This, then, was the basis of the Greek theory of how to structure sentences in a formal speech. (A full exposition of it would be far more elaborate, naturally.) The theory was widely applied in Latin, and not just by orators. For in the ancient world, all reading was reading aloud, and public recitations of poetry and prose works were common. In this context, some writers—notably the perverse genius Tacitus—delighted in disappointing the expectations raised by periodic theory. His Annales starts with what is almost a hexameter line (the classic epic metre), precisely what was not supposed to happen in a well-regulated prose stylist.^* And here is his one-sentence analysis of the decline of history writing during the early Empire:

      TIBERII GAIQVE ET CLAVDII AC NERONIS RES (FLORENTIBVS IPSIS) [OB METVM] FALSAE, (POSTQVAM OCCIDERANT) [RECENTIBVS ODIIS] COMPOSITAE SVNT.

      Of Tiberius and Gaius and Claudius and Nero the events (while they themselves still flourished), [out of fear] were misrepresented, (after they had passed away), [in a setting of recent bitterness] were recounted.³¹

      The translation gives some sense of the rhythm of the whole, but notice how the “limbs” are deliberately mismatched. Three different words for and (underlined) link the four emperors’ names, and then two fairly simple clauses follow, each begun with two short adverbials. In a sense they are parallel (as shown by the parentheses and square backets), but formally they jangle, and the two that ought, by their similar endings (-entibus, -īs), to be parallel and so in contrast, simply are not. Aristotle and Cicero would not have appreciated this monkeying with hard-won stylistic norms. But it only makes sense if readers knew the rules that Tacitus was breaking.

      And to counterweight the periodic balance further, Latin had always had some maxims of its own that emphasized substance over style. They tended to be associated with Cato the Censor (see pp. 70–71). For him, the essence of a fine orator was VIR BONVS DICENDI PERITVS ‘a good man skilled in speaking’. Virtue will out, and damn your technique! Instead of spending too much time planning well-balanced periods, he recommended as preparation for a speech the policy REM TENE: VERBA SEQVENTVR ‘get a grip on the facts and the words will follow’.³²

      Besides these explicit rules on how to put a text together, both the Greeks and the Romans were extremely selective in whose work they accepted as models for good language. They showed extreme prejudice in favour of particular eras, which were thought to have nourished the best writing.

      For Greek, the model was always the language as used by authors writing in the Attic dialect in the fifth and fourth centuries BC. This was the dialect of the city of Athens, and the linguistic mark of quality was naturally discerned in all the writers of this period whose works had survived: they include the tragedians Aeschylus, Sophocles, Euripides; the comedian Aristophanes; the historians Thucydides and Xenophon; the philosophers Plato and Aristotle; and orators such as Demosthenes, Isocrates, and Lysias. “Atticism” has been the ideal that Greeks have striven for in every age over the succeeding twenty-three hundred years until (at long last) the last quarter of the twentieth century. Flexibility and freshness may well have been the virtues that had first made those authors great; they have not been virtues, at least as far as choice of linguistic style is concerned, for any generation of Greek writers until the modern era.

      The value of Atticism remains hard to perceive from outside the charmed circle of the Greek-language community. It really had to do with keeping up technical features of the old language, such as the distinction between genitive and dative cases for nouns, the use of optative mood in inflecting verbs in certain subordinate clauses, and in general the conscious maintenance of words and usages that had come naturally in the fourth century BC, but were stilted already in the first AD. Greeks came to think it a superior way to use the language because of its association with the great works of the past—which none felt able, or perhaps even worthy, to match—and because the ability to use it was the unforgeable proof of a good education. But what to an insider are associations with quality are, to an outsider, difficult to distinguish from sheer obscurantism and snobbery. English has a much shorter continuous written tradition than Greek, but it is as if all modern English writing could only be taken seriously if it employed the spelling and phraseology of the Elizabethan Age.

      For the Romans too the best language was defined ad hominem as the language of the best writers, but they gave themselves somewhat more latitude. Schoolmasters were to define a Golden Age, which covered the century to the death of Augustus in AD 14, and a following Silver Age, which lasted up to AD 150 or 200. The Golden Age included the historians Sallust, Caesar, and Livy, the orator and philosopher Cicero, and the poets Lucretius, Catullus, Virgil, Horace, Propertius, Tibullus, and Ovid. The Silver Age would take in letter writers Seneca the Younger and Pliny the Younger, epic poets Lucan and Statius, epigrammatist Martial, historians Suetonius and Tacitus, satirical poets Persius and Juvenal, novelists Petronius and Apuleius, and the encyclopedists Pliny the Elder and Aulus Gellius. (Curiously, this meant that some of the most widely read authors, the dramatists Plautus and Terence, from the second century BC, were simply timed out.)

      The term classic itself was invented by M. Cornelius Fronto in the middle of the second century AD, on the analogy of the topmost class of citizens: “Go now then, and when you have the time, ask whether ‘QVADRIGAM’ or ‘HARENAS’ was said by anyone from that cohort, at least the more ancient one, whether a significant orator or poet, I mean a significant, ranking [CLASSICVS], landed [ASSIDVVS] writer, not a common sort [PROLETARIVS].”³³

      The focus on a small subset of extant authors as being the sole proponents of good language—and hence fit as models for imitation—was certainly accepted by the Romans; but it was less an