Evidentiality with multiple verbs

[leanancailin]

Does anyone know of a good source, or feel like explaining, how evidentiality functions in more complicated structures than the typical single verb examples? Things like:

* "I know that you have it" / "I made her do it" / "I want him to come with me" - Which verb takes evidentialtiy marking? Do both? Do circumstances in each level impact the evidentiality, or just the highest / lowest level?
* Are there languages where verbs in conjunction are only marked on the first and remain unmarked from there?
* Does evidentiality ever become embedded in a structure the way that tense is in some languages? If I've marked evidentiality on a verb beginning a sentence, are later verbs ever interpreted as a sub-component of that evidentiality, or do they always remain absolute?
* How are mistakes or changes in knowledge marked? If I wanted to say something like "I thought I saw her do it, but it turns out I was wrong", how would that usually work out?

[chris_notts]

You should read a book called "Evidentiality" by Aikhenvald which discusses how it works in different languages. I have a copy somewhere, but my books are a mess since my house move so I can't find it. I did find my grammar of Tariana though, which is a language with evidentiality.

Does anyone know of a good source, or feel like explaining, how evidentiality functions in more complicated structures than the typical single verb examples? Things like:

* "I know that you have it" / "I made her do it" / "I want him to come with me" - Which verb takes evidentialtiy marking? Do both? Do circumstances in each level impact the evidentiality, or just the highest / lowest level?

Comments about these:

"I know that you have it" - Tariana "regularly" uses nominalisations (-ri) with the verb -yeka "know", which I THINK don't show any evidentiality distinctions. The grammar suggests that there are also other options, but I haven't found where they're described yet.

"I made her do it" - Tariana has morphological and serialised causative constructions (depending on transitivity). Either way, there is only one marking of evidentiality in the single causative clause.

"I want him to come with me" - Serial verb construction or nominalisation (depending on S/A coreference). Either way, no distinct evidentiality.

* Are there languages where verbs in conjunction are only marked on the first and remain unmarked from there?

I think it is quite common for some subordinate and complement clauses to be in forms which show reduced or no evidentiality distinctions. However, this probably doesn't normally apply to all subordination and complementation strategies - Tariana, for example, has evidentiality in direct speech (indirect speech is not typically used), and also in explicitly subordinate clauses such as purpose clauses.

* Does evidentiality ever become embedded in a structure the way that tense is in some languages? If I've marked evidentiality on a verb beginning a sentence, are later verbs ever interpreted as a sub-component of that evidentiality, or do they always remain absolute?

Not sure I fully understand the question, but:

1. If evidentiality is marked by some form of clitic, it may be possible to skip it on some verbs. E.g. Basque has a hearsay proclitic / particle "omen" which I don't believe needs to be repeated in front of every verb which is within the scope of the "hearsay".

2. If evidentiality is marked by an affix, it is probably obligatory unless the language has, e.g. clause chaining which neutralises the distinction for non-final verbs in the chain.

* How are mistakes or changes in knowledge marked? If I wanted to say something like "I thought I saw her do it, but it turns out I was wrong", how would that usually work out?

In Tariana, "think" can take a direct speech complement. Therefore, the tense and evidentiality is what would have been thought, e.g.

I thought "it's a jaguar!"

NOT:

I thought it was a jaguar

In general, Tariana appears to avoid finite indirect speech in favour of either direct speech complements or various kinds of nominalised or subordinate clause complementation strategies, many of which neutralise evidentiality distinctions.

[Vardelm]

I think it is quite common for some subordinate and complement clauses to be in forms which show reduced or no evidentiality distinctions. However, this probably doesn't normally apply to all subordination and complementation strategies - Tariana, for example, has evidentiality in direct speech (indirect speech is not typically used), and also in explicitly subordinate clauses such as purpose clauses.

This is really good to know! I'm currently planning on having a "gnomic" evidential that is used in dependent and complement clauses, where evidentiality isn't required because the speaker isn't directly making a proposition about that notion. The truthiness ( ) of the dependent/complement clause is just assumed or considered irrelevant. It also gets used in reported speech where the the speaker is unsure of exactly what was said in the speech being reported, or is paraphrasing.

