190 Austiji Morris Harmon, 



would show itself in the diss3dlal)les, the other in the antepenulti- 

 mates. The presumption is therefore that the two different ratios 

 which we actual!}' find are these two natural ratios of frequency ; 

 and this presumption is not difficult to confirm. 



In order to establish the natural proportion of longs to shorts 

 under accent a count in the dictionary is the first means that sug- 

 gests itself, but not the best, for it does not take into consideration 

 the relative frequency of words. I made a small test, however, on 

 this basis. For the dissyllables I selected the letter F in order to 

 avoid prepositions, which would cause an abnormal proportion of 

 long penults, and found that of the dissyllables beginning with that 

 letter 24 per cent had a short accented syllable. For the antepen- 

 ultimates I counted three pages at random under each letter of the 

 alphabet, and found that 42.6 per cent had a short antepenult. A 

 better means of proof would be afforded by prose writers, could 

 we hit upon one who did not use ciausulae.^ Unfortunately it is 

 only in modern Latinity that we can be sure, without investigation, 

 that quantity is neglected, and I therefore selected Ritschl's Latin 

 speeches {Opusc. V, 627—684) for examination. The count was made 

 in the last two or three words preceding each heavy stop, and 

 yielded 267 dissyllables, of which 25 per cent had a short penult, 

 and 358 words with antepenultimate accent, of which 43 per cent 

 had the syllable short. For the dissyllables it was also possible to 

 make a test in hexameter verse, which admits all the varieties 

 of scansion that they can assume. Choosing for this purpose the 

 first Epistle of Horace, I found that 24 per cent of the diss3ilables 

 occurring in it had a short penult. 



Of course we must not press these figures too close, for they 

 are all of them founded on small counts. But they certainly show 

 clearl}' that the natural proportion of short accented syllables to 

 long is much lower in diss3'llables than in antepenultimate words, 

 and that in the case of dissyllables this natural proportion must be 

 about 3 long to 1 short, in the case of antepenultimates about 3 

 long to 2 short. Consequently the ratios which we found in Am- 

 mianus accord with the natural ratios. 



To remove any lingering doubt we may pursue the investigation 

 a little farther. Up to this point we have lumped all antepenulti- 



1 Tliinking that this might be the case in Tacitus I counted the 

 dissyllables occurring in the chapter-endings of the Histories and found 

 a proportion of 36 per cent short ! This heavy proportion is abnormal — 

 it is appai-ently due to his avoidance of the cretic-trocliaic clausula. 



