The Clausula in Animianus Marcellinus. 219 



cent. Similarly, Latin words inflected after Greek models must be 

 accented as their models are accented ; for instance, Scipiades, Mem- 

 miades with stressed penult. But Greek words taken over into the 

 Latin inflectional system may be treated in two ways ; the}' may 

 be accented either on the Greek basis, as ae'ris^ aethe'ris, Evander, 

 tyrannus, or on the Latin basis, as deris, ae'theris, Evander, tyrdnmts. 

 But although the Greek system may be followed at will, it must 

 be used correctly if at all : in quibiisdam enim nominibus licet vidcre 

 plerosque, recti casus ambiguo tenore deceptos, mendose oblicos pro- 

 ferre, ut qui in patrico casii Evandri et tyranni primam syllabam 

 acuunt potius quam mediant, nullam secuti rationem. This grammarian, 

 then, freely concedes the propriety of accenting Greek loan-words 

 in Greek fashion as Ammianus does, and even acknowledges, though 

 only to censure, the practice of carrying antepenultimate accent 

 through the inflection regardless of the quantity of the final .syllable. 



Donatus and Sergius go eveiT farther than our unknown, for they 

 no longer admit a choice between the Greek and the Latin basis 

 of accentuation, l^ut insist upon the Greek : in the words of Sergius 

 (K. IV. 483. 29) the rule is stated, Graeca autem sins accentibus pro- 

 nuntianda esse noscamus} 



It has seemed best to me to dwell upon the precepts of the gram- 

 marians at some length in order to bring out the value of the passage 

 in the pseudo-Sergius, which has heretofore received little attention. 

 The evidence of the language itself may be despatched briefly. 



A certain amount of information as to the pronunciation of loan- 

 words is obtained from the late poets, in whom we find false quan- 

 tities which can only be due to Greek accentuation. For example, 

 Venantius Fortunatus employs scansions like Cycladas (and often 

 -ades, -idis and the like), Ardtus, eniblema, probleuia, idolutn.'^ But 

 the bulk of the evidence is afforded by the Romance languages, in 

 which the rule holds that words introduced into Latin from the 

 Greek retain the Greek accent in their Romance derivatives, except 

 when the Greek word was oxytone.'^ 



In a general way, then, we learn nothing from Ammianus that 

 is startlingly new. His testimony is chiefly valuable because it 

 emphasizes with enormous weight a feature of late Latin pronunci- 



* Donatus (K. IV. 371. 27) saA^S, sane Graeca verba Graecis accentibus 

 efferimus. This is quoted from Diomed verbatim, but the siguificant 

 proviso of Diomed, si isdem Utteris emintiaveri»tus^ is left off. 



"^ See Leo's metrical index. 



' Meyer-Liibke, Roma?!. Gramm. I. p. 3-4 : Gramm. Storico-comp. delta 

 Ling. Ital. p. 84. 



