the occurrence contrast between VOS and OVS orders

[Harald Hammarström]

>  Anyway, your data is different from all data I have seen so far. For example,
>  In WALS, Dryer provides the following contrast: VOS 26, OVS 9, in his 1228-language data.

This discrepancy can be explained by genetic bias. If you look at the
genera-statistics from the same Dryer data, which discounts for some, but
not all, genetic bias, the difference shrinks to 12 VOS vs. 7 OVS.
However, five of those VOS genera are Atayalic, Sulawesi, Sundic, Oceanic 
and Borneo which are all solidly Austronesian and may have inherited their
VOS order from their ancestor (correct me those who know comp 
Austronesian syntax better). Once this is discounted, we are down at
8 VOS vs. 7 OVS, which is similar to my data and is not statistically
significant.

all the best,

H


>  My interest is that there is a parallel between orders of S, V, O and the
>  orders of R(ecipient), V and T(heme) in terms of frequency.  VTR is 
>  significantly more frequent than TVR. In fact, I havn't found any TVR 
>  language so far.  If somebody can provide me with a data of TVR, I 
>  should appreciate it very much!
>  Based on the parallel of the two contrasts, I conjecture that there is 
>  a common explanation behind the two contrasts.  That is why I asked for 
>  the explanation of the contrast between VOS and OVS.
>  My tentative explanation of the contrast between VOS and OVS is that 
>  Proximity Principle (V-O bounding) does not work on surface, since the 
>  distance between OV and VS in OVS is actually the same on surface, but 
>  it works in VOS.  Any feedback will be most welcome!
>
>  Best
>  Bingfu
>
> Harald Hammarström <harald at BOMBO.SE> wrote:
>  Date: Wed, 16 May 2007 02:40:21 +0200
> From: Harald Hammarström <harald at BOMBO.SE>
> Subject: Re: the occurrence contrast between VOS and OVS orders
> To: LINGTYP at LISTSERV.LINGUISTLIST.ORG
>
>> Dear colleagues,
>>
>> I wonder if anyone has explained the distributional contrast between VOS
>> languages and OVS languiages. The former is significantly frequent
>> than the latter.
>
> Dear Prof. Lu,
> In my data at least, VOS is _not_ significantly more frequent than
> OVS (8/346 vs. 6/346 which is not statistically significant: p > 0.24).
> http://www.cs.chalmers.se/~harald2/alt2007.pdf
> Those figures are controlled for genetic bias, but not for areal bias.
> The VOS languages in that sample were "Xinca", Wari', Mezquital Otomi,
> Inesen~o, Kariri, Washco-Wishram, Cayuvava, Garawa. (However, More recent
> data that I consulted since, shows that Garawa is better classified
> as VSO/VOS.) The OVS languages were Ona, En~epa, Macuna, Urarina,
> Waikuri, Ngarinyin. The only possible cases where one could argue
> areal influence would be En~epa-Macuna among the OVS and Otomi-Xinca
> among the VOS -- that is, there is no areal bias that could account
> for differences in the incidence of OVS vs. VOS languages.
>
> H
>