[50] Ibid. a. 40-43.
You must begin by putting down, along with the matter in hand itself, its definition and its propria; after that, its other predicates; next, those predicates which cannot belong to it; lastly, those other subjects, of which it may itself be predicated. You must classify its various predicates distinguishing the essential, the propria, and the accidental; also distinguishing the true and unquestionable, from the problematical and hypothetical.[51] You must look out for those predicates which belong to it as subject universally, and not to certain portions of it only; since universal propositions are indispensable in syllogistic proof, and indefinite propositions can only be reckoned as particular. When a subject is included in some larger genus — as, for example, man in animal — you must not look for the affirmative or negative predicates which belong to animal universally (since all these will of course belong to man also) but for those which distinguish man from other animals; nor must you, in searching for those lower subjects of which man is the predicate, fix your attention on the higher genus animal; for animal will of course be predicable of all those of which man is predicable. You must collect what pertains to man specially, either as predicate or subject; nor merely that which pertains to him necessarily and universally, but also usually and in the majority of cases; for most of the problems debated belong to this latter class, and the worth of the conclusion will be co-ordinate with that of the premisses.[52] Do not select predicates that are predicable[53] both of the predicate and subject; for no valid affirmative conclusion can be obtained from them.
[51] Analyt. Prior. I. xxvii. p. 43, b. 8: καὶ τούτων ποῖα δοξαστικῶς καὶ ποῖα κατ’ ἀλήθειαν.
[52] Ibid. I. xxvii. p. 43, b. 10-35.
[53] Ibid. b. 36: ἔτι τὰ πᾶσιν ἑπόμενα οὐκ ἐκλεκτέον· οὐ γὰρ ἔσται συλλογισμὸς ἐξ αὐτῶν. The phrase τὰ πᾶσιν ἑπόμενα, as denoting predicates applicable both to the predicate and to the subject, is curious. We should hardly understand it, if it were not explained a little further on, p. 44, b. 21. Both the Scholiast and the modern commentators understand τὰ πᾶσιν ἑπόμενα in this sense; and I do not venture to depart from them. At the same time, when I read six lines afterwards (p. 44, b. 26) the words οἷον εἰ τὰ ἑπόμενα ἑκατέρῳ ταὐτά ἐστιν — in which the same meaning as that which the commentators ascribe to τὰ πᾶσιν ἑπόμενα is given in its own special and appropriate terms, and thus the same supposition unnecessarily repeated — I cannot help suspecting that Aristotle intends τὰ πᾶσιν ἑπόμενα to mean something different; to mean such wide and universal predicates as τὸ ἓν and τὸ ὄν which soar above the Categories and apply to every thing, but denote no real genera.
Thus, when the thesis to be maintained is an universal affirmative (e.g. A is predicable of all E), you will survey all the subjects to which A will apply as predicate, and all the predicates applying to E as subject. If these two lists coincide in any point, a middle term will be found for the construction of a good syllogism in the First figure. Let B represent the list of predicates belonging universally to A; D, the list of predicates which cannot belong to it; C, the list of subjects to which A pertains universally as predicate. Likewise, let F represent the list of predicates belonging universally to E; H, the list of predicates that cannot belong to E; G, the list of subjects to which E is applicable as predicate. If, under these suppositions, there is any coincidence between the list C and the list F, you can construct a syllogism (in Barbara, Fig. 1), demonstrating that A belongs to all E; since the predicate in F belongs to all E, and A universally to the subject in C. If the list C coincides in any point with the list G, you can prove that A belongs to some E, by a syllogism (in Darapti, Fig. 3). If, on the other hand, the list F coincides in any point with the list D, you can prove that A cannot belong to any E: for the predicate in D cannot belong to any A, and therefore (by converting simply the universal negative) A cannot belong as predicate to any D; but D coincides with F, and F belongs to all E; accordingly, a syllogism (in Celarent, Fig. 1) may be constructed, shewing that A cannot belong to any E. So also, if B coincides in any point with H, the same conclusion can be proved; for the predicate in B belongs to all A, but B coincides with H, which belongs to no E; whence you obtain a syllogism (in Camestres, Fig. 2), shewing that no A belongs to E.[54] In collecting the predicates and subjects both of A and of E, the highest and most universal expression of them is to be preferred, as affording the largest grasp for the purpose of obtaining a suitable middle term.[55] It will be seen (as has been declared already) that every syllogism obtained will have three terms and two propositions; and that it will be in one or other of the three figures above described.[56]
[54] Analyt. Prior. I. xxviii. p. 43, b. 39-p. 44, a. 35.
[55] Ibid. p. 44, a. 39. Alexander and Philoponus (Scholia, p. 177, a. 19, 39, Brandis) point out an inconsistency between what Aristotle says here and what he had said in one of the preceding paragraphs, dissuading the inquirer from attending to the highest generalities, and recommending him to look only at both subject and predicate in their special place on the logical scale. Alexander’s way of removing the inconsistency is not successful: I doubt if there be an inconsistency. I understand Aristotle here to mean only that the universal expression KZ (τὸ καθόλου Ζ) is to be preferred to the indefinite or indeterminate (simply Z, ἀδιόριστον), also KΓ (τὸ καθόλου Γ) to simple Γ (ἀδιόριστον). This appears to me not inconsistent with the recommendation which Aristotle had given before.
[56] Ibid. p. 44, b. 6-20.
The way just pointed out is the only way towards obtaining a suitable middle term. If, for example, you find some predicate applicable both to A and E, this will not conduct you to a valid syllogism; you will only obtain a syllogism in the Second figure with two affirmative premisses, which will not warrant any conclusion. Or if you find some predicate which cannot belong either to A or to E, this again will only give you a syllogism in the Second figure with two negative premisses, which leads to nothing. So also, if you have a term of which A can be predicated, but which cannot be predicated of E, you derive from it only a syllogism in the First figure, with its minor negative; and this, too, is invalid. Lastly, if you have a subject, of which neither A nor E can be predicated, your syllogism constructed from these conditions will have both its premisses negative, and will therefore be worthless.[57]