polysynthesis sans polypersonalism?

[Vardelm]

This is an interesting point. Is noun incorporation actually converting arguments into morphemes? I assumed that it was essentially compounding that had various grammatical effects as a result (backgrounding, indefinite, etc.).

How are those two perspectives different? To me it seems fine to consider it as being effectively grammaticalised compounding with semantic and discourse effects.

To me, considering them as morphemes seems to imply that only a certain set of words would be incorporated, basically as a result of being a compound that constitutes Mithun's Type I, which are for an "institutionalized" activity. At issue is whether agents could conceivably be incorporated, so if they aren't (ever?) part of compounds that make up Type I and subsequently grammaticalized, that would rule them out for incorporation.

Reading a bit about Mohawk, the above doesn't seem to be true (aside from agents). Still, for my own learning the question is still a good one to toss out there for clarification.

[Richard W]

I'm not sure of this helps, but we have a sort of type 4 NI for pro-drop Bantu languages (e.g. Swahili), where the class prefixes of the subject are prototypically automatically reflected in the verb. Now, if we extend the number of class prefixes, hey presto, we end up with pukka type 4 NI with the subject being incorporated in the verb.

As a hint to numbers if you feel Type 4 is too limiting, I have a textbook that lists 50-odd classifiers for Thai, saying that another 30 can be dispensed with, and then adds a list of about 50 more words that act as their own classifiers, including body parts.

[Frislander]

The problem with polysynthesis as a term is that it's historically been used more to describe a particular "feel" of a language rather than the presence of any feature in particular, with the end result that it's pretty much useless for actual theoretical purposes, and when people have tried to come up with a consistent definition for what it is it's always ended up excluding languages which pretty much everyone would agree "feel polysynthetic", like requiring noun-incorporation and thereby excluding Eskimo-Aleut, or requiring polypersonalism and thereby excluding Nuu-Chah-Nulth, and so on. As a conlanging term it's fine, because the historical pattern of usage means that people have a pretty consistent intuition about what counts as polysynthetic, but it isn't actually that informative about the language.

[Vardelm]

The problem with polysynthesis as a term is that it's historically been used more to describe a particular "feel" of a language rather than the presence of any feature in particular, with the end result that it's pretty much useless for actual theoretical purposes, and when people have tried to come up with a consistent definition for what it is it's always ended up excluding languages which pretty much everyone would agree "feel polysynthetic", like requiring noun-incorporation and thereby excluding Eskimo-Aleut, or requiring polypersonalism and thereby excluding Nuu-Chah-Nulth, and so on. As a conlanging term it's fine, because the historical pattern of usage means that people have a pretty consistent intuition about what counts as polysynthetic, but it isn't actually that informative about the language.

I found the paper "A structural typology of polysynthesis" by Johanna Mattissen (2015) super helpful in this regard. I think this was linked in Whimemsz's old post. It's at least a good survey of the range "polysynthetic" languages can have.

[Richard W]

There may be a relevant example of agent-backgrounding by incorporation in the old Indic languages. The transformation can go:

agent-NOM patient-ACC verb-FINITE >
agent-INS patient-NOM verb-PASS.PART. >
patient-NOM agent-INS verb-PASS.PART. >
patient-NOM agent-verb-PASS.PART

The second stage is the syntactic origin of Indic split ergativity. The third stage is a run of the mill passive. The "agent-verb-PASS.PART" could serve as your noun-incorporation. It's not a finite verb, but it's what the modern Indic past tenses developed from. The compound is a perfectly standard tatpurusha semantically associated with the instrumental case, e.g. Pali sukā phalaṃ āhariṃsu > sukehi phalaṃ āhaṭaṃ > phalaṃ sukehi āhaṭaṃ > phalaṃ sukāhaṭaṃ "Parrots brought the fruit" > ... > "The fruit is parrot-brought".

[Kuchigakatai]

Based on the link Ars Lande provided, plus some more looking this afternoon, it appears that Haida, Maidu, Arabana, Klamath, and Tlingit are examples where pronouns appear in free form as opposed to bound to the verb. I don't know to what extent each of those is "prototypical" polysynthetic. I think if I follow some examples there and (if needed) add in more noun incorporation, I'll be within the general ball park. I'm not fussy about matching an exact definition of polysynthetic, but it helps to know what the range of natural options are.

and when people have tried to come up with a consistent definition for what [polysynthesis] is it's always ended up excluding languages which pretty much everyone would agree "feel polysynthetic", like requiring noun-incorporation and thereby excluding Eskimo-Aleut, or requiring polypersonalism and thereby excluding Nuu-Chah-Nulth, and so on.

