Zompist Bboard Again

I'd like to discuss evidentiality and its different forms. Kala's system seems misaligned and bulky.

I've looked at various natlangs for inspiration, but I tend to get lost in the minutiae. How does your conlang handle evidentiality (if it does at all)?

I’ll be honest, evidentiality has never really interested me from a conlanging point of view. The only language I’ve ever made that marks evidentiality morphologically is Proto-Western. As I recall, I pulled this system from a reading of Mithun’s The Languages of Native North America, but I have to admit that I was tripping balls something chronic over the 48 hours I came up with that language.

There's a section on evidentiality in ALC.

Your Kala forms seem OK, but it's not at all clear what EVID is supposed to be.

If you have a special form for visual experience, I'd kind of expect special forms for other types of experience (e.g. hearing). To put it another way, how does VIS differ from just directly experiencing something? What do you say if you stumble onto something in the dark?

You say "none of the evidentials distinguish between direct and indirect evidence", but then it's not clear what your HSY category is. If Matt sees something, presumably he uses VIS. He tells Martha, and what does she use? VIS, because it was seen though not directly by her? At what point does someone switch to DUB and why?

You have dubitative listed under Mood as well; I'd leave it out there... you really have several forms of irrealis, which you've grouped together under Evidentiality. That leaves Mood more consistent.

Nishtigian has an evidential system that expresses five levels of evidentiality: visual and non-visual perception, reportative, inferential, and assumptious. The reportative only exists in the perfective aspect and the assumptious only in the future tense.

Visual perception implies the speaker witnessed the event happen firsthand.
Non-visual perception implies the speaker did not witness the event and instead sensed it in some other way.
The reportative implies that the action was reported to have happened, althought hard evidence may not have been provided, e.g. a rumor, confession, etc.
The inferential implies hard or obvious evidence supports the action's occurrence.
The assumptious implies the speaker is going off pure assumption or intuition.

The idea of evidentiality interests me, but I've never seriously used it - possibly because I feel it's something I want to do properly if I do it at all.

Rawàng Ata does, I think, however, have a dedicated idiomatic structure for evidentiality, employing topicalisation (because everything ends up employing topicalisation in Rawàng Ata). I think it'll just have two forms, though: a topic indicating "at my ears" for hearsay, and one indicating "at my eyes" for direct experience, including non-visual. [so, "At my eyes, the dinosaur is cold" = "the dinosaur feels cold to me"/"I can vouch for the dinosaur being cold"]. It would make sense to have at least one more metaphorical form to indicate some sort of intuition or opinion, but I'm not sure what metaphor they'd use, or even how they categorise mental states.

Issues of doubt, certainty, surprise and obviousness, however, will be handled through expressives - so "unexpectedly, it appears the dinosaur is cold" would be expressed along the lines of "wooah, cold dinosaur!", whereas "evidently, the dinosaur is cold" would be "duhhh, cold dinosaur!", and so on.

I dont have any evidential morphemes in any of my conlangs currently, but Poswa is in a state where it could easily develop some. RIght now there are chained verbs that stack onto the present tense of other verbs, so for example pappabe means "you swallow" and pappabevo means "I see that you swallow". Here, the -e- marks present tense 2nd person, and -o marks present tense 1st person. To show the past tense, only the outer vowel changes, because for diachronic reasons this morpheme structure could only appear after a verb conjugated in the present tense. Thus pappabevi means "I saw that you swallowed", and there is no way to use this structure to form a mixed tense sentence such as *I see that you chewed. (In fact, this would be translated with the present tense form and would have to be clarified by context; the outer verb always "wins" when determining the tense of a compound verb.)

Anyway, I dont call these evidentials because the person markers can rotate. Just as pappabevo means "I see that you swallow", pappabeva means "He sees that you swallow", and pappabeve means "you see that you swallow". But in Pabappa, the three person markers -o -a -e merged into just -a, and therefore Pabappa does not mark person on its verbs.

Im not planning on it, but I could decide to preserve the compound suffixes of Poswa in Pabappa, where the person markers of the inner verb would survive due to no longer being word-final, and therefore all verbs would *REQUIRE* an extra morpheme of this type so that the person markers would still be distinct. Thus, essentially all verbs would need to have an evidentiality morpheme, with the default one presumably just being whichever is the shortest and most convenient. The Pabappa cognate of the -vo -ve -va series above would be -da.

