The effects of type and token frequency on word length: a cross-linguistic study

Abstract Inspired by Zipf’s Law of Abbreviation, previous research was mostly directed at the interaction of word length and token frequency. Much less is known about the relationship of word length and type frequency, let alone about the differential impact of type and token frequency on word length. These issues are examined on the basis of a non-representative sample of 10 languages. The token frequency analysis reveals that 8 of the 10 languages show a monotonic decrease in frequency with increasing length while 2 languages reveal a unimodal distribution. By contrast, all 10 languages exhibit a rise followed by a monotonic drop of the frequency curve in the type frequency analysis. There appears to be a notable effect of type frequency on the nature of the token frequency distribution: the greater the average length of the words in the lexicon, the higher the probability of a unimodal distribution. Two principles are required to account for these results—a general dispreference for using long words and a language-particular dispreference for short words in the lexicon.