122 Austin Morris Harmon, 



with entire soundness unless the text itself is entirely sound. ^ Now 

 the text of Ammianus is unfortunately anything but sound, for it 

 rests almost entirely on the evidence of a single manuscript, which 

 is an unfaithful copy of a faulty and defective archetype.^ Conse- 

 quently the conditions in Ammianus are well-nigh the worst possible. 

 This will prevent us from attaining to certainty upon one or two 

 points of his usage, which, however, are of relatively slight import.^ 

 Owing to the uniformity (not to sa}- monotony) of his rhythm, all 

 its essential features jDresent themselves with distinctness even in 

 our imperfect text. 



So great is this uniformit}^ that I have considered it unnecessary 

 to demonstrate my observations by statistical evidence based on the 

 full text of Ammianus. It goes without saying, of course, that I 

 have contirmed them, both in general tenor and in certain particulars, 

 by repeated readings of the Histories, but the statistics which I shall 

 adduce in proof of them are not imposing. They rest upon two 

 distinct bases which include less than 3300 clausulae. One of these 

 is a catalogue of all the sentence-endings (exclusive of direct quo- 

 tations) in the first six and the last three books of our author, 

 amounting in all to 1811. The making of this collection was the 

 first step in my investigation, and in restricting it to sentence-endings 

 I was naturally determined by the fact that here if anywhere the 

 clausula is certain to occur. But, the final cadences, once fixed, 

 those in the interior of the sentence could not be left out of account : 

 hence the second collection, which embraces all the clausulae in 

 Book XXI, numbering 1461. 



The results of my investigation fall under two heads : first, the 

 character of the rhythm in Ammianus, and second, inferences drawn 

 from the rhythm. Under the first head we shall consider first the 

 nature of the clausulae employed by Ammianus (taking up in order 

 the forms, the types and the matter of residual quantity, and clos- 

 ing with a brief stud}^ of their origin], and next the manner of their 

 employment (length of kola, number of kola in sentence, structure 

 of sentence, resi)onsion, etc.). Under the second head fall two groups 

 of observations — those pertaining to linguistic phenomena (matters 



^ Rhythnten d. as/an. w. rdm. Kunstprosa, p. 123. 



^ Vat. Lat. 1873 {V). Gelenius' edition contains some readings from 

 a Hersfeld codex, since lost. This, though it is apparently independent 

 of F", is unquestionably a copy of the same archet\'pe. On these matters 

 see C. U. Clark, The Text Tradition of Amin. Marc, New Haven, 1904. 



* The question whether he absolutely avoided the clausula A (v. p. 175), 

 and the cadence I made in a single word (v. p. 185). 



