Structure and usage do not explain each other: an analysis of German word-initial clusters

Abstract

The present study focuses on German word-initial consonant clusters and asks whether feature-based phonotactic preferences correlate with patterns of type and token frequencies in present-day usage. The corpus-based analyses are based on a comprehensive list of such clusters, representing current usage, and on a number of feature-based phonotactic preferences. Correlating the variables by means of a correlation analysis and a regression analysis leads to a number of observations relevant to the general topic of featural-segmental structures versus usage. First, out of eighteen correlations between (raw and logarithmic) type and token frequencies, and preferred feature patterns, only one significant correlation was found. Second, a regression analysis led to similar results: out of thirteen variables tested, only two contribute to logarithmic type and token frequencies. Only a limited set of cluster properties investigated in the present paper constitutes a relevant predictor of frequency measures. The study thus demonstrates, in accordance with other recent evidence, that preferred phonetic/phonological structures and their usage frequency constitute two separate domains for which distributions may not have to coincide.

Keywords

• frequency
• german
• phonetic features
• phonotactics
• usage-based phonology