Chanson dc Roland. 115 



I said above that the unity of O was ' practically ' perfect. I used 

 the adverb because I acknowledge one inconsistenc}'. In v. 2357 

 Roland is described as dying " desuz un pin." In v. 2375 the same 

 statement is made. But in v. 2874 Charles finds Roland lying " desuz 

 dous arbres." The background of the hero's death has been de- 

 scribed so vividly that the reader cannot but be offended by this 

 alteration of an important detail. May not this alteration, however, 

 be attributed to some copyist ? Three considerations make such 

 attribution plausible : (1) it is O's only instance of a slip of this 

 kind! ; (2) mediaeval copyists were prone to such errors — it is sur- 

 prising that there are not many more of them in O ; (3) the line 

 in which it occurs shows other evidence of having been tampered 

 with— it is incomplete, the last words having been erased. 



Coherence, also, is evident throughout the Oxford manuscript. 

 Each stage of it leads naturally, inevitably, to the next. This is 

 not only true of the longer stages into which the poem separates, it 

 is practically so of the shortest ; that is, of the strophes. With very 

 few exceptions the strophes grow out of, and into, one another. 

 It will be more convenient to illustrate the coherence of the longer 

 stages IdcIow, for we intend, when we arrive at the discussion of 

 their composition, to exemplify their excellence not only as units, 

 but as component units. As to the coherence of the strophes, it 

 would be futile to give an illvistration, for whoever desires one needs 

 only to open the Oxford manuscript at random. 



There are, however, three passages in O which are exceptions 

 to the general rule of coherence. The fact that there are only three 

 is an argument for attributing them to copyists ; but not, in itself 

 alone, a final argument, for our French Homer may well have nodded 

 three times. Let us see if there is confirmatory evidence. The 

 first passage, vv. 761—765, is not in any of the extant redactions 

 except O : almost all editors have therefore attributed it, withovit 

 further ado, to O's copyist. The second passage, vv. 1406-1411, 

 not only contains other technical faults unusual in 0, 2 but also 



• It would be liypercritioal, it seems to me, to class v. 877 and v. 3958 

 with the faulty verse discussed above. In these verses the numbers are 

 more effective, from the artistic point of view, than the arithmetically 

 correct numbers would be. 



- These faults are (1) the inartistic precision of the prophecy : cf. infra., 

 p. 120 ; (2) the padded line. v. 1408 : cf. infra, p. 124. Furthermore, the 

 division of vv. 1396—1411 into two strophes is significant ; there is only 

 one other instance in O of this metrical error : vv. 3780—3792. 



