On the viability of cognitive morphology for explaining language change