Tacitus’ Tiberius: The state of the evidence for the emperor’s Ipsissima Verba in the Annals By

Although Syme never actually tried to test his hypothesis, N. P. Miller did, and published her results in 1968's American Journal of Philology. In "Tiberius Speaks" she compares the number of "solitaries" (that is, words that appear only once in the extant Tacitean corpus) occurring in the speeches, both recta and obliqua, attributed to Tiberius, with the number occurring in the speeches attributed to Nero. She con? cludes, "... we may... with some reason postulate that the high propor? tion of Tiberian solitaries indicates a connection with Tiberius' [actual] speeches" (19). Although Syme's and Miller's hypothesis has not been universally accepted, neither has it ever been fully rejected, and it has continued to haunt a small corner of Tacitean scholarship. All subsequent opinion on the subject, moreover, has assumed at least that Miller's evidence is

even if her conclusions are not. 2 I wish to show here, however, that Miller's methodology in "Tiberius Speaks" is flawed, and that there are absolutely no stylometric data in support of the theory.
I begin with a review of Miller's data. Table 1 is from her 1968 article, in which she compares solitaries from those books of the Annals covering the principates of the emperors Tiberius (books 1-3) and Nero (books 13-16). Miller's data do seem to support the notion that solitaries appear in the speeches of Tiberius with much greater frequency than they do in the speeches of Nero. In fact, she could have made her comparison clearer by leaving out the "emperor-referring" category entirely, and simply comparing the number of solitaries in the speeches to the total number of solitaries in the relevant books. In the Tiberian books, using her own data, 8.5% of all the solitaries appear in Tiberius' speeches, as opposed to a meager 2.7% appearing in Nero's speeches. By either analysis, there is a roughly threefold difference in the frequency of solitaries.
Unfortunately, Miller has failed to account for the difference in the total amount of speech attributed to each emperor. Tiberius has, in fact, nearly three times as many lines of speech, recta and obliqua, as does Nero. Using Miller's own data, it is possible to devise a more accurate and straightforward frequency test, simply comparing the total number of lines spoken in the speeches to the number of solitaries occurring in them. This comparison shows that the frequency of solitaries in Nero's speeches is nearly identical to that in Tiberius' speeches, with Nero's actually having a negligibly higher rate. The threefold difference that Miller uncovered is not, in fact, attributable to any difference in the frequency of solitaries in the speeches of Tiberius and Nero, but is only an indirect measurement of her sample sizes. A comparison of these same speeches with the whole of books 1-3 and 13-16, furthermore, reveals that the speeches of both emperors have a slightly higher frequency of solitaries than the rest of the text (which has a frequency of five solitaries per hundred lines), but the difference is statistically insignificant. 4 The obvious conclusion to be drawn is that solitaries seem to appear in the Annals at a fairly constant rate, regardless of who is speaking.
But the "Tiberian" hypothesis is not based solely on the presence of solitaries: Syme (1958, 700) and Miller (1968, 14) also think that the prevalence of certain abstract nouns "of a Ciceronian type" is an indication that Tiberius' speeches are truly "Tiberian." Since neither author undertook any stylometric tests of this assertion, I did so by counting all abstract nouns in the speeches of Tiberius and Nero. 5 3 In general, the higher the chi square value, the greater the probability that the populations being compared differ from one another because of non-random factors. The p value, which is related to the chi square, is a measure of the probability that the differences between the populations are due to chance alone. Thus, in this sample, the p value indicates that there is a greater than 80% probability that the differences between the speeches of Tiberius and Nero are due to chance. 4 The figures for the line counts and the number of solitaries are taken from Miller 1968, 7, 11, and 19. 5 The speeches of Tiberius appear at Annals 1. 7,11,25,47,69,73,74,78;2.26,36,38,40;3.6,12,24,47,[53][54]56,65,69,73;4.8,14,15,16,30,[37][38]40,52;6.2,3,6,20,29,46. Those for Nero appear at Annals 13.4,14,17,27,51;14.7,11,22,49,[55][56]59,62;15.36;16.7. The criteria I used for identifying "Ciceronian abstracts" are purely morphological. Nouns of the following types were counted: those with nominative ending intas (e.g., iniquitas), those ending in -do (e.g., magnitudo), those ending in -io (e.g., condicio), and those ending inia (e.g., licentia). This is a not altogether satisfactory solution to the problem posed by the notion of "Ciceronian" abstracts, as it necessarily includes some nouns which lack any semantic abstractness (e.g., curia, pecunia), whatever their morphological characteristics, and it likewise excludes a number of nouns which are semantically abstract, but do not fit into those morphological categories. Nevertheless, I believe it is a fair measure of what Syme and Miller are getting at, and that it is the best solution available, first because the semantically non-abstract words that fall into these morphological classes are common enough that they are likely to be evenly distributed throughout the text, and hence should not bias the result, and second, the semantic criteria for "abstractness" are simply too slippery to be fairly and accurately applied.
