Some preliminary remarks on a ‘ weak ’ theory of linearization

One traditional research topic in syntax is the question of linearization of hierarchical structures. Extensive research stemming from the seminal work by Noam Chomsky in the fifties has shown that natural languages have a hierarchical organization, but, since utterances take place in time and not instantaneously, hierarchical structures must be mapped in a linear sequence of words. Linearization can be seen as a process in which a bi-dimensional structure (as illustrated by the traditional branching trees) is conflated in a mono-dimensional one (a linear sequence). Clearly, this process is rule-governed. It is not the case that anything goes, probably because the linearized structure is the input for language acquisition, namely a child has to infer some properties of his/her own language based on a linear input. Generative grammar has systematically dealt with the question of linearization at least from the seventies, when what would get the name of X-bar theory was first elaborated. X-bar theory claims that the each phrase adheres to a universal format according to which the Head (a lexical item) can be attached to a Complement (itself a phrase). To the unit formed by Head and Complement, a further phrase, called Specifier, is attached (adjuncts can further complete the structure). According to a more or less standard version of X-bar theory, the only significant point of systematic cross-linguistic variation in phrase structure is the linear order between Head and Complement, namely languages would come in two big varieties: the [Spec [Head-Compl] ] type, illustrated by Italian and English, and the [Spec [Compl-Head] ], illustrated by Japanese and Turkish. According to this view, linear order is directly encoded in syntax and a child acquiring a language must set the value of the Head parameter for his/her target language. As for Specifiers, the question was less extensively investigated, but at least prototypical Specifiers created by a movement operation (say, Spec,CP or Spec,IP) seemed to be universally left branching, so no Spec Parameter, akin to the Head Parameter, was part of the standard syntactic toolbox. The picture that I have roughly summarized, which was a point of contact between generative theory and the typological tradition stemming from Greenberg’s work (cf. Greenberg [1966]),