Halliday & Matthiessen (1999: 15-6):
In approaching the semantic environment from within grammar, as opposed to approaching it from some postulated cognitive or conceptual plane (i.e. in constructing it as meaning rather than as knowing), we are following the principle from which semantics first evolved in western thought. By the time of Aristotle there had emerged the grammatical concept of a word, and of word classes; for example, 'noun'. This concept was born in the work of the Sophists, in their study of rhetoric, out of the dialectic between form and function. A noun was that 'about which something is said', thus embodying the functional concept of a syntactic (Theme-Rheme) structure; and that 'which inflects for number and case but not for gender', embodying the formal concept of a morphological (case and number) paradigm. The category of noun once established, the question arises of why does a noun appear sometimes in singular sometimes in plural number, sometimes in nominative sometimes in accusative, genitive or dative case? These questions are answered with semantic explanations: a noun is the name of a person, other living creature or inanimate object; a noun is in the plural if it refers to more than one of these entities; it is in the nominative case if it refers to the 'doer' in some kind of action, the accusative if it refers to the 'done-to', and so on. The semantics was construed by exegesis out of the grammar: both the general conception of meaning as a linguistic phenomenon and the specific meanings that were constructed by words, their classes and their variants.