Frequently […] we find that there is a set of morphemes in a language, such that each morpheme in the set is identified by its contrast with all the others in the set. Such morphemes are often arranged in paradigms, and the various crisscrossing relationships among the morphemes of the paradigm are often called categories. Thus there are such Latin morphemes as '–us', '-um', '-ī ', '-ōs' which are defined by their membership in a paradigm, and which are considered as expressing, within the paradigm, such categories as case and number. The presence of these categories is not a happy situation for structural linguistics, which is most useful if it can define everything in terms of some stock of elements (phonemes, morphemes) which are all on a par with each other. It is therefore of interest to note that the categories represented in a paradigm can be set up by means of the very methods which are used to set up the more traditional morphemes. Just as the morphemes 'the' and 'a' can be isolated by comparing, say, 'You’ve lost the job' with 'You’ve lost a job', so also the categories of singular and plural can be isolated by comparing certain utterances, even in a language in which these appear not as distinct morphemes but as categories in a paradigm.
- Harris (1948), Pag. 87