Homing in: on arguing for remote representations*

Linguistic analysis possesses both an ‘arbitrary’ and a ‘natural’ component – on the one hand, methodological principles and various means of organizing and handling data; on the other, empirical studies aimed at exposing linguistic universals through the detailed analysis of specific languages, cross-linguistic comparison, phonetic studies, psycholinguistic experiments, observation of language acquisition, and other sources of pertinent data. In practice, the arbitrary and natural components are intertwined, and each makes use of methods and results from outside linguistics in a narrow sense: the arbitrary component contains many principles and methods which are not peculiar to linguistics at all, but are rather the common property of scientific investigation, while the natural component refers ultimately to aspects of mental and social organization and physical properties of the vocal tract, many of which are independent of speci-fically linguistic behaviours and abilities.