agreement and the Person Licensing Condition
As part of an analysis of PCC effects in a variety of languages, Béjar and Rezac (Reference Béjar, Rezac, Pérez-Leroux and Roberge2003) propose that there are separate agreement probes on v 0 for person and number. The person probe [uπ] searches first, finding the structurally higher argument. If this argument is a clitic with inherent case (e.g., dative), Agree cannot occur, but the clitic moves to v 0, thereby becoming inactive for probing by number. Next, the number probe [u#] may search past the trace of the higher clitic to agree with the theme argument.
Combined with this innovation, they propose the Person-Licensing Condition (PLC) (p. 53):
(10) Person Licensing Condition (PLC) axiom
An interpretable 1st/2nd person feature must be licensed by entering into an Agree relation with a functional category.
This constraint ensures that marked person features must be licensed by the Agree operation. Unlicensed 1st and 2nd person features will therefore cause the derivation to crash. In addition, Béjar and Rezac propose that while full DPs with either inherent case or focus may be shielded from this requirement, unfocused arguments bearing structural case, such as clitics, are susceptible to the PLC.
Given this system, PCC effects between an IO clitic and DO clitic arise when a person probe c-commanding the two encounters the IO clitic first. As outlined above, the IO clitic intervenes, preventing person agreement with the DO clitic. While a third person DO clitic could be featurally unmarked (depending on the language) and not need its person features licensed, a first or second person DO clitic in this scenario would not have its person features licensed, leading to a violation of the PLC and resulting in ungrammaticality.
To explain direct-inverse agreement patterns where internal and external argu- ments are in competition for agreement, Béjar and Rezac (Reference Béjar and Rezac2009) develop a cyclic model of agreement, proposing an analysis that makes use of articulated person features, as shown in Table 5 (following Harley and Ritter Reference Harley and Ritter2002):
Table 5: Articulated person features (Béjar and Rezac Reference Béjar and Rezac2009: 43)
In this system, a person probe may on v 0 may be underspecified, searching only for [π], or may instead be feature-relativized to search for more marked features (e.g., [participant] or [speaker]).Footnote 13 In the latter situation, a probe may enter into an agreement relationship with a less specified internal argument, but then continue searching upward to the external argument to find these features – what Béjar and Rezac (Reference Béjar and Rezac2009: 49) term “cyclic expansion of the search space”. Although the PCC is not their focus, Nevins (Reference Nevins2011) and Anagnostopoulou (Reference Anagnostopoulou, Everaert and van Riemsdijk2017) note that this approach is equally capable of explaining PCC effects in cases where person agreement with one argument bleeds person agreement with another, leaving unlicensed marked features, in violation of the PLC.
Béjar and Rezac (Reference Béjar and Rezac2009: 46) formalize the PLC and PCC as follows:
(11) Person-Licensing Condition (PLC)
A π-feature [F] must be licensed by Agree of some segment in a feature structure of which [F] is a subset.
(12) Person Case Constraint (PCC)
In [α Agr … DP1-oblique … DP2 …], where α includes no other person Agr, DP2 cannot have a marked person feature (1st/2nd, sometimes 3rd animate).
Crucially for our present purposes, PCC effects arise in this type of analysis due to the presence of an intervening argument – in terms of hierarchical structure or derivational timing in the case of cyclic expansion – that prevents the licensing of person features on a second argument.
2.2 Multiple Agree
Another type of approach is offered by Nevins (Reference Nevins2007, Reference Nevins2011) who, following Anagnostopoulou (Reference Anagnostopoulou, Heggie and Ordóñez2005), argues for a system of Multiple Agree, whereby an agreement probe situated high in the structure may enter into an agreement relation with multiple arguments simultaneously.
In his system, PCC effects arise because clitics (but not case-marked DPs) are susceptible to Multiple Agree, and this relation is subject to two conditions called Contiguous Agree and Matched Values (Nevins Reference Nevins2011: 963):
(13) Contiguous Agree
For a relativization R of a feature F on a Probe P, and x ∈ Domain(R(F)), ¬
$\exists $ y, such that y > x and P > y and y ¬ ∈ Domain(R(F))“there can be no interveners between P and x that are not in the domain of relativization that includes x.”
