So... recently I've been dicking around with this largely isolating language called Tǝɣrâ, which is spoken on Telmona by an ethnic group by the name of Tǝɣor (and their conquests). I'm pretty pleased with the argument structure. So here it is.

Person marking and word order in Tǝɣrâ

Stated baldly, the underlying morphosyntactic alignment of Tǝɣrâ is deceptively simple: ergative-absolutive marking with EAV word order. A number of factors, however, conspire to make the surface instantiation of this irksomely complex.

Let us begin with the least marked species of sentance we can- a verb in the imperfect aspect with two animate third person arguments:

lɨɣaolord yeERG tǝshâbeat.IMP kuaboy
The lord is beating the boy.

We can immediately see that there is a deviation from the canonical argument order stated above: we have EVA, not EAV. This is due to the verb-second constraint in Tǝɣrâ. If we look at the same sentence with the verb in the perfect aspect, which is marked by an auxiliary verb, we see the following:

lɨɣaolord yeERG nduPRF kuaboy tâbeat.PRF
The lord has beaten the boy.

(Ignore the tǝshâ~tâ change: a number of verbs in Tǝɣrâ have wholly dissimilar suppletive stems used in the perfect aspect. It does not pertain here.)

Another complication is when we add agent pronouns: canonically these always follow the verb:

kuaboy tǝshâbeat.IMP ṅe1SG
boy beat.IMP 1SG
I am beating the boy.

Here we see AVE order. In the perfect:

kuaboy nduPRF ṅe1SG tâbeat.PRF
boy PRF 1SG beat.PRF
I have beaten the boy.

Note in both sentences that the agent is not marked with the ergative particle ye. This particle is only obligatory with 3rd person agents. When used with first and second person agents, it has a disjunctive force. Additionally, when first or second person pronoun co-occurs with the ergative particle, it behaves in the same manner as a full noun, thus:

ṅe1SG yeERG tǝshâbeat.IMP kuaboy
I am beating the boy.

Incidentally, we've just seen how to emphasise a 1st/2nd person subject there. To emphasise the object, we move it to the left edge of the phrase:

kuaboy tǝshâbeat.IMP lɨɣaolord yeERG waWA
The lord is beating the boy.

Two things to note here: if you're keeping up, you will have expected the AVE word order here, on account of the verb-second constraint. On the other hand, you might not have expected the particle wa at the end of the sentence. We'll explore this in greater depth below, but for now let's just note its position. This wa is one of the few preposed verbal particles- it is not a sentence-final particle. This is more evident in the perfect:

kuaboy nduPRF lɨɣaolord yeERG waWA tâbeat.PRF
The lord has beaten the boy.

It gets just a little bit more complex when we have a pronominal agent rather than a nominal:

kuaboy tǝshâbeat.IMP waWA ṅe1SG
I am beating the boy

kuaboy nduPRF waWA tâbeat.PRF ṅe1SG
I have beaten the boy

So... from this, we might assume that the wa-particle serves to emphasise the object/patient, right?

Wrong. It's the inverse particle, and henceforth we'll be denoting it as INV in the gloss. Tǝɣrâ, like some Athabaskan languages and one or two Tibeto-Burman languages, has a species of direct-inverse argument structure. Wikipedia's as good as any to quote at this stage:

"The definition of a direct–inverse language [...] is widely understood to involve different grammar for transitive predications according to the relative positions of their "subject" and their "object" on a person hierarchy, which in turn is some combination of saliency and animacy specific to a given language. The direct construction is the unmarked one. The direct construction is used when the subject of the transitive clause outranks the object in the person hierarchy, and the inverse is used when the object outranks the subject."

Until we met kua tǝshâ lɨɣao ye wa, we had been dealing with direct constructions. Tǝɣrâ's direct-inverse system does not correspond to the canonical definition given above precisely: in the sentence kua tǝshâ lɨɣao ye wa, the nominal constituents were still marked for agent and patient as they are in a direct construction, rather it is the verbal marking which differed, triggered by the abnormal word order of A preceding E.