Again, contrary to the hypothesis, the frequency of abstract nouns in Tiberius' speeches is actually lower than in Nero's, although the difference is not great enough to be statistically significant. 6 Miller's final argument concerns the use of compound verbs; she says they are "a conspicuous element in the vocabulary of the Tiberian speeches," and lists a number of verbs, five of which are solitaries, commenting, "Some of these verbs come very close together in the text, and the suggestion is strong that they represent a known characteristic of Tiberian language, and may even come from his own speeches" (15-16). I tested this argument by counting all verbs, including participles and periphrastic constructions, in the same passages used for Table 3. The results were as follows: Again, in contradiction to Miller's thesis, the ratio of compound verbs compared to all verbs in Nero's speeches is nearly identical to that in Tiberius'.
Obviously, the speeches of Tiberius and Nero do not differ from one another with respect to the criteria offered. It is possible, however, that Tacitus is employing the same stylistic strategy in the speeches of both emperors, and that the vocabulary of the speeches differs from narrative passages with respect to the criteria identified. 7 To test this hypothesis, I chose six chapters at random from the Annals (3. 36, 6.44, 11.28, 12.60, 13.3, 15.5) and counted the number of abstract nouns and compound verbs: Clearly, Tacitus' preferences for these word types is quite consistent throughout the Annals.
In fairness to Syme's thesis, however, I should point out that his hypothesis about Tacitus' use of the acta is somewhat more restricted than Miller's. He says (1958, 702), in Tacitus, Tacitus began by exploiting the acta in the main as a source for items of public business. He soon grew bolder, but it was not until he reached the end of the year 16 that he decided to reproduce an imperial oration. There are five such renderings in direct discourse.
The speeches that Syme singles out as composed directly from the acta appear at Annals 2. 38, 3.12, 3.53, 4.8, and 4.37. Testing for the frequency of abstract nouns and compound verbs in these speeches, I obtained the following results: Here again, there is no significant difference between the speeches singled out by Syme and the control text, although the frequency of abstract nouns and compound verbs in Syme's group is somewhat lower, not only than that of the control text, but also than that of Tiberius' speeches as a whole (see Table 5).
In sum, not one of the hypotheses advanced concerning "Tiberianisms" in the Annals is supported by the data. The speeches of Tiberius, in fact, maintain a fairly even consistency of style with the rest of the Annals when measured by the criteria suggested. The data presented here do not prove, of course, that Tacitus did not draw upon the acta senatus in composing Tiberius' speeches; they indicate only that the alleged peculiarities of Tiberius' speeches are not measurable by the stylistic criteria proposed. It may still be possible, by subtler and more sophisticated stylometric analysis, to discern a distinct "Tiberian" style in the speeches, but even if this were to be done, we would still be left with troubling questions. How could we tell whether these stylistic peculiarities came from the acta senatus or some other source, such as Tacitus' own literary imagination? Or, assuming that the acta were the source, how would we know whether it was Tiberius' or the senatorial secretary's style that was being reproduced in the speeches? Such problems are not soluble by stylometric means, and, given the state of the evidence so far, further inquiry into the question along these lines would not appear to hold much promise.