^{page 24 note 1} Much of this paper covers the same ground as parts of Steinthal's Geschichte der Sprachwissen-schajt bei den Griechen und Römern. The greater part both of the facts and of the inferences has, I think, been arrived at independently.

^{page 25 note 1} So Aristarchus, quoted by Herodian, on Il. XIII. 103; ν. Lehr's, Aristarchus, p. 261Google Scholar.

^{page 25 note 2} Cf. Pindarion apud Sext. Emp. adu. Gramm. 10. 202: ⋯ναλογ⋯α … ⋯στ⋯ν τε κα⋯ ⋯νομο⋯ον θεωρ⋯α

^{page 25 note 3} de L. L. IX. I.

^{page 25 note 4} This passage is perhaps most accessible in Goetz and Wilmann's edd. of Varro; whether there are adequate grounds for the implied view that the writer draws from Varro I do not know.

^{page 26 note 1} De Ling. Lat. V. 6.

^{page 26 note 2} Ib. VIII. 18.

^{page 26 note 3} Ib. VII. 4.

^{page 26 note 4} Ib. VIII. 54–62.

^{page 26 note 5} Ib. IX. 50.

^{page 27 note 1} L.L. IX. 66.

^{page 27 note 2} Ib. IX. 56.

^{page 27 note 3} Ib. IX. 79.

^{page 27 note 4} Ib. IX. 40.

^{page 27 note 5} Ib. VIII. 68, 69.

^{page 27 note 6} Ib. IX. 43, 44.

^{page 27 note 7} Ib. IX. 39.

^{page 27 note 8} Ib. IX. 108. The controversy seems to have dealt with the noun a great deal more than with the verb. The analogists seem to have argued that formations from the present stem belonged to a different word than formations from the perfect stem. If ‘lego,’ ‘legam,’ ‘legebam,’ follow analogically, and also ‘legi,’ ‘legero,’ ‘legeram,’ we have no right to expect analogy between ‘lego’ and ‘legi.’ If this were once granted, the regularity of the verb in Latin at any rate is very constant, and I suppose offered less target for the anomalists.

^{page 27 note 9} But perhaps Aristarchus himself was under the spell of the nominative, and chose the vocative as being merely a variation of the nominative.

^{page 28 note 1} L.L. X. 21.

^{page 28 note 2} Ib. X. 26.

^{page 28 note 3} Gram. Lat. (Κ) I. 127. I do not think the vagueness of Varro's rule, as compared with that of Charisius, is conclusive that the latter is not as old as Aristophanes. Varro may well have seen that accent at any rate did not fit Latin, and hesitated at putting out a definite Latin rule.

^{page 28 note 4} I print the extract as given in Cram, Artec. Ox. IV. 333, 6. But the colon after σνλλαβ⋯ is clearly a mistake, as also the absence of comma after καταλήξει. The κατάληξις is the exitus. The παραλήϒουσα συλλαβή is the penultimate. The series of ἤ gives the main considerations, the series of ⋯ν subordinate ones.

^{page 28 note 5} The phrase ⋯πιπλοκ⋯ σνμΦώνου means, I suppose, the tendency of the consonant to influence the following vowel.

^{page 29 note 1} Quint. I. 6, 12.

^{page 29 note 2} Quint. I. 6, 13. This example brings us a good way on to the idea that analogy depended on the meaning of words. In general the analogist repudiated this doctrine (L.L. IX. 40), but we find traces of some such noion. Probably at any rate anomalists felt their case to be stronger if they could allege similarity of meaning. The iorce of lepus) (lupus was, I suspect, increased by the fact that they are both wild animals. In L.L. VIII. 10 we have associated with ‘lepus’ and ‘lupus’ the odd word ‘surus.’ This has been amended to ‘s(ci)urus,’ but I should suggest ‘ursus.’

