Approaching language levels and registers in written Chinese with the Menzerath–Altmann Law

Abstract

Abstract The length of language units, such as word length or sentence length, plays a critical role in register classification studies. However, in this line of work, little attention has been paid to the relationship between the lengths of language units at different levels. The Menzerath–Altmann law (MAL) reflects the functional relationship between the lengths of linguistic units at different levels, and its parameters were shown to be register-sensitive. This article focuses on two interrelated questions based on the MAL: (1) whether there are variations in the hierarchical relationships between language units at different levels and (2) whether such variations will influence register classifications. The results based on written Chinese show that (1) the MAL fittings at the ‘sentence > clause > word’ levels outperform that at the ‘clause > word > character’ levels and (2) the classifications based on two registers, i.e., Press (reportage) and Science (academic prose), demonstrate that the fitting parameters at the ‘sentence > clause > word’ levels also outperform those at the ‘clause > word > character’ levels. These indicate that the variations of hierarchical relationships between language units at different levels should be considered in register analysis. Further interpretations were given from perspectives of the information-theoretic principle and language evolution.