recent paper (Dunn et al. in Nature)

[Tom Givon]

Brian-- My guess would be that in IE the adjectival paradigm was derived 
from the genitive paradigm, so the original connection was from OV to 
genitive via VP nominalization, and only then extention from GEN to ADJ 
by analogy. I cannot speak to all UA languages, but for the one I know 
well (Ute) the connection of OV to Gnitive via VP nominalization is the 
same as in IE; but the adjectival paradigm developed through two 
independent channels, one with the verb 'have' ("(he) who has 
whiteness"; with 'whiteness' originally probably a concrete noun that 
has the typical color), and the other directly as a REL-clause without 
the verb 'have' ("(he) who is big"). And--while Genitives are 
pre-nominal (like pre-verbal objects), adjectives are post-nominal, like 
REL clauses. So the position of both GEN and ADJ in the NP is fully 
predicted from their diachronic source. TG

=====================


On 7/11/2011 3:03 PM, Brian MacWhinney wrote:
> Tom, Esa, and Florian,
>
>       I'm having some trouble matching up the email commentary on Dunn et al. with the specific findings of that article.  They show, for example a really tight linkage of genitive-noun order with adjective-noun order for Indo-European and no such linkage at all for Uto-Aztecan.  There are perhaps 10 other lineage specific findings, each of them rather interesting.  To explain this, they invoke "lineage-specific processes" which is, of course, just a restatement of the findings.  Then they attribute this to "cultural evolution".  I have no idea what they might be talking about here.  The movement from hunter-gatherer to feudal societies?
>     I totally agree with Florian that psycholinguistic forces may be at play, with Tom that diachronic pathways may be at play, and with Esa that analogy may be at play.  But can anyone get a bit more specific?  To take a relatively easy one, what is it in Indo-European that links the adjective-noun order to the genitive-noun order.  I would guess it is the presence in Indo-European of modifier-head number-gender agreement along with fusional case-marking, right?  And I assume that this just doesn't hold in Uto-Aztecan, right?  Can any of your folks work out some of this for the less typologically-well-versed of us?  They refer specifically to eight lineage-specific dependencies.  Can each of these be given similar accounts and when and where do we also need to invoke accounts based on learning,  processing, and analogy.
>
> -- Brian MacWhinney
>
> On Jul 11, 2011, at 7:46 AM, Tom Givon wrote:
>
>>
>> I think Florian makes some very good points, and I am looking forward to reading the actual studies he cited. My concern with our (all of us's) functional theories is that they are usually observations about the final (synchronic) product of the protracted diachronic process(es) that create grammatical structures. From my perspective, we need to look at how functional-adaptive factors operate during the process itself. In other words, we need to find a way of studying the mechanisms (and exact loci) where universal principles exert their influence on emerging structures. As we operate now, a lot of our functional observation are both ad-hoc&  post-hoc. This of course reminds me of the "iconicity era" of the 1980s, when we were busy observing that the resultant emerging structures were "iconic", but paid no attention to the biological processes via which such iconicity arose. So for those of you who would like to consider themselves "cognitive", this is, to my mind, the real challenge. And obviously, experimental studies of on-line behavior are a big chunk of trying to understand the mechanisms of emergence. TG
>>
>> ==================
>>
>> On 7/9/2011 1:47 PM, T. Florian Jaeger wrote:
>>> Hi,
>>>
>>> I share Martin's view that the hypothesis that 'universals' are the direct
>>> product of diachronic pathways is absolutely compatible with both functional
>>> and non-functional explanations (by the Croft et al, submitted to LT
>>> discusses the Dunn et al paper and what they do and do not show in detail).
>>> There is now a growing body of work that investigates this claim more
>>> directly by looking at biases operating during the acquisition of artificial
>>> grammars. This work has revealed strong biases to 'regularize' (reduce the
>>> conditional entropy of morphological or word order alternations, e.g.
>>> Hudsan-Kam and Newport, 2005, 2009; Kirby et al. 2008; Smith and Wonnacut,
>>> 2010; Gutman, 2011). This work has replicated Greenbergian universals in the
>>> lab, although most universals remain untested in this terminology
>>> (e.g. Christiansen, 2000; Culbertson and Smolensky, forthcoming a, b; Tily
>>> et al., 2011). In addition and more recently, this work has also directly
>>> addressed whether considerations about processing or communication (not
>>> quite the same) affect the acquisition of word order and case-marking
>>> systems (Fedzechkina et al., 2011, forthcoming). This work is summarized in
>>> a very short commentary on Dunn et al.'s article that Harry Tily and I
>>> submitted to LT (see link below).
>>>
>>> This line of research is beginning to explore the link between acquisition
>>> and diachronic pathways (e.g. via iterated artificial language learning).
>>> Both functional and non-functional explanations for changes are being
>>> explored.
>>>
>>> I also wanted to add that, in addition to Dryer's and Hawkins's
>>> processing-based accounts, there are now also information theoretic accounts
>>> that make predictions about the development of word order (and other)
>>> alternations based on considerations about efficient and robust information
>>> transfer (cf. Shannon, 1948). These formal accounts can be seen as quite
>>> similar to some hypotheses mentioned in your [Tom's] work). See for example,
>>> Maurits et al (2010-NIPS,
>>> http://www.psychology.adelaide.edu.au/personalpages/staff/amyperfors/papers/mauritsetal10nips-wordorderuid.pdf).
>>> These accounts test the predictions of a framework laid out in Genzel and
>>> Charniak (2002), Aylett and Turk (2004), Jaeger (2006, 2010) and Levy and
>>> Jaeger (2007). In this work, choices in production are linked to
>>> considerations about efficient and robust communication through a noisy
>>> channel. Most of this work has focused on reduction phenomena (incl.
>>> relativizer and complementizers omission, contraction of auxiliaries,
>>> phonetic reduction, argument omission, prononominalization, etc.; for a
>>> overview and references, see Jaeger, 2010,
>>> http://linkinghub.elsevier.com/retrieve/pii/S0010028510000083), but Maurits
>>> et al provide the first extension to word order choices (see also Gallo
>>> (2011,
>>> https://urresearch.rochester.edu/fileDownloadForInstitutionalItem.action?itemId=13759&itemFileId=31899)
>>> for the same principle at work beyond intra-clausal planning.
>>>
>>> Florian
>>>
>>>
>>> Links to papers that I have links to are given below. The first paper
>>> contains all references mentioned above:
>>>
>>> Tily and Jaeger (submitted commentary on Dunn et al):
>>> http://rochester.academia.edu/tiflo/Papers/674181/Tily_H._and_Jaeger_T.F._submitted._Complementing_quantitative_typology_with_behavioral_approaches_Evidence_for_typological_universals
>>>
>>> Fedzechkina et al (2011):
>>> http://rochester.academia.edu/tiflo/Papers/674181/Tily_H._and_Jaeger_T.F._submitted._Complementing_quantitative_typology_with_behavioral_approaches_Evidence_for_typological_universals
>>>
>>> Tily et al (2011):
>>> http://rochester.academia.edu/tiflo/Papers/674181/Tily_H._and_Jaeger_T.F._submitted._Complementing_quantitative_typology_with_behavioral_approaches_Evidence_for_typological_universals
>>
>