(14) Matched Values
For a relativization R of a feature F,
$\exists $α, α ∈ {+,−},$\forall $ x, x ∈ Domain(R(F)), val(x,F) = α“all elements within the domain of relativization must contain the same value for the feature F being agreed with.”
Furthermore, to account for crosslinguistic variation in PCC patterning, Nevins (Reference Nevins2007) assumes, following Béjar (Reference Béjar2003), that probes may be relativized with respect to which features they seek to value.
As a result of Contiguous Agree, a probe relativized to search for a marked value of a feature, such as [+participant], will be able to enter into a Multiple Agreement relationship with IO and DO clitics bearing the features [+participant] and [−participant], respectively (assuming that the IO clitic c-commands the DO clitic) (although see the discussion regarding Matched Values below). However, the opposite feature specification, such as a 3 > 1/2 person combination, would lead to a violation of Contiguous Agree, since while the probe and DO clitic match in features, the intervening IO clitic bears the opposite feature, as illustrated in (15).
(15) Example of a violation of Contiguous Agree
The other condition, Matched Values, plays a crucial role in explaining the Strong PCC in this system by ruling out 1 > 2 and 2 > 1 combinations. According to Nevins (Reference Nevins2007), in Strong PCC languages the probe is relativized to search for contrastively specified values of the feature [Author].Footnote 14 Since combinations of two local persons will yield opposite values for this feature (i.e., [+Author] and [−Author]), such combinations are also predicted to be excluded, as the probe would only be able to enter into an agreement relation with the higher argument.
PCC effects thus arise in such a system because an opposing feature value breaks the contiguity of agreement between the probe and the DO. As in Béjar and Rezac's approach, it is the presence of a second, structurally higher clitic in the same search space as the DO that yields the PCC effect.
3. Against a PCC analysis of Person Complementarity in Inuit
Despite exhibiting an apparently prototypical PCC pattern, the instances of person complementarity found in Inuit are unlike the PCC in several important ways. This section argues that these phenomena are distinct from the PCC, both empirically and theoretically.
3.1 Intransitives
Given that the canonical PCC effect involves φ-features of an IO bleeding agreement with a DO, as in French, or less canonically those of a subject bleeding agreement with an object (as proposed for Kichean by Preminger (Reference Preminger2014), a number of authors have explained it in terms of either intervention (Béjar and Rezac Reference Béjar, Rezac, Pérez-Leroux and Roberge2003, Reference Béjar and Rezac2009; Preminger Reference Preminger2014) or as a lack of matching features between the two arguments with respect to a higher probe (Anagnostopoulou Reference Anagnostopoulou, Heggie and Ordóñez2005; Nevins Reference Nevins2007, Reference Nevins2011).
Ergative subjects in Inuit are presumably structurally higher than absolutive ob- jects (given both word order and binding facts), and thus might intervene between an agreement probe and absolutive DOs, particularly given that agreement with both ergative and absolutive in Inuit has been argued to occur high: on Infl and C according to Bittner and Hale (Reference Bittner and Hale1996), on C according to Compton (Reference Compton2014, Reference Compton2016), and in an Agr projection at the clausal periphery according to Yuan (Reference Yuan2018). In addition, ergative case is treated as an inherent case in a number of languages, and thus we might expect it to share properties with dative indirect objects, which act as (defective) interveners crosslinguistically.
A reasonable interim hypothesis is that the ergative subject in Labrador Inuttut intervenes between a π probe and the absolutive object, either blocking agreement with local person features and yielding a violation of the PLC or preventing Multiple Agree, thereby causing the ungrammaticality of the 3 > 1/2 declarative/participial forms.
However, one problem with an intervention-based approach using either the PLC or Contiguous Agree is that in fact, local person agreement in intransitives is also ungrammatical in this clause-type:
This state of affairs is unexpected under the PCC accounts outlined above. With only one argument – the subject – there are no other arguments to act as potential interveners or to interrupt an operation such as Multiple Agree, and yet the otherwise expected local person forms are ungrammatical.Footnote 15 Thus, while these forms evince a type of person complementarity between local and non-local persons in this dialect, these constructions are fundamentally unlike the PCC in that there may be only one argument involved, and the derivation does not follow naturally from the accounts of the PCC presented above.
Instead, a more precise descriptive generalization would be that absolutives in general in this construction cannot bear local person; as the exponence of either object-indexing markers in transitives or subject-indexing markers in intransitives. An account based on this generalization is proposed in section 4.
