Abstract
The Glottalic Theory of Proto-Indo-European, whereby the traditional voiced series of stops *b, *d, *g has been replaced by the ejectives *p’, *t’, *k’, has been argued for mainly on typological grounds. What I will try to show here is that since any proposed sound change must satisfy the condition of naturalness, every reconstructed segment or class of segments, no matter how typologically plausible, must be derivable in each daughter language in conformity with this fundamental principle. To illustrate this, I will present a case from Proto-Algonkian showing how both the comparative method and typological considerations are insufficient to help us choose whether should be reconstructed since these two factors are not necessarily determinative of naturalness.