through some sort of calculation.^58 The Diaries clearly imply in these cases
that sunset and moonset, and therefore the sun and the moon,shouldhave
been visible on that evening if not for adverse atmospheric conditions; thus
what was being predicted (or postdicted) in these cases was not only the length
of the sunset–moonset lag but also, effectively, thefirst appearance of the new
moon. This leads us to conclude that some months began with an actual new
moon sighting, and others, with a new moon prediction.
The accuracy of new moon sighting and prediction in the Astronomical
Diaries and their relevance to the calendar have been discussed in detail in an
article (Stern 2008) which I shall present here with minor corrections and only
in summary. The lunar accuracy of new moon sightings and predictions in the
Diaries can be established on the basis on modern astronomy.^59 A small
proportion (c.1.5%) of new moon sightings occurred one day before the new
moon should have been visible; these were, presumably, false sightings.^60 More
frequent are new moon sightings that are one day late (c.6.5%), presumably as
a result of bad weather on the previous evening.^61 These results are evenly
spread throughout the period of the Diaries, but mainly during the fourth–first
centuriesBCE.
Of the non-sighted, predicted new moons, late predictions seem not to be
attested. Predictions that are one day early (i.e. before the new moon should
have been visible according to modern astronomy) amount to 6.5–10% of
cases. This suggests a tendency for early new moon prediction;^62 which in turn
suggests, very importantly, that new moon predictions are unlikely to have
been backed up, as in the neo-Assyrian period, by new moon sightings from
other localities.^63 Calendar months beginning on an evening when, according
to the Diary, the new moon was not sighted but only predicted are therefore
(^58) Sachs and Hunger (1988–2006) i. 21–2 refer only to‘predictions’, but the possibility
sometimes of corrective postdictions cannot be entirely excluded. How sunset–moonset–lag
predictions were made is unknown, but various models have been proposed by Brack-Bernsen
(1997), (1999) and Brown (2000) 173–89, and one text, TU11, lays down a range of prediction
rules (Brack-Bernsen and Hunger 2002). 59
For an explanation of the methodology, see Stern (2008) 38–9 nn. 17–20.
(^60) On the possibility of false sightings, see Doggett and Schaefer (1994), Stern (2001) 110–11,
and Hoffman in http://chem.ch.huji.ac.il/nmr/foo/ppt/nmoon_files/frame.htm; it must be as-
sumed, however, that the Babylonian astrologers were well trained to avoid them. One of the
‘early’sightings listed in Stern (2008) 24 is erroneous: in–194 (= 195BCE) the entry should be
7 June (not 6; my error), and thus the moon sighting was on time, not early; this brings down
the percentage to about 1.2%.
(^61) One of the‘late’sightings listed in Stern (2008) 26 is erroneous: in–145 (= 146BCE) the
entry should be 7 February (not 8; Sachs and Hunger’s error), and thus the moon sighting was on
time, not late; this brings down the percentage to about 6.2%. I am grateful to Victor Reijs for
alerting me to both these cases.
(^62) The criterion (or criteria) used by Babylonian astronomers in the Diaries for predicting
new moon visibility remain(s), however, unknown; see discussion above, nn. 19 63 – 20.
With one possible exception: see below, n. 71.
The Babylonian Calendar 89