Towards a Brain Compatible Theory of Syntax Based on Local Testability
Valentino Braitenberg, Stefano Crespi Reghizzi · Topics in computer mathematics · 2003
Chomsky’s theory of syntax came after criticism of probabilistic associative models of word order in sentences. Immediate constituent structures are plausible but their description by generative grammars has met with difficulties. The type 2 (context-free) grammars account for constituent structure, but already trespass the mathematical capacity required by language, because they generate unnatural mathematical sets: a consequence of being based on recursive function theory. Abstract associative models investigated by formal language theoreticians (Schutzenberger, McNaughton, Papert, Brzozowsky, Simon) are known as locally testable models. A combination of locally testable and constituent structure models is proposed under the name of Associative Language Description, arguing that it equals type 2 grammars in explanatory adequacy, yet is compatible with brain models. Two versions of ALD are exemplified and discussed: one based on modulation, the other on pattern rules. A sketch of brain organization in terms of cell assemblies and synfire chains concludes.