What are some strategies for making boundaries between top-level clauses obvious?

I think the most straightforward strategy is some kind of collection of sentence-final particles, but I'm curious about others.

I'm particularly interested in features that would enable discontinuous clauses.

I was reading a paper recently about prolific agreement in some Nakh-Daghestanian languages. The noun class that the absolutive argument belongs to can be repeated many times throughout the sentence:

Aˤli-l      qːatːa          b-ullaj      b-ur
Ali(I)-ERG  house(III).ABS  III-do.PROG  III-AUX

Here's an excerpt from Foley's paper Agreement in the Languages of the Caucasus showing a curious fact about this agreement. There are many kinds of words that can exhibit agreement with the absolutive argument, but agreement isn't uniform within a word class.

So far we’ve seen that NEC languages allow verbs of all stripes to agree — lexical and auxiliary, finite and nonfinite. But agreement is not limited to verbs. A remarkable property of NEC agreement is its ‘promiscuity’: a wide range of lexical categories can participate in gender agreement, all controlled by an absolutive clausemate. ... Yet while agreement in NEC may be promiscuous, it is also spotty. For any given lexical category that can potentially agree, typically only a minority of lexical items in that category actually do. For example, just 32% of Archi verb stems participate in agreement (Chumakina & Bond 2016:111) ...

A feature like this could be useful for making clause boundaries obvious if the repeated feature changes frequently from clause to clause. I think the absolutive argument is probably pretty variable, but I am not certain.

For instance, suppose we have two sentences. Following the notation of the paper, I'm using ROOT(n) to mean that ROOT has the nth noun class. (I) and (II) don't have any independent expression.

1) 1sg-ERG cat(I)-ABS see-PRES-I
   I see the cat.
2) 1sg-ERG mouse(II)-ABS NEG-see-PRES-II
   I see the mouse

Combining the sentences in the following way is possible, since it's clear which verb goes with which argument

3) 1sg-ERG cat(I)-ABS mouse(II)-ABS see-PRES-I NEG-see-PRES-II
   I see the cat, but don't see the mouse.

What are some ideas for features that could be used to support disentangling discontinuous top-level clauses?

  • 1
    Should it be ABS(I) and ABS(II)? The way you have it now makes it look like there are different cat and mouse roots but one absolutive case marker.
    – curiousdannii
    May 6, 2020 at 5:20
  • There is a single absolutive case marker in the example. In Foley's paper, ROOT(I) through ROOT(IV) are used to mark the noun class of a given noun. Example (3) is supposed to contain two interleaved clauses. As far as I know, the Nakh-Daghestanian languages don't actually permit this; it's a hypothetical feature inspired by the prolific agreement that they do have. May 6, 2020 at 5:29
  • 1
    Ah, I thought you were proposing there would be two sets of ABS markers to go with the two clauses and two verbs. So is ROOT(I) mean there's an extra marker? But it's not just considered a morpheme?
    – curiousdannii
    May 6, 2020 at 5:42
  • No, I wasn't proposing two parallel sets of case markers. Although that is a good idea. ROOT(I) is meant just to be an indication that ROOT is in class I. May 6, 2020 at 20:32
  • 1
    Is qːatːa really correct or do you mean qaːtaː? I'm surprised to see a word start with a geminate plosive. Sep 7, 2021 at 20:26

1 Answer 1


What are some ideas for features that could be used to support disentangling discontinuous top-level clauses?

Interesting question! For a start, some languages have ‘verbal classifiers’ (Aikhenvald 2000). These are a set of incorporated nouns or dedicated affixes which agree with a verbal argument, most commonly the S/O argument. Such a system will certainly be mostly or entirely orthogonal to the gender system, so is useful for further disambiguation. Even 5 or so verbal classifiers — a very small number — is enough to give a reasonable number of class/classifier combinations: (I denote the noun classes in uppercase and verbal classifiers in lowercase)

1sg-ERG cat(I,i)-ABS mouse(II,i)-ABS i-see-PRS-I NEG-i-see-PRS-II
I see the cat, but don't see the mouse

By the way, I see no reason why the verb should agree only with the absolutive argument in noun class (cf Bantu), so let’s add some polypersonal agreement for further disambiguation:

1sg-ERG cat(I,i)-ABS mouse(II,i)-ABS i-see-PRS-I-1sg NEG-i-see-PRS-II-1sg
I see the cat, but don't see the mouse.

Let’s go further. In some languages, alignment is lexically determined by the verb used. Though this usually occurs with agreement affixes, one can imagine a system where there are several possibilities for case-marking of intransitive and transitive verbs, with the verb retaining its transitivity. This adds extra information when two verbs are juxtaposed with their arguments interleaved. In this case, it seems reasonable for a verb like ‘see’ to take absolutive–dative marking (or absolutive–accusative, or nominative–accusative, if you like):

1sg-ABS cat(I,i)-DAT mouse(II,i)-DAT i-see-PRS-I-1sg NEG-i-see-PRS-II-1sg
I see the cat, but don't see the mouse.

Let’s go further. It is common for converbal constructions to specify whether the verb is same-subject or different-subject with respect to the main verb, or following clause in the case of clause chaining. Of course, the converbal affix only attaches to the verb itself. However, the Australian language Kayardild has a rather odd construction in which an oblique case is suffixed to nearly every single word of a different-subject complement clause (Evans 1995). This gives me enough precedent to postulate a system wherein a same-subject or different-subject affix is added to every word of the relevant clause:

1sg-ABS-SS cat(I,i)-DAT-SS mouse(II,i)-DAT i-see-PRS-I-SS NEG-i-see-PRS-II-1sg
I see the cat, but don't see the mouse.

(Incidentally, converbs already allow a certain amount of discontinuity in the main clause: e.g. the man, being a traitor, was tried by the king.)

Let’s go further. Negative concord allows the negative to be marked several different times within the clause (e.g. dialectal I ain’t got no money). It is straightforward to state that e.g. the negative prefix must be applied to both the absolutive argument and the verb:

1sg-ABS-SS cat(I,i)-DAT-SS NEG-mouse(II,i)-DAT i-see-PRS-I-SS NEG-i-see-PRS-II-1sg
I see the cat, but don't see the mouse.

All these together then give enough leeway to do stuff like:

mouse(II,i)-ABS-SS NEG-mouse(II,i)-ABS-DS i-chase-PST-II-I-SS NEG-i-catch-PST-II-I-DS cat(I,i)-ERG-SS, 1sg-ABS-SS rat(II,ii)-DAT-SS cat(I,i)-ABS-DS ii-see-PST-II-SS i-tell-PST-I-1sg-DS, ii-catch-PST-II-I-SS i-get.bored-PST-I rat(II,ii)-ABS-SS cat(I,i)-ERG-SS.

The cat chased a mouse, but didn’t catch it; I saw a rat, and told the cat. The cat caught the rat but got bored.

(Warning: I’m a bit unsure as to whether some of the nouns should take SS or DS marking, but it should be fine as is.)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.