Oh, that's great guys. I didn't (and haven't) read Ars Lande's link because I'm sure it goes to a recent linguistics books, and I try to avoid Google Books reserving it instead to those moments where I absolutely need to get a passage in a rather unavailable copyrighted book, because of view limits.

All of these posts so far have addressed traditional object incorporation, but i still want to see example of subject incorporation when you develop the language to the point where you can show examples. i imagine this will be much moire difficult which is why i am so interested in it.

It may well not be feasiable. I freely admit I don't really know WTF I'm doing with incorporation, so this might be like watching a conlang trainwreck. I'll post whatever I come up with though, so feel free to drop in, point, & laugh.

Subject incorporation is attested in many languages. The thing is that it's nearly only attested with unaccusative verbs, that is, verbs whose subjects lack control or are patient-like or are inanimate, and even the extremely rare examples of incorporated [+control] animate subjects still don't involve humans or spirits, but animals, so still "inanimate" to the extent they're irrational. Some examples:

　我頭痛。wǒ tóutòng [Mandarin]
　1SG head-hurt
　'My head hurts.' (bradrn's abstract example in page one)

You can tell this involves an incorporated subject because (1) 痛 tòng also exists as a clearly intransitive verb meaning 'hurt' (很痛喲！hěn tòng yo! 'It hurts!'), (2) Mandarin has SAdvV order for degree adverbs and here they appear before 頭 tóu 'head' (我很頭痛 wǒ hěn tóutòng 'my head hurts a lot', lit. "1SG very head-hurt"), and (3) the moment you use an un-incorporatable disyllabic noun or a modified monosyllabic one then degree adverbs can't go in that position anymore, as Mandarin mostly incorporates only monosyllabic nouns into verbs (我眼睛痛 wǒ yǎnjing tòng 'my eyes hurt', ungrammatical *我很眼睛痛 wǒ hěn yǎnjing tòng with intended meaning 'my eyes hurt a lot', grammatical 我眼睛很痛 wǒ yǎnjing hěn tòng). Although I guess you could still argue this involves an incorporated locative... Anyway, to continue onto more examples:

　kǫ́ 'fire' [Hare North Slavey, Rice 1991: 55]

　gozhíkǫ́dawé
　go-zhí--kǫ́-da-wé
　area-into--fire-(inflection)-happen
　'A forest fire is starting.' (lit. "It nearby-fire-happens.")


　ətləg-in ətlˀa wˀi-gˀi [Chukchi, Spencer 1995: 450, citing Polinskaja and Nedjalkov 1987: 259]
　father-GEN mother.ABS die-3SG.INTR
　'My father's mother died.'

　ətləg-ən ətlˀa-wˀe-gˀe
　father-ABS mother-die-3SG.INTR
　'My father's mother died on him.' (lit. "Father mother-died.")

Notice the vowel harmony effect of the incorporated ətlˀa- in the second sentence.


　ko 'vomit (noun)' [Ahtna, Rice 2008: 381]

　ti-ko-si-ni-ɬ-taen
　out-vomit-1SG.DO-(aspect)-causative-lie.SG.ANIM
　'I went out to vomit.' (lit. "Vomit-made-lie-me out.", with ""animate"" vomit inside the verb)

　tsula 'tongue' [Ahtna, Rice 2008: 384]

　de-zaa ts’a-na-tsula-l-tses
　REFL-mouth (adverb)-ITERATIVE-tongue-(voice)-move.flexible.thing
　'His tongue goes in and out of his mouth.'

Notice how the iterative aspect of 'move [flexible thing]' appears before the incorporated noun.

　ɬi- [Ahtna, Rice 2008: 382]
　'dog' (word root)

　ɬi-y-az-’aɬ
　dog-3DO-(aspect)-bite
　'A dog bit him/her once.' (lit. "Dog-bit-her.")

Rice notes she only finds this incorporation of an animate in control for the noun 'dog'.


- Rice, Keren. 1991. "Intransitives in Slave." In: International Journal of American Linguistics. Vol. 57, no. 1 (January), pp. 51-69. Pub.: University of Chicago Press.
- Rice, Keren. 2008. "On incorporation in Athapaskan languages". In: Grammatical Change and Linguistic Theory: The Rosendal Papers. Ed.: Thórhallur Eythórsson. Pub.: John Benjamins.
- Spencer, Andrew. 1995. "Incorporation in Chukchi." In: Language. Vol. 71, no. 3 (September), pp. 439-489. Pub: Linguistic Society of America.

