or its predecessor (which we assume, as Ver. and Put., to have averaged some eighteen letters to a line): PILVMVOCABANTTRIBVS EXVEXILLISCONSTABAT ORDOSEXAGENOSMILIT ESIICENTVRIONESVEX ILLARIVMVNVMHABEBAT VEXILLVMCLXXXVIHOM The ending -bat in constabat and habebat led to the omission of the three lines after constabat in the next copy. These were then noted at the top or foot of the page, or in the margin. When this page was itself copied, the scribe endeavouring to restore them was guided by the first word ordo to insert them after ordines instruebantur in § 3, and thus ordo and uexillum were both cut off from their proper verbs, to make nonsense by combining with others. This process of the reinsertion of omitted1 matter in the wrong place may be seen, to quote examples from this Decade alone, in the following other passages; nor do we suppose that the list is at all complete: Praef. § 1: 2.38. 2; 2. 39. 2-3; 2. 40. 8; 2. 43. 5; 4. 2. 2-10; 4. 7. 10-11; 4. 25. 4; 4. 28. 2; 6. 18. 5; 6. 40. 1-16; 7. 27. 1; 8. 10. 13; 8. 31. 6-7; 8. 37. 12; 9. 9. 17; 9. 18. 14; 9. 30. 6-8; 9. 34. 16; 10. 13. 13; 10. 21. I; 10. 36. 2. Some of these we have discussed in Class. Quart. IV. (1910) 274, and ibid. V. p. I seq., some earlier in this article. Now the MSS. even as they stand do not leave us without warning where the corruption has taken place. In § 4 ordo is placed before instruebantur by L (after it by L1), and altogether omitted by O. In the same way in 4. 7. 6 the words ut . . . irae are completely lost in M, but have lost only ut in O. In 7. 27. I the words ciuitatem . . . decem-(uiris) preserved in MPFUpO had fallen into the margin of the archetype of HTDLA; H restores them twice, in the right place and in a wrong place, but in each case omits ciuitatem; L restores them in the right place, but with the same omission; on the other hand TDA omit everything save ciuitatem, but add to it the tell-tale his (see p. 3 sup. footnote). For a list of other examples of partial restoration see our note above on 6. 18. 5. Further, in M after ordo stands or stood a mark of separation (;) high up in the line, now mainly erased (the corresponding mark in the archetype of DA has been interpreted by D1 or D2 and A as i, i.e. in, so that they give 1 On omissions in the uncial archetype see further the second part of this paper (on 8. 31. 5 and 9. 40. 4) and the Introduction to our second volume of the text (VI.-X.); and to show that the first Decade is not alone in this respect, we may mention, e.g. 27. 32. 7, where Put. reads 'postero die castellum Phyrcum uocant | copias omnis eduxit.' But Spirensis (acc. to Rhenanus) gave postero die omnes copias ad propinquum (possibly -qum) Eliorum (19 letters) castellum Pyrgum uocant eduxit. This means that a predecessor of Put. omitted two whole lines (ad propinquum ... uocant) (40 letters or less), but copied out one or both in the margin, whence Put. took only one and that in the wrong place. ordo in sexagenos). In M again at the point of junction (constabat uexillum) the letters-at uex-are written 'pleniore calamo,' i.e. with much blacker ink, not merely than the letters which precede, but than the letters which follow, and therefore show that there was something before the scribe in his exemplar which caused him to hesitate and leave a blank to be filled up later; examples of this (nineteen in number) will be found in our note on 1. 41. 4,1 to which add this passage and 10. 37. 15). Before and after uexillum there is a punct added in OHA and by M or M2; this is often another trace of invading marginalia; see the examples in our note on 6. 42. 13 above and on 2. 32. 10. Therefore both at the place whence we suppose that the words came and at the place to which we suppose that they were taken the MSS. show signs of the disturbance. What is then the result if the transposition be regarded as established? Nothing can alter the fact that Livy supposed each of the three lines to contain 15 divisions (quindecim in § 5, triginta and quindecim in § 7), whereas Polybius (6. 24. 3) states that there were only 10. Again Livy tells us that each maniple had 20 light-armed men, whereas according to Polybius (6. 24. 4) the whole of the ypoopoμáxoi were divided equally between all the 30 maniples of the three lines, which, since he gives the number of Hastati, Principes, and Triarii respectively (ib. 21. 