New Papers from Late 2015
New papers which I already announced on the website, but for which I only uploaded my drafts, are now officially available. The first is a paper on Chinese dialects in which I test the minimal lateral network (MLN) approach (List et al. 2014a) on a set of 23 Chinese dialect varieties. In contrast to previous studies (List et al. 2014b on MLNs, I test the potential of minimal lateral networks to reconstruct ancestral states, that is, to infer which of several cognate sets reflecting one meaning slot was used in teh ancestral language to express the concept. It turns out, that the methods generally do not work very well. However, the use of the reference phylogeny and the use of the MLN criterion enhance the analysis. It seems that the main problem for this particular failure of the MLN method for the task of ancestral state reconstruction is the binary representation of cognate sets, where a multi-state representation would make much more sense. It is further questionable whether the "vocabulary size distribution" is really a useful criterion for the selection of the weight-ratio in parsimony approaches.
The second paper is the result of a contribution with Gerhard Jäger (University Tübingen), and which we presented at the QITL-6 conference in Tübingen (in fact, Gerhard presented it, since I was on another conference by then). The paper investigates the potential of automatic approaches to cognate detection and phonetic alignment for the purpose of family tree reconstruction. The results lag behind alternative methods which are often much simpler regarding their workflow, yet it has the advantage of providing characters which can be employed by character-based methods like parsimony, Maximum Likelihood or Bayesian Inference.