Last week I presented a talk to the 8th International Conference on Oceanic Linguistics on the probability of protoforms:

Many papers about language subgrouping make the argument that if a word form is present in two language subgroups, then it probably reflects their common proto-language. This fairly reasonable assumption has taken center-stage in the reconstruction of Proto-Central-Pacific and Proto-Oceanic, and subsequent inferences about these respective societies. In this talk I will introduce a method that calculates the actual probability that a given form reflects the proto-language. This method builds on the linguistic comparative method, and provides a powerful tool for testing and confirming reconstructions in a probabilistic framework.