¹ I am most grateful to Professor C.O. Brink for reading an earlier draft of this paper and making various helpful comments, to Mr. R. H. Martin and Mr. J. N. Adams for certain corrections, and to my wife for checking some of the figures in the statistical appendix. The paper has also benefited from discussion after it was read to the Oxford Philological Society.

² Wölfflin, E., Philologus 25 (1867), 92–134Google Scholar; 26 (1867), 92–166; 27 (1868), 113–49.

³ Nutting, H. C., ‘The use of “forem” in Tacitus,’ Univ. of California Publ. in Class. Phil. 7, no. 6 (1923), 209 ffGoogle Scholar.

⁴ Löfstedt, E., Syntactica 2, Lund 1933, 276–90Google Scholar.

⁵ Eriksson, N., Studien zu den Annalen des Tacitus, Lund 1934Google Scholar, passim.

⁶ Syme, R., Tacitus, Oxford 1958, 711–45Google Scholar. It is appropriate to record here how much I am indebted to Syme's treatment of the subject, both for the information supplied and the ideas suggested.

⁷ It is possible to represent this development by a graph, showing the chronological sequence of Tacitus' writings and movement from ‘normality’ to σεμνότης. But, as my argument will imply, any such presentation would be far too schematic and over simplified to be of serious use.

⁸ R. H. Martin has raised one or two pertinent objections to the view propounded by Löfstedt and Eriksson. See notes 11 and 13 below.

⁹ This too is far too simplified a picture to be wholly acceptable. There is not an even spread of poeticisms in Tacitus; rather he accumulates them in particular passages (e.g. Ann. 1, 65) for special effect. Much the same applies to Livy, as R. M. Ogilvie has suggested in various notes on Livy 1–5. Nevertheless, granted that poeticisms tend to accumulate in particular passages, it may still be possible to find a general change in their use. The analogy of developments in Livy deserves further attention, There, too, no simple formula seems likely to present the truth.

¹⁰ For a fuller examination of the distribution of quis and quibus see R. H. Martin in CR 1968. He shows that the evidence is not so straightforward as has been supposed.

¹¹ A very little perhaps. See Lindholm, E., Stilistische Studien. Zur Erweiterung der Satzglieder im Lateinischen, Lund 1931, 196 ffGoogle Scholar.

¹² Martin, R. H., ‘Variatio and the development of Tacitus' style’, Eranos 51 (1953), 89–96Google Scholar.

¹³ Syme maintains that, though much Sallustian influence may still be detected in 13–16, its extent has diminished. But he produces no wholly cogent evidence for such diminution. For instance, as to one matter he mentions, while it is true that there are character-sketches in 13–16 not wholly in the manner of Sallust, so there are elsewhere in Tacitus' writings. On the other hand nothing could be more Sallustian than the characterization of Poppaea at 13, 45, 2–4.

¹⁴ Martin, R. H., ‘-ere and -erunt in Tacitus’, CR 60 (1946), 17–19Google Scholar.

¹⁵ Many more examples of various kinds may be found in Syme's appendixes 45 and 46.

¹⁶ It seemed unnecessary to take account of the influence of the position in which the examples occur, i.e. whether before a vowel or a consonant, because, since so many examples are involved, any such influence is likely to even out. I have noticed a slight tendency for examples of nec and neque to cluster (probably through no other reason than subconscious harking back to a word used shortly before—the phenomenon is, of course, known elsewhere), but these clusters cancel one another out and do not affect the validity of figures taken from very numerous examples.

¹⁷ To what extent change of source may have contributed to change of style is hard to assess. Certainly Tacitus' historical sources could sometimes serve as stylistic models—at least Tacitus seems prepared to borrow from them an epigram or turn of phrase. But I doubt whether change of source could possibly affect the basic material of his language. And, though a partial change of source is likely after book 12, we know nothing of possible stylistic differences between Tacitus' sources.

¹⁸ Cf. Sherwin-White's commentary on Pliny, p. 123.

¹⁹ Much still remains to be investigated before any really firm conclusion can be obtained. Amongst matters which might reward further exploration and supply further means to gauge stylistic change are Tacitus' use of brevity and ellipse, of asyndeton, and of the ablative absolute.

²⁰ I have excluded words which occur ten times or more in Tacitus and a few which, though they occur less than ten times in Tacitus, are particularly common elsewhere.

²¹ The sample is of all ‘new’ words beginning with the letters a, b, c, d, e, and s. I have excluded all ‘new’ words which are highly specialized, i.e. by their nature likely to occur only very rarely. ‘New’ means ‘not attested in any earlier part of the writings of Tacitus’.

²² One might expect a progressive diminution in the number of ‘new’ words and, in so far as we do not find this in 13–16, see evidence here for a substantial change of vocabulary in 13–16. But most of the very common and basic words have appeared already in the Histories and earlier. My figures relate not to them, but to words which are not indispensable and therefore tend to be occasional in occurrence.