zompist bboard

So I've heard the accepted wisdom is that, generally, SOV-type languages have lots of cases, while SVO-type languages have few or none. So I did a small study a little while back to see just how true that was.

This was the result.

The attachment is a visual summary of my findings.

Basically, most languages in general have 1 (that is, no) case, with SVO languages making up the majority of those, and SOV languages did make up the bulk of multi-case languages (then again, considering SOV and SVO languages combined constitute like 90% of world languages...). However, these were trends, rather than absolutes. Personally, I figure that if a language has a single dominant word order, cases don't really matter, because the subject and object always show up in their prescribed place.

A couple of notes (also noted in the document): first, for the graph, having "1" case basically means there's no case system -- having "0" cases are those languages which have, to quote the WALS, "borderline nominal case," which is something that probably means several different things depending on the language in question; secondly, this was a small sample, drawn from the WALS samples for word order and number of cases -- that is, languages that fell on both lists.

I don't know if anyone will find this useful at all -- mostly it's for the natlangers. Kinda feel like this was more or less just reaffirming general assumptions, but like with a twist.

vergil wrote:So I've heard the accepted wisdom is that, generally, SOV-type languages have lots of cases, while SVO-type languages have few or none.

It's not as far as I'm aware.

Vijay wrote:

vergil wrote:So I've heard the accepted wisdom is that, generally, SOV-type languages have lots of cases, while SVO-type languages have few or none.

It's not as far as I'm aware.

It's definitely part of conlanging lore though. I've seen conlangers making such a claim over the years at various points.

vergil wrote:Basically, most languages in general have 1 (that is, no) case, with SVO languages making up the majority of those, and SOV languages did make up the bulk of multi-case languages (then again, considering SOV and SVO languages combined constitute like 90% of world languages...). However, these were trends, rather than absolutes. Personally, I figure that if a language has a single dominant word order, cases don't really matter, because the subject and object always show up in their prescribed place.

Perhaps, but then redundancy is a ridiculous common and useful feature of language (because it helps ensure important grammatical information is passed through the noisy channel of speech).

There's also the additional issue of some languages being classified as being f one word-order type when in fact the situation is much more complicated, for instance Japanese is often referred to as an SOV language when in fact the driving factor with regards to word order is its topic-comment structure (I think both Tibetan and Quechua are also like this).

A couple of notes (also noted in the document): first, for the graph, having "1" case basically means there's no case system -- having "0" cases are those languages which have, to quote the WALS, "borderline nominal case," which is something that probably means several different things depending on the language in question; secondly, this was a small sample, drawn from the WALS samples for word order and number of cases -- that is, languages that fell on both lists.

The thing with WALS' "borderline" category with that map is that it's for any languages that have cases but don't use them to distinguish core arguments, so it's things like the Algonquian locative. It's a rather artificial distinction though because they do count those cases in the other languages, and it doesn't distinguish languages which have multiple such non-local cases (like say both a locative and an instrumental) from those with only one.

Also counting those "borderline" cases as 0 and those languages with no case marking at all as 1 is about as unintuitive as it gets.

Yeah, that's fair.

Probably it'd be best just to cut out all the "borderline nominal cases" instances -- particularly since WALS fails to specify how many of these borderline cases the languages in question actually have.

At least then categorizing a language with no cases as having 1 case makes a little more sense. (The logic behind that, at least, is that these languages have a kind of "omnicase" that's used in all environments. I'm not sure what having 0 cases would be like under that strange definition -- maybe pro-drop?)

Frislander wrote:There's also the additional issue of some languages being classified as being f one word-order type when in fact the situation is much more complicated, for instance Japanese is often referred to as an SOV language when in fact the driving factor with regards to word order is its topic-comment structure (I think both Tibetan and Quechua are also like this).

Pretty sure that's not true of Quechua. Quechua is pretty solidly SOV, and indicates topic with a clitic, not word order.

zompist wrote:

Frislander wrote:There's also the additional issue of some languages being classified as being f one word-order type when in fact the situation is much more complicated, for instance Japanese is often referred to as an SOV language when in fact the driving factor with regards to word order is its topic-comment structure (I think both Tibetan and Quechua are also like this).

Pretty sure that's not true of Quechua. Quechua is pretty solidly SOV, and indicates topic with a clitic, not word order.

You're probably right, I don't know that much about Quechua, just like I'd seen the topic clitic lumped with the case suffixes, probably on Wikipedia (which has since been edited).

vergil wrote: The attachment is a visual summary of my findings.

Using a linear graph to visualize a discrete scale (e.g. a number of grammatical cases) is a bad practice. Consider using a different type of a graph.

Basically, most languages in general have 1 (that is, no) case, with SVO languages making up the majority of those, and SOV languages did make up the bulk of multi-case languages (then again, considering SOV and SVO languages combined constitute like 90% of world languages...). However, these were trends, rather than absolutes. Personally, I figure that if a language has a single dominant word order, cases don't really matter, because the subject and object always show up in their prescribed place.

I'd assume it's because OV languages tend to be head-last, which implies postpositions — on the other hand VO correlates with head-first and prepositions. While prepositions and postpositions are symmetrical to a degree, prefixing and suffixing are not. As a result, postposing systems (likely SOV) are much more likely to become case systems than preposing systems (likely SVO) are.

Why is there an asymmetry in affixing? I assume it's easier to parse it that way.

What would VSO as the primary word order of a free word order system suggest to you?

Sounds a bit like Romani, actually. (Okay, Romani doesn't actually have a clear primary word order other than simply VO, but VSO is very common in Romani, as is SVO).