Subject autonomy marking in Macro-Tani and the typology of middle voice

3.1.1 Geography and genealogy

Macro-Tani languages are primarily spoken in the far Eastern Himalaya, in the northeast Indian states of Arunachal Pradesh and Assam, as well as in small pockets across the international border with Tibet (Figure 1).

Figure 1:

Approximate geographical concentration of Macro-Tani languages, showing rough locations of languages discussed in this study.

The precise genealogical status of Macro-Tani languages is uncertain. While clearly falling within the Trans-Himalayan family, their internal position within the phylum remains uncertain. Most recent studies place Macro-Tani languages in a high-branching position, and do not generally link them with the subgroups discussed in Section 2.2 (Blench and Post 2014; Sagart et al. 2019; Zhang et al. 2019). This implies that, if the markers we will discuss in this section are indeed cognate with those discussed in Section 2.2, the origins of middle or middle-like marking in Trans-Himalayan languages must be placed at a considerable time-depth within the family.^[2]

Internally, Macro-Tani languages consist of at least a Tani subgroup – whose shared ancestor Proto-Tani has been partially reconstructed by Sun (1993) – plus the Milang language. Although Milang seems to fall outside of Tani proper (Post and Modi 2011), it is nonetheless very similar to Tani languages, and seems likely to form a genealogical unit with Tani languages at a higher level; hence, “Macro-Tani.” The authors have conducted extensive fieldwork on both Milang and the several Tani languages that will be discussed below; except when noted, all generalizations that we will make regarding middle-like marking in Macro-Tani languages appear to hold for all languages in the group, and can be safely reconstructed to at least the level of their most recent shared ancestor. We believe that cognate constructions will also be found in some neighboring languages which are generally thought of as being genealogically close to Macro-Tani languages – in particular, in the variously labelled “Mishmic” or “Lohitic” languages Kera’a and Tawrã, which are spoken to the immediate east of the Macro-Tani area. Unfortunately however, our data are not sufficient to enable inclusion of those languages into the present discussion.

3.1.2 Core grammatical features

In this section, we outline a few relevant features of Macro-Tani grammars; readers who are familiar with (Macro-)Tani grammars may safely skip ahead to the following Section 3.2.

Macro-Tani languages are verb-final and agglutinating, with an expansive predicate morphology. The structure of a finite predicate can be divided into an obligatory root or stem, any optional derivations, and at least one obligatory inflection, in that order. Inflections terminate the grammatical predicate word, and are often followed by one or more focus-marking enclitics (“particles”). Predicate derivations tend to mark functions related to manner, result, direction, aktionsart, modality and valence manipulation, however these functional categories are not lexically watertight, and it is common to find derivations coding more than one such function (10).

Upper Minyong (Adi)

“Subject” and “object” are well-coded grammatical relations in Macro-Tani languages; in addition to relational marking, evidence for subjecthood is found in imperative constructions, nominalization-based constructions, cross-clause co-referentiality constraints, and egophoricity marking (Post 2007: Section 14.1, 2017: 98-99). Relational marking is nominative-accusative, with differential object marking based on animacy and referentiality. Subjects are unmarked for relationality, as are indefinite and generic objects. Broadly speaking (details vary somewhat from language to language), definite objects are marked either by an accusative case suffix (in the cases of pronouns and demonstratives), a definite+accusative enclitic (in the cases of common nominal nps, cf. (10)), or a non-agentive postposition (in the cases of human or human-like proper names and other high animacy + discourse-prominent nps).

We next turn to middle-like marking in Macro-Tani languages.

3.2.1 Markers

Middle-like markers in Macro-Tani languages occur within the derivational area of the predicate, where they can co-occur with one, two, or perhaps as many as five or six other derivations. In principle, middle-like markers in Macro-Tani languages might be aligned with (other) markers of valence manipulation, including causatives and a wide range of applicatives; like markers of valence manipulation more generally, middle-like markers tend to occur to the right of other derivations, close to the inflectional slot, and exhibit a relatively wide leftward scope over the event semantics as expressed via the predicate stem; see again Example (10), which illustrates placement of the Minyong middle-like marker -hɨ (glossed ‘saut’). Middle-like markers in Tani languages are mostly of the form -si, as in most varieties of Lower Adi, or sometimes -su ∼ -ɕu, as in the Padam variety of Lower Adi. In the Siyom River area, a Galo/Minyong (Adi) contact zone, we usually find -hi or -hɨ. In the Subansiri River area, a Galo/Nyishi/Tagin contact zone, we usually find -ɕi or -ɕɨ. Milang shows a form -ču [tɕu]. Excluding Milang, the Tani forms seem to reconstruct to a Proto-Tani form along the lines of *-ɕu; Proto-Macro-Tani thus seems likely to have had an affricated form *-ču [tɕu].

3.2.2 Functions

In true reflexive scenarios, in which the subject is both actor and affected, middle-like marking is obligatory in all known Macro-Tani languages. In the Milang Example (11), the reflexive meaning of the construction is underscored by the presence of a reflexive/emphatic noun adaŋ ‘self.’ This construction is unacceptable without the middle-like marker -ču (12).

