Now, it is not necessary for us to agree on our idea of the nature of a phoneme whether we are to understand it as a class of sounds (each sound being itself a slice out of a continuum of sound), or regard it as some new entity containing a ‘characteristic’ sound plus an on-glide and an off-glide. For linguistic work it suffices to know how to recognize the phonemes of a language. But Trubetzkoy offers a specific picture of the phoneme as a ‘functional’ sound: ‘The phonologist considers in the sound only that which fills a specific function in the language system'.
- Harris (1941), a pag.345

Since many linguistic workers in America may want to have some idea of Trubetzkoy’s method, a few of its lines will be indicated here. Phonemes are points in a network of contrasts. Two phonemes which have no features in common […] cannot be contrasted. Two phonemes which have in common some feature which no other phonemes has are in UNIDIMENSIONAL contrast. Two phonemes whose common feature is also common to some other phoneme are in PLURIDIMENSIONAL contrast […]. Pairs of phonemes having similar contrasts between them may be equated in a PROPORTIONAL formula […]. Two phonemes in particularly close and limited contrast form a RELATION-PAIR. The difference between them is a RELATION-MARKER. The two phonemes are considered identical except that one has the marker of their private relation while the other does not; they would be represented not as A : B but as A : (A+a).
- Harris (1941), a pag.346