would havefirst become visible. Thus the lunar, planetary, and stellar posi-
tions supplied and dated by the Diaries according to the Babylonian calendar
enable us to calculate that in the vast majority of cases the months began when
the new moon wasfirst visible.
Sighting and predictionIn the Babylonian calendar, the length of the month was only either 29 or 30
days (other month-lengths are unattested in the sources). The beginning of the
new month could therefore occur on only one of two days, the determination
of which depended, as explained above, on thefirst appearance of the new
moon. The problem was how to determine when the new moonfirst appeared.
This could be achieved empirically, by scanning the evening sky above the
western horizon in the area of the path of the moon, 29 days after the last new
moon had been sighted. But new moon sighting was not always possible, for
example, in poor weather conditions. In such cases, new moon visibility could
instead have been predicted.
The practice of predicting new moon visibility, or rather, the length of
forthcoming months (itself dependent, presumably, on new moon visibility
prediction) goes back to an early period. In seventh-century astrological texts,
month- length predictions are already well attested^16 and seem to have been
inferred, rather inaccurately, from the dates of previous new and full moons.^17
A fourth-centuryBCEtext suggests that month-lengths may have been similar-
ly inferred from the dates of lunar eclipses (which occur at the full moon),
predicted in advance at 19-year intervals.^18 By this period, however, more
advanced methods had been developed. One source lays down various rules
for inferring month-lengths from predicted sunset–moonset lags on thefirst
evening of the month.^19
It is not known which of these rules—if, indeed, any at all—were ever used
in practice for predicting the new moons and hence the lengths of Babylonian
(^16) e.g. Parpola (1970–83) i, nos. 45,70, ii. 54; Hunger (1992) nos. 46–7, 58–60, 83, 257, 267,
516.
(^17) Parpola (1970–83) ii. 53; Beaulieu (1993) esp. 67, 72–6; Brown (2000) 198–200.
(^18) Neugebauer and Sachs (1967) 205; see discussion in Stern (2008) 36–7 n. 10.
(^19) This text is TU11 (Brack-Bernsen and Hunger 2002), which comes from late-3rd-c.BCE
Uruk but is thought to be of pre-Seleucid origin (ibid. 6). It includes a number of rules for
predicting month-lengths which are mostly based on predicted sunset–moonset lags, and are
clearly related to new moon visibility: thus the rule that if the sunset–moonset lag is less than 10º,
the 1st of the month will be postponed to the next evening (and the old month will be full: obv.
33, 37, see ibid. pp. 35, 43–5), is clearly based on an assumption that if the lag is less than 10º, the
new moon will not be visible. According to Brack-Bernsen and Hunger, this rule would achieve
an accuracy of about 95% (pp. 45, 48–50 and n. 51), which is good but still means that one
prediction in twenty would have been inaccurate. See further Stern (2008) 37 n. 11.
The Babylonian Calendar 77