Milang

If the reflexive noun is omitted, the usual context-free interpretation of this construction is still reflexive. However, it is also possible to derive an active transitive interpretation from the same construction (13). The active transitive interpretation can be forced through inclusion of a non-subject-coreferential np (14).

Milang

Examples such as (14) thus demonstrate that Macro-Tani middle-like markers do not function to detransitivize a clause. Furthermore, they do not indicate a relatively higher degree of subject affectedness. Instead, the function of a Macro-Tani middle-like marker in a clause such as (14) appears to be to indicate a relatively heightened degree of subject autonomy. In other words, the sense of middle-like marking in the albeit constructed sentence in (14) is one in which the subject referent is construed as acting volitionally, with control and responsibility, and without being assisted or manipulated in its performance of the marked event (possibly in contrast to a standing expectation). Example (15) is from a Milang text in which a speaker is recalling a childhood memory.^[3] In this narrative, a group of men were returning to a village from an arrow poison collecting expedition, and distributing what food they did not consume on the trip to some watching children. In the example, the definite direct object of ram- ‘give’ is marked in the accusative, and while the indirect object Recipients are not overtly expressed in the clause, they could be (marked in an accusative or non-agentive case). The function of middle-like marking in (15) is thus not to detransitivize, nor to highlight subject affectedness, but rather to highlight the volitional and purposeful nature of the subject’s action.

Milang

Similarly, in the Tangam ditransitive clause in (16), the presence of middle-like marking functions not to detransitivize the clause, nor to highlight the subject’s affectedness, but rather to highlight the subject’s – in this case Mohandas Karamchand Gandhi’s – relatively heightened degree of independence and agency.

Tangam

Unlike in some languages, middle-like marking in Macro-Tani languages is not restricted to lexically transitive stem types; it is also common on lexically intransitive stem types, where it has similar semantic effects. In the following Milang minimal pair, subject-autonomy marking highlights the volition and control of the subject, and might be uttered by a child who is no longer being put to bed by their parents (17)–(18). In Example (19), from a Lare Galo text, subject-autonomy marking in -hí works together with the reflexive/emphatic noun aíɨ (analogous to Milang adaŋ; see above) to indicate that a door appeared to open by itself, without the involvement of a causer.

Milang

Lare Galo

Perhaps even more unusually, Macro-Tani middle-like markers can occur in predicates headed by adjectival stems. As with intransitive verbs, and because of the lack of potential ambiguity with reflexives that we find with transitive stem types, the subject autonomy function of Macro-Tani middle-like markers is especially clear with adjectives. While such adjectives are not reported by our consultants as feeling marked or unusual, they are nonetheless rare in our corpus; (20) and (21) were constructed during research for this article.

Lower Adi

These differences notwithstanding, middle-like marking in Macro-Tani languages continues to occur in constructions which are generally associated with middle marking as it has been previously described in Trans-Himalayan languages. Macro-Tani middle-like marking is common in part-reflexives (“possessor raising”), in which an affected entity is construed as a part of the subject – typically, a body part. In the attested Milang example (22), middle-like marking works together with a speaker-proximate demonstrative to bias an understanding that a referenced shoulder belongs to the subject, in this case referring to the speaker. Yet Milang -ču does not force such an interpretation; if it is made clear that the referenced shoulder belongs to a different person, the predicate need not change its form (23). In the latter case, and in fact in both examples, the core function of Milang -ču is again to highlight subject autonomy.

Milang

In core Tani languages, middle-like marking is obligatory in reciprocal constructions (24)–(25). Milang differs in this respect, and does not make use of its middle-like marker -ču in reciprocal constructions (cf. Modi 2017: Section 9.2.5).

Lare Galo

In Lower Adi as in most other Eastern Tani languages, middle-marking is an obligatory component of the passive construction. In a Lower Adi passive, the subject is understood as a patient, the actor is expressed as a syncretically ablative/instrumental-marked oblique, and the predicate is marked by two morphemes: -ko ‘pass’ and the middle-like marker -si (26). In the case of a Lower Adi passive, middle-like marking might be said to at least coincide with reduced transitivity. However, it would not appear reasonable to attribute the transitivity reduction component of a Lower Adi passive to the middle-like marker per se; more likely perhaps, it should be attributed to the passive marker -ko ‘pass.’ Furthermore, although the subject is uncontroversially “affected” within a Lower Adi passive construction, this does not seem to be a factor contributing to the appearance of middle-like marking within this construction. Rather, what is criterial here is the expression of a semantic patient as subject. While patient subjecthood is sometimes described as less fundamental to the form and function of passive than is “demotion” of an erstwhile agent (Givón 2001: 125; Shibatani 1985), in Macro-Tani languages a passive in -ko-hi ‘pass-saut’ is in fact canonically associated with an expression of volition or control on the part of the patient. In (26), our Lower Adi consultants are quite clear that the subject referent is understood as somehow responsible for the misfortune it has undergone, which seems likely to have come about as a consequence of its own actions. This interpretation of a Lower Adi passive is thus consistent with our claim that the principal function of Macro-Tani middle-like markers is to mark subject autonomy rather than subject affectedness.^[4]