3.2 Clause-type
Until now, we have confined our discussion of the observed person complemen- tarity in φ-marking on verbs to the declarative/participial clause-type. This clause-type is the default form of marking on declarative matrix clauses in Canadian Inuit dialects. It co-exists with what is called the indicative clause-type, whose use varies across the Inuit dialect continuum. In both Siglitun and Uummarmiutun, in the Western Canadian Arctic, the indicative clause-type is “very restricted in usage” and “very rarely used” (Lowe Reference Lowe1985a,Reference Loweb), and typically follows a word meaning ‘finally’ or ‘in the end’, suggesting it conveys additional meaning.Footnote 16 In Eastern Canadian dialects, its usage is also considered marked, with Swift (Reference Swift2004: 16) describing it as the “marked (focus, surprise) form” in the Nunavik (Arctic Quebec) dialect. Conversely, in West Greenlandic, it is the indicative that is the default form for matrix declarative clauses, with the declarative/participial form being used with predicates acting as modifiers (Fortescue Reference Fortescue1984).
Somewhat unexpectedly under a PCC account, the person complementarity observed in the declarative/participial clause-type in Labrador disappears in the otherwise more marked indicative clause-type (as noted by Johns and Kučerová Reference Johns, Kučerová, Coon, Massam and Travis2017, but also discussed in Johns Reference Johns, Burgess, Dziwirek and Gerdts1995, Reference Johns2000):
Furthermore, person complementarity effects are not found in the other clause- types in the Labrador dialect (see also examples throughout Smith Reference Smith1977):Footnote 17
Assuming some version of the accounts of the PCC outlined in section 2, it is not clear why this should be the case. If the PCC effects in the Labrador dialect in the declarative/participial clause-type were due to intervention between clitics asso- ciated with the ergative subject and the absolutive object, changing the clause-type should not alleviate the situation, in particular given that the relative case positions of the arguments have been maintained and the exponents of φ-marking are for the most part identical across clause-types. Compare, for instance, the declarative and indicative forms in (17) and (18) above.
Under a defective intervention account, if case alignment is held constant, and thus the relative positions of the ergative and absolutive arguments remain the same, a change in clause-type is not expected to alleviate a PCC effect. The higher argument should still act as an intervener, blocking agreement with the lower argument.
Similarly, under Nevins's (Reference Nevins2007, Reference Nevins2011) account of the PCC, such a change in clause-type should not prevent 3 > 1/2 clitic combinations from violating Contiguous Agree, assuming the same hierarchical positions for the arguments, and for the clitics themselves. Once again, the relevant φ-markers, including portmanteau forms, are the same in these two clause types in related dialects, as is their relative position to other verbal morphology.
It is noteworthy that in other languages that exhibit the PCC, the effect is not known to depend on different values of tense, mood, or clause-type. Although the effect has been known to disappear in non-finite clauses, the relevant clause-types in Labrador are presumably finite, permitting not only agreement but also tense.
In sum, all else being equal, PCC effects are not expected to vary by clause-type when case positions and the exponence of φ-marking are held constant. This type of person complementarity is, once again, unlike the PCC.
3.3 Innovative periphrastic possession
As introduced above, the periphrastic possessor-marking found in South Baffin Inuk- titut can be seen as an additional instance of person complementarity in Inuit, insofar as local-person possessor marking is banned on nouns marked with oblique case (which we can consider to be third person), as illustrated in the (21), modified from (9) above.Footnote 18
Under a PCC approach to the ungrammaticality of the earlier synthetic forms, a person probe on the functional head responsible for possessor marking first finds the oblique-marked noun itself, which would act as an intervener blocking subsequent person agreement with the possessor. Abstracting away for the moment from whether φ-markers are clitics or agreement, under a Béjar and Rezac-style approach, possessors with local person features would trigger a violation of the PLC. In contrast, under a Contiguous Agree approach, there would be a lack of featural contiguity between the [−participant] possessum and the [+participant] possessor.