9) as 1200, 1200, and 600, and the total of the legion as 4200 at the least (rising to 5000 on special occasions) would be at least 1200 (the proportion was the same whatever the total of the legion, as Polybius explicitly states, c. 21. 10); and 1200 divided by 30 gives 40, not 20, to each maniple. Livy's total of light-armed men is larger; if we disregard the centurions, he reckoned 15 times 60 to the rorarii, the same for the accensi, and 20 for each of 30 maniples of the Hastati and Principes, or 900+ 900 + 600, 2400 in all, double the total of Polybius. This in itself is not incredible, since the number of light-armed men is likely to have been considerably smaller in the second century B.C. than in the fourth; after Marius they had disappeared from the legion altogether. But even so a difficulty remains as to the numbers of Hastati and Principes themselves; Livy's reckoning of 2400 light-armed and 900 triarii proper (3300 together), if the total of the legion was 4200, leaves only 900 heavy-armed soldiers for his 30 maniples, i.e. 30 for each maniple; if we take his own figure of 5000 for the legion, the Hastati and Principes are left with 1700, which gives 56%. Both totals are certainly too small (the triarii had 60), and the second is a little too fractional for daily use! These considerations no doubt decided Livy to say nothing about the number in a maniple; his authorities left him in the dark, and he does not profess to know. Madvig acutely guessed (Emend. p. 189) that his difficulty arose from applying to the manipular part of his fourth-century army the total of 30 maniples. This was a true total for 1 See also § 33 (a) (b) (c) of the Preface to Vol. I. of the text in our edition (Oxford, 1914). 14 RESTORATIONS AND EMENDATIONS IN LIVY VI.-X. the later (Polybian) army when the Triarii also were included in the manipular system. It seems probable therefore that in the army which Livy was describing there were only 10 maniples of Hastati, 10 of Principes, and 10 ordines of the third division (Triarii + Rorarii + Accensi). Then the numeration becomes clear; the third division has 3 times 600, i.e. 1800; each of the first two has 1600, each maniple running to 160. This gives 1800 +2 (1600) = 5000, Livy's total. In each of the maniples of 160 Livy counts 20 light-armed men. This may be true, but it seems likely that Livy has applied to a maniple the figure which his authority meant to belong to a half-maniple or century—a word which neither Livy nor Polybius uses at all in the description, and whose Greek equivalent (perhaps Táyμa) may well have been ambiguous. If so, each maniple of the Hastati and Triarii had 40 light-armed and 120 heavyarmed soldiers, the latter number being just double that of each section of the Triarii, as in the Polybian system. These two misunderstandings forced Livy to leave his account incomplete; but if our rearrangement of the text be sound, the details he does give are perfectly consistent with one another. Madvig's conjectures as to other confusions may now be set aside. There remain two or three minor points. Since in M the letters -anges similes in § 3 are written 'pleniore calamo' (see above), it is just conceivable that they represent an alteration (in M's exemplar) from phalanx similes as Luterbacher proposes, but this is hardly necessary. A point where corruption is more likely is the last letter of the word postremo, after which there is a punct in M. Either postrema, which Klock proposed (but which we should take as neuter plural, leaving instruebantur unchanged), or postremi, which most editors adopt from Ortmann's conjecture, seems to us extremely probable; the phrase postrema acies is used at the end of § 8. In § 6 frons in acie is the reading of a responsible corrector in D (D3) also of F3A4, for whom there is less to be said; Up gives frons in aciem; but MPF(?)OTD(?)A have foris in acie; H has foris in ace; L has in aciem foris. Madvig conjectured primae frontis acies; it is perhaps critically more probable to read prima sors in acie (for a similar use of prima sors cf. 22. 29. 9, and Verg. Georg. 4. 