Long story short, I havent done it yet, but I could derive evidentials from independent verbs that lost their person marking and therefore became locked to the first person meaning. "To see" would be just one of several, but it might end up degrading into the unmarked form and therefore I would need to come up with new ones.

High Lulani has modal adverbs. Placed at the end of an utterance, they convey things like emotional reaction, evidentiality or likelihood. They will most likely evolve into full-blown evidentiality clitics or suffixes by the time this language becomes Koine Fezhlê.

Evidentiality in my languages tends to come across as a bit of an afterthought that I don't pay a huge amount of attention to. Though to be honest I've realised recently that even with things like tense and aspect I tend not to go into the details of how they are used.

Viksen has some evidential modifiers, of which the "most important" (read: the only ones I've actually detailed) are tjsji /ʈʂi/ and tadédjzj /tadiɖʐ/, marking "direct" and "indirect" evidentiality respectively. Neither is required and this can be seen as a rather ungrammaticalised evidentiality system.

eda xasan tjsji
3pp eat DIR
“[I’ve seen myself that] they’re eating”

xasan tadédjzj eda
eat INDIR 3pp
“they’re eating [or so I’m told]”

(Note also different word order; SOV is "indicative", SVO is "subjunctive", and intransitive arguments pattern as O.)

Greater Atlian has "direct" and "hearsay" evidentialities that are marked on every finite verb. The basic difference is that hearsay verbs include a morpheme -lle /ɬe/ or -le /le/. Most often this is suffixed after the tense/agreement endings, although in several parts of the paradigm there's something else going on as well. I'd give a paradigm table if formatting it for the board wasn't such a bother.

Thank you everyone, this info is immensely helpful.

In its current state Kišta has two composite reportatives based on the finite conjugated imperfective and irrealis participles. The irrealis form marks reported information that's counter to the speaker's expectations. It can also be used as a simple direct mirative but I'll probably make that kind of use pretty marked and use some contrastive focus particle as the main mirative strategy.

masako:

I recently gave a bunch of optional evidentials to Professor Victoria Fromkin's Pakuni language. It was a huge conlang made for a big hit kids tv show from the 70s in the USA.

Originally, I had them be CVCV like most of Pakuni's words and marked them with -Adverb and put them sentence-final, like Chinese's mood or whatever particles. Recently, I've been cutting them to CV because I think they might have the frequency and identifiableness of prepositions.

I used that Typology book by Whalen plus some of my own vague memories to make up my set of evidentials.

The conlang's word order and everything except phonology are very much like English and the evidentials and noun grammatical gender for pronouns are just enough exoticness to keep me busy. I've made some other optional grammatical features but haven't gotten them all online. I might do some more exploring toward that end.

I decided to add evidentials because I haven't used them much before and wanted to read that section of the linguistics typology book. Evidentials interested me a lot when I was in college, we used the Contemporary Linguistics textbook. I've rarely seen them i the languages I study, they seem like mostly a certain Native American languages thing. I don't think Iroquoian or 1600s Massachusett has them.

Whimemsz: Wow, cool response to my reply. Horray for Ojibwe!

Something useful I found out recently from The Oxford Handbook of Grammaticalization - evidential categories often develop from "ordinary", non-evidential TAM forms. This typically happens when a TAM form takes on an evidential implication or shade of meaning. This can remain stable of course, but sometimes it may then become the primary or exclusive meaning of the verb form, with no TAM meaning anymore.

E.g. One could imagine the English have-perfect gaining a reportative or hearsay-like meaning ("He went to the store?" "Well, I *heard* he had gone to the store"), and this could eventually become its primary meaning, thus potentially creating an evidentiality contrast in the past tense ("He died." "You saw?!" vs. "He had died." "How'd you hear?"). This could eventually become its sole meaning, with no semantic connection to time at all. ("Currently, she had intended to hire three new employees.")

"The Oxford Handbook of Grammaticalization" - that sounds cool. Now I have the Cambridge Encyclopedia of Ancient Languages and the 3-volume Oxford Encyclopedia of Ancient Egypt. That last one is extremely useful and well-done. I hope I can make extensive use of it.

I have a book "The World Lexicon of Grammaticalization" that I use a lot when conlanging. It's nice to have because they're not intuitive.