One problem for such an approach is that the periphrastic construction (as well as other innovative possessor-marking forms, discussed below) correlates with a pro- cess of regular phonological change in Inuit, whereby the possibility of heterorganic consonant clusters becomes more and more constrained towards the eastern part of the dialect continuum. Contrasts in such clusters had previously differentiated per- son. Compare, for instance, the following possessed case-marked forms in the North Baffin dialect:Footnote 19
The overt contrast between many such forms distinguishing first and second person can be seen in the difference between /nn/ and /ŋn/ sequences. However, as part of a language-wide, west-to-east cline of regressive place assimilation (Bobaljik Reference Bobaljik1996), /ŋn/ clusters have become illicit in South Baffin, Nunavik, Labrador, and West Greenlandic. Thus, without any modification to the system, such forms would have become ambiguous between first and second person, as observed by Dorais (Reference Dorais2003). The strategy for resolving the ambiguity in South Baffin has been to employ an overt strong pronoun and use the third person form of possession on the noun (which Compton Reference Compton2014, Reference Compton2016 argues is default agreement).
However, we might posit that this phonological change is merely a coincidence, and that there really is a PCC effect that bans local person features in the context of oblique cases. Fortuitously, other dialects have also made changes to their posses- sive marking, but have instead either (i) simply opted to regularize their paradigms to avoid the geminate consonants in question or (ii) have created innovative endings – but crucially, such changes target only persons that have become subject to ambiguity. Thus, in Labrador and Tarramiut Nunavik dialects, the exponence of first person marking in the absolutive has been extended to the oblique cases, as in (24), while in the Itivimiut Nunavik dialect, innovative forms involving suffixation of the first person pronoun have been adopted, as in (25), as have similar regularizations of second person endings (not shown).
Crucially, the second person forms in Labrador and Tarramiut were left unaffected, leaving the (now assimilated) geminate clusters to indicate second person. If such changes were indeed motivated by a Strong PCC effect analogous to the transitive verbal pattern involving the feature [participant], we likely would have expected all [participant] forms to be affected.Footnote 20
Moreover, it is striking that this cluster of phenomena (periphrastic possessor- marking, paradigm regularization, and innovative pronominal suffixes) all converge on the very dialects that have lost the original phonological contrasts that separated first and second person forms.Footnote 21
In sum, the innovative periphrastic possessor construction in South Baffin is not due to a PCC effect, but is instead the result of a strategy to compensate for a loss of contrast brought about by regular phonological change, like parallel strategies in the other affected dialects. This is further supported by the fact that such innovations specifically target forms at risk of becoming ambiguous. Once again, the person complementarity observed in such constructions is not due to the interaction of features on two arguments, but is instead phonologically motivated, and thus is uncharacteristic of the PCC.Footnote 22
4. A probe-based account of person complementarity
In this section I offer an alternative explanation of the person complementarity found in φ-marking on verbs in Labrador Inuttut. In sections 1 and 3, we observed that person complementarity in the declarative clause-type in this dialect only restricts φ-marking of the absolutive argument – in both transitives and intransitives. As argued above, syntactic accounts of the PCC, though they could exclude the ungrammatical transitive forms in Inuit, would not explain why intransitives with local person subjects should also be ungrammatical, or why the effect depends on clause-type.
Instead, I propose that the probe responsible for π-agreement with absolutive arguments in this clause-type is either defective or absent entirely, assuming separate probes for person and number, as in Béjar and Rezac (Reference Béjar2003, Reference Béjar and Rezac2009). As a result, agreement is not possible with absolutive arguments bearing local person features. Crucially, though, number (#) agreement is possible with both ergative and absolutive arguments, as illustrated here with a plural object:
(26) taku-ja-kka
see-tr.part-1sg.3pl
‘I see them.’
This suggests that while the probe responsible for absolutive person agreement is defective or absent in this clause-type, #-probes for the two arguments are available, allowing both transitive and intransitive #-agreement for ergative and absolutive ar- guments, in addition to ergative π-agreement, as schematized in (27). The subscripts for ergative and absolutive are used only for expository convenience; they identify the arguments that each probe will end up agreeing with as a result of the proposed mechanics of the Agree operations.Footnote 23
(27) φ-probes available on declarative/participial C in Labrador Inuttut CP
For concreteness, I assume Béjar and Rezac's (Reference Béjar and Rezac2009) system for the sample derivations discussed below, particularly insofar as person features are articulated into [π], [participant], and [speaker]. I further assume that the latter two features must be licensed as required by the PLC. Finally, I assume that the person probes in Labrador Inuttut are underspecified – relativized to search only for the feature [π] – and that they stop searching after encountering this feature.
