On the feasability of biolinguistics : Koster's word-based challenge and our 'natural computation' alternative