A simultaneous-equation regression model of language complexity trade-offs

In this paper we develop a statistical model of language complexity trade-offs using four typological measures (related to phonology, morphology, syntax and lexicon). The data come from the 100-language sample that appears in the World Atlas of Language Structures (WALS), and the trade-offs are calculated using different types of correlation coefficients. All those coefficients are statistically insignificant when they are computed using a standard (product-moment) methodology, but they become significant when we use simultaneous-equation regression methods, especially the ones based on seemingly unrelated regressions (SUR) and three-stage least squares (3SLS). These results are related to ideas suggested in the theoretical literature, especially in the one about language as a complex adaptive system.