In an intransitive declarative/participial clause in Labrador Inuttut, the only available agreement probe, by hypothesis, is u#abs. In the presence of a third person argument, the u#abs probe will find the sole subject argument and value its number. Since third person arguments have no marked person features that need to be licensed by the PLC, no person agreement is required. However, in the case of a first or second person subject (i.e., typically covert pro in Inuit), the marked person features [participant] and potentially also [speaker] would not be valued by a person probe, leading to a violation of the PLC. Local person subjects of intransitives are thus correctly predicted to be ungrammatical in this clause-type.
Moving now to transitives in this clause-type, I propose that they make available two additional probes: one for person and one for number. The person probe uπerg searches first, finding the structurally higher ergative argument—on the assumption that the ergative argument is merged in Spec,vP and that the relevant head for agreement case-checking with both ergatives and absolutives is C. Since the probe is relativized only for [π], it will enter into an agreement relation with the ergative argument regardless of its person features, and will not probe further. Next, the number probe u#erg will search, finding and valuing the number of the ergative argument. Finally, a second number probe u#abs will search its c-command domain for a goal. The ergative argument, having already had its number valued, will be invisible to this second number probe, allowing the second probe to find and agree with the structurally lower absolutive argument. For a third person absolutive, this will be sufficient. However, for the reasons outlined above for intransitives, a local person absolutive, not having had its marked person features licensed, will be ruled out by the PLC.
In sum, given that no π-probe agrees with absolutive arguments, (covert) local person pronouns will not be able to occur in this case position, assuming either that covert pro obtains case through agreement or that its marked π-features are subject to the PLC.
This analysis is similar to that of Johns (Reference Johns2000), who associates the two clause- types and local and non-local person features with different-sized structures. For her, local person features occur at T and third person occurs at Asp, with indicative verbs moving higher to T, and declarative/participial verbs only moving to Asp. For reasons presented in Compton (Reference Compton2014, Reference Compton2016) (including interactions with clause-typing morphology), it is assumed here that agreement occurs instead on C.
One might instead think that both clause types are simply collapsing into one – a possibility suggested in Johns (Reference Johns2000: 9). Such an analysis would have the advantage of simplicity, as well as naturalness in terms of the types of diachronic change we might expect. However, the behaviour of verb-like adjectives in Nunavik and Labrador dialects suggests that this is not correct. Whereas in most other dialects these adjectives are compatible with either clause-type, since they act in most re- spects like verbs, in Nunavik and Labrador dialects they cannot combine directly with the indicative marker.
Furthermore, they are not directly compatible with local persons in either of these clause-types. Instead, to be used in first or second person, both the declar- ative/participial marker and the copula must precede the indicative marker – essentially a nominalization construction (see 28b).
If we were merely dealing with the merger of the two clause types (i.e., either a loss of featural contrast in the presence of local person features, or a neutralization of their vocabulary items for the clause-type markers in the context of these persons), we would not expect the grammar to require them to be stacked in this way. It therefore appears that they maintain a contrast in grammatical function.
In sum, person complementarity in Labrador Inuttut arises from the availability and configuration of the probes themselves, not as an effect of intervention or mismatched features between the relevant arguments. This conclusion is further supported by the fact that when the identity of the probing head is changed (i.e., to variants of the C head associated with other clause-types), the effect disappears. Locating the cause of the observed person complementarity on the probing head also explains why the effect extends to intransitives of the same clause-type.
5. Residual issues
As part of their argument for the periphrastic possession construction being an in- stance of the PCC, Johns and Kučerová (Reference Johns, Kučerová, Coon, Massam and Travis2017: 404), using data from Yuan (Reference Yuan, Weber and Chen2015), note that whereas local person of a possessor cannot be marked in the synthetic inflectional forms on oblique-case marked nouns in the South Baffin dialect, number can be. Johns and Kučerová take this number marking to be a repair involving a clitic, contending that if this marker were truly a default reflex of the Agree operation, and thus potentially an instance of agreement (following Preminger Reference Preminger2014), we would expect it to be third singular, not third plural, as illustrated in (30).Footnote 24
However, the form -ngin- is crucially not indexing the plurality of the posses- sor (whose local person features are at issue), but rather is the form otherwise used to indicate a third person possessor with a plural possessum. In fact, this example is an instance of default (third person singular) agreement with the possessor, with the form also indicating the number of the possessum. Furthermore, even if this morpheme had been indicating the number of the possessor, the fact that in both Béjar and Rezac's (Reference Béjar and Rezac2009) and Preminger's (Reference Preminger2014) accounts, π and # probe sepa- rately, and may yield separate exponents, would lead us to expect the possibility of default agreement for one probe and not the other. This is what Preminger refers to as the granularity of agreement (versus the featural coarseness of clitic doubling). The possibility of number marking in periphrastic possession is therefore not indicative of the PCC.
