y-recorded years, but also those of the three years of
1886, 1889, and 1891, for, although it would obviously have been
inaccurate to utilize these incomplete records when calculating the
yearly rhythm, there seems no objection to making use of them in the
present section of the inquiry. It is hardly necessary to remark that the
terms "first day of the month," "second day," "third day," etc., are to be
understood as denoting "new-moon day," "day after new moon," "third lunar
day," and so on; but it should be explained that, since these discharges
occur at night, I have adopted the astronomical, instead of the civil,
day; so that a new moon occurring between noon yesterday and noon to-day
is reckoned as occurring yesterday, and yesterday is regarded as the first
lunar day: thus, a discharge occurring in the night between December 31st
and January 1st is tabulated as occurring on December 31st, and, in the
present discussion, is assigned to the lunar day comprised between noon of
December 31st and noon of January 1st.

Since it is obvious that the number of discharges in any one
year--averaging, as they do, only 1.25 per day--are far too few to yield a
curve of any value, I have combined my data in two series. The dotted
curve on Chart 9 is obtained by combining the results of the years
1886-92: two of these years are incompletely recorded, and there are no
records for 1890; the total number of observations was 179. The broken
curve is obtained by combining those of the years 1893-97, the total
number of observations being 185. Even so, the data are far too scanty to
yield a really characteristic curve; but the _continuous_ curve, which
sums up the results of the eleven years, is more reliable, and obviously
more satisfactory.

If the two former curves be compared, it will be seen that, on the whole,
they display a general concordance, such differences as exist being
attributable chiefly to two facts: (1) that the second curve is more even
throughout, neither maximum nor minimum being so strongly marked as in the
first; and (2) that the main maximum occurs in the middle of the month
instead of on the second lunar day, and the absence of the marked initial
maximum alters the character of the first week or so of this curve. It is,
however, scarcely fair to lay any great stress on the characters of curves
obtained from such scanty data, and we will, therefore, pass to the
continuous curve, the study of which will prove more