Canela dialect

Vowels

Consonants

• Stops /p, t, k/ can be voiced [b, d, ɡ] in syllable-initial position within unstressed syllables, following voicing and in syllable-final position following voicing.
• The velar nasal /ŋ/ can fluctuate in free variation with a prenasalized and voiced plosive [ŋɡ~ɡ] between dialects.
• In certain dialects, the affricate /t͡s/ becomes post-alveolar [t͡ʃ].
• /n, m/ are prestopped [ᵈn, ᵇm] when after oral vowels.
• /v/ is only realized as such syllable-initial position, elsewhere it is pronounced as [w].
• /j/ is heard as [z] when in final position of consonant clusters and in initial position within stressed syllables, and is nasalized as [z̃] before nasal vowels. It is heard as [j] elsewhere in syllable-final and unstressed syllable-initial positions.
• /l/ is heard as a lateral flap [ɺ] when in intervocalic positions, or following consonants.
• /h/ is heard as a glottal stop [ʔ] in syllable-final position when preceding consonants, and as a velar [x] before high-oral vowels.^[5]

Finiteness morphology

In Canela, like in all Northern Jê languages, verbs inflect for finiteness and thus have a basic opposition between a finite form and a nonfinite form. Finite forms are used in matrix clauses only, whereas nonfinite forms are used in all types of subordinate clauses as well as in some matrix clauses (such as recent past clauses and any clauses which contain modal, aspectual, or polar operators).^[6]: 450 Nonfinite forms are most often formed via suffixation and/or prefix substitution. Some verbs (including all descriptives with the exception of cato 'to exit', whose nonfinite form is cator) lack an overt finiteness distinction.

The following nonfinite suffixes have been attested: -r (the most common option, found in many transitive and intransitive verbs), -n (found in some transitive verbs), as well as -c and -m (found in a handful of intransitive verbs which take a nominative subject when finite).^{[6]: 448–9 [7]}

In Proto-Northern Jê, a handful of verbs, all of which ended in an underlying stop, formed their finite form by means of leniting the stem-final consonant (*-t, *-c, *-k → *-r, *-j, *-r); in turn, the nonfinite form received no overt marking.^[8]: 544 At least three verbs still follow this pattern in Canela.^[7]

Person inflection and case

In all Northern Jê languages verbs, postpositions, and relational nouns inflect for person of their internal argument by taking absolutive ("internal") or accusative person prefixes. The accusative series is required by a subclass of transitive verbs (in finite clauses only) as well as by some postpositions; the absolutive series is the default one and is found with most transitive and all intransitive verbs in finite clauses, with all verbs in nonfinite clauses, with all relational nouns, and with some postpositions.^[8] External arguments of verbs are not indexed by person prefixes but are rather encoded by nominative/agentive (unmarked) noun phrases (including personal pronouns) in finite clauses, or by ergative phrases in nonfinite clauses.

In the table below, the label class II refers to a subclass of vowel-initial stems which take the thematic consonant */ĵ-/ in the basic (uninflected) form as well as in some inflected forms (e.g. *∅-j-arkwa ‘my mouth’, *ba-j-arkwa ‘our mouths’, *rop j-arkwa ‘the jaguar's mouth’) but not in others (*g-arkwa ‘your mouth’, *c-arkwa ‘his/her/its mouth’). The archaic allomorphs *∅-/ĵ-/ (first person, class II) and *g- (second person, class II) are only marginally preserved across Northern Jê: the former is preserved in Pykobjê (as in j-apackre ‘my ear’),^{[9]: 47, 97} whereas the reflexes of the latter have been found in Kĩsêdjê, Canela, Pykobjê^{[3]: 213–4} and in the triadic kinship terms of Mẽbêngôkre.^{[3]: 217–8}

Voice

Two valency-reducing operations are encoded by prefixes in Northern Jê: the anticausative voice (finite *a(j)-, nonfinite *bi(t)-) and the antipassive voice (finite *a(p)-, nonfinite *jə-/*ju-).^{[8]: 541, 544}

Nominal number

In most Northern Jê languages, nouns which denote human beings may receive an overt collective plural suffix (Proto-Northern Jê -jê). Its reflexes have been attested in Kĩsêdjê (-jê), Tapayúna (-jê),^[10] Parkatêjê (-jê), Pykobjê (-ji), Canela (-jê), among others. In fact, this suffix is part of many names of Northern Jê peoples, as in Kĩsêdjê, Parkatêjê, Pykobjê (self-denomination Pyhcopji), Apànjêkra, and is the ultimate origin of the term Jê.

Morphosyntactic alignment

Different main clause constructions present different combinations of alignment patterns, including split-S (default), ergative–absolutive (recent past), and nominative–absolutive (evaluative, progressive, continuous, completive, and negated clauses). In contrast, subordinate clauses are always ergative–absolutive.

Prototypically, finite matrix clauses in Canela have a split-S alignment pattern, whereby the agents of transitive verbs (A) and the sole arguments of a subclass of intransitive verbs (S_A) receive the nominative case (also called agentive case^[3]), whereas the patients of transitive verbs (P) and the sole arguments of the remaining intransitive predicates (S_P) receive the absolutive case (also called internal case^[3]).^[6] In addition, transitive verbs are subdivided into two classes according to whether the third person patient is indexed as absolutive (allomorphs h-, ih-, im-, in-, i-, ∅-) or accusative (cu-),^[6] which has been described as an instance of a split-P alignment.^[3]: 272 There are only several dozen of transitive verbs which take an accusative patient, all of which are monosyllabic and have distinct finite and nonfinite forms. It has been suggested that all transitive verbs which satisfy both conditions (monosyllabicity and a formal finiteness distinction), and only them, select for accusative patients, while all remaining transitive verbs take absolutive patients in Canela and all other Northern Jê languages.^[8]: 538

All subordinate clauses as well as recent past clauses (which are historically derived from subordinate clauses and are headed by a nonfinite verb) are ergatively organized: the agents of transitive verbs (A) are encoded by ergative postpositional phrases, whereas the patients of transitive verbs (P) and the sole arguments of all intransitive predicates (S) receive the absolutive case (also called internal case^[3]).^[6]

Evaluative, progressive, continuous, completive, and negated clauses (which are historically derived from former biclausal constructions with an ergatively organized subordinate clause and a split-S matrix clause) in Canela have the cross-linguistically rare nominative-absolutive alignment pattern.^{[11][12][6][13]} An example of this alignment type in negated clauses is given below.^[11]: 162

In nominative–absolutive clauses, the sole argument of an intransitive verb (S) is aligned with the agent argument of a transitive verb (A) in that both may be expressed by nominative pronouns, such as wa 'I.NOM' or ca 'you.NOM' (nouns do not take case inflection in Canela), which occupy the same position in a phrase (in the example above, both precede the irrealis marker ha). At the same time, the sole argument of an intransitive verb (S) is aligned with the patient argument ('direct object') of a transitive verb (P) in that both may be indexed on the verb by person prefixes of the absolutive series ( such as i- 'I.ABS' or a- 'you.ABS'). There are no elements which pattern as ergative or accusative in this type of clauses in Canela.

The historical origin of the nominative–absolutive clauses in Canela has been shown to be a reanalysis of former biclausal constructions (a split-S matrix clause, headed by the auxiliary, and an ergative–absolutive embedded clause, headed by the lexical verb) as monoclausal, with the loss of the ergative.^[6][12]

Classes of predicates

The following table summarizes the proposed classes of predicates in Canela.