The coding of target alignment and scaling in pitch accent transcription

This paper provides data and discussion on the need for teasing apart information that is to be considered part of a phonological transcription of pitch accents and properties that may be coded at a phonetic level. The discussion stems from the study of both production and perception data on the variety of Italian spoken in Pisa, with particular attention to the interplay of alignment and scaling information in the classification of pitch accents. Acoustic measurements and perception tests show that in some cases, in particular for two accents showing the same starred tone (i.e., H*) in Pisa Italian, the scaling appears to be important for distinguishing pitch accents and for identifying them. This information, both phonetically grounded and significant in distinguishing the patterns, could then be considered important for underscoring the different properties of the pitch accents. Conversely, the presence in both patterns of a low tonal target preceding the peak does not appear to interfere with the subjects’ judgments in identifying the two patterns. In the paper it is argued that trying to code all the systematic and significant phonetic properties at the phonological level would lead to a highly complex and redundant system, in which detailed information on scaling could also be included. It will be underscored that in assigning phonological labels, even if they are phonetically transparent, one goal may be to avoid redundancy as far as possible, and to encode only information that distinguishes elements within the system; this would mean that phonetic transcriptions would code all the phonetically relevant characteristics, i.e. the presence of all targets and the significant scaling differences. Within the transcription, then, phonological and phonetic information may be differentiated.