^{page 30 note 1} ‘Sed ii qui in loquendo partim sequi iubent nos consuetudinem partim rationem, non tam discrepant, quod consuetudo et analogia coniunctiores sunt inter se quam iei credunt, quodest nata ex quadam consuetudine analogia.’ The words ‘ii—credunt’ are taken by Steinthal and Sir John Sandys to indicate a third party, of which they justly say that ‘till partim is defined, it carries us no farther.’ I take the words rather, 'these disputants (anomalists and analogists), who some of them bid us follow reason, while the others bid us follow usage, are not so far apart as they think.

^{page 30 note 2} I. 6. 16.

^{page 31 note 1} IV.-IX.

^{page 31 note 2} The most definite instance I know of is the case of Sisenna, who used ‘adsentio’ for ‘adsentior’ in the senate, and was followed by others (, Gellius, N.A. II. 25Google Scholar, and Quint. 1.5, 13, from Varro). This is not a very flagrant defiance of usage. SirSandys, John (Hist, of Class. Schol. I., p. 142)Google Scholar says of , Trypho ‘that the titles of several of his books show that he was a strict adherent of analogy.’ But analogy is often a mere name for declension (ν. p. 25)Google Scholar, and to write on it does not show the views of the writer. Nor do the fragments, as given by Velsen, show much analogistic tendency. He certainly postulated δορ and δονρ as nominatives for δορ⋯ς— δουρ⋯ς, and remarked that the plural and singular of έϒϰελυς were differently declined, but is not recorded as saying anything that savours of the views mentioned by Sextus. Caesar, who wrote ‘de analogia,’ and should on the same principle be a ‘strict adherent of analogy,’ evidently argued against some analogist views (Gell, N.A. I. 10, and XIX. 8)Google Scholar.

^{page 31 note 3} Adu. Gram. X. 195.

^{page 31 note 4} I. 6, 27.

^{page 31 note 5} L.L. IX. 5, 6.

^{page 31 note 6} I. 6, 21: ‘Recta est haec uia. quis negat? sed adiacet et mollior et magis trita.’

^{page 31 note 7} Cf. ‘Consuetudo non ratione analogiae sed uiribus par … multorum consensione conualuit, ita tamen ut illi artis ratio non accedat sed indulgeat,’ Gramm. Lat. (K.) I. 50.

^{page 32 note 1} Adu. Gramm. X. 222. I do not understand the word ⋯πλ ⋯πλ⋯. All these words are surely compound, and Sextus could hardly fail to see it. It may be objected to my assumption that he refers to the name Εὐμεν⋯ς that the analogists would have said that analogy could not be expected between the uοсαьuΙum εὐμεѵ⋯ς and the nomen Eὐμεѵ⋯ς. But it does not follow from the evidence that either (1) they all accepted this distinction, or (2) that they held it valid in the case of two identical words.

^{page 32 note 2} I. 6, 4.

^{page 32 note 3} I myself, by using the phrase ‘none of your correspondents have,’ became the subject of quite an animated controversy in a local newspaper.

^{page 32 note 4} I. 6, 1.

^{page 33 note 1} Gram. Lat. (Κ) I. 50 and 439. Cf. Quint. I. 5: ‘Simplices uoces prima positione, id est natura sua, constant,’ though there the prima positio is not the nominative or first person present, but the uncompounded noun, as opposed to the compound.

