Algorithmic generation of random languages argues for syntax as a source of phylogenetic information

The Parametric Comparison Method (PCM, Longobardi and Guardiano 2009) uses syntactic parameters (Chomsky 1981, Baker 2001, Biberauer and Roberts 2012) to study phylogenetic relationships between languages. This method has already successfully generated phylogenies of Indo-European (IE) languages (Longobardi et al. 2013) which are competitive with those produced by the classical comparative method (Durie and Ross 1996), lexicostatistics (Dyen et al. 1992), or Bayesian phylogenetics (Bouckaert et al. 2012). A question raised by the PCM framework, and indeed by all these methods, is whether their conclusions about language relatedness are secure against chance similarities between languages. As far as the PCM goes, using a randomly simulated distribution of parametric distances between languages (which are defined to range between 0 and 1), it is possible to perform statistical tests of the hypothesis that the distances observed in the real world are unlikely to arise by chance, and thus to motivate judgments of relatedness based on syntax. Bortolussi et al. (B+; 2011) have proposed an algorithm to enumerate the possible languages defined by a system of heavily interdependent parameters and to sample randomly from that set. We propose an improvement to this algorithm, thus validating the PCM as a method to formally study the relationship between languages and populations.