As cited by Spencer:

- Polinskaja, Maria; Nedjalkov, Vladimir P. 1987. "Contrasting the absolutive in Chukchee." Lingua. Vol. 71, pp. 239-269.



(I would've liked to add a Coptic example of this, since this is also found there, but I've run out of time to write this post.)

[Richard W]

　我頭痛。wǒ tóutòng [Mandarin]
　1SG head-hurt
　'My head hurts.' (bradrn's abstract example in page one)
...
Although I guess you could still argue this involves an incorporated locative...

I've always thought of the similar Thai structure (e.g. word-for-word I arm broke = "I broke my arm") as a native bahuvrihi.

[Kuchigakatai]

I've always thought of the similar Thai structure (e.g. word-for-word I arm broke = "I broke my arm") as a native bahuvrihi.

I know I'm getting off-topic here, but what do people mean when they use "bahuvrihi" in general linguistics sometimes, as you seem to have done there? You don't mean the Sanskrit definition of bahuvrihi, since "arm-break" is still in the same word class as "break" (verb), and doesn't have an idiomatic meaning but is rather an endocentric compound. Or is this from a misunderstanding of what bahuvrihi means? But I'm pretty sure this isn't the first time I see bahuvrihi used for... something non-idiomatic, endocentric and in the same word class, even though you'd think that'd be a tatpurusha.

Incidentally, I never see people with an interest in general linguistics using "tatpurusha". Maybe it is a non-distinction of bahuvrihi and tatpurusha after all, so basically anything that's not a dvandva or compound that repeats a root?

[Richard W]

I've always thought of the similar Thai structure (e.g. word-for-word I arm broke = "I broke my arm") as a native bahuvrihi.

I know I'm getting off-topic here, but what do people mean when they use "bahuvrihi" in general linguistics sometimes, as you seem to have done there? You don't mean the Sanskrit definition of bahuvrihi, since "arm-break" is still in the same word class as "break" (verb), and doesn't have an idiomatic meaning but is rather an endocentric compound.

Bahuvrihis are basically adjectival, and I see it as being an 'adjective' as Thai grammarians do use that term for their own language, even though there is a widespread view that what they call 'adjectives' are actually stative verbs. Now, as a phrase rather than a clause, "arm" is the 'head' and the verb "break" serves as an 'adjective' describing it, or voice-neutral participle, so "arm break" means "broken arm" (the word for 'break' I had in mind is intransitive), and thus one gets the bahuvrihi with meaning "having a broken arm". (One doesn't really need the concept of 'participle' here if one pushes through the concept of 'adjectives' being stative verbs.)

Note that I don't agree that "arm-break" is in the same subclass as "break"; I see it as a stative verb. For reference, the Thai sentence is ผมแขนหัก (RTGS: phom khaen hak.)

Or is this from a misunderstanding of what bahuvrihi means? But I'm pretty sure this isn't the first time I see bahuvrihi used for... something non-idiomatic, endocentric and in the same word class, even though you'd think that'd be a tatpurusha.

Sameness of class is no guide. Sanskrit rājaputra can be a tatpurusha meaning 'a king's son', or a bahuvrihi meaning 'having kings for sons'. In Vedic Sanskrit, they are distinguished by accent.

I found the definitions of the terms 'endocentric' and 'exocentric' confusing. I found it much easier to remember the meanings of the exemplars - tatpurusha = 'His man' ('His' = 'the king's) and bahuvrihi = '(with) much rice'. (The English use of 'with' for circumstances gives wider coverage than it may in other languages.) A Pali example of a bahuvrihi I have before me is chinnahattha = 'with hands cut off'. So this is an exocentric compound. Note that most bahuvrihis translate quite literally into English - one just has to realise that it is a bahuvrihi.

Incidentally, I never see people with an interest in general linguistics using "tatpurusha". Maybe it is a non-distinction of bahuvrihi and tatpurusha after all, so basically anything that's not a dvandva or compound that repeats a root?

It may be that tatpurushas are relatively unremarkable to English speakers - we effectively use them in noun stacks (e.g. car engine), hyphenated adjectives like store-bought or man-eater, and the related karmadharayas exist in words like blackbird. On the other hand, bahuvrihis are fairly rare - we normally add an extra element, as in blue-eyed.