A larger, but related debate has to do with whether φ-markers in Inuit are agree- ment markers or clitics (Compton Reference Compton2014, Reference Compton2016; Yuan Reference Yuan, Weber and Chen2015, Reference Yuan2017; Johns and Kučerová Reference Johns, Kučerová, Coon, Massam and Travis2017). While this topic goes beyond the scope of the present paper, the existence of PCC effects is relevant to the debate, as Nevins (Reference Nevins2011) has proposed that the PCC holds of clitics and not of genuine agreement. In his system, clitics, but not case- marked DPs, are subject to Multiple Agree, and thus in a given language may also need to obey Contiguous Agree. Furthermore, since IO and DO clitics are assumed to move to T in a number of languages, they may enter into the same agreement domain. For Nevins, then, PCC effects are expected with pairs of D0 clitics, not with genuine agreement, which is the exponence of the Agree relation on a functional head.
As I have argued, the person complementarity effects in Inuit do not pattern like the PCC, in that they do not seem to involve intervention and are instead motivated by the availability of probes or by phonological factors. This complementarity in person-marking therefore cannot be used to diagnose the status of φ-markers as clitics.Footnote 25
6. Conclusion
In this article, I have argued that two types of person complementarity in Inuit – a ban on local person agreement with absolutive arguments in the declarative clause-type in Labrador Inuttut, and periphrastic possession of obliques in South Baffin Inuktitut – are not instances of the PCC. While the PCC is characterized by intervention and is driven by featural differences between arguments with respect to probes, the phenomena examined here were shown to be able to occur without intervening arguments (in the case of intransitives) and even to have a phonological motivation (in the case of periphrastic possession). These phenomena are therefore unlike the PCC.
Instead, it was argued that a defective or missing π-agreement probe on the C head of the declarative/participial clause-type is responsible for the person complementarity observed in Labrador. As a result, this clause-type never permits an absolutive argument to be first or second person. As expected given such an analysis, the effect extends to intransitives, but may be alleviated when the identity of the probing C head is changed (to another clause-type, with a distinct C that has a different set of probes).
The periphrastic possessor construction in South Baffin Inuktitut was argued to be the result of a regular phonological change in Eastern Canadian dialects that would have rendered a number of possessed oblique-case marked forms ambiguous between first and second person. It was shown that neighbouring dialects also affected by this change in the phonology have implemented alternative strategies to avoid a loss of contrast, some simply regularizing paradigms. The shared phonological motivation for these changes, and the non-PCC character of these alternative strategies supports the contention that the periphrastic possession construction is unrelated to the PCC. See Yuan (Reference Yuan, Weber and Chen2015) for a non-PCC analysis of this construction using feature movement.
Given that neither can be attributed to PCC, these phenomena cannot be used to diagnose the status of φ-markers as clitics or agreement based on Nevins's (Reference Nevins2011) analysis of the PCC crucially involving clitics.
Finally, with respect to the competing analyses of the PCC presented here, it is curious that an effect so similar to the PCC occurs in Labrador Inuttut, albeit due to a different syntactic configuration. The proposal made here ties this person complementarity to a defective or missing probe, with the result that local person cannot be valued on absolutive arguments. While distinct from both the Split/Cyclic-Agree and Multiple Agree accounts of the PCC, in that the effect does not arise due to features on an intervening argument, the current proposal shares with these analyses the property that such effects arise from the operation of agreement in the narrow syntax. In other words, the machinery used in these two accounts can be extended to account for the data presented here. This suggests that syntactically oriented approaches are on the right track, insofar as they may extend more naturally to related phenomena, including person complementarity in Inuit.
