This is a pedantic post (aren’t they all?). I was reading the relatively recent Budé edition of Symmachus’ speeches and relationes, wherein Jean-Pierre Callu rounds off the first complete translation of Symmachus into French or indeed any modern language (he published the first of the five volumes in 1972 and the last in 2009). Callu dates Symmachus’ first speech to 26 February 368; the 1883 edition of Seeck had placed it on 25 February 369. The principal dating criterion is that the speech explicitly celebrates the fifth anniversary of Valentinian’s accession in February 364, his Quinquennalia (Or. 1.16 lustrum imperialium iam condis annorum, “you now bring to their close a lustrum of imperial years”). It is clear that the whole of the fifth (and tenth, and fifteenth) year of an emperor’s rule was celebrated, so it could be any time within that period from February 368 to February 369. But it is likeliest to belong to the anniversary itself, and judging by parallel cases, the bigger celebration is likelier to be at the beginning of the year than at the end. So we should probably, though without complete confidence, follow Callu and count the five years inclusively: 364, 365, 366, 367, 368.
But on which day, 25 or 26 February? Reference works usually date Valentinian’s accession in 364 to 26 February. Seeck’s invaluable Regesten (published 38 years after his Symmachus edition) places it on 26 February. So does The Prosopography of the Later Roman Empire and the Cambridge Ancient History.
Two sources give us the date: the chronicle known as Consularia Constantinopolitana gives die. v k. Mar. (“the fifth day before the Kalends [the first] of March”, counting inclusively). Ammianus tells us that the Roman empire was without a ruler for 10 days after the emperor Jovian died (which we know from other sources to have been on 17 February). But he also shows why counting inclusively from 17 to 26 February is not a straightforward answer. 364 was a leap year, and Ammianus takes the opportuinitry for a faux-learned digression on the leap year (26.1.8-14). The Romans added the leap-day not at the end of the month of February, but as a second 24 February – or in Latin, bis sextum kalendas Martias. It was therefore called the bissextile day. This was a day of ill-omen, and although Valentinian had arrived in Nicaea in time to be acclaimed on that day, he waited till the day after (26.1.7). If we choose to mark Valentinian’s dies imperii as the Romans did, it will be (as the Constantinopolitan Chronicle tells us), the fifth day before the Kalends of March, or 25 February. If we choose to translate to the modern Gregorian calendar, it will be 26 February. But we should remember that the anniversary would have been celebrated willy-nilly on the fifth day from the Kalends of March: 25 February in non-leap years, the day which we call 26 February only in leap years. So if Symmachus gave his speech in 368, you could describe it as being either 25 or 26 February, depending on your perspective. If he gave it in 369, it was on 25 February.
One final thought. Ammianus devotes a good deal of attention to depicting Valentinian’s reluctance to be acclaimed on a traditional ill-omen. On the fifth day before the Kalends of April (a month later, in Roman terms, 28 March in ours), Valentinian appointed his brother Valens as co-emperor at the suburb of Hebdomon, before processing with him back into Constantinople. As Neil McLynn has pointed out, this was also a significant day: Palm Sunday. Is Ammianus’ emphasis on Valentinian’s significant choice of the day for his accession influenced by his distaste for the day chosen for Valens?