'substantial surprise'

[Tom Givon]

Dear Esa,

I don't have the original Evans/Levinson BBS article, which I will try 
to obtain ASAP. So the following is based on the assumption you quoted 
them--and yourself-- accurately.

I must say that these picky laments about (the demise of) universals go 
back to (at least) Bloomfield, and are really dependent on apriori 
definitions, or a philosophical stance. The extreme Bloomfieldian 
position assumes that universals must be concrete SURFACE  FEATURES of 
language, rather than CONTROLLING/EXPLANATORY PRINCIPLES that govern the 
distribution of surface features. The extreme Chomsakian position makes 
the facts themselves so abstract and so formal (as against 
substantive/functional) that "universals" are observed--really, 
posited--virtually by fiat. Still, there is one useful element one could 
extract from Chomsky's extremism: The insistence that universal are not 
just concrete surface features that must appear in all languages, but 
rather controlling "principles & parameters". This is fully in line with 
Carnap's, in my view apt, distinction between "factual generalizations" 
and "theoretical statements". So far so good. The problem is, of course, 
that Chomsky's P&Ps are abstract, non-substantive and divorced from 
real  (Carnapian) factual generalizations. Above all, Chomskian 
universals are of  EXTANT STATE states rather than DEVELOPMENT.

It would be useful to point out that there is a middle-ground, 
biologically-based alternative to our view of universals, a middle 
ground that has existed at least since H. Paul, and has been espoused, 
at least in late life, by Joe Greenberg: That universals are NOT 
concrete surface features ("all living beings have cells"; "all 
languages have verbs"), but rather more general principles that underlie 
the phylogeny and ontogeny--and for language, the diachrony--of the 
distribution of all surface forms. Evolutionary and molecular biology 
are constructed to  elucidate those principles in biology. In biology, 
these universals of development account for the distribution of surface 
forms. In linguistics, in addition to evolution & child language 
development, the most powerful developmental process that explains the 
distribution of  concrete surface forms is DIACHRONY. So talking about 
the distribution of Tamil forms without understanding the diachrony of 
those forms is, to my mind,  somewhat self-defeating.

One more philosophical/definistional element in this discussion is the 
notion that universals must be 100%, either or. A single exception 
throws out a strong statistical tendency (say 99.9%). This again flies 
in the face of what is known about biologically-based  complex systems, 
where most surface-feature generalizations are strong statistical 
tendencies rather than 100%. This is so because biological systems & 
behaviors are adaptive compromises between conflicting  tho equally 
valid general ('universal') adaptive principles.  As Ernst Mayr observed 
long ago, only physics & mathematics have exceptionless  laws. What is 
more, there is usually more than one possible solution to an adaptive 
imperative, and related species may adopt different solutions, each 
having both advantages & drawbacks. Language typology reveals exactly 
that--a range of possible ways of solving the same 
adaptive-communicative task, each having adaptive advantages and 
disadvantages.

Again, what is universal is/are not 100% generalizations about surface 
features, but the controlling principles of the developmental processes 
that create complex systems. Diachronic change behaves, in this regard, 
very much like biological evolution (see ch. 3 of my 2008 book "The 
Genesis of Syntactic Complexity"), and produces seemingly-conflicting, 
oft puzzling surface facts. Speaking of old books, ch. 6 of my "On 
Understanding Grammar" (1979) "Where does crazy syntax comes from?" is 
dedicated to showing how diachrony can often produced counter-universal 
surface features. My 2000 paper "Internal reconstruction:" As method, as 
theory" (in a TSL volume edited by Spike Gildea) discusses a monumental 
case of seeming violations of well-known synchronic generalization in 
the Athabaskan (Tolowa) verb complex. Bern Heine has also some recent 
discussion on the subject.

In sum, I think the position you attribute to Evans/Levinson is a 
regrettable throwback to non-biologically- based, non-process-based 
approach to "surface universals" that has been dogging linguistics for a 
long time. For as long as functionally-oriented linguists do not strive 
to produce, in Carnap's words, "theoretical generalizations" that are 
explanatory rather than mere summaries of the fact (thus really Carnap's 
"factual generalizations"), I am afraid we will continue to cede the 
field of  "theory" to the Chomskians, who have of course botched it 
royally, for well-understood reasons. And as long as the only 
alternatives availably to us are the two extreme positions of Bloomfield 
and Chomsky, a real understanding of universals will remain a 
tantalizing mirage.

Best,  TG

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


Esa Itkonen wrote:
> Dear Funknetters: Nicholas Evans and Stephen Levinson have written an important article 'The myth of language universals' (Behavioral and Brain Sciences). Their thesis is that our knowledge of linguistic diversity is far from complete. Even the most self-evident generalizations may be, and are, falsified. On two occasions they mention the fact that one way to express sentence negation in Ancient Tamil is by means of zero (i.e. lack of tense marker). In brief, "almost every new language description still guarantees substantial surprises".
>
> In fact, it need not even be a NEW language description which produces a substantial surprise, it can also be an old but neglected one. Again, interestingly, we have to do with Ancient Tamil. Proto-Tamil (from which Ancient Tamil had descended) is assumed to have had at least four inflecting cases, Ancient Tamil had seven (or eight if vocative counts too), and Modern Tamil has eight (if benefactive in N-DAT-aaka does not count). Against this rather unexceptional background it is rather surprising to learn that case-endings could be, and were, either interchanged or simply dropped in Ancient Tamil. There can be no doubt about this fact. It is directly documented by all existing texts; it is confirmed by the (two thousand years old) grammar Tolkaappiyam; and it is further confirmed by such living authorities on Ancient Tamil as Thomas Lehmann and Asko Parpola. This phenomenon of suffix-dropping extends to non-finite verbs as well, with the result that a typical sentence is jus
> t a string of uninflected roots, with the final word (= finite verb) as the only inflecting one.
>
> How can this phenomenon be explained? Is it due to the fact that all existing texts are poems? When asked, Lehmann replied: "This could be the answer", while Parpola shrugged: "Nobody knows."
>
> In any case, this phenomenon is surely interesting enough to be more widely known (or so I naively thought). Hence, I wrote a paper (in 2003) and sent it to a typological journal. It was rejected by two referees both of whom declared the phenomenon in question to be impossible.(And, believe it or not, one of the referees ALSO claimed it to be thoroughly common, and even one that occurs in the author's, i.e. my, native language, i.e. Finnish, which borders on insanity.)
>
> This was not the first time that my ideology has clashed with that of referees. I do not see the value of repeating what has been said hundreds of times before; rather, I see the value of inventing/discovering something new. 90% of the referees with whom I have dealt with during the last 35 years or so, hold the opposite view. This is why, if my contribution has been accepted at all, it has more often than not been accepted by the editor and contary to the referees' opinion.
>
> To sum up, this phenomenon is real; it provides a "genuine surprise" à la Evans & Levinson; and it can be read on my home-page (click below) under the title "A case system with interchangeable and optional endings" (2003).
>
> Esa 
>
> Homepage: http://users.utu.fi/eitkonen
>
>