[Lingtyp] languages with accusative/ergative alternation

Tue Jul 2 23:49:19 UTC 2024

Thanks Martin for your commentary. Just to clear it up, you ask:

>It seems to me that the way Tony labels the arguments here is not well-motivated: Why is ergative-marked "angute-m" an A in the first sentence, but absolutive-marked "angun" an S in the second sentence? Why is 'seal' a P in both sentences?

I was directly quoting Osahito’s example, and the labeling was his. I imagine he was assuming that all ergatives are A and never S since they occur in the construction where two arguments are indexed in the verb; and that all absolutive intransitive subjects are S since they occur where only one argument is indexed in the verb. Meanwhile, he applied a more semantic criterion for P. But I agree it seems inconsistent.

And you ask:

>It's logically possible to say that both these sentences are transitive, each with an A and a P, but do we actually want to say that? Do we want to say that the Yupik ABM ("ablative-modalis") is an accusative case? I'm not sure, so I asked whether any language had been described in this way (does Miyaoka 2012 say that?)

Yes, Inuktitut has been described this way.  There is a longstanding tradition to call the Modalis case ‘Accusative’; and to regard the  Absolutive - Modalis construction as an alternative way of expressing semantic transitivity alongside the cognate of the Yup'ik Ergative - Absolutive construction.  In this tradition, neither of the two is taken as basic. An early and influential example is:

Kalmár, Ivan. 1979. Case and context in Inuktitut (Eskimo). (National Museum of Man Mercury Series, Canadian Ethnology Service Paper No. 49.) Ottawa: National Museums of Canada.

I think a lot of the confusion is that for Yupik-Inuit, ‘transitive' and ‘intransitive’ are understood entirely as labels for constructions where the verb indexes, respectively, two vs. one argument; and this leads one to consider the construction where the patient or undergoer is in the Modalis / Accusative case as “intransitive,” even though it still has two quite canonical arguments.  I think Kalmár’s and others' point of calling the Modalis “Accusative” was to support the idea that the two constructions should be considered as equal alternatives for an underlying or semantic transitivity but with different discourse properties. For Yup'ik, Osahito in his grammar is very nuanced about the alternatives, pointing out discursive differences according to the argument structures of different verbs.

Tony Woodbury

From: Martin Haspelmath <martin_haspelmath at eva.mpg.de<mailto:martin_haspelmath at eva.mpg.de>>
Subject: Re: [Lingtyp] languages with accusative/ergative alternation
Date: July 2, 2024 at 12:33:58 AM CDT
To: Matthew Dryer <dryer at buffalo.edu<mailto:dryer at buffalo.edu>>, LINGTYP LINGTYP <LINGTYP at listserv.linguistlist.org<mailto:LINGTYP at listserv.linguistlist.org>>

On 01.07.24 19:00, Matthew Dryer wrote:
Martin,

I don’t understand why you say “by definition, a transitive pattern is a dominant one (occurring in more than two thirds of the cases)”. Why can there not be two transitive patterns, neither of which is dominant?

Matthew
I'm not entirely sure (which is why I posted the query), but it seems to me that the definition of "transitive" requires that there be a single transitive pattern. In English, for example, (i) is dominant over (ii), so we do not say that both are transitive, and that "at" is an accusative preposition. Instead, we say that "at" is an oblique marker.
(i) They shot the bear.
(ii) They shot at the bear.
Lazard (2002) explains how the notion of transitivity can be grounded in the notion of "the major biactant construction" of a language. He does not say explicitly that there can only be one such construction, but it seems to be presupposed. In my (2011) paper (on S, A, P, T, R), I rely on Lazard, and I also mostly assume that there is just one transitive construction. I was unsure what to do with cases like Yupik (mentioned by Tony Woodbury in this thread), so I mostly ignored them (but I briefly mentioned Tagalog in n. 12).
Angute-m(A)    nayiq(P)         ner-aa
man-ERG.SG     seal.ABS.SG  eat-IND.3SG.3SG
’The man is eating /has (just) eaten the seal’

Angun(S)       nayir-mek(P)   ner’-uq.
man.ABS.SG   seal-ABM.SG     eat-IND.3SG
’The man is/has (just) eaten a/the seal’

It seems to me that the way Tony labels the arguments here is not well-motivated: Why is ergative-marked "angute-m" an A in the first sentence, but absolutive-marked "angun" an S in the second sentence? Why is 'seal' a P in both sentences?

It's logically possible to say that both these sentences are transitive, each with an A and a P, but do we actually want to say that? Do we want to say that the Yupik ABM ("ablative-modalis") is an accusative case? I'm not sure, so I asked whether any language had been described in this way (does Miyaoka 2012 say that?). Maybe one problem is that making a distinction between a pattern with a dominant member (as in English (i) and (ii)) and a pattern where there is not clearly a dominant member (as in Yupik) boils down to frequency, and linguists are often reluctant to make such decisions on the basis of frequency of use.

Denis Creissels cites the example of Balinese (from Udayana 2013):

Cang n-yemak baju ento.
I ACT-take shirt DEM
'I look the shirt.' (Actor Voice, accusative alignment(?))

Baju ento jemak cang.
shirt DEM PAT.take I
'I look the shirt.' (Patient Voice, ergative alignment(?))

However, there is no argument flagging here (so the "alignment" concerns only word order), and the Actor Voice is characterized by a voice prefix, so it's not an uncoded alternation (unlike the English indirective/secundative alternation, and unlike the Yupik alternation cited above).

Perhaps the issue boils down to how exactly we individuate the relevant constructions. For example, Creissels (2024) often talks about "variants of the transitive construction", which seems to be in line with Lazard's (2002) presupposition that there is one "major biactant construction", but do we want to say that the two Yupik sentences cited by Tony Woodbury are "variants of the Yupik transitive construction"? I'm not sure.

Thanks for the discussion!

Martin

References

Creissels, Denis. 2024. Transitivity, valency and voice. Oxford: Oxford University Press (to appear).
Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15(3). 535–567.
Lazard, Gilbert. 2002. Transitivity revisited as an example of a more strict approach in typological research. Folia Linguistica 36(3–4). 141–190. (doi:10.1515/flin.2002.36.3-4.141<https://doi.org/10.1515/flin.2002.36.3-4.141>)
Miyaoka, Osahito. 2012. A grammar of Central Alaskan Yupik. Berlin: De Gruyter Mouton.
Udayana, I Nyoman. 2013. Voice and reflexives in Balinese. Austin: University of Texas at Austin. (PhD dissertation.)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20240702/f6513b58/attachment.htm>