| Abstract | [Introduction]
This paper is a progress report on part of a biologically-based approach to the understanding of spoken language presently in preparation (Lindblom, MacNeilage, & Studdert-Kennedy, 1984a). Our aim is to view language in the perspective of the evolution of adaptive functions. Our strategy is to begin by focusing on aspects of language which are closest to the transmission process, namely, the production and perception of the sound pattern of language; in linguistic terms, the phonological level. Our rationale is partly practical-these more peripheral aspects are closer to direct observation than more central aspects associated with meaning and thought. But it is also partly theoretical-we believe the constraints of the transmission process have played a crucial role in molding the form of language functions as a whole.
The topic of this paper is the serial organization of language output-the process whereby some intention, which is itself not serially organized, is converted into a rule governed sequence of linguistic symbols. If we look directly at the result of this serial organization process-if we immediately encounter the central paradox of speech research. We find that the parts of this output that are supposed to signal the string of consonants and vowels given us by the linguist are neither context-free, nor marked off discretely from segment to segment. The context-sensitive representation of a given consonant or vowel in the transmission process is termed the Invariance problem. The absence of obvious temporal boundaries to the representation of segments is termed the Segmentation problem (see Lindbolm, 1982). Collectively these two problems constitute what we can call the Nonisomorphism Paradox. |
|---|
