The Menzerath-Altmann law in syntactic structure revisited

According to the Menzerath-Altmann law, there is inverse proportionality between sizes of language units and their constituents (i.e., longer language units are composed of shorter constituents, and vice versa). The validity of the law was confirmed many times for the relation between lengths of a word and its syllables. However, the relation between lengths of sentences (measured in clauses) and clauses (measured in words) is problematic. In this paper, a new language unit – linear dependency segment – is introduced with the motivation to avoid some problems connected to the Menzerath-Altmann law on the syntactic level. The new unit is intermediate between clause and word and its definition takes into account both the linearity of language and dependency syntactic structure. It is shown that the relation between sentence length in clauses and clause length measured in linear dependency segments abides by the Menzerath-Altmann law in two Czech dependency treebanks.