from String analysis of sentence structure by Zellig Harris, Mouton Co. 1962.

String analysis characterizes the sentences of a language as follows: Each sentence consists of one elementary sentence (its center), plus zero or more elementary adjuncts, i.e. word-sequences of particular structure which are not themselves sentences and which are adjoined immediately to the right or to the left of an elementary sentence or adjunct, or of a stated segment of an elementary sentence or adjunct, or of any one of these with adjuncts adjoined to it. An elementary sentence or adjunct is a string of words, the words (or particular sequences of them) being its successive segments. Each word is assigned (on the basis of its affixes or its position in elementary sentences and adjuncts) to one or more word-categories (rarely, word-sequence categories). Hence we can replace each word of a string by the symbol of its category, thus obtaining a string of category-symbols (called a string formula) as a representation of the word-string in question
[ . . . ]
it is possible to decompose each sentence into elementary strings which combine (to form a sentence) in accordance with specified rules. If in a given sentence we find a sequence of words which cannot be assigned to any known string formula occurring in it in accordance with some known rule, then a new string or rule of occurrence has to be set up. The intention is that a few classes of strings, with simple rules describing how they occur in relation to each other, will suffice to characterize all sentences of the language (pp. 9-10).