The person-animacy connection: Evidence from Algonquian and Dene

1. Introduction

This squib presents evidence from the Algonquian and Dene language families to support a connection between person and animacy. A range of morphosyntactic patterns in these languages, including pronoun inventories, agreement restrictions, and hierarchy effects, are argued to indicate that inanimate nominals lack formal person features. This proposal allows the morphosyntactic patterning of inanimates to fall out from grammatical principles that are independently required to account for person effects. We conclude that the often-assumed model in which third persons are “personless” must be revised to allow for languages in which only inanimate third persons lack formal person features.

The squib is organized as follows. Section 1 provides background on person features and the notion of personlessness. Section 2 shows that various patterns in Algonquian and Dene morphosyntax follow from an analysis in which inanimate third persons are personless but animate third persons are specified for person. Section 3 considers whether the proposed person-animacy connection is conditioned by semantic animacy or grammatical animacy.

2. Background: Person and personlessness

Since the work of Benveniste (1966, 1971), it has been a common assumption that the representation of speech-act participants – first and second person – differs from third persons in that only participants bear a formal person feature. The absence of person features on non-participants has been shown to account for a range of phenomena, from pronoun forms to agreement morphology to licensing of other features (Kayne 2000, 2002; Harley and Ritter 2002). In recent work, however, a finer-grained distinction has arisen. It has been proposed for various languages that non-participants are not a unified syntactic class: the absence of person features applies not to all third persons, but only to inanimate third persons (Rooryck 2000; Alexiadou and Anagnostopoulou 2006; Piriyawiboon 2007, 2013; Adger and Harbour 2007; Demonte et al. 2011; Richards 2014; Ghomeshi and Massam 2015; Bartošová and Kučerová 2015).

In this squib, we test this proposal against data from Algonquian and Dene, two language families in which the morphosyntactic prominence of person features makes it particularly straightforward to assess the status of third persons. Using morphological agreement patterns and their failure with inanimate arguments as diagnostics, we show that the Algonquian and Dene data strongly support the proposed connection between animacy and person: animate and inanimate third persons consistently pattern differently, with only animate third persons taking part in the same patterns as first and second persons. This difference follows if animate third persons have person features while inanimate third persons do not.

For concreteness, we adopt the model of person features shown in Table 1, in which the articulation of the person specification corresponds to proximity to the deictic centre (Béjar 2003, Béjar and Rezac 2009, Lochbihler 2012).

Table 1: Person feature specifications in Algonquian and Dene

In these representations, first, second and third person animate nominals are specified for person features, but inanimate nominals are not.

3. The patterning of person and animacy in Algonquian and Dene

Person plays a significant role in various morphosyntactic patterns in Algonquian and Dene, including pronouns (Algonquian, section 3.1), verbal agreement (both families, sections 3.2 and 3.3), obviation (Dene, section 3.4), copula insertion (Dene, section 3.5), and direct-inverse marking (Algonquian, section 3.6). Each pattern provides evidence for the proposed connection between animacy and person.

3.1 Personal pronouns in Algonquian

Most Algonquian languages have a set of pronominal person prefixes that occur in both possessed nouns and emphatic personal pronouns, as illustrated for Innu in (1) (Drapeau 2014: 56, 95).¹

Importantly, the Innu third-person prefix u- can only be used to index animate third persons in both possessive constructions and emphatic pronouns (Drapeau 2014: 55, 87). There is no inanimate third-person prefix; if pronominal reference to an inanimate is desired, the only option is to use a deictic demonstrative such as neme ‘that (inan.)’ (Drapeau 2014: 87). If the pronominal person prefixes ni-, tshi-, u- are analyzed as expressing person features, then the exclusion of inanimates from the person prefix paradigm follows directly from the absence of person features in the specification of inanimates, as in Table 1 above.²

