Zipf's Law Extended to Small Samples of Adult Speech

Abstract

Zipf's law relates the frequencies of words found in speech samples to the numbers of different words at each frequency. The present study extended this law to very small samples (400 words) of both written and oral speech. A secondary purpose was to investigate whether oral and written samples from the same speakers deviate from Zipf's law in systematic ways, allowing correct identifications of speakers across the two modalities. Zipf's law was strongly confirmed, but systematic deviations from the generalization proved difficult to demonstrate.