Dominant words rise to the top by positive frequency-dependent selection

Mark Pagel*, Mark Beaumont, Andrew Meade, Annemarie Verkerk, Andreea Calude

*Corresponding author for this work

Research output: Contribution to journalArticle (Academic Journal)peer-review

10 Citations (Scopus)


A puzzle of language is how speakers come to use the same words for particular meanings, given that there are often many competing alternatives (e.g., “sofa,” “couch,” “settee”), and there is seldom a necessary connection between a word and its meaning. The well-known process of random drift—roughly corresponding in this context to “say what you hear”—can cause the frequencies of alternative words to fluctuate over time, and it is even possible for one of the words to replace all others, without any form of selection being involved. However, is drift alone an adequate explanation of a shared vocabulary? Darwin thought not. Here, we apply models of neutral drift, directional selection, and positive frequency-dependent selection to explain over 417,000 word-use choices for 418 meanings in two natural populations of speakers. We find that neutral drift does not in general explain word use. Instead, some form of selection governs word choice in over 91% of the meanings we studied. In cases where one word dominates all others for a particular meaning—such as is typical of the words in the core lexicon of a language—word choice is guided by positive frequency-dependent selection—a bias that makes speakers disproportionately likely to use the words that most others use. This bias grants an increasing advantage to the common form as it becomes more popular and provides a mechanism to explain how a shared vocabulary can spontaneously self-organize and then be maintained for centuries or even millennia, despite new words continually entering the lexicon.

Original languageEnglish
Pages (from-to)7397-7402
Number of pages6
JournalProceedings of the National Academy of Sciences of the United States of America
Issue number15
Publication statusPublished - 9 Apr 2019


  • Approximate Bayesian computation
  • Frequency-dependent selection
  • Language evolution
  • Neutral drift
  • Shared vocabulary


Dive into the research topics of 'Dominant words rise to the top by positive frequency-dependent selection'. Together they form a unique fingerprint.

Cite this