6. According to Geminus's elaborate report (
Gemini elementa astronomiae, ed. and transl. Manitius C. (Leipzig, 1898), 110ff), the astronomers of the school of Euctemon, Philippus and Callippus, after having stated the faultiness of the octaëteris, “brought forth as another period the 19-year cycle. They had found from observation (
), namely, that in 19 years, 6940 days or 235 months are contained, including the intercalary ones, and that there are 7 intercalary in these 19 years.” It is not clear what those astronomers, in order to test the equation “19 years = 235 months”, had actually observed and found equal to a round number of days, 6940: The length of 19 years or that of 235 months. As for the term ‘year’, it goes without saying that it is always the year defined by the recurring seasons which is intended. Since, however, this ‘tropical’ year is only 20 minutes shorter than the sidereal, it requires very careful observations to recognize that the two are not identical. In order to assess the relation between
tropical years and days, considering that a number of days without fraction was found, the Greek astronomers must have selected equinoxes or solstices 19 (or
n × 19) years apart. This is very improbable. At any rate the year length resulting, 365 5/19 = 365·2632
d, is even higher than that of the sidereal (365·2564
d) and comes close to the Babylonian sidereal year of System A:
PA = 365·2606 (see below). This might indicate that actually not tropical but sidereal phenomena (heliacal risings, etc.) had been observed and that observations of the cardinal points (as Euctemon's of the summer solstice of —431) served only to establish an approximate connection between the seasons and the sidereal calendar, such as attested by the
parapegmata ascribed to Euctemon, Eudoxus and Callippus. If, however, they were to have observed the following two pairs of lunar eclipses immediately preceding Euctemon's solstice observation of —431: —454 Nov. 9, 19;20
h M.T. Athens —435 Nov. 9, 18;45 M.T. Athens Δ = 6940
d—35m —450 Mar. 4, 21;00 M.T. Athens —431 Mar. 4, 21;20 M.T. Athens Δ = 6940
d+20m they would have felt entitled to vindicate a strict identity between the numbers of years, days and months as mentioned by Geminus since the excess of —35
m and +20
m over 6940
d is not observable. Indeed there is an overwhelming number of such pairs (or rather triplets since there are always three, apparently never more, in one series), but not all of them of course yield the same result. Thus that of —448 Aug. 6, 7;40
h (maximum phase not observable in Athens) was preceded by another on —467 Aug. 6, 19;00
h, the difference being only
c. 6939·5
d, from which a year length of 365·235
d (≈ 365 4/17) would have resulted. In the same way the Babylonians could find the named period relation roughly confirmed by a similar triplet of lunar eclipses, the last of which occurred on —502 Jan. 10,
c. 75
d before the introduction of the 19-year cycle: —540 Jan. 10, 4;51
h M.T. Babylon —521 Jan. 10, 4;32 M.T. Babylon Δ = 6940
d—19m —502 Jan. 10, 4;56 M.T. Babylon Δ = 6940
d+24
m Contrary to the Greek procedure, however, their mathematical analysis was based exclusively on the exact length of the month, which saved them the detours characterizing the Greek approach. On the other hand, from 235 months = 6940
d there follows the faulty month length, not mentioned explicitly, of 29·5319
d, while Geminus at the outset (pp. 100–1) gives the much better value 29 ½
3 = 29·5303
d, and thereafter (par. 43; the passage may be a later insertion by a scholiast) the exact Babylonian value, 29;31,50,8,20
d = 29·530594
d. Only then, after having told us about the efforts made to construct a lunisolar calendar based on a system of intercalations and omissions of days fitting these inaccurate values, Geminus states that it was found at discord with the “time of the year” (
i.e. the year determined by the seasons), which, as established by “observations stretching over a number of years (
)”, comprises 365 ¼ days. This figure, suspicious on account of the simple fraction 1/4, was eventually accepted by the school of Callippus as the correct length of the (tropical) year. Nineteen such years equal 6939 ¾ days; by division by 235 (not carried out by Geminus) a new (indeed better, though nowhere else attested) value for the length of the month: 29·53085
d, was obtained. The Callippic 76-year cycle as a final result of these repeated modifications of the underlying parameters worked well enough, but the way which led to it seems to evidence an undeniable lack of understanding of the actual problems involved. By contrast, Hipparchus's approach, as demonstrated in the above, is strictly different. He first derives a value for the tropical year, excellent for his time, from his observations of equinoxes and solstices and of the change in longitude of stars near the ecliptic. Only then does he state that, using the best available value (the Babylonian) for the length of the month, the 19-year cycle furnishes the same tropical year length as that resulting from his observations.