I’ve been thinking recently about agreement patterns, and I think there’s an interesting one at the intersection of Chukchi–Kamchatkan style agreement and the transimpersonal constructions that are one potential source of split (or active-stative) intransitivity.

The Chukchi – Itelmen Pattern

Languages in the Chukchi and Itelmen family often show an agreement pattern where verbal agreement prefixes are always controlled on a nominative S-A basis, but suffixes seem to show an absolutive alignment. This produces a pattern where transitive verbs agree with A and P via separate morphological slots, but many intransitive agreement possibilities are marked by circumfixes. Compare the following Itelmen examples from this article:

The complicating factor here is that the suffixes for the same person do vary with transitivity, so an intransitive verb generally doesn’t look like a reflexive transitive verb even if it has circumfixal agreement.

Transimpersonals and Split Intransitivity

Moving on to transimpersonals, in The Typology of Semantic Alignment, Andrej Malchukov argues that an agentive-patientive or active-stative split in intransitives can arise from transitive verbs with a non-referential, impersonal or generic subject. The following West Greenlandic example literally means ‘it stormed us’:

In some Yupik varieties this has gone further, so that the transitive inflection with indefinite subject can be used generally with intransitive verbs to mark lack of control and spontaneous or unintended actions:

The second example literally means something like “it started to crawl the child”.

In languages where transitivity is not otherwise overtly marked and third person actor agreement is zero, a similar pattern can result in ambiguity between a transitive and intransitive clause and reanalysis as a patientive intransitive. For example, in Ika, there is no structural difference between a patientive intransitive and a transitive verb with a third person agent:

Notice that this pattern assimilates some underlying intransitives to transitive agreement, but only the patientive ones. The agentive ones may still be distinct, especially because in order for there to be ambiguity with an agentive transitive under this null morpheme pattern:

• There would have to be zero marking of patient objects
• And zero marking of transitivity

… which would probably create ambiguous agreement forms which could mean “I Xed him/her” and “I Xed (myself)”.

The (Simple) Idea

The obvious way to fix this and fully assimilate all intransitives to the transitive paradigm would be to make agentive intransitives inflect like transitive reflexive verbs. What you basically have then is one agreement slot that always agrees with the agent if there is one (or a generic impersonal agent if not), and one that always agrees with the most affected entity (which can’t be impersonal).

Let’s demonstrate this with some made-up agreement forms based on the conlang phonology I posted previously. Here are the forms:

Person	Agent Prefix	Undergoer Suffix
1st singular	m-	– xʷ
1st plural	k-	-m(ə)n
2nd singular	t-	-t
3rd animate singular	i- ~ j-	-(V)r
2nd/3rd animate plural	xa(r)-	-ntʃ
3rd inanimate/impersonal singular	–	–
3rd inanimate plural	–	-n
3rd animate and 2nd plural reflexive	N/A	-ʃ

Under the logic that agentive intransitives inflect like reflexives and patientives inflect like transitives with an impersonal subject, you get examples like the following:

Person	Agentive Intransitive (-pana-)	Patientive Intransitive (-min-)
1st singular	mə-pana-xʷ	min-xʷ
1st plural	kə-pana-mən	min-mən
2nd singular	tə-pana-t	min-t
3rd singular animate	i-pana-ʃ	min-ir
2nd/3rd animate plural	xa-pana-ʃ	min-əntʃ
3rd singular inanimate	pana	min
3rd inanimate plural	pana-n	min-ən

Emergent patterns:

• For first person and second person singular presence or absence of the agentive suffix is the only marker of agentivity
• For plural second and third person animate, a difference in the undergoer suffix from reflexive -s to non-reflexive agreement also marks the distinction
• For third person inanimate, there is no difference (it’s hard to regard inanimate objects as agentive in any case)

This system is clearly not maximally efficient (most agentive intransitives end up with double agreement morphology), but it’s no more inefficient than other systems attested in the wild (e.g. in Chukchi and Itelmen), and arguably more efficient since the circumfixal agreement in those languages doesn’t mark the agentive ~ patientive distinction.

Addendum: Reflexives and Middles

Another point worth making is the obvious one that in some languages a subset of common intransitive verbs are formally reflexive transitives. It’s fairly common for a reflexive construction to evolve middle voice functions, where middle covers a diverse array of meanings involving lowered transitivity or the affectedness of the subject. The list from Grammatical Voice by Zúñiga and Kittila includes:

• Anti-causatives (e.g. vanish, recover)
• Natural auto-benefactives (e.g. acquire, request)
• Natural reciprocals (embrace, converse)
• Emotion middles (be angry, be frightened)
• Emotive speech actions (complain, lament)
• Cognition middles (believe, meditate)
• Bodily action verbs (wash, shave)
• Non-translational motion (turn, bow)
• Translational motion (climb up, fly)
• Change in body posture (sit down, rise)

All of these are situations where a participant is affected and the cause of the action is either irrelevant or internal to the affected participant. Most of them are therefore agentive in nature, although not all. It doesn’t seem impossible for a language with both this and the transimpersonal construction to expand them until the entire intransitive domain is consumed by other or the other (or both if there are multiple construals available).

Addendum 2: Merger of Agentive Intransitives and Reflexives

It’s also worth pointing out that many languages show a partial merger of reflexives / reciprocals and active intransitives, although it’s usually expressed by making reflexives formally intransitive instead of by making active intransitives formally transitive. English has this for many verbs where the reflexive is the most common by usage: intransitive “wash” means “wash oneself”, intransitive “fight” means “fight each other” etc. But this is not the universal pattern, and there are some reflexives in English which are structurally distinct to an active intransitive.

Grammatical Voice claims that Leko, a language from Bolivia, has this as its dominant / only strategy for forming reflexives, and gives the following examples:

I would be surprised if there were truly no more explicit reflexive construction at all, but even so it makes sense for some language to use intransitive agreement or coding of a semantically transitive verb to express a reflexive or reciprocal meaning.

Another example of a more marked identity between a derived subset of active intransitives and reflexives is a polysemous antipassive~reflexive marker. The following examples are from Warrungu:

More examples of antipassive~reflexive polysemy can be found here.