[Kuchigakatai]

Bahuvrihis are basically adjectival, and I see it as being an 'adjective' as Thai grammarians do use that term for their own language, even though there is a widespread view that what they call 'adjectives' are actually stative verbs. Now, as a phrase rather than a clause, "arm" is the 'head' and the verb "break" serves as an 'adjective' describing it, or voice-neutral participle, so "arm break" means "broken arm" (the word for 'break' I had in mind is intransitive), and thus one gets the bahuvrihi with meaning "having a broken arm". (One doesn't really need the concept of 'participle' here if one pushes through the concept of 'adjectives' being stative verbs.)

Note that I don't agree that "arm-break" is in the same subclass as "break"; I see it as a stative verb. For reference, the Thai sentence is ผมแขนหัก (RTGS: phom khaen hak.)

Ahhh... with "arm" as the head of the compound! I see!

Well, that was interesting. You could argue that about Mandarin too, although I still get tempted towards subject incorporation because 痛 tòng 'hurt' is also pretty semantically stative regardless (unlike 破 pò 'break (intrans.)'). Sadly, Mandarin doesn't have "arm-break" as far as I can tell, it seems people usually say "I break-PFV arm" or "I arm [insert adverb] break-PFV".

Note that most bahuvrihis translate quite literally into English - one just has to realise that it is a bahuvrihi.

"Broken-armed", "king-son-ed", "high-riced", "cut-handed"?

[bradrn]

Note that most bahuvrihis translate quite literally into English - one just has to realise that it is a bahuvrihi.

"Broken-armed", "king-son-ed", "high-riced", "cut-handed"?

“Broken-armed” is actually fine for me as a native speaker, although the other ones aren’t.

[Richard W]

Note that most bahuvrihis translate quite literally into English - one just has to realise that it is a bahuvrihi.

"Broken-armed", "king-son-ed", "high-riced", "cut-handed"?

'With' works better. Of course, if the individual morphemes are hard to translate quickly, then that challenge will be inherited. But 'with broken arm', 'with king son', 'with much rice' and 'with cut (off) hand' are mostly intelligible. 'With king son' is on the border of grammaticality, but 'with the king her son' seems unremarkable English. As with many literal translations, one does have to clean the translation up, but this rough translation gets the sense across. It's not unlike the recasting that generally needs to be done when translating Latin ablative absolutes into English.

[Kuchigakatai]

"Broken-armed", "king-son-ed", "high-riced", "cut-handed"?

'With' works better. Of course, if the individual morphemes are hard to translate quickly, then that challenge will be inherited. But 'with broken arm', 'with king son', 'with much rice' and 'with cut (off) hand' are mostly intelligible. 'With king son' is on the border of grammaticality, but 'with the king her son' seems unremarkable English.

I understood what you meant, and was just joking there. BTW I love the use of "with the king her son"; it renders rājaputra well and yet it also sounds so economic and natural. I wouldn't've thought of that one. Thanks for your good explanation of bahuvrihi --my total ignorance of anything to do with ancient (or modern) Indic shows.

[Richard W]

Well, that was interesting. You could argue that about Mandarin too, although I still get tempted towards subject incorporation because 痛 tòng 'hurt' is also pretty semantically stative regardless (unlike 破 pò 'break (intrans.)'). Sadly, Mandarin doesn't have "arm-break" as far as I can tell, it seems people usually say "I break-PFV arm" or "I arm [insert adverb] break-PFV".

Thai has some verbs that can be used either way round - ผมเจ็บแขน - word for word "I ache arm" and also ผมแขนเจ็บ = word for word "I arm ache". The word เจ็บ is stative by the khwaam v. kaan test - 11:1 ratio in raw google hits - but the Royal Institute Dictionary lists it as a verb, as opposed to an adjective. I rationalised the sufferer-verb-organ sequence as having an 'accusative of respect', which is equivalent to your disliked alternative of having a locative for the original Mandarin example.

[Frislander]

"Subject incorporation" is a bad term here, because what's typically being referred to is specifically the incorporation of transitive agents, not any "subject" (however it's defined in any given language). This is because intransitive subject incorporation is pretty common among languages that have incorporation (Iroquoian being particularly known for it), but transitive subjects are very rarely incorporated at all. The reason given for this is that transitive agents tend to be animate, specific and definite, specially in natural speech (where we tend to prefer to keep the main subject of the narrative as the subject, hence why passives exist), both things incorporated objects are typically not, and so are strongly disfavoured as targets of incorporation. Plus another possible reason occurs to me that if you allow transitive verbs to incorporate either participant there is the possibility of ambiguity as to which role is being incorporated in any given instance.

