On Mon, Aug 13, 2012 at 6:58 PM, Hugo Cesar de Castro Carneiro <
[log in to unmask]> wrote:

> Russell Tomlin obtained the following frequency distribution for 402
> languages of the world. SOV 45%, SVO 42%, VSO 9%, VOS 3%, OVS 1% and OSV
> 0%. A similar distribution can be seen in WALS data.
> Thinking abouit this distribution I realized that the commonest two
> features are:
>    1. objects usually are adjacent to verbs (96%);
>    2. subjects tend to come before objects (91% -- In languages that
>    objects and verbs are adjacent to each other, subjects are the first
>    constituent of a clause -- 87%)

FWIW you have the percentages mixed up; it's OV-adjacent 91% and SO-order
96%.  SO-order is the most common thing cross-linguistically in S-O-V

> Matthew Dryer, the author of the word order data in WALS proposes to divide
> the languages into a 4-way typology, instead of a 6-way tyopology. He
> proposes "SV & OV", "SV & VO", "VS & VO" and "VS & OV". He says that VSO
> and VOS languages are not two different sets, as VSO languages do use VOS
> sentences and vice-versa. So they are verb-initial languages.
> But according to his proposal SOV and OSV languages are the same thing and
> this doesn't work for me.

It makes a certain sense from a purely syntactical point of view:  You have
predominantly right-headed, left-branching languages ("SV & OV"),
left-headed, right-branching ones ("VS & VO") and ones with the verb in the
middle (SVO and OVS).  Maybe I am wrong, but I don't think it's worth
making separate categories for the latter two; OVS is only 1% and the main
idea I think is that the verb is in the middle and thus case roles tend
more to be determined by word order.

> Can someone explain me how the information flow works for each word order?
> What about the verb-initial languages? And OVS and OSV?

That's a pretty big can of worms.  Topic and focus are indicated all sorts
of ways, including in English clefting and intonation.  I have no idea if
anyone can make general correlations between information flow and word
order, but I'd be interested to hear from anyone who could.

On Mon, Aug 13, 2012 at 8:56 PM, Hugo Cesar de Castro Carneiro <
[log in to unmask]> wrote:

> Does anyone know if the object-initial languages (OVS and OSV) tend to be
> ergative-absolutive?
> WALS has a pretty cool feature combination function:

Unfortunately in this combination there is not a lot of overlap between the
two datasets.  No OSV languages are indicated; however the 3 OVS languages
indicated are either neutral alignment (2 instances) or nom-acc (1
instance).  If what Matthew Dryer says is of any relevance then
theoretically, maybe OSV are something like SOV and OVS is like SVO.  In
that case, SVO (and OSV?) are, perhaps not surprisingly, predominantly
neutral alignment, and SOV (and OSV?) more varied, with nom-acc still
somewhat more represented than erg-abs.