zompist bboard

A lot of linguistic literature seems to take a division between core arguments and obliques for granted. But last night, I stumbled onto wondering about whether the distinction is universal while googling on a topic of recent interest - applicatives. In particular, I am still very interested in languages where applicatives, and verbal agreement with what would be non-core NPs in a language like English, are very common and very extensive. Such languages do seem to exist, with some Northern Caucasian languages looking like good examples from the limited information I have on them.

Anyway, according to "Applicative Constructions" by David Peterson, Marianne Mithun has argued that in some American Indian languages, like Tuscarora, applicatives are commonly used to avoid the need for multiple clauses. In such a language, the choice is not applicative vs oblique adpositional phrase, but between applicative vs introducing another clause complete with a fully inflected verb. The article referenced is "Understanding and Explaining Applicatives", published by the Chicago Linguistic Society, volume 37. I would be extremely grateful to anyone who could provide me with a copy of this article, since I wasn't able to find a copy.

I was, however, able to find this:

www.eva.mpg.de/~cschmidt/SWL1/handouts/Mithun.pdf

This gives a similar argument for Mohawk, complete with examples of how what would be obliques in English end up as core arguments that the verb agrees with in Mohawk. So, the question is: to what extend is the split between core and oblique universal, and if it is universal what is the minimal extent that the two are differentiated by?

Another place where I think we could look for languages with minimal differentiation between core and oblique arguments would be languages with highly productive symmetric serial verb constructions. In SVCs, the extra verbs can increase the number of arguments of the construction as a whole, and it's not obvious that any small subset of those arguments is privileged beyond the shared actor that you typically find.

However, most languages with productive SVCs also have some previously independent verbs which act more like adpositions. Typically some verbs get specialised to express adposition-like meanings and mostly occur in combination with other verbs. What I'm not sure about is whether objects of these specialised verbs normally behave any differently to objects of other verbs in SVCs.

I suspect there isn't a simple answer to this question, because the facts of natural language tend to be messy. Different roles in different constructions often show some degree of similarity without being exactly identical - languages with applicatives are actually a good example of this. People talk a lot about 'symmetric' and 'asymmetric' applicative languages, based on whether the applicative object exclusively possesses the object properties or whether the original object also retains them, but in fact it seems that many languages have complicated mixtures where different object-like arguments each have some, but not all, properties of objects in non-applicative transitive clauses. In the languages Peterson talks about, there are even different distributions of object properties for different applicative constructions in the same language.

So I guess the same thing will happen here - by some criteria a language won't have a robust core/oblique distinction, but there will probably be subtle differences between different constructions that could be used to argue for a split of some kind.

Anyway, this ramble is just basically a way of asking for your thoughts (and the Mithun article if you have a copy!).

_________________
Try the online version of the HaSC sound change applier:

Even in English various verbs require PP or other positional/directional complements rather than NP ones, or in addition to them. For instance, you cannot merely put a thing; you have to put it somewhere - omitting the destination information usually produces an unacceptable clause. This sort of thing suggests to me that what's core to one verb is not to another, even within the same language, and that the notion should be extended to look at types of information rather than just types of NP.


Also, FWIW on languages agreeing with non-core participants, Haida has an interesting system. It does not use agreement, as such - verbs are not marked for person/number/etc as an inflective paradigm. Rather, full nominals are left alone, but just about any pronouns in the sentence are moved to the front of the verb and cliticize there. There is no core/oblique distinction observed in this - pronouns (for anything) are attached the verb, simple as that. Even possessive pronouns, and even when their possessees are not close to the verb. I have not finished reading all 1300 glorious pages of Haida Syntax yet, but I have read the chapters on constituent order, NP anaphora, and the complement/adjunct distinction (among others), and nowhere yet is there any substantial treatment of a core/oblique distinction. Except for the author having noted in passing a couple times that it appears almost congruent to the pronoun/noun distinction: pronouns are generally core and nouns are generally oblique. (Something similar works in a number of polysynthetic languages: the arguments are pronouns and any nominals that happen to be present are all oblique/adverbial. But Haida can't be called a poly lang, and the pronouns on the verb do not co-occur with co-indexed nominals - e.g. the subject will be a noun somewhere in the sentence or else a pronoun on the verb, never both at once.)

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

R_S's example with "put" is more about the fact that obliques may be arguments instead of adjuncts, and arguments may be obliques instead of core. If you have to mention something when using a particular verb, that something is one of the verb's arguments; if you don't even have to have one implicitly in mind, it's an adjunct.

