When we wish to know the analysis of a particular utterance, it is impossible merely to replace each morpheme by its class symbol (e.g. 'I know it = N⁴ V¹ N⁴ = N⁴ V¹') since many morphemes may be members of several classes. W. F. Twaddell has suggested that such analyses of utterances be carried out by repeated substitution tests on the basis of the equations, in what he termed ‘experimental substitution at all levels’. To carry this out, we would ask what substitutions are permitted by the equations for each morpheme or morpheme sequence of our utterance, in the class environment which it has in that utterance. This is repeated until we know unambiguously to what class each occurrence of each morpheme in our utterance belongs.
- Harris (1946), a pag.181