Is that true? ( edit ): WALS.info seems to back it up, so I guess it is. ( / edit )xingoxa wrote:There seems to be some kind of relationship between word order and the prevalence of ergativity.
Verb-final SOV languages are quite often ergative.
Verb-initial VSO languages are sometimes ergative.
Verb-medial SVO languages are rarely (perhaps some would claim never) ergative.
For some reason.
I checked WALS.info.
The Word-Order features having to do with Verbs are 81, 82, 83, and 84.
The Alignment features are 98, 99, and 100.
150 languages have both features recorded;
78 (52%) are verb-final, 52 (35%) are verb-medial, and 20 (13%) are verb-initial;
20 (13%) are ergative, and 130 (87%) are non-ergative.
Out of the 78 verb-final languages, 16 (21%) are ergative.
Out of the 52 verb-medial languages, 1 (2%) is ergative.
Out of the 20 verb-initial languages, 3 (15%) are ergative.
Out of the 20 ergative languages, 16 (80%) are verb-final, 3 (15%) are verb-initial, and 1 (5%) is verb-medial.
Out of the 130 non-ergative languages, 62 (48%) are verb-final, 51 (39%) are verb-medial, and 17 (13%) are verb-initial.
You'd have to do a chi-square test to see whether the correlation "verb-medial means non-ergative" is statistically significant.
Or to decide whether the statistical correlation between verb-final and ergative is statistically significant.
If word-order and morphosyntactic alignment were statistically independent
you would expect to see 2% of the languages be verb-initial and ergative, just as you do,
and expect to see 11% of them be verb-initial and non-ergative, as compared with the 12% that are.
But you'd expect to see 5% be verb-medial and ergative, instead of the 1% that is;
you'd expect to see 7% be verb-final and ergative, instead of the 11% that are;
you'd expect to see 30% be verb-medial and non-ergative, instead of the 34% that are;
and you'd expect to see 45% be verb-initial and non-ergative, instead of the 41% that are.
Is that significant?
So; why is it that verb-medial languages tend not to be ergative and ergative languages tend not to be verb-medial?
http://wals.info/feature/combined/99A/81A and http://wals.info/feature/combined/100A/81A show just about the same correlation, ANAICT.
http://wals.info/feature/combined/98A/82A, http://wals.info/feature/combined/99A/82A, and http://wals.info/feature/combined/100A/82A don't show any immediately-obvious-to-me correlation between alignment and the "SV or VS" feature.
But http://wals.info/feature/combined/98A/83A and http://wals.info/feature/combined/99A/83A seem to show that OV languages outnumber VO languages about 5 to 1 among ergative languages, while there're almost as meany VO languages as OV languages among all alignments taken together.
http://wals.info/feature/combined/100A/83A, http://wals.info/feature/combined/98A/84A, http://wals.info/feature/combined/99A/84A, and http://wals.info/feature/combined/100A/84A, probably shed less light on the subject; or if they shed light on it I don't see it yet.