The bulk of the synchronistic data used to recover Roman chronology can be classified as "seasonal". Virtually all of the chronological data for the third century B.C. is of this type. All of these synchronisms come from literary sources, notably most Polybius and Livy, but some also come from Cicero, Caesar and Plutarch.
Two types of source statements fall into this category:
An event is explicitly tied to a point in a season: the beginning of spring, the onset of winter etc.
- An event is related to another event which is driven by the solar year, most notably the harvest or the sailing season.
Since the Julian calendar is a solar calendar, these correlations allow us to estimate an approximate range of Julian dates for the associated event. The event in turn can be tied, usually somewhat loosely, to a Roman civil date, most commonly the turn of the consular year.
The second class of data is the less problematic. Assuming that the Mediterranean climate has not dramatically changed since ancient times (although it certainly has changed somewhat), it is possible to determine harvest dates and sailing conditions from modern experience. As a cross-check, we also have some ancient agricultural manuals, such as those of Varro, Columella and Cato. In general, the Mediterranean was closed to sailing from mid November to mid March, though in some areas different conditions might apply. Harvest dates vary from area to area, being earlier in more southern regions: early June in Sicily, mid June to early July in central Italy, mid July in Gaul.
Direct seasonal statements are more problematic. One must first establish what the ancient authors understood the seasons to be. Varro, De Re Rustica I 28, writing in A.U.C. 717 = 37, gives two definitions, and equivalent dates according to the Caesarian calendar. Varro first proposes a definition based on the sun's motion through the zodiac, with the seasons starting when the sun is in the 23rd degree of Aquarius, Taurus, Leo and Scorpio, respectively, resulting in dates and durations as follows:
Spring: 7 February -- 91 days
- Summer: 9 May -- 94 days
- Autumn: 11 August -- 91 days
- Winter: 10 November -- 89 days
But he then provides a second definition, based mostly on astral events which are much easier to observe, which divides the year into 8 parts with different dates:
From the Favonius (a spring breeze) to the vernal equinox: 45 days (7 February to 24 March)
- From the vernal equinox to the rising of the Pleiades: 44 days (24 March to 7 May)
- From the rising of the Pleiades to the summer solstice: 48 days (7 May to 24 June)
- From the summer solstice to the rising of the dog star: 27 days (24 June to 21 July)
- From the rising of the dog star to the autumnal equinox: 67 days (21 July to 26 September)
- From the autumnal equinox to the setting of the Pleiades: 32 days (26 September to 28 October)
- From the setting of the Pleiades to the winter solstice: 57 days (28 October to 24 December)
- From the winter solstice to the Favonius: 45 days (24 December to 7 February)
Clearly these definitions are not aligned. (And the lengths of segments 6 and 7 seem to me to be in error since the setting of the Pleiades is usually dated 7/8 November -- suggesting the numbers should be 42+47 rather than 32+57 days.) The second set of definitions, however, appear the more likely to have been used in practice.
Polybius, on the other hand, only discusses two seasons, usually translated as Summer and Winter. According to L. Pedech, La méthode historique de Polybe, 462, these were delimited by the rising and the setting of the Pleiades. While this date is generally used, V. W. Warrior, AJAH 6 (1981) 1 at 24, has argued that Polybius' seasons were more loosely defined than this, reflecting the actual conditions conducive to military and political activity in any given year.
One result of this difference in practice is that an author, such as Livy, may draw on sources using different seasonal definitions. Hence, great care must be taken to determine which source he is using and, where possible, what that source understood the season to mean.
Other issues can complicate interpretation. A seasonal statement which is not clearly tied to a seasonal event (e.g. "at the beginning of spring the consuls....") may be formulaic rather than chronological. An author may extrapolate seasonal data based on false assumptions. For example, Plutarch's description of the circumstances of the battle of Pydna in 168 are clearly embroidery based on a misapprehension that Livy's date, a.d. III Non. Sept, was roughly equivalent to the Julian date, and hence in high summer (early September), not early summer (mid June), when it actually fell.
Finally, it is very common for an event that can be related to the Roman civil year not to be directly associated with a seasonal indication. Hence it is often necessary to make some ancillary estimates. As a typical example, we might be told that a consul was in place A at the harvest, where he requisitioned supplies, and then marched his army to place B to fight a battle, and then returned to Rome to conduct the election for the next year. The points of synchronism are, on the Julian side, the harvest (e.g. early June) and, on the Roman side, the election (usually around Ianuarius when the consular term started on Id. Mart.). However, in order to establish the actual synchronism, it is first necessary to estimate how long it would have taken for the consul to march from A to B, fight the battle and return to Rome. This kind of estimate is typically based on assumed statistics such as the ability of an army to march 30-35 km a day (Vegetius 1.8).
Clearly, this type of data is inherently imprecise. Given the variability of Roman intercalation, it is simply not possible to use it to determine exact year lengths. However, a seasonal datum can usually be limited to a period of a few weeks, which is about the size of a Roman intercalation. For this reason, such synchronisms are frequently sufficiently precise to allow the number of intercalations between the datum and some other point to be estimated with a fairly high degree of probability.
Roman chronologists typically estimate the number of intercalations between two successive seasonal correlations, essentially assuming that the Julian distance between the two events can be estimated by the distance between the mean Julian dates for each event. This is generally reasonable but is clearly subject to the risk that each event may have been closer to an extreme limit of the possible range. Additionally, in building a chronology on the basis of such incremental determinations, an error made in estimating any individual increment may propagate through all earlier events.
In order to minimise this risk, and to improve the probability of the estimates being correct, I have here used a slightly different technique, where possible: estimating the total number of intercalations between each seasonal synchronism and the nearest Roman year whose Julian dates are certain. In practice, this can usually only be done for the third century B.C. where it means estimating the number of intercalations between a given year and A.U.C. 564 = 190. Given a series of such estimates, the reconstruction is then derived from the differences between them. In this way, an error in any individual synchronism is prevented from propagating through the entire series. This technique also has the advantage that it allows the chronologist to estimate directly the error range in Roman/Julian conversion for any given model of intercalations in the year to which a synchronism applies. Because a Roman intercalation was either 22 or 23 days long, each additional intercalation adds an extra day of uncertainty to the Roman side of a synchronism.
The standard practice of Roman chronologists is to assume that intercalations are alternately 22 or 23 days, as described by Censorinus and Macrobius. As far as I can determine there is no clear historical justification for this. However, this procedure is still used here, since it does give the median correlation in the error range of possible correlations; and it may in fact represent third century practice.
Once the error range exceeds 22 days, the aggregate number of intercalations between a seasonal synchronism and a fixed point is theoretically subject to an additional source of uncertainty, since 22 intercalations of 23 days is the same amount of time as 23 intercalations of 22 days. Since there are only two seasonal synchronisms in the period 168-78, both are very loose, and both are quite far removed from fixed points, this factor means that it is pointless to use seasonal synchronisms to estimate the number of intercalations to a fixed point, or to provide error bounds, in this period. However, this problem is greatly mitigated in the third century by the fact that the seasonal synchronisms are typically only separated by a few years from each other. Since the error in the estimate of the Julian side of the equation is typically of the same order as the Roman side at this point, closely related Julian synchronisms act as a rough governor to limit the growth of the combined error.
Website © Chris Bennett, 2001-2011 -- All rights reserved