Global Cocktail Parties and an Arms-Race in Language Evolution

Thomas HillsUniversity of Warwick
James AdelmanUniversity of Warwick


Many species change their signals in response to crowding, consistent with information theoretic accounts of signalling in the presence of noise. In this article we explore the hypothesis that the statistical structure of language has evolved to enhance reception and processing in response to competition in the language market. In particular, we investigate how concreteness has changed over the last 200 years. We take a big data approach, combining large text corpora (including the Google Ngram corpus, the Corpus of Historical American English, and presidential speeches) with the recent collection of concreteness norms for over 40,000 English words (Brysbaert et al., 2013). Across corpora we find that concreteness has steadily increased since the 1800s. This takes place both within and across word classes, indicating that the rise in concreteness is systemic and not limited to changes in grammar. By comparing recent concreteness norms with older norms, we show that the observed changes in concreteness are not due to a bleaching effect caused by the loss of concreteness as words age. We further investigate how the statistical distribution of other properties of words change in a way that may indicate that language is becoming more distinctive. We discuss a number of potential explanations and implications of these changes, including changes in literacy, gender, the Flynn effect, and the influence of competition in the marketplace of ideas.


Global Cocktail Parties and an Arms-Race in Language Evolution (181 KB)

Back to Table of Contents