Scaling Laws for Phonotactic Complexity in Spoken English Language Data

Area of Science:

• Linguistics
• Computational Linguistics
• Psycholinguistics

Background:

• Zipf's law and Heaps' law are fundamental statistical principles observed in complex systems, including language.
• Phonotactics, the study of sound sequences in spoken language, has not been extensively analyzed through the lens of these scaling laws.

Purpose of the Study:

• To investigate the applicability of Zipf's law and Heaps' law to phonotactic sequences.
• To determine if boundary-spanning phonotactic sequences influence the relationship between these two statistical laws.
• To assess the goodness-of-fit and scaling coefficients for Zipf's law concerning phonotactic sequences of varying lengths.

Main Methods:

• Analysis of phonotactic sequences of different lengths from the Buckeye corpus of spoken English.
• Inclusion of both within-word and across-word boundary phonotactic sequences.
• Statistical evaluation of Zipf's law and Heaps' law adherence.

Main Results:

• The expected relationship between Zipf's law and Heaps' law in phonotactics is only observed when boundary-spanning sequences are included.
• Zipf's law demonstrates a high goodness-of-fit and a significant scaling coefficient when sequences of more than two sounds are considered.
• Phonotactic cognition appears to utilize information from boundary-spanning sound sequences.

Conclusions:

• The study confirms the relevance of Zipf's law and Heaps' law to phonotactic analysis.
• Incorporating boundary-spanning sequences is crucial for accurately modeling phonotactic statistical properties.
• Findings suggest that cognitive processes for phonotactics involve processing sound sequences that extend beyond individual word boundaries.