Correlation between Linguistic Measures: An Extended Analysis

Abstract

In this paper we reconsider the results that appeared originally in Coloma (2017) about the possible existence of negative correlation between linguistic measures, using a newly-assembled database of 80 languages for which we have the same text (which is the fable known as “The North Wind and the Sun”). Most conclusions of the original paper become reinforced, especially the ones related to the existence of language complexity trade-offs. This is particularly clear when we look at partial correlation coefficients between three linguistic ratios (phonemes per syllable, syllables per word, and words per clause), especially when we use simultaneous-equation regression methods and instrumental variables.