have survived,^11 it can be reasonably reconstructed as a cycle offive years or 62
months, with 29- and 30-day months occurring mostly in alternation, which
indicates right away a lunar calendar scheme. Every year consists of twelve
months (in most years,five 29-day and seven 30-day months), but two
intercalary months of 30 days are inserted at equal intervals within thefive-
year cycle. The month is divided into two halves, of 15 days and 14 or 15 days;
it would appear that the year is also divided into two halves, summer and
winter, at the beginning of which the intercalary months are placed.^12
The precise reconstruction of the calendar is subject to debate. If we
extrapolate from the surviving fragments and assume the same distribution
of months every year of the cycle, i.e.five 29-day and seven 30-day months,
with an intercalary 30-day month at the beginning of year 1 and in the middle
of year 3, the resulting calendar is quite inaccurate in relation to the moon and
to the seasons. The average year length is 367 days, thus excessive by nearly
two days (more precisely, 1^3 / 4 days), and hence by the end of thefive-year cycle
has an excess of about nine days. The average month length is also excessive:
because of the preponderance of 30-day months, thefive-year cycle ends with
just over four excessive days. These discrepancies are so gross that modern
scholars have generally treated this simple reconstruction as unlikely. After
only four cycles or twenty years, indeed, the calendar will have fallen more
than one month behind the annual seasons; whilst the average month length
will have accumulated a 16-day discrepancy, thus shifting the calendar’s
nominal new moon to the real full moon (and vice-versa).
In order to eliminate the four-day (per cycle) discrepancy from the lunar
phases, scholars have long suggested that in years 2 and 4 of the cycle, the
month of Equos would have counted only 28 days (instead of 30 days, as in the
other years of the cycle). This conjecture is possible because the end of Equos
in years 2 and 4 is not extant in any of the fragments. In support of this
conjecture, it has been pointed out that the month of Equos is designated as
‘not good’(anmatu) in spite of counting 30 days, whereas throughout the
cycle, 30-day months are consistently designated as‘good’(matu) and 29-day
months as‘not good’(anmatu).^13 The designation of Equos asanmatuwould
reflect the fact that in some years it did not count 30 days, but only 28.^14 The
argument for specifically 28 days (and not 29 as in otheranmatumonths) is
stronger in year 2, where the number of festival days at the beginning of the
(^11) Duval and Pinault (1986) 31, Olmsted (1992) xi.
(^12) On the half-month and possible half-year divisions, see Duval and Pinault (1986) 270,
404 – 7.
(^13) On these designations, which in this context should perhaps be given the sense of
‘complete’and‘incomplete’, see Duval and Pinault (1986) 19, 270–5, 409; Olmsted (1992)
12 – 13, 32–3.
(^14) This conjecture, generally attributed to the early 20th-c. scholar Eóin MacNeill, is endorsed
by Duval and Pinault (1986) 406, 411–15; see also Olmsted (1992) 14.
304 Calendars in Antiquity