COSMO's “motor theory” is not the motor theory of Liberman, Cooper, and Mattingly

The sensorimotor model of speech proposed in the target article has considerable value. The Bayesian implementation captures the probabilistic nature of between-person communicative interactions involving speech and the integrality of the speech perception and production systems. In the context of a simulation of between-person communication, this integrality fosters emergence of some aspects of existing phonological systems, including the tendency for vowel systems to reflect a dispersion principle, and of phonological systems to exploit the nonlinearity of the articulation-to-acoustic mapping. However, the article fails as a test between theories of speech perception because the modelers conflated two goals: one to test COSMO against subparts of itself, and another to test COSMO against existing motor and auditory theories. For the first goal, findings show that the full model performs better than the motor and auditory subcomponents. However, for the second goal, the testing fails at least with respect to the motor theory of speech perception, because that theory is not the same as the model's motor subset.