Archive for September, 2021

Resolving the probability waveform of dependent clauses

Monday, September 27th, 2021

This is not the post I thought it was going to be. Early last week I had a moment of epiphany in the shower in which I thought I'd solved everything in just the kind of elegant, first-principles way that I had previously announced was vain and impossible. In deference to my ecstatic past self I'll briefly explain my discovery:

First, it occurred to me that finiteness itself is a category which I should not have assumed to be meaningful in Koa, but had sort of calquingly smuggled in accidentally from the language families of my closest acquaintance. It then hung around trying to circularly justify its existence. If eliminated, there could instead be a distinction of dependent versus nondependent clauses, marked by the particles u and i respectively. This would then yield the following clausal phrase structure, at last common to all clauses:


...with of course the refinements that if you have a nominal subject then you would omit the pronoun and vice versa, and also the dependent marker is optional and typically omitted with pronominal subjects. This gave example sentences like

ni kulu [le Óleka u loha le Iuli]
"I heard that Olga loves Julie"

le Iuli sa ka mina [le Óleka u loha]*
"Julie is the woman Olga loves"

le Óleka sa [ka mina u loha le Iuli]
"Olga is the woman who loves Julie"

It also gave me the rigorously motivated ability to optionally omit the equivalent of a complementizer or relative pronoun for dependent clauses with pronominal subjects:

ka mina (u) ni loha*
"the woman I love"

ai se ilo (u) ni loha se?
"do you know that I love you?"

This was so beautiful that I was willing to overlook the potential unintuitiveness, but alas, it was not to be. For one thing, I started to wonder why speakers really truly had to keep track of whether a clauses was dependent or not to know how to form it -- that is, why they couldn't just be modular -- in this theoretical IAL. That was a design philosophy that hadn't been articulated before but maybe should have been? But also, there were errors. In the asterisked sentences above, heads had been fronted out of their clauses, breaking the lovely pattern I had been so excited about. If we were really going to grit our teeth and do this thing, they ought to be

le Iuli sa [le Óleka u loha ka mina]
"Julie is the woman Olga loves"

(u) ni loha ka mina
"the woman I love"

That just clearly will not work. I mean, thinking about it now, I guess one could say that the shared argument could remain in situ in the matrix clause, but then how is that better than traditional relatives? What really killed it, though, was that serial verbs make the clarity of the structures very murky very fast:

ta si sano [le Malia u halu tai me se i(?) pea]
"she said Maria still wanted to be with you"

The serial verbs are halu/pea but they have different dependence marking! But if the above had u pea, wouldn't that be taking that clause down an additional level? But if it's i pea, how do you know which verb it's making a complex with? ta si sano i pea or halu i pea? I don't know the answer, and in a way that makes me feel like I shouldn't be asking the question in the first place. This is just a mess

To make matters worse, I would also lose that te tai ko... structure that for some reason had become a hill I was willing to die on. For a minute I thought that maybe ko could be an optional specifier for clauses used in nominal positions, but the battle was already lost. It was just needlessly complicated.

So what would be so bad about holding onto one potentially important realization -- that finiteness can probably be eliminated as a meaningful category in Koa, inasmuch as it's basically synonymous with verbal usage of a predicate -- while optimizing for simplicity and modularity?

Here, then, is my answer to Koa dependent clauses, hopefully the really truly final answer this time:

  • Clauses used as predicates are syntactically and morphologically identical to independent clauses
  • Clauses used as adjectives are preceded by ve when the head is not the subject of the dependent clause, optionally otherwise
  • Clauses used as nouns are preceded by either ko or ve
...and the example sentences now look like this:

ni kulu ko/ve le Óleka i loha le Iuli
"I heard that Olga loves Julie"

le Iuli sa ka mina ve le Óleka i loha
"Julie is the woman Olga loves"

le Óleka sa ka mina (ve) loha le Iuli
"Olga is the woman who loves Julie"

And borrowing one from past years,

pai ve ka mama i na ma mai koa
"a mommy-not-feeling-well day"

ti pai i ve ka mama i na ma mai koa
"this day is mommy-not-feeling-well-ish"

Maybe it's not quite as elegant as the whole business with u, but (1) as Marisa said, in mathematics elegance (i.e. concision) is the opposite of clarity, (2) this is much, much, much, much easier to learn and use than any previously conceived system, (3) this lets me use u as a plural definite article if I still want to, and (4) it saves te tai ko, surely a flag of victory.

One further related note, which I found forgotten in my dictionary but nicely articulates the difference between sivu vihe "green leaves" and sivu ve vihe "leaves that are green": "ve shifts pragmatic relevance ('emphasis') to the modifier rather than the head." I'm grateful to past Julie for leaving me those breadcrumbs back to a clear understanding of this one...

Detail #417: Some ideas about relativization

Sunday, September 26th, 2021

I assume everyone knows of the relativization hierarchy by now. 

1. A Second Relativization Hierarchy

Let's instead imagine a species with a language faculty that creates two relativization hierarchies, but also permits for a systematic exception.

The first hierarchy is familiar - the relativization hierarchy. I will not even modify it for this idea.

The second hierarchy is an "external" relativization hierarchy. It, and the first one, have implications between them. I will have the same order for that hierarchy:

Subject > Direct Object > Indirect Object > Oblique > Genitive > Object of comparative

What the second hierarchy tells us is which roles an NP of an external clause can be relativized as.

Thus, if a genitive in the main clause can take a subclause in which it corresponds to the oblique, then so can also the oblique, the indirect object, the direct object and the subject. 

One could also imagine extreme things like "only the subject in the main clause can take relative clauses" or "only the subject in the main clause can correlate with anything but the subject in the subclause".

