^{page 159 note 2} This word is convenient, but of course inaccurate.

^{page 159 note 3} Most, if not all, of my ‘tracings’ have been done at least twice, in many cases at an interval of years.

^{page 160 note 1} See his Bacchylides, passim.

^{page 160 note 2} E.g. Sappho und Simonides, p. 49.

^{page 160 note 3} E.g. 1231.

^{page 160 note 4} It is possible, I think, that Hibeh Papyrus 17 might be more easily read in a photograph taken with a ‘panchromatic’ plate.

^{page 160 note 5} Cf. ProfessorHousman, , Proc. Class. Assoc., 1921, p. 77Google Scholar: ‘There is one foolish sort of conjecture…. The practice is, if you have persuaded yourself that a text is corrupt, to alter a letter or two and see what happens.’

^{page 160 note 6} ναβαισατρεῦ.

^{page 160 note 7} Alexis, 2. 385K; his own suggestion accounts for three words out of two and a half lines at the cost of an unlikely elision.

^{page 160 note 8} Schol. on the Triballian episode Ar. Av. 1629.

^{page 161 note 1} Some of Hoffman's conclusions have been proved wrong by later discoveries, but as a collection of the evidence available in 1893 his book has never been superseded.

^{page 161 note 2} Note that Johannes Grammaticus' ends with the words .

^{page 161 note 3} Such as occur in the same MSS.

^{page 161 note 4} Though this is hardly necessary in view of Thessalian and Homeric parallels; cf. ἐσκέραμεν, Alc. 156. 1, which escaped Mr. Lobel.

^{page 161 note 5} The list, one of the results of six weeks' study of the Bacchylides Papyri, is before me as I write—e.g. [σέυον]το, 3. 5 (so also Süss in Teubner text) should certainly be [ἵεν]το.

A more detailed reply will be printed in the Proc. Camb. Philol. Soc.