Model fits are the Aikake information criteria of models predicting the linguistic feature from just the language family, from population alone, and from the three demographic variables, respectively.

Smaller values indicate better fits. § = Demographics and geographic location predict typology better than geographic location alone (χ2 model comparison (p<.05). $ = Predictive power of population is reduced (significantly larger residual deviations) by randomly shuffling languages within their families. Indicates that reported effects generalize to within language families. ** = Reported pattern is significant (p between 0.05 and 10–11) after controlling for language family. ◊ = Pattern no longer significant (p≥.05) after controlling for language family. 1 = Area and Number of Neighbors are significant predictors controlling for population. . = Consistent with the pattern reported, but not significant. x = Pattern after controlling for geographic covariates is non-significantly inconsistent with the pattern observed without controlling for geographic location.