However, I imagine it could be likely for any NP to also permit an "echo" of its own role in the subclause, which would create a systematic exception. This type of exception I'd like to term a "linear" exception.

2. Subdivisions of the relativization hierarchy

One could imagine, for instance, that inanimate nouns have a stricter hierarchy than animate nouns.

3. Questions about the relativization hierarchy

3.1 Do we know where secundatives are with regards to the relativization hierarchy?

3.2 Is there any research or even any hypotheses around as to whether there's any roles that go to the right of objects of comparisons, or between the known elements?

Detail #416: Disjunctive Reciprocals

Thursday, September 23rd, 2021

It is conceivable that a language distinguishes the following meaning-structures in the morphology of its reciprocals:

A acts on B and B acts on A > A and B act on each other

A acts on B or B acts on A > A or B act on each other_disjunctive. 

One could even imagine, then, that partial negations also could fit into this:

A but not B act on each other_disjunctive. 

This could be done by differential case or by some marker that is entirely separate from case. 

Detail #415: Reciprocals and Collective Nouns

Thursday, September 23rd, 2021

Normally, reciprocals take a plural noun (or several coordinated noun phrases) and make the entities that make up an agent (or comparable role) act upon each other.

A rather interesting situation is the use of collective nouns and reciprocals:

?the team watched each other in astonishment

*laget såg förvånat på var-andra (literally: team-the saw surprisedly on each-other)

Yeah, no, the ? mark there is probably wrong, for a huge majority it probably genuinely is *. I am convinced that some languages permit, without hesitation, constructions analogous to the one above.

Historically, "one another" and the Russian construction "drug druga" (drug = nom, druga = acc) probably consist of two bits - one ("one", "drug") encoding the subject, and the other ("another", "druga") the object. From a purely abstract look at it, it feels like this construction might be marginally more tolerable in English? In Russian, I have it on a fairly reliable source that this is entirely acceptable.

*the team watched one another in astonishment

para obnimala drug druga 

In Swedish, some passive verbs primarily have a reciprocal meaning:

vi kysstes ('we were kissing')

vi slogs ('we were fighting (each other, but can also mean 'we fought random people')

Now, most Swedes accept

paret kysstes ('the couple were kissing')

despite this being a reciprocal with singular subject. 

However, I can't just let it stand at "this is okay in a language so there you go", can I?

We could imagine, for instance, that a language does get any number of congruence or discongruence-phenomena with this.

1) Singular forms of 'each other'.

In Swedish at least, 'varandra' is morphologically plural. One could imagine a system where a singular form appears with a singular subject. In Russian, 'drug druga' is morphologically singular, but one could imagine a partitive genitive instead on the subject noun - getting us something like 'druga druga'.

Here, an unrelated idea appears: in the dual, a language could very well have a reciprocal singular pronoun: 'both saw the other', but 'the thief and the policemen saw each other' .

2) Plural verb forms and adjectives with singular subjects.

Some varieties of English permit plural verb forms with collective singulars, but one could imagine a situation where plural verb forms only are used with singulars in situations where the singular is blocked by the presence of a reciprocal pronoun.

3) Congruentially Forced Plurals

The subject could exhibit other behaviors that are typical of plurals, including morphological or congruence-related behaviors. Maybe the reciprocality forces an explicit plural marking, so that "the groups saw each other" can signify both 'the members of the group saw each other' or 'the different groups' members saw the other groups'. Maybe adjectives need plural congruence. Maybe they are forced to take a plural article.



Sending a Message

Friday, September 3rd, 2021

I have an ever-growing backlog of concepts for which I want to create new Kílta words. It gives me all the stress you expect from a todo list. One benefit of the delay, however, is that I regularly think up better derivations, or better nuance, if I have time to let a concept percolate a while before I commit to it.

One concept I've been thinking about is a way to indicate if a state of affairs communicates some other message. For example, if someone stops answering your phone calls, that says something. Often we can rely on Grice and experience to help us figure out when other messages are being communicated by someone's actions, but I wanted a way to be a bit more explicit about it. This sat on my todo list quite a while, and then just yesterday a good way to handle this presented itself: an auxiliary with a converb.

The verb ráno means signal, make a sign, as well as point out.

Eman në tátiën mai ráno.
child TOP dog LAT point-out.PFV
The child pointed at the dog

But yesterday it occurred to me that my send a message sense matches with this nicely. Now, a general converb followed by ráno marks that the state of affairs also communicates some other message.

Ha kë mës mítët ráno.
1SG DAT NEG speak.CVB.PFV signal.PFV
She didn't talk to me (which makes some other point, too)

Often an overt translation of this into English is going to be a bit clunky, but I've got growing pile of those in Kílta, too.

In any case, rather than creating a new clause-final particle or entirely new lexeme, I've just added to Kílta's substantial battery of auxiliary verbs.

Complete Grammar of the Yajéé Language

Wednesday, September 1st, 2021

P. A. Lewis is a professional classical oboist and produces conlangs as a hobby. He has been interested in conlanging since 2018, and has been a member of the conlang community since 2020. His primary interest in language is in historical linguistics, and thus his conlangs are all spoken in a single conworld, Omnia (website coming soon). Some conlangs he has created include (in order of how proud he is of them): Yajéé, Andva, Radoza, and Chiset.


This is a complete grammar of the Yajéé language, featuring an extensive overview of its phonology and morphosyntax in its current state. The grammar includes a robust discussion of the pitch accent system employed by the language. Other notable features include: a telicity-based derivational system which impacts the semantics of the aspect under suffixes, umlaut and other phonological changes which result in multiple stems for nearly every noun and verb, and rampant pronoun dropping despite having no verb agreement.

Version History

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.