Vol. 22:2 (2023) ► pp.154–188
Evidence of Zipfian distributions in three sign languages
One striking commonality between languages is their Zipfian distributions: A power-law distribution of word frequency. This distribution is found across languages, speech genres, and within different parts of speech. The recurrence of such distributions is thought to reflect cognitive and/or communicative pressures and to facilitate language learning. However, research on Zipfian distributions has mostly been limited to spoken languages. In this study, we ask whether Zipfian distributions are also found across signed languages, as expected if they reflect a universal property of human language. We find that sign frequencies and ranks in three sign language corpora (BSL, DGS and NGT) show a Zipfian relationship, similar to that found in spoken languages. These findings highlight the commonalities between spoken and signed languages, add to our understanding of the use of signs, and show the prevalence of Zipfian distributions across language modalities, supporting the idea that they facilitate language learning and communication.
Article outline
- Introduction
- Lexical frequency in sign languages
- Method
- Corpora
- The British Sign Language (BSL) corpus project
- The DGS-korpus project
- Corpus NGT
- Coding
- ELAN
- Exclusion and inclusion criteria for sign categories
- Sign categories included
- Fully lexical signs (core lexicon)
- Depicting constructions
- Pointing signs
- Buoys
- Gestures
- Excluded sign categories
- Uncertain signs
- Mouthing
- Extra-linguistic manual activity
- Fingerspelling
- Names
- Cued speech and initializations
- Collapsing over specific tokens within a sign type
- Creating a frequency distribution and assessing the fit to a “Zipfian” one
- Corpora
- Results
- What do the most frequent signs look like across the three corpora?
- Discussion
- Conclusions
- Acknowledgements
- Data and code availability
- Notes
-
References