Clause Reform

The Clause Reform is a proposal which aims to make clause boundaries and the attachment of subordinate clauses more intuitive. It supercedes the Subclause Reform.

Motivation

In official Toaq, verbs are allowed to be underfilled. This raises two big questions:

• How do we know where one clause ends and the next clause begins?
• What are the semantics of an underfilled verb?

Toaq's official answer to the first question is that clauses are greedy; they eat up as many arguments and adjuncts as they can get. However, experience with Toaq Delta leads me to believe that greediness impedes the language's usability. Because subordinate clauses look just like non-subordinate clauses, the grammar presents learners with a beautifully simple lie: that you can underfill any clause and include trailing adjuncts in any clause. Being perceptive loglangers, most will come to realize that the "real" grammar, the one that enforces greediness, looks more complicated, but this gap between the real grammar and the intuitive grammar that works 98% of the time is fairly large. Greediness becomes something you watch out for, not quite something you internalize.

The theory behind this proposal (and likewise the Subclause Reform) is that giving up a little bit of convenience in order to simplify the grammar can be a very worthwhile tradeoff. This was what happened with auto-hóa, for instance: for all the tinkering that was done, it turned out that a little bit of verbosity was what gave us the most intuitive and usable grammar.

As for the semantics of underfilled verbs, there are some sensible answers out there, such as filling the verb's implicit arguments via existential closure or imagining them to be an implicit 'ké', but I'm a bit pessimistic that speakers won't fall into the trap of using an ambiguous mix of the two. The Toaq Dzu Way seems to be to make your implicit variables explicit, in which case we probably shouldn't agonize over this question.

Clearer clause boundaries

Verbs may not be underfilled; they always require as many arguments as their definition calls for.

Shao tú poq, lä sho zaomıa já.
Everybody wants to become famous.

*Shao tú poq.
Everybody wants. (Everybody wants what?)

Consequently, nouns and adjectives must be intransitive, because Toaq's syntax gives them a single covert argument.

*Ké jea
The buyer (buyer of what?)

Ké jea lô chao
The buyer of the vehicle

*Tú poq buı
Everybody outside (outside what?)

Tú poq nı̣buı
Everybody outside this place

Nouns and adjectives also must not end with a verb that forms serials.

*Ké kuqnu
The alleged (not: "the allegation")

Ké kuqnu juna
The alleged fact

When determiners bind variables, the variable name is the full serial verb of the determiner's complement, excluding any adjectives. So sá kuqnu toaqpoq lıq would bind the variable kúqnu toaqpoq.

The adjective head kı- is covert most of the time. In official Toaq, the rule is that kı- must be overt whenever the preceding word could form a serial verb. In this proposal, the rule is instead that kı- must be overt in the verbal complex; everywhere else, it gets to be covert.

Kea kı̣pıao ní gıaqchuo.
This instrument is an elaborate machine.

Gıaqchuo ní kea pıao.
This elaborate machine is an instrument.

A big advantage of these changes is that sentence fences (well specifically, the words ꝡa, da, and móq) may now always be covert.

Pu tam gaı jí sá arane. Sao hó ꝡeı!
I saw a spider. It was huge!

Jua hú. Duashao jí, ꝡä luı faq hí raı.
That's strange. I wonder what happened.

Hoı déo, soa súna máma ba.
Kids, come help your mom.

Because ꝡa is no longer such a load-bearing word, this also a good opportunity to relax the phonotactic restrictions surrounding /w~j/, allowing this phoneme to appear even in word-medial positions.

Dynamic antecedents

Officially, relative clauses are generated in-place inside determiner phrases, adjoining to an 𝘯P. Under this proposal, relative clauses become extraposed. They can either appear at the very end of a clause, or just before a cleft such as nä or bï.

Gaı tú poq súq, ꝡë chụmlaoshue póq shâq ké jıo.
Everybody who was waiting in front of the building saw you.

Tú poq, ꝡë chụmlaoshue hóa shâq ké jıo, nä gaı hóa súq, jü chıa hạchıaısı hóa!
Everybody waiting in front of the building saw you, who need to be more careful!

The antecedent of a relative clause is determined by the variables and anaphoric pronouns that it references. The nearest noun that binds one of these variables or pronouns is understood as the antecedent.

Fana ní coa ké aomo, ꝡë jıu tî máq jí.
This bridge leads to the island on which I was born.

Or, if the relative clause references the resumptive pronoun hóa, the antecedent is simply the nearest noun.

Chıaq jí sáq tue, ꝡë ea kuıca hóa ní saꝡaq.
I am aware of three circumstances that could explain this unfortunate situation.

The effect of this change is that center embedding becomes illegal. This is attractive because it lets us fully preserve the Embedding Property and also avoid having center embedding be a source of confusion for listeners in the first place. The story becomes: if you have a heavy constituent, either front it or extrapose it.

*Gaı tú poq, ꝡë chụmlaoshue hóa shâq ké jıo, súq.

Unified grammar for subclauses

Content clauses and event accessor clauses are generated in exactly the same positions as relative clauses; they can either be fronted with nä or bï, or extraposed to the end of the clause. Like relative clauses, they have antecedents. A content clause's antecedent is the nearest content-bearing noun (such as júna), and an event accessor clause's antecedent is the nearest event noun (such as fáq).

He teaca júna jí, ꝡä bu sula kı̣udua báq tue po jíadaq.
The fact scares me that matters of the future can never be known.

Táosıo, lä seraq nhâna kú râo núaq já, bï chı duı hıam jí hóq.
I think the plan to attack them at night is too dangerous.

Jôı tú faq mabala, ë dea nháo láqbıo, nä jeaq rueqmoe jí.
With each horrible striking of the bell, I grew more tense.

Finally, another possible antecedent for any relative clause, content clause, or event accessor clause is the word có. This word is essentially a lightweight placeholder for content to come at the end of the post-field, much like the 'it' in English "It delights me that they tried". The semantics of có are probably like sá raı, but with low scope.

Jaıca có jí, ꝡä leo nháo hú.
It delights me that they tried it.

Kaq jí có, ë marao súq.
I saw you dance.

Cho jí có, ꝡë bo jí hóa.
I like that which I have. (?)

When có is used as the object of a preposition or the final verbal argument of a clause, it may be covert. This is how the subordinators ꝡä, lä, etc. come to still appear as if they have no antecedent most of the time.

Bu dua áma, ꝡä gáma nä hıa gáma…
We don't know what the camel is up to…

Tua jí, ꝡä naı ceo sho bua súq ní, ꝡë daqbuaı bue kú hóa jí.
I cause you to now begin to inhabit this place, which has long been my home.

Tıshue jí ní chôe, ë rıu'aona súq jí nha.
I'll stay here until you come back to me.

An alternative

If center embedding turns out to be impractical to give up, here's the other possibility I was imagining for this proposal.

Center-embedded clauses cannot have trailing adjuncts, and so we give up the Embedding Property. They are generated on the right periphery of clausal arguments. (I.e. in sá poq po kúe, ꝡë hao hóa, the relative clause is neither a constituent of the poq DP nor the kúe DP, but rather something that sits off to the side, as if it has undergone a minimal amount of extraposition.) The word có must be overt whenever it follows a center-embedded clause, almost as if it's a terminator. This is necessary in order to disambiguate between Aojaı póq, ꝡë cho hó sío, có, ꝡä hao júna, ꝡä luq tú and Aojaı póq, ꝡë cho hó sío, ꝡä hao júna, có, ꝡä luq tú.