networks & systems laboratory> research> current projects> a computational evolutionary approach to colour term typology

A Computational Evolutionary Approach to Colour Term Typology
Smart Internet Technology Research Group

Introduction

- Colour Terms vary between languages.

- Some languages have only 2 basic colour terms – others 11.

- But the variation is far from random – so what causes these patterns?

Cross-Linguistic Evidence

- Berlin and Kay (1969) – languages evolve from simple to complex, and gradually add colour terms over time.

- The best examples of each colour term fall into clusters, both for different speakers of the same language, and for speakers of different languages.

- The World Colour Survey looked at 110 languages – and found that most languages lie somewhere along the evolutionary trajectory shown below (Kay and Maffi, 1999). The top of the hierarchy represents a language with only two color terms such as Dani. From this starting point, colour term systems can evolve until each unique hue is represented by a separate colour term.

- Purple, pink and brown are less predictable. Purple usually but not always emerges before orange.

- We never see Turquoise or Lime basic terms.

Explaining Language Typology

- What causes this kind of typological pattern?

- Individual psychology?

- Properties of the mind/brain constrain the range of possible human languages.

- Or an interaction of psychological and social processes?

- Colour terms systems may have developed as the result of a process of cultural evolution.

References

- Berlin, B. & Kay, P. (1969). Basic Colour Terms. Berkeley, CA: University of California Press.

- Chomsky, N. (1972). Language and Mind. New York, NY: Harcourt Brace Jovanovich Inc.

- Hurford, J. R. (1987). Language and Number The Emergence of a Cognitive System. New York, NY: Basil Blackwell.

- Kay, P. & Maffi, L. (1999). Colour Appearance and the Emergence and Evolution of Basic Colour Lexicons. American Anthropologist, Volume 101, pages 743-760.

An Expression-Induction Model
Ten artificial people were created.
They could learn colour word denotations by observing other artificial people talking.
The learning mechanism was a form of Bayesian inference, and the red, yellow, green and blue unique hues were made especially salient.
Conversations consisting only of naming colours were simulated over twenty generations of speakers, in 425 separate simulations.

Simulated Languages Show the Same Patterns seen in Real Languages

- Most of the colour terms which emerge in the simulations are of the same type as those in Kay and Maffi’s (1999) evolutionary trajectory.

- The correlation between the results of the World Colour Survey and the evolutionary model is 0.959 (Pearson’s product moment coefficient, P«0.01).

- The emergent colour term systems as a whole tend to be of the same types as those found in the World Colour Survey.

Contacts

Mike Dowman
Associate Professor Judy Kay

*This research is sponsored by the University of Sydney and the Australian Government through IPA and IPRS scholarships.
 
University of SydneyDesigned by eliu