Lower Adi

Middle-like marking in Macro-Tani languages is frequently found on verbs denoting situations that are prototypically associated with middle voice crosslinguistically. These may include grooming actions, changes in body posture, emotions, cognitive actions, and perception (cf. Kemmer 1994: 195). Middle marking is not usually obligatory with such events, however, and when it does occur it tends to correlate with an enhanced focus on the centrality of the subject.^[5]

Grooming verbs in Macro-Tani languages are lexically transitive, and prototypically associate with a particular type of object noun – usually, one denoting a body part. As with the part-reflexives discussed above in this section, the appearance of middle-like marking in a clause denoting a Macro-Tani grooming event typically supports a reading in which the object noun pertains to the subject (27)–(28). Yet here too, it is possible to force a non-reflexive reading through insertion of a non-subject-coreferential genitive possessor (29). While a sense of subject-autonomy is present in both cases, it may be more marked in the second case.

Lower Adi

We have only been able to attest a single exception to our generalization that middle-like markers do not function to detransitivize stems in Macro-Tani languages. This exception is with what seems to be the most prototypical “middle verb” in the grooming class, and perhaps overall: ‘wash (a body).’ In Eastern Tani languages and in Milang, verbs with this meaning tend to be based upon a transitive verb root (30), yet show reduced transitivity with middle-like marking (31)–(32). In fact, the corresponding forms in Tangam (hɨɨhɨ) and Milang (hamcu) represent the only true middle deponents yet attested in Macro-Tani languages; neither can occur in their etymological root forms. In the Tangam case, it is clear that the modern-day deponent no longer reflects a productive composition, the vowel of the etymological middle-like suffix having irregularly harmonized with the preceding root nucleus (Proto-Tani *ɦɨr-ɕu  > Tangam hɨɨhɨ ; the usual Tangam correspondence is -hi, cf. Example (16) above). In both cases, the resulting stems are lexically intransitive.

Lower Adi

One final remark on distribution: middle-like marking has the very unusual property among Macro-Tani predicate derivations of being able to occur twice within at least some types of stem: once closer to the root, and a second time closer to inflections. In such cases, only the leftmost marker can be ambiguous between (part-)reflexive or autonomous readings (33); the rightmost marker can only have an autonomous reading (34).^[6]

Lower Adi

3.2.3 Discourse distribution and frequency

As discussed in Section 2.1, middle voice constructions are structurally and functionally marked, and generally apply to a restricted subset of the verbal lexicons of middle marking languages. Although Macro-Tani middle-like marking is likewise marked relative to active voice constructions, we have been unable to detect any lexical restrictions at all in its occurrence. As was discussed in Section 3.2.2. Macro-Tani middle-like markers appear to be able to occur on all manner of transitive and intransitive verbs, and even mark at least some types of adjectives. We are unaware of a corpus study that addresses the relative frequency of active vs. middle-marked clauses crosslinguistically, and we have not been able to conduct one for the present study of Macro-Tani languages. Impressionistically, however, we feel that the frequency of middle-like marking in Macro-Tani languages is likely to be significantly higher than may be the case in languages whose middle marking hews more closely to the received typological prototype. As a very rough initial index of frequency, we found a total of 93 occurrences of the middle-like marker -hi in a Galo corpus comprising 3,170 lines of text, at a rate of 2.9% (or once every 34 lines of text). In a smaller Tangam corpus of 634 lines, the middle-like marker -hi occurred more frequently, at 49 mentions or a rate of 7.7%. Of these mentions in both languages, only a tiny number occurred in reflexive or part-reflexive clauses; the overwhelming majority (>95%) functioned purely to highlight the autonomy of clause subjects. Thus, we found that texts in which “self-reliance” was a significant theme showed a higher proportion of mentions. One such text is examined here:

In a Tangam text entitled Kids’ Future (Text 4 of Post 2017: 284–293), speaker Baduk Tamut discusses the most significant lifestyle choice that Tangam-speaking children will make: to remain in the Tangam-speaking area, or to move down to the Adi- and Hindi-speaking lowlands. His discussion contrasts the conditions of upland peoples and lowland peoples by talking about what an individual can do in each place, what sorts of resources they will have access to, and how self-reliant they will be able to be. Out of a total of 81 predicates that were uttered across the 39 lines of this text,^[7] we found that 18 clauses showed middle-like marking, or nearly 22%. A few lines are re-presented here to give something of the flavor of this narrative. Note that in the first two lines (35)–(36), which discuss life in the non-Tangam-speaking lowlands, middle-like marking is absent. When the perspective shifts to that of the Tangam-speaking area, the self-reliance and autonomy of Tangam speakers’ traditional lifestyles, and likewise their control over the means of production in their area, are highlighted through predicate marking in -hi in every line, and in nearly every clause (37)–(41).

Tangam

We of course realize that these are simply a few lines of a single text in only one of many Macro-Tani languages. However, we judge this sequence to be representative of at least the subject autonomy-highlighting function of middle-like marking in Macro-Tani as it can be deployed in discourse.^[8]