240 ^Instill Morris Harmon, 



merits some emphasis. It is onh' natural, to be sure, that in a 

 geographical excursus, for instance, he should recast the data 

 which he obtained from pul:)lished sources ; but that he should 

 rewrite all the letters and sjjeeches which his work incorporates is 

 a matter of more moment. A recent writer has ventured to assert 

 that " Ammianus, unlike other Latin hi.storians whom we have read, 

 does not make speeches for his characters to deliver." ^ Yet both 

 speeches and letters conform throughout to the manner of Ammianus 

 and cannot be authentic except perhaps in their general tenor. An 

 interesting letter may l)e instanced — the enigmatic warning of Prc- 

 copius to Ursicinus, originally written in short-hand, which in Am- 

 mianus' version runs as follows (I. 160. 22 ff.) 



Amendatis procul Graiorum legatis I 



forsitan et necandis III 



rex longaevus non contentus Hellesponto II 



iunctis Granici et Rhyndaci ponti1:)us II 



Asiam cum numerosis populis pervasurus adveniet, II 



suopte ingenio II 



inritabilis et asperrimus, IV 



auctore et incensore III 



Hadriani quondam Romani j)rincipis successore : III 



actum et conclamatum est IV 



ni caverit Graecia. II 



We may be sure that this is not the way in which Procopius ex- 

 pressed himself. Other good examples are the letter of Sapor to 

 Constantius (I. 122. 16 ff.) and the reply of the emperor (I. 223. 22 ff.). 

 As for speeches, the reader will find in Book XXI the addresses of 

 Julian and of Constantius to their respective fcirces, both in the 

 style of Ammianus. A detailed analysis of the speeches of Con- 

 stantius, Julian and Valentinian brings out the fact that all three 

 agree with one another and with Ammianus not only in the ca- 

 dences which they employ but fstill more significantl}) in the relative 

 frequency of the various cadences. This is brought out in the fol- 

 lowing table, based on three speeches of Constantius (I. 64. 29 ff. ; 

 143. 7 ff.; 255. 13 ff.), three of Julian (I. 99. 20 ff . ; 236. 28 ff . ; II. 37. 

 21 ff.), and two of Valentinian (II. 67. 4 ff. ; 106. 28 ff.j. The results 

 are expressed in percentages, and compared with the normal ratios 

 in Ammianus.^ 



' Grlover, Life and Letters in the Fourth Century^ p. 29. 

 * See p. 205. 



