The assumption that a certain number of errors are original, that is to say, go back either to the author’s own autograph or to the transcript first made from it, is in itself probable: we know in fact that some which appear in every copy, without exception, of the first and second recensions at length receive correction by erasure in Fairfax 3. So far as we can judge, the text of the _Confessio Amantis_ during its first years exhibited a steady tendency to rid itself of error, and the process of corruption in the ordinary sense can hardly be said to have set in until after the death of the author. There are a large number of various readings in the case of which we find on the one side the great majority of first recension MSS., and on the other a small number of this same type together with practically the whole of the second and third recensions, as, for example[AF]: i. 2836 to H₁XERCLB₂ do AJMG, SAdBΔΛ, FWH₃ 2847 be _om._ H₁XGERCLB₂ _ins._ AJM, SAdBΔ, FWH₃ 2953 wele H ... B₂ weie AJM, SAd BΔ, FWH₃ 3027 preieth H₁ ... B₂, W braieth, AJM, S ... ΔΛ, FH₃ 3374 an Erl hier H₁ ... B₂, Λ mad a Pier AJM, SAdBΔ, FW (H₃ def.) 3381 place H₁ ... B₂, BΛ maide AJM, SAdΔ, FW (H₃ defective) ii. 833 that diere H₁ ... B₂, B that other AJ(M), SAdΔΛ, FWH₃ iii. 12 euermore H₁ ... B₂ enemy AJM, SAdBTΔ, FWH₃ 354 I may H₁ ... B₂ he may AJM, SAdBTΔ, FWH₃ iv. 109 day H₁ ... B₂, H₃ lay AJM, SBTΔ, FW (Ad def.) v. 316 thanne (than) H₁ ... B₂, Δ hom AJM, SAdBTΔ, FWH₃ 368 And for no drede now wol I wonde H₁ ... B₂, Λ In helle thou schalt understonde AJM, S ... Δ, FWH₃ cp. 394, 424, 786, &c. 2694 Whan that sche was bot of ȝong age For good ERCLB₂ That only for thilke avantage Of good AJMH₁XG, S ... ΔΛ, FWH₃ 2771 nyh _om._ ERCLB₂ _ins._ AJMH₁XG, S ... Δ, FWH₃ 3110 burned as the silver ERCLB₂ burned was as selver AJMH₁XG, S ... ΔΛ, FWH₃ cp. 3032, 3246, &c.