Categoriae 5. 2b 37 – 3a 6
ἔτι αἱ πρῶται οὐσίαι διὰ τὸ τοῖς ἄλλοις ἅπασιν ὑποκεῖσθαι κυριωτατα οὐσίαι λέγονται. ὡς δέ γε αἱ πρῶται οὐσίαι πρὸς τὰ ἄλλα πάντα ἔχουσιν, οὕτω τὰ εἴδη καὶ τὰ γένη τῶν πρώτων οὐσιῶν πρὸς τὰ λοιπὰ πάντα ἔχει· κατὰ τούτων γὰρ πάντα τὰ λοιπὰ κατηγορεῖται. τὸν γάρ τινα ἄνθρωπον ἐρεῖς γραμματικόν· οὐκοῦν καὶ ἄνθρωπον καὶ ζῷον γραμματικὸν ἐρεῖς, ὡσαύτως δὲ καὶ ἐπὶ τῶν ἄλλων.
Further, primary essences are said to be essence most properly through their underlying all other things. And as primary essences are towards the others, thus the kind [eidos] and the type [genos] are towards the remaining things. For against these all the remaining are alleged. Say you call a certain man grammatical: accordingly, you are calling both a man and an animal grammatical.
οὐσία: literally ‘being’. A technical term in Aristotle, often translated (for reasons I have never understood) as ‘substance’. I have used ‘essence’ here to avoid the implications of something material, since it refers to properties as much as to things.
What is interesting about Aristotle's choice of examples is that while the statement 'a [particular] animal is grammatical' is true in terms of both biology and logic, semantically it is odd. ‘Grammatical’ is not usually a word which is relevant in relation to ‘animal.’
Lexical relations do not work quite the same way as logic. There are a couple of ways this can be understood. One is that words can be classed into hierarchical (subordinate/hyponymic or superordinate/hypernymic) relations with each other. Aristotle's observation here, that the more specific term includes the meaning of the more general one (known as ‘transitivity’), is essentially correct. However, hyponymy works from the top downwards. That is, characteristics which apply to a higher part of the tree also apply to lower ones, but not necessarily in the reverse order, since lower members of the tree acquire further defining characteristics. As a result, there are certain restrictions about what words can be predicated to a particular concept.
Another way we can understand the oddness of this collocation (at least for the modern reader) is to look at binary features. One major semantic distinction we make is between human/nonhuman. 'Animal' is marked [-HUMAN], while one of the implications of the adjective 'grammatical' is 'able to use language', and this is ordinarily limited to humans. Because we tend to conceptualize the world in such a way that humans are separate from the rest of the animal kingdom, there is a clash in meaning; the components of the two words contradict each other. The question to ask at this point is: Would this also have been true for the ancient Greek, or did their understanding of ζῷον include humans as well?