On the Distributional Analysis of Pitch Phenomena

Both the present analytical techniques and methods of structuring pitch data in language are not entirely satisfactory. Neither traditional procedures, essentially aimed at didactical applications nor the more theoretical work done on transformational-generative lines have reached sufficient depth of insight or fully accounted for available data. For general methodological reasons it is preferable to try to elaborate a distributional theory of linguistic “tone” and intonation rather than develop further the existing TG-type of analysis. By means of examples taken from Swedish and English an attempt is made to show how strictly distributional criteria allowing no previous semantic or grammatical information may be employed to define the difference between tone and non-tone languages and to structure pitch data in a non-tone language. Methods of empirically testing hypotheses, including instrumental analysis and psychophonetic experiments, are suggested.