Direct-Inverse constructions in Towwu pũ saho

Towwu pũ saho has a fairly interesting syntactical system. It acts much like a direct inverse language, though the information is carried in a particle between the nouns, rather than on the verb.

To understand how this works, you first need to understand its word order. Tps is an SOV language, but there is little necessary relation between subject and agent. Rather, the subject position is ordinarily held by the topic (when the topic is the agent or patient), which in turn is generally the most definite (technically the difference is referring vs non-referring expressions but it's been a while since I've worked on this so I need to brush up on the difference again) or proximate argument. Then there's an animacy hierarchy which determines word order absent an unusual topic or differences in definiteness. The most animate argument comes first followed by less animate arguments. The hierarchy is as follows:

Animacy table from most to least animate
1st person	2nd person	3rd person	4th person	Human	Animal/ Moving Force	Inanimate (natural objects)	Inanimate (Artificial)	Abstract

After the word order is properly established, one of eight particles is chosen. This clarifies the semantic roles and the definiteness (since there are no articles) of the main arguments. A direct particle is used when the more animate (regardless of position in the sentence) argument is the agent and the inverse when the less animate argument is the agent. When the arguments have the same animacy, if the subject position is held by the agent, use the direct and use the inverse when the patient holds that position. The table below shows the role particles.

Grammatical relations in transitive sentences
“Voice”		Direct		Inverse
Agent		Referential	Non-referring	Referential	Non-referring
Patient	Referring	go	mã	i	lu
	Non-referring	e	bo	sa	nẽ

This is probably best shown with a series of examples. The following words are ebe "man", ho'o "hat", caupe "to put on, to wear", tẽmẽ "to see", ũcẽ "woman", uxxale "snake".

Ebe go ho'o caupe "The man puts on the hat". Here both arguments are definite, so the most animate goes first and a direct marker is used.

Ebe e ho'o caupe "The man puts on a hat". Still very straightforward

Ho'o mã ebe caupe "A man puts on the hat". Since the less animate argument is definite while the more animate argument is not, the less animate argument is moved to the beginning of the sentence. It still uses a direct marker though because the agent is the more animate argument. A more natural translation might be "The hat was put on by a man". If you want to make "a man" the topic (for some reason) you could say Rĩ ebe mã ho'o caupe or just Ebe mã ho'o caupe.

Ebe bo ho'o caupe "A man wears a hat". Not a very illuminating sentence, but it works. Since they have the same definiteness regular animacy rules apply.

Now for the inverses.

Ebe i uxxale tẽmẽ "The snake sees the man". Same definiteness, so the more animate argument comes first. But the agent is the less animate argument, so we use the inverse.

Ebe lu uxxale tẽmẽ "A snake sees the man". A very strange sentence that would be more likely translated "The man is seen by the snake". However, this does fall the normal rules for animacy

Uxxale sa ebe tẽmẽ "The snake sees a man". Note that while the agent is in the subject spot, you still use the inverse.

Ebe nẽ uxxale tẽmẽ "A snake sees a man". Pretty straightforward.

When the arguments are on the same level:

Ebe go ũcẽ tẽmẽ "The man sees the woman"

Ũcẽ go ebe tẽmẽ "The woman sees the man"

Ebe i ũcẽ tẽmẽ "The man is seen by the woman" or "The man, the woman sees him"

Ũcẽ i ebe tẽmẽ "The woman is seen by the man" or "The woman, the man sees her"

These are all kind of weird examples, many seeming quite unnatural. So now I'll give one example for each (not necessarily related to each other) with TAM markers and other particles to make the sentences work better.

Hã go ba ngĩ tẽmẽ "I just saw him"

Igea mã uxxale ku ngõnã "A snake ate the egg"

Ba e igea ijji ĩxũ "She might like eggs"

Uxxale bo igea ngĩ ngõnã "Snakes like eggs"

Hã i onã fu fũxã tẽmẽ? "Have you ever seen me before?"

Sei ebe lu uxxale ãxõũ ngõnã ella "(As you know, I wish) a snake would eat that man over there" This sentence has a lot going on. Sei is a distal, visible determiner. Ãxõũ marks the sentence as a desire of the speaker (even though the speaker is never mentioned in the sentence). Ella at the end of a sentence marks the entire sentence as something that should be obvious to the discourse participants.

Hau, uxxale sa be'oi uwẽ vasi ngõnã "Agreed, the snake could be a man-killer" lit. "Agreed, the snake could frequently eat people"

Ebe nẽ uxxale ijji ã ngõnã "A snake could be eating a man"

Antipassives, Ergative Verbs and Nominalizations in Kélojùù

I have not worked on this language for a long time. And then a little while ago someone was asking about someone else's conlang, specifically if the agent nominialization could fit on unaccusative verbs (something like that). And that got me thinking. Then a few days later I was looking up stuff on antipassives and learned that while they normally aren't in nom-acc languages, it does happen in some Nilo-Saharan langs. So I thought about how to shove it in mine.

