Identifying graphematic units

Abstract

It is a matter of debate how far the description of a writing system should be based on the units and categories of the respective spoken language. The present paper pursues the idea of relative autonomy: accordingly, writing systems should be based on as little phonological information as possible. Otherwise, existing structures may be superimposed by structures from the spoken language and not be discovered. As a necessary step in this direction, the present paper proposes a procedure to identify vowel and consonant letters across languages without reverting to phonology. This is achieved by making use of the different distribution of vowel and consonant letters. The proposed identification procedure is shown to work for English, Dutch, and German. Keywords: letter distribution; written minimal pairs; multi-dimensional scaling