^{page 33 note 2} , Usener in Kleine Schriften II. p. 297Google Scholar, moved, I think, by an unfortunate German propensity to find in competent writers an unintelligent and servile use of earlier authorities, supposes that in the formula ‘natura, analogia, consuetudo, auctoritas,’ both etymologia and uetustas might be interchanged with natura. Quintilian adopted uetustas from Palaemon, but knowing etymologia from other lists inserted that too! Now I do not think any such theory is necessary, for though Quintilian's distinction between uetustas and auctoritas may be a little over-fine, it is quite intelligible. To defend an archaic word on the ground that it was in the past in good use is distinguishable from defending it because it is used by some accepted author. But even if we agree with Usener that the distinction is absurd, I cannot accept his theory as it stands. For while it is possible, I think, that ‘etymology’ (used loosely for the processes into which etymology enquires) may have been a synonym for natura, and that Quintilian adopted it while giving the meaning a very different turn, I cannot understand (nor does Usener give, to my mind, an adequate explanation) how uetustas can have been so used. There is a modification, however, of the theory which I think is possible. This would be to regard uetustas as the opponent of natura, just as consuetudo is the opponent of analogia. This would find some support from Varro, who says ‘neque omnis impositio uerborum exstat quod uetustas quasdam deleuit,’ and later ‘uetustas pauca non deprauat, multa tollit.’ On this view the formula ran ‘natura (or etymologia), uetustas, analogia, consuetudo, auctoritas,’ and the meaning is that the true words, or ἔτѵμα are given us by nature, then modified by time (uetustas) till we get the prima positio of the noun or verb, the ‘scribo’ of Diomedes. This prima positio is inflected by analogy, which again is modified by usage and literary authority. I certainly think it is worth considering whether Quintilian did not find this formula, and either misunderstood it or deliberately gave it a new meaning, though the acceptance of this view would probably entail the abandonment of the theory of Quint. I. 6, which I put forward in Class. Quart., January, 1914. When I wrote that paper I had not read Usener's article.

^{page 33 note 3} This view, however, may be as old as Varro. Cf. Gramm. Lat. (Κ) I. 141: ‘consuetudini et suauitati aurium censet summam uim esse tribuendam.’ The subject of ‘censet’ is, I think, clearly ‘Varro,’ and not, as Steinthal and Sandys, ‘Piinius.’

^{page 34 note 1} Adu. Gramm. (Κ) X. 221, etc.

^{page 34 note 2} On Ignatius ad Smyrn. 8.

^{page 35 note 1} Perhaps we may add that the καθολіκά were felt to be the safeguards of the heritage of ‘Ελληѵіσμός, much as catholicity protected the 'faith once delivered.’ Cf. Sext. Emp. adu. Gramm. X. 179.

^{page 35 note 2} , Souter, History of the Canon, p. 179Google Scholar, but he does not give the references.

^{page 35 note 3} As it is frequently supposed, I think, that this Church use of ‘canon’ is derived from, or at any rate preceded by, the literary use of the word by the Alexandrine grammarians for the list of the classical authors in any department, I take this opportunity to comment on a doubtful piece of lexicography. The Paris Stephanus s.u. καѵώѵ has the following: ‘Similiter καѵώѵ dicitur catalogus auctorum classicorum, praeceteris legendorum, a grammaticis Alexandrinis confectus, v. Ruhnken, Histor. Or. p. xciv.’ Evidently following this Liddell and Scott have s.u., ‘in Alexandrine Grammar collections of the supold Greek authors were called καѵ⋯ѵες, as being models of excellence, classics, , Ruhnken, Hist. Or. Crit. Graec. p. xcivGoogle Scholar. Cf. Quintilian X. 1, 54 and 59.’ But Quintilian never suggests the word καѵώѵ, and when I turn to Ruhnken I find that he says nothing of the kind. Not only does he give no evidence, but he never even suggests that such lists were called καѵόѵες. He himself calls them ‘canons,’ but that is all.

A rather different view is given by Westcott, History of the Canon, Appendix A: ‘One instance of the metaphorical use of the word requires special notice. The Alexandrine grammarians spoke of the Classic Greek authors as a whole as ό καѵώѵ, the absolute standard of pure language, a perfect model of composition, v. Redepenning, Origines I. 12.’ Here again we seem to have the same confusion between т⋯ σήμαіѵоѵ and т⋯ σημαіѵόμεѵоѵ. For Redepenning says nothing the sort. He merely remarks, referring to Quint. X. 1, 54, that no living authors were included by the Alexandrines in ‘dem bald allgemein anerkannten Kanon.’ The fact that Quintilian uses the word ordo of these lists sugests that the Greek name was τάξіς which confirmed by some of the later evidence quoted by Ruhnken.