[…] each setting up of a component of more than one-phoneme length takes care of at least one limitation of phonemic occurrence; this is equally true of the automatic-sequence components […] which replace the highly limited distibution of phonemic pitch and stress. In some cases the components can be so selected that practically every possible combination and sequence of the components actually occurs. Any combinations and sequences that do not occur will, of course, have to be stated.
- Harris (1944), a pag.202