Is a 'History and Geography of Human Syntax' meaningful?

In addition to its theoretical impact, the development of molecular biology has brought about the possibility
of extraordinary scientific progress in the historical study of classification and geographical distribution of
different species and different human populations
(cf. Cavalli Sforza et al., 1994).
We want to suggest that parametric theories of linguistic variation in generative grammar can prompt analogous
progress in the study of the history and geographical
distribution of different language families.
Thus, this work aims at unifying two traditionally
unrelated lines of investigation:
- the formal study of syntactic variation in generative grammar
- the reconstruction of historical relations among
languages (phylogenetic taxonomy)
The pursuit of this approach will be argued to seriously question the traditional belief in the orthogonality of grammatical typology and language diachrony/genealogy,
enabling us to tentatively suggest a positive answer to the following problem, that we conventionally label ‘Humboldt’s problem’:
Are the typological and the genealogical (phylogenetic)
classifications of languages significantly isomorphous?
From the second half of the 19th century on, three different types of enterprises attempted to classify languages and/or populations separated for centuries or millennia into historically significant families:
A) for relatively shallow time depths and in particularly favored cases, the classical linguistic comparative
method, based on inspection of the lexicon, can
often provide sharp taxonomic conclusions, immune
from the need of serious probabilistic evaluation, since the relatedness hypotheses are warranted by few patently highly improbable phenomena, most notably recurrent (optimally ‘regular’) sound
correspondences. These properties largely solve the problem of a safe choice of comparanda, allowing
for solid conclusions.
B) Beyond the classical one, the only other linguistic method so far proposed is Joseph Greenberg’s (e.g. 1987, 2000) mass comparison; still based on the lexicon, it suggests more far-reaching but much
less rigorous and widely acceptable taxonomic
conclusions, because the very choice of the compared
entities, based on pretheoretical resemblance among arrays of words, is less safe from chance. Obvious
probabilistic questions, often of unmanageable
complexity, arise and receive controversial answers.
C) A third comparative practice stems from a different discipline, population genetics (cf. Cavalli Sforza et al., op. cit.): no question arises here about the
comparability of the basic entities, since they are drawn from a finite and universal list of biological options: a blood group must be compared to the same blood group in another population, obviously, not to other sorts of genetic polymorphisms. The only issue concerns the statistical and empirical
significance of the similarities discovered. This is why population genetics is considered so useful to complement linguistics in the task of classifying
populations and languages.
We propose that the contribution of linguistics proper to such issues can be completely renovated on the grounds
of parametric generative theories, which in principle
allow one to bring modern cognitive science to bear on issues of cultural variation and historical explanation.
Since parameters form a finite and universal list of
discrete biological (though culturally set) options, they resemble the set of polymorphisms studied by population
genetics and potentially enjoy similar (and perhaps even greater) formal advantages, overcoming in principle all questions on the choice of comparanda affecting linguistic
methods based on the vocabulary. On the other side, the a priori probative value of parametric comparison is
mathematically very high: e.g. just 30 binary independent
parameters generate 230 languages = 1,073,741,824. The probability for two languages to coincide in the values of 30 independent parameters with binary equiprobable values = 1/230, of three languages = (1/230)2, i.e. less than one in one billion billions.
We will test and exploit such a potential by establishing exact comparisons of parameter values among some languages whose degree of cognation is independently known, in order to prove the effectiveness of the method to provide historically correct taxonomic insights before applying it to controversial cases.
For the past fifteen years, a number of scholars have
studied the parametric variation of the structure of nominal
phrases in several languages. Relying on this and new specific work and following the MGP method of
Longobardi (2003: relatively many parameters in relatively
many languages in a single limited subdomain), we have worked out a preliminary list of 50 binary parameters
affecting DP-internal syntax and tested their values in over 20 ancient and contemporary varieties drawn from several
Indoeuropean and non-Indoeuropean subfamilies
(including Modern Italian, French, European Portuguese,
Latin, Classical Greek, Modern Greek, Gothic, Old
English, Modern English, German, Bulgarian, Serbo-Croat, Arabic, Hebrew, Hungarian, Wolof, among others). Each
relevant parameter has been tentatively set for such
languages, obtaining up to 50 precise correspondence sets of parameter values, and for every pair of languages we could
arithmetically count identities and differences. In our
formalism, the relative distance between any two languages
is expressed by a coefficient, which consists of an ordered pair of positive integers <i,d>, where i is the number of identities and d the number of differences. This procedure of lexically blind comparison, coupled with especially
designed empirical and statistical methods, will be argued
to prove adequate to generate the essentially correct phylogenetic tree, as resulting from traditional methods of lexical comparison. This is why the answer to
Humboldt’s problem above appears (perhaps surprisingly) tendentially positive.
The elaboration of a compact variation model of a whole syntactic subdomain, even though for the mere purposes of historical taxonomy, begins, in turn, to yield a number
of insights for parameter theory itself from both a
synchronic and a diachronic point of view:
- we will show how parameters, even in a limited area of grammar, appear to be tightly interrelated and not freely orthogonal to each other (cf. Fodor
2000): some parameters become irrelevant (in
different senses, to be technically distinguished) in a language due to the setting of other parameters (or to variation in the composition of the lexicon: Kayne 2003);
- parameters can be evaluated according to their relative historical weight:
1) certain parameters may understate or overstate
identities and differences because they trigger a high number of irrelevant settings in the other parameters.
Their diachronic resetting will be much more
consequential (catastrophic in Lightfoot’s 1999 sense)
than that of others;
2) some parameters appear to be taxonomically more significant than others because they are diachronically more stable. The possibility that this may be due to different degrees of exposure to external triggers and other modules at the interface, along the lines of Keenan’s insight about Inertia, will be explored.
This line of investigation will be construed as a peculiar
way to progress ‘beyond explanatory adequacy’ in
generative grammar (Chomsky 2001), leading to a better
understanding of both classical and new issues in the cultural and natural history of the language faculty.
References
Cavalli Sforza, L. P. Menozzi, and A. Piazza (1994). The History and Geography of Human Genes.
Princeton: Princeton University Press.
Chomsky, N. (2001). Beyond explanatory adequacy.
Eliminating labels. MIT Working Papers in
Linguistics.
Fodor, J. (2000) The Mind doesn’t Work that Way :The Scope and Limits of Computational Psychology. Cambridge: MIT Press.
Greenberg, J. (1987). Language in the Americas.
Stanford: Stanford University Press.
Greenberg, J. (2000). Indo-European and Its Closest Relatives: The Eurasiatic Language Family. Volume I: Grammar. Stanford: Stanford University Press.
Kayne, R. 2003. “Antisymmetry and Japanese”, English Linguistics 20. 1-40.
Lightfoot, D. (1999) The Development of Language:
Acquisition, Change, and Evolution. Oxford:
Blackwell.
Longobardi, G. (2003) Methods in Parametric
Linguistics and Cognitive History. Linguistic Variation
Yearbook 3, pp. 103-140.