208 Austin Morris Hannon, 



like II. 39.7 ciiain li'tiu iucrepabat\ 53.27 ctia)]i ti'nn i)i Italia and 

 204. 9 etiaiii ti'im praepeditmu), but some in which the two words 

 are transposed in order, as in I. 247. 24 turn etiaui amiciim} 



No other new compounds have come to m}' notice in Ammianus, 

 l)ut I tind in him vahialjle testimony in regard to the irregular 

 accentuation of several familiar compounds, wdiich in most cases 

 agrees with the testimon}^ of grammarians. 



The grammarians have asserted that the accent of adJiuc was 

 oxytone, and their statement has been accepted by modern linguists."^ 

 Ammianus confirms the fact (II. 124. 12 adhi'tc apud Persas; I. 191. 12 

 adhiic imperator; cf. I. 16. 15; 164. 4; 262. 9; II. 218. 11 ; 253. 13). 

 But antehac has antepenultimate accent (I, 285. 4 antchac inclytain), 

 and illuc penultimate (11. 191. 9 illuc irigressus). 



Furthermore the grammarians tell us that in compounds of -iiide 

 the accent is not penultimate but antepenultimate.-^ Here again 

 the evidence of the clausula aifords full confirmation, and also 

 shows that deinceps falls into the same category. Thus : 



siibinde : I. 333.3 suhinde comnwantibus ; 241.21 subinde prae- 

 dicabat; 31. 21 subinde per aidaeuni; cf. I. 98. 18; 132. 23; 

 306. 9; 322. 25; II. 14. 6; 55. 22 

 deinde: I. 130.32 deinde sunt progressi; II. 21.26 deinde strue- 

 bafur; 126.20 deinde Tusciam; cf. I. 172.13; II. 132.20; 

 218. 27 

 proinde : I. 214. 3 proindc iactans (this word is necessarily rare 



in the clausula) 



deinceps: I. 10. 7 observatwn est deinceps; H. 208. 19 deinceps 



socius; I. 176.6 deinceps nm/tis; II. 10.12 deinceps Iiorta- 



batur; I. 178. 30 deinceps proliiberent; I. 202. 13 deinceps 



arcerentur. 



The accent of exinde, however, is ambiguous, in some cases it 



has an accented ante]:)enult like deinde (I. 80. 13 niotus est exinde; 



110. 23 exinde Roniani ; 164. 12 exinde flumine ; II. 137. 25 exinde rapuit), 



^ The word amicjim here is corrupt : V reads initium. J. Hermann 

 suggested inthmun^ which is perhaps the best of many attempts to better 

 the I'eading, but it is far from satisfactoiy. It is possible that Ammianus 

 wrote lyxi^ov : on this hypothesis the confusion would have come in 

 early, while the text was still in capitals. 



* Lindsay, Lai. Langjiage. p. 163. 



' Servius ad Am. VI. 743 ; Priscian, XI\'. 10 ; cf. also Summer, p). 103. 

 Seelman {/hisspr,. p. 41) is Avrong in interpretiuft- Serv. in Do?!. IV. 444. 

 26 K against the usage, which Servius admits as a fact (dic/iinis). and 

 criticizes theoretically'. 



