recent paper (Dunn et al. in Nature)

Sat Jul 9 09:21:19 UTC 2011

It seems to me that (at least for FUNKNET readers) the crucial question 
is to what extent general functional factors, in particular 
processing-ease considerations as proposed by Dryer (1992) and Hawkins 
(1990, 1994, 2004, 2007), can be taken to explain cross-linguistic word 
order tendencies. For extraposition due to heaviness, probably nobody 
would doubt the relevance of processing ease. But Dryer and Hawkins have 
argued that processing ease also explains the Greenbergian correlations.

Tom Givón’s proposal that these correlations are the “direct product of 
diachronic pathways of grammaticalization” does not contradict the 
functional Dryer-Hawkins account. It could be that the relevant 
diachronic changes are functionally constrained, i.e. that those 
grammaticalizations that yield the most processable structures occur 
most often.

There are at least three ways in which synchronic patterns could be 
explained in “evolutionary” terms:

(1) unconstrained evolutionary: Synchronic states are the result of 
diachronic changes, but there are no particular constraints on the latter

(2) constrained evolutionary: Synchronic states are the result of 
diachronic changes, and only certain kinds of diachronic changes are 
possible, but the contraints on changes are not functional (e.g. Blevins 
2004)

(3) diachronic-functional: Diachronic changes are typically functionally 
constrained, and hence synchronic states can be said to be functionally 
adapted

I always thought that Tom Givón’s diachronic view of synchronic states 
was of type (3) (and similar views have been expressed by Greenberg, 
Bybee, Croft, and many others).

By contrast, the paper by Dunn et al. that Givón mentions (see 
http://language.psy.auckland.ac.nz/wordorder/, 
http://www.nature.com/nature/journal/v473/n7345/full/nature09923.html) 
says that it found only family-specific linkages of traits, and the 
authors do not endorse a diachronic-functional account of the 
Greenbergian correlations. One gets the impression that Dunn et al. opt 
for view (1) above.

However, I think that the available evidence is still fully compatible 
with the Dryer-Hawkins explanation in functional terms (and hence with 
view (3)), and I know of no alternative explanation of the Greenbergian 
correlations (the generative headedness parameter has been abandoned by 
the genertivists themselves). It is true that we would expect to find 
systematic evidence for linked changes in diachrony if we look 
systematically (e.g. adposition-noun order changing once verb-object 
order changes, or vice versa). Dunn et al. did not find good evidence of 
this kind, but that doesn’t mean that it doesn’t exist – maybe it exists 
only at greater time depths than is provided by the four families they 
look at.

Consider the following data cited by Matthew Dryer in a recent 
presentation in Leipzig: Looking only at families (most of them 
isolates) that have only a single type, we find the following distribution:

OVVO

Postp883

Prep026

We thus see a very striking synchronic correlation, which cannot be due 
to accident. We cannot have certainty that prepositional OV languages 
and postpositional VO languages are disfavoured for functional reasons, 
and clearly there are other important factors such as language contact, 
but the functional account still seems to be the best story that is 
available.

(It seems that the primary goal of the Dunn et al. paper was to apply 
their phylogentic methods, and that because of these new methods and the 
paper’s anti-Chomskyan conclusion, startling to some readers, the paper 
was accepted by Nature. For typologists and functionalists, there is 
nothing surprising there, as far as I can tell.)

Martin

****************

References

Dr Blevins, Juliette. 2004. /Evolutionary phonology: the emergence of 
sound patterns/. New York: Cambridge University Press.

Dryer, Matthew S. 1992. The Greenbergian word order correlations. 
/Language/ 68(1). 81–138.

Dunn, Michael, Simon J Greenhill, Stephen C Levinson & Russell D Gray. 
2011. Evolved structure of language shows lineage-specific trends in 
word-order universals. /Nature/ 473:79-82

Hawkins, John A. 1990. A parsing theory of word order universals. 
/Linguistic Inquiry/ 21(2). 223–261.

Hawkins, John A. 1994. /A performance theory of order and constituency/. 
Cambridge: Cambridge University Press.

Hawkins, John A. 2004. /Efficiency and complexity in grammars/. Oxford: 
Oxford University Press.

Hawkins, John A. 2007. Processing typology and why psychologists need to 
know about it. /New Ideas in Psychology/ 25(2). 124–144.

On 08/07/2011 21:08, Tom Givon wrote:
>
> Dear FUNK folks,
>
> A month ago David Kronenfeld sent me a recently-published paper 
> ("Evolved structure of language shows lineage-specific trends in 
> word-order universals", Nature, 473:79-82, by M. Dunn, S.J. Greenhill, 
> S. C. Levinson & R.D. Gray) that made some interesting claims about 
> the cross-language distribution of word-order universals (henceforth 
> "Greenberg correlations"). David asked me to comment on the paper, 
> which is not all that easy to interpret--primarily because of 
> methodology and  terminology imported from quantitative evolutionary 
> biology. However, since one of the co-authors is a well-known & 
> thoughtful linguist (Steve Levinson, MPI-Nijmegen), I thought that the 
> effort might be worth while.  I am still not sure I understand the 
> paper's conclusions correctly. But I see, tentatively, a way of 
> interpreting them that would make sense.
>     The paper notes first that the "standard" functional-cognitive 
> explanation of Greenberg's correlation did not pan out, be they 
> Lehmann's "harmony", Vennemann's "operator-operand",  or their formal 
> equivalents (X-bar, GB parameters). Alas ignoring a well-established 
> alternative explanation (see below), the paper then shows that 
> statistically, word-order-cum-morphology correlations are 
> lineage-specific, i.e. family-specific. Using data from four 
> families--Indo-European, Austronesian Bantu (a sub-family of 
> Niger-Congo) and Uto-Aztecan, the paper concludes that only within  
> historically-related groups or sub-groups can one find predictable 
> "Greenberg correlations". The conclusion the authors draw is that 
> "Greenberg correlations" are not universal, but depend on "cultural 
> evolution". Or, de-jargonized, that languages that share more of their 
> diachronic history also share more of their "Greenberg correlations".
>     For the past 40 years (Givon 1971, 1974, 1979 chs 5-6-7, 2001 ch. 
> 5, 2009 chs 3-4-5), and following the illustrious tradition of F. 
> Bopp, H. Paul and A. Meillet, I have attempted, apparently in vain, to 
> convince y'all that word-order-cum-morphology "Greenberg correlations" 
> are the direct product of diachronic pathways of grammaticalization. 
> And that apparent exception to those correlations are due to two major 
> factors: (a) the existence of alternative grammaticalization patterns 
> for the same construction or morpheme; and (b) word-order change that 
> leaves recalcitrant old morphology "harmonized" with the old 
> word-order,  thus "incompatible" with the current word-order. The 
> overall conclusion is that synchronic typology is the direct and 
> straight-forward product of diachrony, and that typological universals 
> are mediated by diachrony (as well as, to a lesser extent, by 
> acquisition and evolution).
>     Of course, it may well be that I have misinterpreted the thrust of 
> the Nature  paper altogether, but if it means anything coherent  to 
> me, then it simply re-states well-know diachronic observations.
>
>
>

-- 
Martin Haspelmath (haspelmath at eva.mpg.de)
Max-Planck-Institut fuer evolutionaere Anthropologie, Deutscher Platz 6	
D-04103 Leipzig
Tel. (MPI) +49-341-3550 307, (priv.) +49-341-980 1616