Astronomicon

corrected by a hand somewhat resembling L². But it had many corrections which L² has not, it lacked many which L² exhibits, and its corrections were often different from those of L²; and the primitive substance of its text was derived not from L but from α. For example in iv 318 'extremam Erigonae tribuit' L has erigonem, G trigone. Neither is derived from the other; they are equidistant from the truth, and point to erigone or erigonē in the common source.

To determine in such cases the reading of α we can sometimes employ the joint witness of M to the reading of the archetype. iv 414:

α had ibi touit or something of the same length; γ could not read it and left the proper space; L read it wrong and reproduced it inexactly. L² wrote ouit* overhead, of which more anon.

III 432 munere duty/gift] munero M, numero L, munera G. α, like M, had munero, which in L and G is corrupted separately and differently.

IV 519 feminei incedunt nec female they walk nor] feminea iuceat nec M, feminea iace (space of 3 letters) nec G, feminea iacet nec L. α had iaceat (as L² corrects) or iaceant.

IV 894 mundus in ipsis L, mundus ipsis GM and almost certainly α; for it is evidently much more probable that the unmetrical reading was in the archetype and that L made the obvious correction, than that in was twice omitted. "If it could be shown that M and G agree in corruptions which are corrected by first-hand conjecture in L, then I would be proven wrong," said Mr Thielscher Philol. LXXXII p. 173.

The many cases—far more than are registered in my apparatus criticus or need to be registered in any—where GM agree in truth and L is in error are not certain proof of G's separate derivation from α; for it will appear hereafter that G has derived true readings from β or some source other than α, and some of its true readings are such as may be due to conjecture. But a valid argument may be drawn from sundry agreements in orthography.

G's orthography in general, compared with L's, is modernised and vulgarised. It tends to assimilate the preposition in compounds where L does not, and to give -es in accusatives plural where L gives -is. It is natural to infer that in such places the spelling of α was that of L and not of G; and the inference becomes more than natural, it becomes virtually certain, in the many places where M agrees