Some base things to keep in mind. First of all, most transitive verbs in Klj are ambitransitive, but they are ergative verbs, even though in English they are often accusative verbs. Why? Such is life. Anyway, this means that the subject of an intransitive verb (well, detransitivized) is treated as the patient rather than the agent. So while in English we can do "He cooks food" -> "He cooks" the same process (simple deletion of the object) in Klj means "He is cooked/He cooks (like a cake does)". The next thing to consider is that Klj has two basic denominalization processes. One is the action noun (much like a gerund or infinitive in English) and the other is concrete noun. The concrete noun can work like an agentive/patientive/result, basically some sort of more concrete object or idea.  The interpretation of these is based on the role of the subject of the original verb, which is important because generally only intransitive verbs can take these in Klj. And so comes the need for the antipassive

The antipassive (-nú) demotes the object of a transitive verb, while keeping the old agent in subject position. Consider the verb wííza "to break (something)". You might have a sentence like mọ́llééṃò zawíízajù "I broke the pot". To say "The pot broke" you could say  mọ̀llééṃò ìwíízajù or use the passive/reflexive/general detransivizer -ḍà giving us mọ̀llééṃò ìwíízadàjù "The pot broke/was broken". "I broke (things)" is zawíízánújù and "someone who breaks things" is wíízánúsaw but "a broken thing" is just wíízasaw with no passive required.
Other than nominalizations, when is the intransitive important? For one, the intransitive form of ambitransitive verbs in the present tense often has a habitual meaning. So (to use a slightly silly example), zawííza means "I break (apart) a lot". The antipassive allows us to keep the subject as an agent in these habitual clauses, zawíízánú "I be breaking (things)". Habitual intransitives can take a genitive complement to reintroduce the former direct object as in mọ̀òlleek zawíízánú "I be breaking pots" (had the singular been used there, it would mean "I keep breaking this same pot"). 
So what's even the point of having a passive (as seen above) if all the verbs are ergative? Well, a large part of it is to clarify that that action was intentionally done/caused by the agent. In the ergative (technically unergative?) form, no intention is drawn to the fact that it was caused or done. This implies either an accident or something causeless. Pots can just randomly break, yannow? A passive with a reintroduced agent complement (with the dative postposition) means that agent intentionally did the verb. Even one without a reintroduced agent could be taken as a volative act, though because it could also be reflexive or something else, usually an indefinite pronoun is brought in regardless. The next question is, "if the act was volative and the agent known, why use the passive and not the active voice?" Further research is needed, the likely answer has something to do with bringing attention to the patient/result of the action. So the difference between mọ̀llééṃò ìwíízajù, mọ̀llééṃò ìwíízadàjù, and mọ̀llééṃò zlàkọ ìwíízadàjù is something like "The pot (has) broke(n)/Somebody (accidentally) broke the pot" "The pot was broken (by someone)" and "The pot was broken by me!" respectively/

The antipassive works in the opposite (ish) way. An antipassive with a reintroduced patient (also in the dative) implies that the agent did the verb on accident. So while mọ́llééṃò zawíízajù means "I have broken the pot", mọ́llééṃò kọ zawíízánújù means "I accidentally broke the pot".

Pragmatically, this means that people, when accused of doing something, often reply with an antipassive. Mọ́llééṃò ìwíízajù! "You broke the pot!"(Note that this is technically ambiguous and could mean "The pot broke!") nẹẹh, zawíízánújù "It was an accident!" (lit. "no, I broke (something)"). This is a way of admitting fault while trying to absolve yourself of full guilt.

Obviously, there's still a lot to work out here/discover. But it is a work in progress and more progress I've gotten on this language than I have all year. Plus, it was like actual work combing morphology, syntax, and pragmatics, which feels good.

TbKt conjunctions

So I was reading about syntax and figured I could do a mini post on TbKt conjunctions. We'll look at about 4 or 5 today (all particles...except they can inflect so whatever).

The first is ī. This is "and" and links together multiple nouns within a phrase to a single subject/head. The key thing to remember here is that it works on a phrasal level instead of a clausal one and keeps both the noun and the verb the same. For example, the sentence Yān oxdīc rōxub ī zhōluq "I hit the dog and the cat". This also implies some sort of unity of coherence in the verb. Ī can link nouns separated by relative clauses. For example, Yān oxdīc rōxub vit anmōs-anmōs ī zhōluq "I hit the dog, which ran away, and the cat".

