A special problem of morpheme division may be mentioned here. In some cases, not only does a morpheme unit have a unique alternant which occurs only when it is next to a particular second unit, but the second unit also has a unique alternant when it adjoins the first; e.g. 'children', if it is divided into /čild/, alternant of /čajld/ only before '–ren', and '–ren', alternant of /-z/ ‘plural’ only after 'child'. Such situations often result from vowel contraction; e.g. Menomini morpheme units ending in /ε/ have alternants with /y/ (instead of /ε/) before morpheme units beginning with /o/; and units beginning with /o/ have alternants with /ā/ (instead of /o/) after units ending in /ε/: instead of having the sequence / … ε-o…/ we have / …y- ā…/. Each morpheme functions as the environment which determines the alternant of the other. In such cases it is sometimes hard to decide where to put the division between the two alternants. Thus 'children' could alternantively be divided into /čildr/ and '–en'; from the point of view of grammatical arrangement each of the two points of division has advantages and disadvantages. In another Menomini contraction, the sequence of certain morpheme units ending in /aw/ followed by certain other units beginning in /ε/ has not / …aw- ε…/ but / …ō…/. We could say that the unit ending in /aw/ had an alternant ending in / ō/, and the one beginning in /ε/ had an alternant without the /ε/; or we could divide differently. The choice is immaterial here, and can be decided only by seeing which division would be more similar to the division of other morpheme sequences.
- Harris (1942a), a pag.176