[Vardelm]

"Subject incorporation" is a bad term here, because what's typically being referred to is specifically the incorporation of transitive agents, not any "subject" (however it's defined in any given language).

I think this is a good point, and is in line with what I'm thinking. With that said....

Plus another possible reason occurs to me that if you allow transitive verbs to incorporate either participant there is the possibility of ambiguity as to which role is being incorporated in any given instance.

I've noted that some languages reduce the valence of their verb marking when they incorporate a noun, and that's what I'm leaning towards. I will have voice/valence markers as a suffix on the verb. If that marker calls for a stative (patient) participant and you then see that stative/patient noun/pronoun, then you have a good idea that the incorporated noun in an agent.

So maybe I'm not incorporating agents on a transitive verb, I'm incorporating the noun that was / would have been the agent on a transitive verb and now is part of an intransitive, passive verb.

[Kuchigakatai]

"Subject incorporation" is a bad term here, because what's typically being referred to is specifically the incorporation of transitive agents, not any "subject" (however it's defined in any given language). This is because intransitive subject incorporation is pretty common among languages that have incorporation (Iroquoian being particularly known for it), but transitive subjects are very rarely incorporated at all. The reason given for this is that transitive agents tend to be animate, specific and definite, specially in natural speech (where we tend to prefer to keep the main subject of the narrative as the subject, hence why passives exist), both things incorporated objects are typically not, and so are strongly disfavoured as targets of incorporation.

All of that is true. But it's interesting to see that incorporation of animate and specific nouns is attested in intransitives as long as they are [-control], like the "Father mother-died" = 'My father's mother died on him' example I gave above. You'd think this would also allow e.g.

• I mother-hit my brother = 'I got my mother to hit my brother (without my mother being in control)', 'My mother inadvertently hit my brother in my favour'
• It king-listened my father.ACC (with "it" as an impersonal marker, not an inanimate pronoun) = 'The king listened to my father' (cf. the Ahtna example above with 'vomit')

but curiously it doesn't.

Maybe it does come down to ambiguity: why shouldn't "I mother-hit my brother" mean, for example, 'I hit my brother for my mother', with "mother-" as a recipient, or 'I hit my brother like I was his mother', with "mother-" as an equative/essive?

But then I wonder why there isn't distinct morphology to disambiguate different types of incorporation... Surely you can still justify you have incorporation even if you have multiple incorporation types, say, if the aspect marking consistently appears on the other side with strict order, phonetically unstressed for good measure (1SG ITER-PRF-mother-hit my brother).

[Vardelm]

Are there any grammars of specific languages anyone would recommend that discuss noun incorporation? I have "A Grammar of Akwesasne Mohawk" by Nancy Bonvillain, which has a decent section on it. However, the other grammars I've found so far are either languages that don't feature noun incorporation much or just don't cover it in the grammars I've seen. I'd be particularly interested in any of the Salishan or Wakashan languages if they have NI. I've seen info that they do, but nothing further. It seems to be difficult to even track down which languages have NI for sure.

[Frislander]

Are there any grammars of specific languages anyone would recommend that discuss noun incorporation? I have "A Grammar of Akwesasne Mohawk" by Nancy Bonvillain, which has a decent section on it. However, the other grammars I've found so far are either languages that don't feature noun incorporation much or just don't cover it in the grammars I've seen. I'd be particularly interested in any of the Salishan or Wakashan languages if they have NI. I've seen info that they do, but nothing further. It seems to be difficult to even track down which languages have NI for sure.

They only have "incorporation" if you count "verbalising suffixes that attach to nouns with semantics like verb roots in other languages" as being incorporation, but because the verbal suffix isn't an independent root in itself it's generally not (the same for Eskimo-Aleut), though the functions are often similar.

Also if you want papers I would suggest the classic Mithun The Evolution of Noun Incorporation just because it has lots of diverse examples.

[bradrn]

Also if you want papers I would suggest the classic Mithun The Evolution of Noun Incorporation just because it has lots of diverse examples.

I can recommend that as well! I read it a few days ago and found it really interesting — the only reason why I didn’t recommend it then was because I had the impression that Vardelm already read it.