The second is jasā "and/or". It isn't commonly used and more or less replaces ī. Yān oxdīc rōxub jasā zhōluq "I hit the dog and/or the cat"

Then there is ocāk "and". Unlike ī, ocāk operates on a clausal level. More specifically, it links two (otherwise unrelated) clauses together, and draws attention to the fact that the subject changed. Indeed, in common usage it almost acts as a DS switch reference marker. This is very useful when the object cannot be promoted to subject position (perhaps because of indefiniteness) yet there is some sort of continuity between the clauses. Use of ocāk followed by a verb but no arguments indicates that the subject has changed, probably to the former object (though arguments can of course be added, especially for clarity of emphasis). For example Yān oxdīc rōxub ocāk anmōs-anmōs  "I hit a dog and the dog ran (away)". If the dog had already been previously referenced, then Rōxub oxaxdic yān ūmpa anmōs-anmōs "I hit the dog and it ran away" would be more appropriate.

Which brings us to ūmpa "and". This also works on a clausal level, but specifically marks the subject as staying the same between the clauses, like a SS SR marker. Sometimes it translates to something very similar to ī (despite working on different levels) but implies disunity of the verb, or some sort of sequence. This is especially true when the new "clause" is only an arugment, meaning that the subject and verb have been carried over. For example Yān oxdīc rōxub ūmpa zhōluq is literally "I hit (the) dog and (the) cat" but would be understood as "I hit the dog and then the cat" or "I hit the dog and (I hit) the cat (but with something to make the acts of hitting be considered separate)".

Agis means "but" and works as a replacement for ocāk. It simply marks surprise or unexpectedness of the next clause and can be emphasized with the suffix -(a)x. This suffix can also be added to the other conjunctions, with the same sort of meaning. Some examples. Rōxub oxaxdīc ūmpax mōnak anmōs-anmōs "I hit the dog but it didn't run away". Yān oxdīc rōxub īx zhōluq "I hit a dog AND a cat".

Anyway, that's an intro to conjunctions in TbKt. To think that I was accidentally doing pivot and pseudo-switch reference long before I ever knew them.

Symmetrical Voice in Toúījāb Kīkxot

So I read an article about the Totoli language of Sulawesi today (For reference it is "Symmetrical Voice and Applicative Alternations: Evidence from Totoli" by Nikolaus P. Himmelmann and Sonja Riesberg). Interesting stuff, mostly focusing on how Totoli has significant features of both symmetrical voice languages (its own typology) and Philippine-type languages (a relative of the symmetrical voice languages of course). Some things I liked included the authors getting slightly annoyed with Philippinists (for assuming everything is like Philippine languages, I guess) and the general overall topic. I'm always looking for info about obscure languages of Indonesia. My general feeling is that it is really hard to find such info, especially online. A lot of the examples used seemed pretty natural to me as an Indonesian speaker, so I guess I do understand this applicative suffixes after all :p. I was interested that the benefactive/instrumental suffix (-kan in indonesian, -an in Totoli, for the actor voice) can have a iterative function...which is a function of the locative suffix (-i in both languages) in Indonesian. Or maybe I don't, as I didn't quite get what exactly their conclusion that there was a locative voice in opposition to an undergoer voice with a goal applicative suffix meant, though the proposal seemed reasonable enough.

Anyway, this reminded me a lot of Toúījāb Kīkxot and how it came to have the typology it has. When I was learning Indonesian I thought the voicing system of Indonesian (which I later learned is called symmetrical voice) was cool and a feature I didn't see often, if ever in conglangs (sure, austronesian alignment appears lots, but Indonesian is pretty much ignored by everyone so yeah). I also decided I wanted to do something with triconsonantal roots (but had no knowledge of Arabic at the time and no access into any materials, so we get what wonderful mess we have), but that's not super relevant right now.

Now, with my Indonesian grammar book (The Sneddon one, I highly recommend it), I saw that suffixes like -i and -kan could do cool things with objects and marking arguments. I didn't quite get it at the time (let alone know that these are called applicatives) but I decided that Toúījāb Kīkxot should have them too. Originally there were 2-3 (locative, "benefactive" and an optional patient), later I added a third instrumental/causative. They were pretty much as classic applicative suffixes, as far as I can tell, marking the role of the direct object (often raising the transitivity of the verb, requiring a change of verb form) or the subject in the undergoer voice. I'll cover what they actually do in another post, since the benefactive form is especially confusing, but that's how they came about. It's one part of the language I am really proud of, as it gives a very different flavor than English and can do some pretty cool things. Plus it really helps with focus and showing what is most important in a sentence.