[chris_notts]

I think it is quite common for some subordinate and complement clauses to be in forms which show reduced or no evidentiality distinctions. However, this probably doesn't normally apply to all subordination and complementation strategies - Tariana, for example, has evidentiality in direct speech (indirect speech is not typically used), and also in explicitly subordinate clauses such as purpose clauses.

This is really good to know! I'm currently planning on having a "gnomic" evidential that is used in dependent and complement clauses, where evidentiality isn't required because the speaker isn't directly making a proposition about that notion. The truthiness ( ) of the dependent/complement clause is just assumed or considered irrelevant. It also gets used in reported speech where the the speaker is unsure of exactly what was said in the speech being reported, or is paraphrasing.

I think it's generally true that the more integrated the two verbs are, the less likely they are to be independently and fully inflected. This is discussed in "Subordination" by Cristafaro, who looks at whether balancing or deranking strategies are used for different main verbs in various languages. Verbs would be more integrated if:

1. they share an S/A argument
2. The main verb partially or completely determines the tense, aspect or mood of the dependent verb because of its semantics

A difference in the degree of integration often motivates splits in coding. For example, the complement of want is almost always unrealised (irrealis) and in the relative future, so there is a high degree of TAM integration. In most languages, the complement of want is only optionally marked for TAM or never marked for TAM. Further, many languages show a split depending on whether the S/A of want is the same as the S/A of the complement, with the least independent form of the complement used when the S/A are the same. E.g:

SPANISH:

S/A same: infinitive complement
S/A different: subjunctive complement inflected for subject/tense
Note that past wishes also occur in the subjunctive, but the particle "ojalá" is typically used instead of a main verb.

Quiero ir = I want to go (infinitive)
Quiero que vayas = I want you to go (subjunctive)
Ojalá que hubiera ido = I wish I'd gone (subjunctive)

ENGLISH:

S/A same: deletion of second subject
S/A different: no deletion
Past wishes involve a finite complement

I want to go
I want you to go
I wish I had gone

TARIANA:

S/A same: serial verb construction (no independent marking of A/S, TAM, evidentiality, ...)
S/A different: nominalised complement (and maybe other complement types???)
I have no clue how unrealised wishes are expressed in Tariana.

In my conlang Mɛdíṭṣai, which also has evidentiality, whether it occurs in subordinate/complement clauses or not depends how integrated the two clauses are. The following table is from my grammar, and is based on the work of Cristafaro:

Code: Select all

Main Verb Type        Examples                   Complement Type
Modal                 be able to                 SVC
Phasal                start, finish              SVC
Try                   try, aempt, manage        SVC
Desire                want, hope                 SVC, ṇɔi
Manipulation          make, let, order           SVC, ṇɔi
Aitude               like, fear, enjoy          ɕi, ṇɔi
Aention              see, hear, discover, find  ɕi
Knowledge             know, understand, realise  direct speech
Thinking              think, assume, suppose     direct speech
Speaking              say                        direct speech

ṇɔi is the marker of purpose clauses, SVC is serial verb construction and ɕi is a subordinate clause marker which could be translated "when", "if", ... etc. In terms of independent marking of TAM and agreement they can be ordered as follows: SVC < ṇɔi < ɕi < direct speech. The following demonstrates in English what I mean about how my conlang works:

I [be.able.to go] (SVC)
I [start go] (SVC)
I [try go] (SVC)
I [want go] (SVC) / I want [in order that he go] (Purpose) (depending on coreference)
I [make happy] you (SVC) / I make [in order that you carry it] (Purpose) (depending on transitivity)
[When I danced] I like (When) / I like [in order to swim] (Purpose) (specific event vs activity)
[When you came] I saw (When)
I know "he's here" (Direct speech)
I think "he's here" (Direct speech)
I say "he's here" (Direct speech)

Note that it's only the complement types that involve direct speech which show a full range of evidentiality distinctions in my conlang. Of course, that doesn't necessarily help you directly, but maybe the hierarchy in the table is useful to you in thinking about where you might want to put the split between evidential marked complements and non-evidentially marked complements?