^{page 3 note 1} Plato (notably in Apol. and Laws) and Xenophon (Mem. I 1, §§ 2, 5) use the two expressions as synonymous. I waive here the questions whether they (a) were mistaken, (b) mislead us (pointlessly), (c) conspired (in a very complicated manner) to do so.

^{page 3 note 2} For the statistics given I admit full responsibility, though I have made some use of Todd's index to Aristophanes and von Essen's to Thucydides. I have had to decide silently a Hew cases (none of importance) where there is room for disagreement. [E.g., Sophocles fr. 86 (Nauck)—not in point here—seems to me T, but C to L. and S.] But I should be glad to supply my full references to any scholar who may wish to use or criticize my results.

^{page 4 note 1} I find in Herodotus 123 certain instances of ν=62 TA (=54 with an oblique inf. expressed + 3 where oblique tlvai is understood + 5 others) + 19 TP (=10 with thai expressed + 8 with understood + I abs.) + 28 CA (=12 with a dependent inf. expressed + I with a dependent inf. understood +9 with the accusative of a noun +3 with the accusative of a pronoun + 3 with the dative of a noun) + 14 CP. I am indebted to Mr. J. Enoch Powell for a small correction.

^{page 4 note 2} They also cite Plato Laws 879c, where I should suppose that that is understood ; but in any case it is clearly outweighed by my figures for earlier authors.

^{page 4 note 3} In ‘Lysias’ I count pi ν.+78 certain TA [=61 with an oblique inf. expressed (22 times this inf. is ε⋯ναι) +15 where oblique ναι is understood +2 others] +8TP+2CP + the 3 instances discussed above (all of which I should class as TA with oblique nvai understood). It certainly looks as if in Lysias [as in Antiphon (10 ν=4 TA+6 CP) and Andocides (27 ν.= 27 T)j the active voice of νμἱζειν always means ‘think’; 1 suspect that this is true of all genuinely Attic prose.

^{page 5 note 1} In Aristoph. I count 56 ν.=43 TA+4 TP+ 1 CA (with a dependent inf., V. 1196) +8 CP. Of the 43 TA, 19 have an oblique inf. expressed (8 times this inf. is ε⋯ναι), and 25 imply an oblique ε⋯ναι understood. That makes 44; but I have counted IV. 1049 twice, avipε⋯ναι since пον⋯αι is an instance of expressed oblique inf., and κα⋯ proves that (not ώς, by the way) is understood with . More proof that ε⋯ναι is understood in such sentences is afforded by comparing (e.g.) R. 776, v. with PI. 776, thai ν. χρησλµον (where could have been omitted without change of sense).

Compare the figures for ηγεĪσθαι in Aristoph. Of 29 occurrences 9 mean ‘lead’ and 20 ‘think.’ Of these 20, 5 have an oblique ε⋯ναι expressed, and the remaining 15 imply an oblique ε⋯ναι understood.

^{page 5 note 2} Hel. 919 is of value (I think) for Тà θεȋα μη ν. of Diopithes' decree. Belief in the validity of religion, the reality of the gods and their ordinances, is the point.

^{page 5 note 3} A good instance of νομἱζειν and ήγεîσθαι synonymous, suggesting that, for the sense ‘think,’ they should have received parallel treatment in L. and S.

^{page 5 note 4} In Euripides (ignoring frr. dub. et spur.) I find 73 ν.=43 TA+17 TP+4 CA (of which 2 have a dependent inf. and 2 the accusative of a noun) + 9 CP. Of the 43 TA, there are 20 instances with an_ oblique inf. expressed (in 2 of these the inf. is ε⋯ναι), 22 where an oblique ε⋯ναι is understood, and 2 others. The addition is correct, since I have counted Heraclid. 1039 twice for the same reason as Aristoph. N. 1049 was counted twice in n. 6 supra. That ε⋯ναι is understood in the 22 cases of TA mentioned supra is proved by this line, and also by comparing (e.g.) Med. 527, with Heracles 484, .

I add the statistics for Thucydides. Certain occurrences of ν=279=264 TA+6 TP+4 CA (of which 3 govern a dative and 1 has a dependent inf.) + 5 CP. Of the 264 TA, there are 210 instances with an oblique inf. expressed (in 66 of these an oblique tlvat expressed is involved), 48 where an oblique *tvai is understood, and 6 others, That oblique tivai is understood in the 48 cases mentioned follows from (a) the fact that an oblique ttvai is clearly understood in phrases of the type of VIII 56, 4, (cf. Aristoph. N. 1049, n. 6 supra), (b) from a comparison of (e.g.) II 21, 3, пερικλέα αίγιον ν with IV 82, . Note (a) that there is no warrant in Thucydides for the Herodotean usage (CA) with the accusative of a noun; (b) that there are plenty of instances which illustrate the ease with which an oblique can be omitted with νομἱζειν(TA), e.g., έν (which is as dose a parallel to θεονς;—εĪναι being understood—as one might reasonably expect to find).

^{page 5 note 5} I count this as an example of TP with εîναι Understood. Of this type there are 5 instances in Thuc, 17 in Eurip., 7 in ‘Lysias,’ 3 in Aristoph., and 8 in Herod. Although this is usually omitted, there is justification for saying that it is really implied. Herodotus inserts or omits it with equal readiness; of TP with an expressed εἶναι he exhibits 10 examples, whereas ‘Lysias’ and Aristoph. have only 1 each, and Thuc. and Eurip. none.

^{page 6 note 1} Better ‘instances’ than any of the above might have been alleged, e.g., Eurip. Heraclid.201, . But this is, of course, an instance of TP with εĪναι understood.

^{page 6 note 2} [Lysias] VI 17; cf. XXX 17, àσεβεĪν ιέγονТα. The point is more fully developed in my article ‘Plato, Socrates and the Myths,’ C.Q. XXX (1936), p. 142.