Abstract
Languages are more likely to have lexical items for some concepts (e.g. CHILD and SUN) than others (e.g. PARENT and WEATHER). We present a large-scale quantitative analysis of factors that predict lexicalization across languages, and test in particular the hypothesis that usage frequency predicts lexicalization. We consider 854 concepts and demonstrate that average usage frequency across 46 languages from the Language Documentation Reference Corpus is a relatively good predictor of the typological prevalence of lexicalization across the 41 languages in the World Loanword Database (β = 0.25, 95% CI = [0.16, 0.33]) and across 262 languages from the Intercontinental Dictionary Series (β = 0.30, 95% CI = [0.23, 0.38]). This finding implies that cross-linguistic regularities in lexicalization can be attributed in part to shared usage patterns across cultures.