3.2 Verbal agreement in Algonquian

In the inflectional paradigm known as the conjunct order, an Algonquian verb inflects with a string of suffixes, one of which is a person/number agreement marker called the central suffix (Goddard 1979, Nichols 1980). When a transitive verb takes an animate third-person object, the central suffix can be a portmanteau form that expresses features of both the subject and the object simultaneously. For example, the Ojibwe forms with animate objects in the first column of Table 2 display the portmanteau central suffixes -ak ‘1sg>3sg’ and -at ‘2sg>3sg’, which are dedicated to these particular subject-object combinations. When a transitive verb takes an inanimate third-person object, however, portmanteau central suffixes never appear. Instead, the central suffix always patterns exactly as it does in an intransitive verb, indexing only the subject. For example, the Ojibwe forms with inanimate objects in the second column of Table 2 display exactly the same subject-marking central suffixes -aan ‘1sg’ and -an ‘2sg’ that occur in the intransitive forms in the third column.³

Table 2: Agreement in Ojibwe negative conjunct forms (Nichols 1980)

There is ample evidence that transitive verbs with inanimate objects are syntactically fully transitive in Ojibwe (Lochbihler 2012: 71), such as the fact that the object is obligatory and can only be suppressed through the addition of derivational antipassive morphology. The intransitive appearance of the central suffixes in the inanimate-object forms in Table 2 thus cannot be attributed to the absence of a syntactic direct object. It can, however, be attributed to the absence of a person feature on the object, as in our proposal. The personlessness of the inanimate object means that the subject is the only argument that bears person features and is thus the only possible goal for person agreement, just as in an intransitive. Consequently, any agreement position that tracks person features – such as, by hypothesis, the Ojibwe central suffix – will be able to index an animate third-person object but not an inanimate third-person object.⁴

3.3 Verbal agreement in Dene

In Dene, as in Algonquian, there is evidence that inanimate third persons are ignored by verbal agreement. Singular third-person subject agreement is zero in Dene languages; thus, singular subjects do not provide a usable diagnostic for distinguishing third persons from non-persons, except in the case of adjectival predicates, for which see section 3.5.

Plural agreement, however, is overt, and appears only with animate subjects. The Tłı̨chǫ data in (2) shows that an animate plural subject triggers agreement on the verb (prefix ge- 3pl) in (2a) while an inanimate plural subject does not, as shown in (2b).

In the same way, transitive verbs show agreement with animate objects (prefixes we- 3sg.obj and gı- 3pl.obj) in (3a) but not with inanimate objects, as shown in (3b).

The patterning of we- and gı- thus precisely parallels the third-person pronominal prefix in Algonquian languages discussed in 3.1.

As in Algonquian, the failure of agreement with inanimates in Tłı̨chǫ follows if the relevant agreement positions track person features and inanimates lack person features.

3.4 Obviation in Dene

In Dene languages, the usual object agreement markers cannot appear when both the subject and the object are third person, as shown in (4a), but an obviative marker signalling non-coreference of the two third person arguments is possible, as in (4b).⁵

The obviative marker has a plural form go- that occurs with an animate object, as in (5).

However, the plural form of the obviative marker cannot occur with a plural inanimate object, as shown in (6a). The general obviative marker must instead be used, as in (6b).

The inadmissibility of the plural obviative marker with inanimates reinforces the observation from the preceding section that inanimates cannot trigger agreement on the verb, a fact that can be accounted for if agreement seeks person features and inanimates lack person features. The general obviative marker ye- is the only object marker that can track an inanimate object, since it does not mark agreement but rather non-coreference with an antecedent.

3.5 Copula insertion in Dene

Some Dene languages have a small class of non-inflecting predicative adjectives, which appear without a copula as predicates of inanimate subjects, as in (7a) versus (7b), but require the insertion of a copula in order to be grammatical with animate subjects, as in (7c) versus (7d) (Tłı̨chǫ; MLBW2012).

This phenomenon reduces to the occurrence of person agreement with animate but not inanimate subjects. With an animate subject, the copula is inserted to host the morphology that results from person agreement with the subject. With an inanimate subject, no person agreement takes place and hence the copula need not be inserted.