The saliency hierarchy in Tǝɣrâ is 1st/2nd person > 3rd person animate > 3rd person inanimate. If the two arguments of a verb both occupy the same "rank" on the hierarchy and we have unmarked word order, a direct construction is used. If the agent outranks the patient and we have unmarked word order, a direct construction is used. However, if the patient outranks the agent, or if there is a marked word order, an inverse construction is used.

For example, taking the sentence "the lord is beating me", we might expect the Tǝɣrâ

*lɨɣaolord yeERG tǝshâbeat.IMP ṅe1SG
*The lord is beating me.

based on the sentence with a nominal patient seen above:

lɨɣaolord yeERG tǝshâbeat.IMP kuaboy
The lord is beating the boy.

We would be wrong to expect this. As the 1st person outranks the 3rd person, we should rather expect an inverse construction:

lɨɣaolord yeERG tǝshâbeat.IMP waINV ṅe1SG
The lord is beating me.

Given that "animate" in Tǝɣrâ thought indicates "capable of language", animals are outranked by humans. So in a sentence where we have a non-human agent and a human patient, we need the inverse construction:

rɨsordog yeERG ñjuabite.IMP lɨɣaolord waINV
A dog is biting the lord.

Moving on, like many languages with absolutive alignment, Tǝɣrâ lacks a "passive voice", allowing instead free deletion of agents in transitive clauses. However, in such a construction, while the agent is no longer overtly instantiated, it is considered to still be underlyingly present. And an unstated agent requires the inverse construction:

lɨɣaolord ñjuabite.IMP waINV
The lord is being bitten.

That is, unless an instrument is stated. (Obviously.) Let us take the straightforward direct construction "the lord beats the boy with a stick".

lɨɣaolord yeERG tǝshâbeat.IMP lǝlaṃstick taINST kuaboy
The lord is beating the boy with a stick.

(Note the unmarked position of the instrument here: between E and A.)

To say "the boy is being beaten with a stick", we delete the agent as we did with the lord getting bitten, but instead of using the normal inverse particle, we meet a new particle: mɛṃ (which for the sake of argument we'll call the "instrument focus", or IFO):

lǝlaṃstick taINST tǝshâbeat.IMP kuaboy mɛṃIFO
The boy is being beaten with a stick.

This instrument focus particle also crops up when an instrumental phrase is left-dislocated for emphasis:

lǝlaṃstick taINST tǝshâbeat.IMP lɨɣaolord yeERG kuaboy mɛṃIFO
The lord is beating the boy with a stick (and not e.g. another boy).

The same particle also occurs with intransitive verbs of motion where the destination or origin is emphasised:

suafield seALL cugo.IMP lɨɣaolord mɛṃIFO
The lord is going to the field.

Now, what about transitive verbs lacking a stated patient? (I hear you cry.) Unexceptionally, Tǝɣrâ has an antipassive for this kind of thing. It is expressed by a particle from the same family as wa and mɛṃ. This is mo, which promotes the agent to a patient (i.e. it loses its ergative marker), and cannot co-occur with another particle from the same family.

lɨɣaolord tǝshâbeat.IMP moANT
The lord is beating somebody.

The complication here is that the antipassive particle is found with a number of verbs which (to English sensibilities) are semantically intransitive anyway. As an example:

lɨɣaolord hohsleep.IMP moANT
The lord goes to sleep. (On account of being exhausted from all the beating, presumably.)

This is because this same set of verbs are actually transitives in Tǝɣrâ, with causative meaning:

lɨɣaolord yeERG hohsleep.IMP kuaboy
The lord puts the boy to sleep.

Finally, because this isn't complex enough, there are actually two antipassive particles: the already-encountered mo, and yi. The distinction between the two is one of control or agency: mo indicates a deliberate act on the part of the agent (and has an inceptive/inchoative nuance), while yi indicates a lack of intention or volition:

lɨɣaolord hohsleep.IMP yiANT
The lord is sleeping.

Now, if you're keeping up, all this should be fairly straightforward. Happily, negation and relativisation add a further layer of complexity, and will be explored in a later post.