From as early as the 2nd century A.H. (8th century A.D.), scholars of Arabic had noted the presence in the language of correlations between phonetics and semantics. Leading philologists tried to describe the phenomenon, drawing up lists of the words concerned; and in time summary outlines taking up no more than a few pages in grammatical works evolved into detailed monographs. Not until Ibn Jinnī (d.392/1002), however, did a theory of lexical substitution emerge. But, for reasons which I will explain in this paper, the theoretical framework which this pioneering philologist developed in his Kitāb al-Khaṣā’iṣ won no support from either contemporary or later Arabic grammarians nor, indeed, from Orientalists. The 20th-century notion of phonetic features, unknown as it was to Ibn Jinnī, has changed our outlook somewhat since phonemes are now written in a matrix of features. Even so, there has been no change in the way minimal lexical units are conceived, and the relationship between sound and meaning is still regarded as arbitrary in modern linguistics. Reappraisal of Ibn Jinnī’s theory has only become possible since a group of modern French linguisticians, of whom I am one, demonstrated, through matrix and etymon theory, that minimal lexical units are simultaneously formal and notional in nature. This paper will give a detailed account of matrix and etymon theory. The object of the theory is to reorganize the lexicon not only of Arabic but also of other Semitic languages by explaining relationships of homonymy, polysemy and enantiosemy. Informed as it is by contemporary scholarly perceptions and analyses of analogy, this theory is, however, equally an outgrowth of the endeavours of early Arabic lexicography.