Séminaire de Recherche en Linguistique

The cartographic project is specifically concerned with the detailed reconstruction of the fine hierarchical structure of the clause and its major phrases (which has to be discovered despite the enormous cross-linguistic variation in surface word order). In my presentation I want to discuss a novel type of evidence for establishing finer-grained hierarchical structures, based on a particular hypothesis concerning cross-linguistic word order variation. The hypothesis is that the attested orders of n elements, which invariably appear to be a subset of the mathematically possible ones, can be derived (without deriving the unattested orders) if only the head of a projection can move in one of the ways we know movement can take place. So, for example, the attested orders of demonstrative, numeral, adjective and noun (14 out of the 24 mathematically possible combinations of the 4 elements) can be derived, without deriving the 10 orders which are not attested as the unique order of the language, if only the head N (or phrases containing the head noun) can move.

Now, if we consider the attested orders of other nominal elements (as, for example, the triplets composed of numeral, classifier, noun; multiplier, base, noun; degree adverb, adjective, noun; measure phrase, adjective, noun), we find that only a subset of the mathematically possible combinations is attested (4 out of 6 – factorial 3). The attested ones again turn out be derivable (without deriving the unattested ones) if only the element that constitutes the head can move. This provides a new way of establishing the correct constituency of these triplets of nominal elements, whose constituent structure is debated in the literature: whether, for example, the constituent structure of numeral, classifier, noun is [numeral [classifier [noun]]] or [[numeral classifier] noun]; that of multiplier, base, noun is [multiplier [base [noun]]] or [[multiplier base] noun].

In turn, the more precise hierarchical structure that we can reach in this way will permit a more restrictive theory of word order variation.