The major difference between core arguments and oblique arguments, AIUI, is that:
* Core arguments are marked for the grammatical and syntactic relations and functions rather than, or at least moreso than, their semantic roles. A core argument's semantic role is likelier to be shown by voice-marking or agreement-marking on the verb.
* Obliques (both arguments and adjuncts) are marked for their semantic roles, not for their grammatical or syntactic functions or relations.

Everything in the core is (theoretically?) an argument, and every adjunct is (theoretically?) oblique; but some arguments can also be oblique, like the "where would you put it?" prepositional phrase in "put" clauses.

-------------------------------------------------------------------------------------

Whether or not the core vs oblique distinction among arguments is actually considered universal, may depend on the school of the linguist you ask.

If it is universal, it means that in every language at least some clauses have at least one core argument, which means that in every language at least some clauses have a subject.

Clearly some languages don't have any core arguments besides subjects.

Also, there's a lot of agreement between, on the one hand, those linguists who think the distinction is universal but is important in some languages and unimportant in others, and, on the other hand, those linguists who think the distinction occurs in many languages but doesn't occur in many others.

At the extremes, there are linguists who think it's universal, and there are linguists who think it occurs only in some languages and is virtually useless even in those.

-------------------------------------------------------------------------------------

As for applicatives:
Note that it's about promoting oblique arguments to "direct or primary object", not about promoting adjuncts.

In some schools, among them "MAPping theory", there are considered to be in many languages such things as "Morphosyntactically-Assigned Argument Positions" or "MAP"s. And one of the main typological parameters for languages is thought to be "How many MAPs does this language have?"

I know you're aware of Klaiman's supertype of "Derived Voice Systems", in which the main function of grammatical voice marking of a verb is to promote various arguments into and demote various arguments out of various grammatical or syntactic relations or functions.

According to Mapping Theory (and some other schools), such valency-rearranging operations (whether or not they raise or lower valency) tend to target either the most-syntactically-privileged MAP or the least-syntactically-privileged MAP. That is, they either move something into or out of the top core position, or into or out of the bottom core position.

Applicativization moves oblique arguments into the second MAP, and may move whatever was already in the second MAP, if it was already occupied, into an oblique position, or else make it implicit.

Whether or not what Mapping Theory says in the paragraph before the last is unconditionally universally true, it's a very strong statistical correlation; Applicativization tends to be restricted to 2-MAP languages (languages with just two grammatical or syntactic relations or functions).

(BTW: Actually, the same is true of dechticaetiativity.)

Languages with three MAPs tend to have valency-rearranging operations like "dative applicativization" (promoting some oblique argument into the 3rd MAP), and/or "dative movement" (promoting what's in the 3rd MAP into the 2nd MAP).

-------------------------------------------------------------------------------------

If a language such as Tagalog, or what David Peterson says the average conlanger thinks is meant by "trigger language", has only one MAP (i.e. the Subject), then the only way to promote an oblique argument into the core is to promote it directly to Subject. So such languages may be "voice-prominent", since the Subject may fill one of many semantic roles and the verb's voice is what tells the addressee which role the Subject fills.

2-MAP languages with applicativization often require two steps to get an oblique argument into the Subject position; applicativization moves it into the Object position, and then passivization moves it from there into the Subject position.

3-MAP languages, OTOH, often need two steps to get an oblique argument into the primary (or direct) object position; "dative applicativization" moves it into the secondary (or indirect) object position, then "dative movement" moves it into the primary (or direct) object position. Whether or not it's then possible to passivize the result and thus move what was originally an oblique argument into the subject position, I don't know; if so, I imagine it depends on the language.

-------------------------------------------------------------------------------------

Obviously all the mapping-theory stuff I spoke about applies only to languages which have what Klaiman called "a Derived Voice system". And, if there's more to mapping-theory than what I already said, I haven't found out about it yet.

But, even leaving aside the voice system, it's still possible that a language has one or more MAPs, though it might be that it's hard to conclude how many it has. The "subject properties list" and the "correlates of the absolutive" in your pbwiki and in the Kneequickie are good guides to at least two of them; but note they don't say every language has one.

As you can see from reading them, though, the syntactic subject, if there is one, is highly likely to be a core term (probably the same goes for the absolutive); and, if there are any core terms, it's highly likely that one of them is either a syntactic subject or an absolutive.

