It turns out that Caxton's text is of great value. He followed a MS. (now lost) which is, in some places, even more correct than MS. C. The following readings are of great importance, as they correct MSS. C. and A. (I denote Caxton's edition by the symbol Cx.)
Bk. i. met. 4. 7. Cx. writheth. (Cf. p. xlii. above, l. 6.)
Bk. i. pr. 6. 54. Cx. funden (_but read_ founden).
Bk. i. pr. 6. 65. Cx. norissing. (Perhaps better than _norisshinges_, as in the MSS.; for the Lat. text has the sing. _fomitem_.) Cf. Bk. iii. met. 11. 27.
Bk. ii. pr. 3. 59. Cx. seeld (_better_ selde). It is clear that _yelde_ in MS. A. arose from a reading _[gh]elde_, which really meant _zelde_, the Southern form of _selde_. See below.
Bk. ii. met. 3. 14. Cx. selde (_correctly_). And so again in Bk. ii. pr. 6. 15.
Bk. ii. pr. 6. 63. Cx. may I most. (MSS. C. A. _omit_ I.)
Bk. ii. pr. 8. 17. Cx. wyndy (which is right; see note, p. 434).
Bk. iii. pr. 1. 26. Cx. thyne (_better_ thyn, _as in_ Thynne).
Bk. iii. pr. 10. 10. Cx. denyed (_or read_ deneyed).
Bk. iii. pr. 10. 51. Cx. that the fader. (MSS. that this prince.) Caxton's translation is closer; Lat. text, _patrem_.
Bk. iii. pr. 11. 152. Cx. maistow (Thynne _has_ mayst thou) MS. C. _omits_ thou; and MS. A. is defective.
Bk. iv. pr. 6. 188, 189. Cx. and some dispyse that they mowe not here (_misprint for_ bere). MSS. C. and A. omit this clause.
Bk. v. pr. 1. 9, 10. Cx. assoilen to the the dette (where the former _the_ = thee).
In a few places, Caxton's text is somewhat fuller than that of the MSS. Thus in Bk. ii. pr. 3. 8, Cx. has: thei ben herd _and sowne in eeres_ thei, &c. However, the Lat. text has merely: 'cum audiuntur.' And again, only 9 lines lower (l. 17), Cx. inserts _and ajuste_ after _moeve_; but the Lat. text has merely: 'admouebo.' In some cases, it is closer to the Latin text; as, e. g. in Bk. i. met. 3. 9, where Cx. has _kaue_ (Lat. _antro_), whereas MSS. C. and A. have the pl. _kaues_. In Bk. i. pr. 3. 41, where C. has the E. form _Sorans_, Cx. preserves the Latin form _Soranos_.
It thus appears that a collation with Caxton's text is of considerable service.
THYNNE. Thynne's edition of Chaucer, printed in 1532, contains Boethius. I suspect that Thynne simply reprinted Caxton's text, without consulting any other authority; for it is hard to detect any difference, except that his spellings are somewhat less archaic. Hence this text, by a lucky accident, is an extremely good one, and I have constantly referred to it in all cases of difficulty. Readings from this edition are marked in the foot-notes with the symbol 'Ed.'