Language Evolution and Computation Bibliography

In order to make distinctive speech sounds, it is necessary to control two separate acoustic cavities. There has been a longstanding debate about whether a lowered larynx is essential for this. Lieberman and Crelin have used it as an argument against speech in Neanderthals. This claim is controversial, not only for paleontological reasons, but also because researchers do not agree on the need of a lowered larynx for distinctive speech. Researchers using similar methods (computer modeling) arrive at opposite conclusions. The problem is that one needs to take articulatory and anatomical constraints into account when investigating the acoustic implications of vocal tract morphology.

In order to study what the effect of lowering the larynx is, a reimplementation of Mermelstein s vocal tract model has been made. This is a computer model of the geometry of the (human male) vocal tract, whose controls correspond to the actions of the muscles involved in speech. This model was used to explore the possible articulations and the corresponding acoustic signals of different vocal tract geometries.

Experiments were run with the original male model, a model of the female vocal tract and a model that is a combination of these two tracts. It was also found that the female vocal tract is better than the male one. This observation was confirmed by a reanalysis of the data from the Peterson and Barney study. This establishes an evolutionary advantage of a vocal tract that has a pharyngeal and oral cavity of equal length (as in the female tract). It has a larger signaling space than the male tract. Males probably had evolutionary advantage from size exaggeration, as proposed by Fitch. It must be noted however, that the differences found so far are significant but small.