But I don't know that a language has to have either a syntactic subject or an absolutive, nor do I know if one has to have a core argument. If a language does have core arguments, I don't know that it must also have oblique arguments.

I highly suspect, though, that it's psychologically easier to speak a language that has either syntactic subjects or absolutives or both, than one that has no such thing. There may be common diachronic forces that frequently cause such things to evolve into core terms or MAPs; I suspect there are, but I doubt they always operate.

In every language, very many clauses cross-linguistically semantically seem to have more than one argument. For a language with only one MAP, such a clause would be easier to express if the remaining arguments could be oblique arguments.

In most languages, a numerous minority of clauses cross-linguistically semantically seem to have more than two arguments. For a language with only two MAPs, such a clause would be easier to express if the remaining arguments could be oblique arguments.

In several languages, a few clauses cross-linguistically semantically seem to have more than three arguments. For a language with only three MAPs, such a clause would be easier to express if the remaining arguments could be oblique arguments.

I've never read any report or rumor of any language anyone has said has more than four MAPs. I also do not know any clause which, no matter what language it's in, would semantically require five or more arguments, though I vaguely remember reading that someone claimed they had one. If that's so, then to express any clause that needs five or more arguments, it's going to be nearly impossible to do it without some of the arguments being obliques.

------------------------------

HTH?

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

_________________
Basilius

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

Tom, following on from your post, I will try to illustrate what I mean. I'm not sure if I explained myself very well, and as I am not an expert when it comes to MAPs I may have misinterpreted something.

My point was that there are some applicatives that promote don't always demote the existing object. I'll give some examples from some languages Peterson talks about.

Let's start with Bukusu. It is a 'symmetric' Bantu language - this means that the applicative object (promoted) and the original object are syntactically and morphologically treated in a similar way. For example:

1. The verb can agree with either object

wanjala a-mu-kul-il-a sii-tabu
Wanjala 3sS-CL1O-buy-APP-FV CL7-book
Wajala bought her the book (agreement with 'her')

wanjala a-si-kul-il-a omu-xasi
Wanjala 3sS-CL7O-buy-APP-FV CL1-woman
Wanjala bought it for the woman (agreement with 'book')

2. Either object can be the subject of a passive version of the applicative verb

omu-xasi a-kul-il-w-a sii-tabu nee-wanjala
CL1-woman 3sS-buy-APP-PASS-FV CL7-book by-Wanjala
'The woman was bought the book by Wanjala'

sii-tabu sy-a-kul-il-w-a omu-xasi (?nee-wanjala)
CL7-book CL7S-TENSE-buy-APP-PASS-FV CL1-woman (?by-Wanjala)
'The book was bought for the woman (?by Wanjala)'

3. Either object can be relativised on using the normal direct object relativisation strategy

omu-xasi ni-ye wanjala a-kul-il-a sii-tabu
CL1-woman REL-CL1 Wanjala 3sS-buy-APP-FV CL7-book
'the woman who Wanjala bought the book for'

sii-tabu ni-sy-o wanjala a-kul-il-a omu-xasi
CL7-book REL-CL7-REL Wanjala 3sS-buy-APP-FV CL1-woman

Sometimes, each object may display different object properties at the same time. In the following sentence, one object has been promoted by the passive, while the other is represented by object agreement on the passivised verb:

omu-xasi a-si-kul-il-w-a nee-Wanjala
CL1-woman 3sS-CL7O-buy-APP-PASS-FV by-Wanjala
'The woman was bought it by Wanjala'

So as you can see, while the applicative object has been promoted, it doesn't follow that the original object has necessarily been fully demoted. There is, however, one way in which the two objects differ: if both are represented by NPs in the clause, the immediately post-verbal one must be the applicative object, not the original object.

So: most of the major morpho-syntactic properties of objects are available to both the base and applicative object in Bukusu. The only exception is that the applicative object has first dibs on post-verbal position if both are represented as NPs.

Later (maybe tomorrow) I will post a summary of Peterson's other major example, Hakha Lai. That is an even more interesting example, with a much muddier picture of how object properties are distributed in applicative clauses.

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

_________________
Try the online version of the HaSC sound change applier:

I followed this thread with interest (although some of the concepts discussed are ones that I'm still getting my head around), but above all, I wanted to thank Chris for including the link to in his original post; I had wondered how a language like Mohawk handled some of these constructions, and the handout provides some good examples. Thanks again!

p@ (a deep bow with a flourish),
Glenn