165 sunt quibus ad portas cecidit custodia sorti). Then the confusion at this point might be due to the effort of some corrector to replace the difficult sors by the more obvious frons. LONDON. October, 1917. (To be continued.) C. F. WALTers, R. S. CONWAY. GREEK COMPOUND ADJECTIVES WITH A VERBAL ELEMENT IN TRAGEDY. A GENERAL treatment of Greek compounds seems much to be desired. It would have to be undertaken by one who had an up-to-date philological equipment, to which I cannot lay claim. But rather with the hope of eliciting discussion on the subject and learning from others I offer the following observations, and in further study of the subject should be grateful to anyone who would advise as to the exact statistics that may be desirable over and above what I give below. I was led to the subject by a feeling that the treatment of many individual compounds by editors was far from satisfactory, and that possibly a collection of the material might help to bring out the exact meaning of some of the well-known difficulties in the Tragedians. One is dealing here with a highly developed and somewhat arbitrary poetic idiom, and it may perhaps be impossible, as one must admit from the outset, always to make precise the poet's meaning, but it is worth while to make the attempt. Homer is full of compound adjectives of the ornate and standing epithet type in the main, though of the class represented by βούθυτοι τιμαί or πάθεα δακρυοπετῆ οι δημηγόροι στροφαί (to take three instances from Aeschylus Supplices) he has hardly any. He has compounds in sufficient number, in which the verbal meaning is clear, and a large number have the verb in the first place-τερπικέραυνος, ἐχέφρων (the parallels of ὀλοόφρων ταλασίφρων, etc., show that the second element is to be regarded as substantival), μevedńios, Bariáveipa. These we must regard as belonging to the most primitive stratum, as Sanskrit parallels show. His proper names exhibit the Sanskrit categories clearly. Of the Tat-purusha type we have 'Aσrvávaž; of the Kama-dhāraya type, Karoiλios and nearly Kaλλikoλóvn; of the Bahu-vrīhi type, kaλλiуúνaixa. But from the first Greek seems to specialize, for whereas the same word may be either KD or BV, and either TP or BV in Sanskrit, in Greek this does not seem to be the case, though of course there are καλλίπαις = καλὴ παῖς in Orestes 964, καλλίπολις Rep. 527C. άγκυλόμητις might theoretically be KD, a crooked counsellor, but in practice is an epithet of Kpóvos. Nevertheless it may be of value to look back on these origins in treating expressions of tragedy. One is surprised to find how few Epic compounds are used by Pindar and the Tragedians in comparison with the total number of compounds they employ. On a rough calculation there are 96 compound adjectives in the Olympian odes. Of these only 14 are found in Homer, Hesiod, and the There is one other surprising fact. Only 28 have any definite ement, and 57 are of the BV type. It is worth while to give the το ements: r Initial ἀλιτο-, ἀναξι-, δαμασι-, ἐρασι-, ὀρσι-, φύγο-, tot counting nào, which is in no sense verbal). (2) Final in -os, -ns, αστος. όκος, ελάτης, θαλμιος, -πετης, πορθος, -ρεπης, -φορος; in της. ώματος, ελατος, -φατος. The absence of the last type is comparson with Aeschylus. The few that call for comment are are the possibility of the adjective is created by the substantive This buburuaTos seems in itself impossible, a combination of sostantive to make an adjective; but when coupled with LOs being that part of the appa which by its lightness makes to assicu, we have, reduced to prosaic form, 'quickly racing cars of light Ποικιλοφόρμιγξ ἀοιδή is the complicated strain of the lyre, ......s much a case of transference of the epithet as νείκος ξύναιμον KULITOS only becomes a possible word by combination with A not conceivably be used predicatively. One or two of a may be added. depins xúpa should not be regarded as 11 land which has a fence (created by) the sea.' doλxýρeтμOS ss 3V. just as much as peλáμßρотos yn (Eur. fr.), but it seems en type, an extension of the simple 'much-rice land,' Aerland (of those who use) long oars.' These observations can be considerably developed in dealing with the by ceret to make a complete list of compound adjectives in toes a specimen play from each from about the middle of and the result of that was as follows: ecllowing statistics, which refer to verbal compounds tion between the three Tragedians which a treatment would bring out. ee compounds, in which the verbal element comes .. being omitted as before), shows that they emer, and were not indebted to each other. Furipides 15. They are: A. deği-, S. aλe§-, . Ε. αρχες, Α. ἀ- στεργο, S. ἀτιμ-, 5. δακες, όλαος Ε. έχει, S. ζευξι-, Ε. θελξι-, Α. καμψι-, -, ΑΕ. μιξο-, Α. μισο-, Α. μνησι-, S. νικο-, AE. Αν και ε. Α. περσι-. Ε. πλησι, Α. ρεψ, Α. ῥυσι-, 1. ou. The primitive type survives, but |