Copula insertion also provides a crucial piece of evidence against the possibility that inanimates possess a person feature that is realized by a Ø-morpheme. As mentioned in 3.3, verbal subject agreement is phonologically null in Dene languages for third person singular, regardless of the animacy of the subject. It would be tempting to conclude that animates and inanimates alike trigger zero agreement, though both carry person. However, if this were the case, one would expect copulas to be inserted with all adjectival predicates, since the realization of any agreement morpheme would require a copula. That copulas are absent with adjectival predicates of inanimate subjects demonstrates that person agreement with these subjects is also truly absent, rather than present but spelled out by a silent morpheme.⁶

3.6 Inverse marking in Algonquian

Algonquian languages show a pattern of direct-inverse marking, in which the morphosyntactic alignment of agreement on transitive verbs is determined according to the person hierarchy in (8) (e.g., Pentland 1999: 235, Valentine 2001: 268).

1. (8) Algonquian person hierarchy

   1st/2nd > animate proximate 3rd > animate obviative 3rd > inanimate 3rd

As part of this system, a verb is marked with an “inverse” suffix whenever the object outranks the subject on the person hierarchy, as in the Meskwaki forms in (9) (Goddard 1994).⁷

Béjar and Rezac (2009) derive the distribution of the inverse marker from the articulated person features in Table 1 above: since each step down the person hierarchy corresponds with one less degree of articulation in the person specification, the inverse marker can be understood as occurring whenever the object has a richer person specification than the subject.

Béjar and Rezac do not discuss the fact that inanimates are ranked below animates on the person hierarchy, as indicated by inverse marking in inanimate-subject forms such as (9c). Under the traditional view in which animacy is simply a gender feature (e.g., Goddard 2002), it is surprising that animacy plays a role on a hierarchy that is otherwise derived purely by the richness of the person feature.⁸ If we take inanimates to lack person features, however, their ranking at the bottom of the person hierarchy is exactly what we expect. In any form with an inanimate subject and an animate object, the animate object has person features while the inanimate subject does not. Under Béjar and Rezac's analysis, in which the inverse theme sign appears whenever the object is a richer goal for person agreement, it follows that all inanimate-subject forms should be inverse, as the animate object will always be a richer goal than the personless subject.

3.7 Summary

The proposal that inanimates lack person features accounts for a range of morphosyntactic patterns in the Dene and Algonquian families, as summarized in Table 3. The proposed connection between animacy and person enables a simple and unified analysis of each pattern. If this connection were not recognized, we would be forced to stipulate that each pattern in Table 1 is conditioned by both person and animacy together, a more complex analysis that does not explain why the connection between person and animacy is so pervasive in Algonquian and Dene morphosyntax.

Table 3: Patterns accounted for by the person-animacy connection

4. Semantic and grammatical animacy

The correspondence between animacy and person in Algonquian and Dene is robust, but there is variation in the precise division languages make between nouns that bear [person] and nouns that do not. In the Dene languages, this division is semantically based: speakers of Ts’úùt’ínà and Tłı̨chǫ divide humans and animals (animate) from plants and non-living things (inanimate), as demonstrated by the plural agreement triggered by the object ‘rabbits’ in (5) but not by ‘oranges’ in (6b).⁹ The neighbouring Dënesųłıné makes the “animacy cut” in a different place from Tłı̨chǫ, separating humans from all other animals, plants, and non-living things, as demonstrated by the adjectival predicates in (10). In the Dënesųłıné form in (10a), ‘dog’ patterns as inanimate, failing to trigger agreement, but in the equivalent Tłı̨chǫ form in (10b), ‘dog’ patterns as animate, obligatorily triggering agreement and copula insertion (see the ungrammatical form without agreement in (10c)).¹⁰

