^{page 453 note 1} I count the following incidence per total lines in the relevant odes:

1.31 (28 B.C.), 1: 10; 27, 1: 12; 35, 1: 13.3; 29, 1: 16; 37 (30 B.c.), 1: 16; 9, 1: 24; 16, 1: 28; 17, t: 28; 26 and 34, none.

2.9,1:24;200:241 3, 1:28; 7,1:28;14, 1: 28; 17, 1: 32; 19, 1: 32; 1, 1: 40; 5, 1 1, 13, 15, none.

3. 1, 1: 24; 3, 1: 36; 5, 1: 56; 29, 1: 64; 4, 1: 80; 2, 6, 17, 21, 23, 26, none.

4. 4, 9, 14, 15, none.

The progressive diminution of incidence book by book is remarkable. What is the explanation?

^{page 453 note 2} I am grateful to Mr. R. G. M. Nisbet for drawing my attention to an article on the chronology of the early odes by Wilkinson, L. P., who writes (Hermes lxxxiv [1956/1917], 497): ‘It looks as though something had happened in 30 which had impelled [Horace] to give to the world everything he had written hitherto that he thought worth preserving, before embarking on a new poetic enterprise.’ The truth of the words in (my) italics cannot be presumed. Though I accept the probability that the Odes did not overlap the Epodes by much, it is likely that Horace's decision to switch from iambics and hexameters to lyrics (signalized by the publication of the Epodes and Satires 2 in 30) would have been preceded by some experimentation in the lyric metres; and there is no apparent reason why Horace should have published these first beginnings (which he obviously thought successful and worth following up) at the same time. One cannot argue that, because Horace published no lyrics in 30 B.C., therefore he had not yet written any that he prized.Google Scholar

Wilkinson's principal object in this article (which I think he achieves) is to refute a very early date (35-33 B.C.) for Odes 1. 14 (o navis). He supports Torrentius' suggestion of 29/28 B.C. The period between Actium and the summer of 30 must have been one of great hopes balanced by equally great anxieties: perhaps Odes 1. 14 should rather be placed here, between Epode 9 and Odes 1.37 (nunc est bibendum).