In the three sentences above, the forms that immediately follow the noun kar ā€˜doctorā€™ are clearly in a relation of substitution as they have identical linguistic function viz., that of marking the relation between kar ā€˜doctorā€™ the noun and ta ā€˜to goā€™ the verb. But by the criterion of potential pause, -yi is a nonword [a suffix], kotu ā€˜uptoā€™ a word, a postposition and he ā€˜nearā€™ more a word, a postposition than a nonword [a suffix]. Substitutability is not sufficient on another score viz. that it not only can not tell words from affixes as [e.g.a. 31 a-c] show, but also cannot tell words from phrase, since phrases are substitutable for words.
 

The criteria are dissonant rather than consonant giving rise to the need to hiearchise them so as to arrive at the discrete Mao word. Isolability, for instance, conflicts with Potential Mobility, Potential Pause and Elliptibility; an isolable unit need not be potentially mobile or elliptible or be marked by potential pause. But the resort to the expedient of hiearchisation is of no avail because while Insolubility/potential pause override all the other criteria whenever the two conflict with the latter, there are cases where Isolability overrides Potential pause and cases where Potential Pause overrides Isolability: le, the future tense auxiliary is isolable but there is no pause marking its boundary with the preceding linguistic element, but it is a word ; postpositions have potential pause marking their boundary with the nouns that precede them, but postpositions are not isolable, but are words. The result : there seems to be no discrete Mao word, available for a rigorous global definition, although in particular instances it is possible to identify the word in the majority of cases. For instance, Potential Pause is decisive except when it is overridden by Isolability.