The semantic nature of the trigger is made clear by minimal pairs such as (11), where a verb's agreement with its subject varies according to the interpretation of the subject as living (11a) or dead (11b). Plural agreement is only possible with a living referent (11c).¹¹

Regardless of exactly where the cut is made, the distribution of [person] in Dene is rooted directly in semantics: all nominals that pass a certain threshold of semantic animacy bear person features. In Algonquian the situation is more complex, as there is a grammatical gender contrast between animate and inanimate nominals (Dahlstrom 1995, Goddard 2002). The gender contrast is grounded in semantic animacy: all nouns with semantically animate referents (humans, animals, spirits) belong to the animate gender. However, the animate gender also includes many nouns with semantically inanimate referents, and the gender classification of such nouns is not predictable from their semantics alone. In Meskwaki, for example, ‘snow’, ‘potato’, and ‘raspberry’ are animate, while ‘fire’, ‘squash’, and ‘strawberry’ are inanimate (Dahlstrom 1995: 60). The status of such nouns as grammatically animate or inanimate must be arbitrarily specified in the lexicon. For Algonquian, then, our proposal faces an important question: which notion of “animacy” is involved in the person-animacy connection? Is [person] borne only by nominals with semantically animate referents, or by all grammatically animate nominals, whether or not the referent is semantically animate?

A full examination of this issue is beyond the scope of this squib. However, taking Innu as an example, we observe that the answer appears to depend on which diagnostic we consider. For personal pronouns (section 3.1), it is semantic animacy that is important: the Innu third-person pronoun uiǹ can refer only to a human or animal (Drapeau 2014: 87). In contrast, for person agreement (section 3.2), it is grammatical animacy that matters: portmanteau person agreement suffixes such as -ak ‘1sg>3sg’ occur whenever the object is a grammatically animate third person, regardless of its semantic animacy, as shown for the grammatically animate but semantically inanimate object ‘helicopter’ in (12).¹²

In Innu, then, the feature [person] patterns differently on pronouns and lexical nouns: a pronoun with the feature [person], such as uiǹ ‘s/he’, must be interpreted as having a semantically animate referent, but this is not the case for a lexical noun with the feature [person], such as sheuekatshu ‘helicopter’. We suggest that in Innu, as in Tłı̨chǫ and other Dene languages, the feature [person] is ultimately tied to semantic animacy, a connection that can be transparently observed in the interpretation of Innu pronouns. Unlike in Tłı̨chǫ, however, Innu also allows the feature [person] to be idiosyncratically included in the lexical entry for particular nouns, even when the referent is not semantically animate, as is the case for sheuekatshu ‘helicopter’. The result, on the surface, is a blurring of the connection between [person] and semantic animacy in Innu, in contrast to the tight connection shown in Tłı̨chǫ. We must leave further investigation along these lines to future research.

5. Conclusion

Algonquian and Dene, the largest two language families of North America, provide compelling evidence that [person] is specified on animate nominals but not on inanimate nominals. This proposal allows a wide variety of animacy effects in pronominal forms, agreement, copula insertion, direct-inverse marking, and obviation to follow straightforwardly as specific cases of more general person effects. The absence of inanimate personal pronouns, the failure of agreement with inanimate arguments, and the lack of copula insertion with inanimate subjects are all consequences of the absence of a [person] feature to realize or agree with, while the ranking of inanimates at the bottom of the person hierarchy reflects the fact that all other members of the hierarchy have [person] while inanimates do not.

Our analysis upholds the longstanding idea that some nominals can be “personless” but suggests that the precise nature of the personless class is subject to crosslinguistic variation. The Benveniste approach in which all non-participants lack person features may be correct for some languages, but in others the personless class may consist of only a subset of third persons, such as Algonquian and Dene inanimates, Persian inanimates (Bayanati and Toivonen 2019), Spanish se (Nevins 2007), and English null objects (Massam et al. 2017). The extensive connections between animacy and person in Algonquian and Dene morphosyntax provide a thorough demonstration of the utility of allowing the specification of person features to vary in this way.