FIGURE 4.

BEAST tree from the 94 syntactic parameters set in the 69 languages of our sample. The best model that we determined is a Gamma Site Model with Substitution Rate = 1, a Mutation Death Model with death p = 0.1, a Relaxed Clock (Logarithmic) with clock rate = 1, and a uniform Yule model for the birth rate. The Monte Carlo Markov Chain produced 10,000,000 trees, 25% of which were used for the burn-in and discarded for the purpose of the calculation of the consensus tree. The tree is a consensus tree of 7500 different trees sampled through the 7,500,000 trees (with a sample stored every 1000 generated trees) produced by the Monte Carlo procedure.