not only in the twenty-five years, the nine years, the five years, and
the four years, but also, and to an exaggerated extent, in the
typical single year. The higher-order effects vary very much both in
magnitude and in position, but the second-order effect is prominent
in all the curves, and, except in the very exceptional years whose
abnormality t it shares, it has its maxima oscillating over the first ten
days of February and August respectively. This oscillation again
shows itself in the gradual decrease of the amplitude of the second-
order effect as the number of years increases whose mean curve is
analysed. Thus for the single year the amplitude is 3'57 F. ; for the
five years it is 2 -06 F. ; for the nine years 1'95 F. ; and for the
twenty-five years, 1'38 F. Thus Table I is in itself a sufficiently
striking demonstration of the existence of a second-order meteoro-
logical effect which combines with the primary solar effect to give us
our normal periodic variation of atmospheric temperature.
The gradual diminution of the amplitude of the second-order oscillation with the extension of the time over which the means are calculated, might perhaps be taken to indicate that if the means were sufficiently extended it would disappear altogether ; and the fact that in hundred-year means for Vienna (which will be referred to later), the second-order amplitude is reduced to one degree, whereas the first-order amplitude is 19'8 F., might be regarded as a further indication of the same fact. In the absence of data extending over the whole period it is, of course, impossible to form a definite opinion, but if the arrange- ment were so that the effect which appears in twenty-five-year means is obliterated in a hundred years, it would be very interesting to know how the second -order effect appeared in the other three groups of twenty-five years of which the hundred are made up. Some light might be thrown upon the matter by the analysis of all the com- ponent years of the original twenty-five, but this has not yet been carried out.
The fact that the maximum of the second order for Vienna comes on March 24 shows that the remaining effect there is different in character from that shown in the curves for the British stations.
Comparison of the Effect at Kew with that at the other Stations.
Reference has hitherto only been made to the Kew curves, since it is the Kew temperatures which have been studied in detail, but the analysis of the twenty-five-year mean temperature curves for Aberdeen, Valencia, and Falmouth are also given in Table I. The second-order curve is in each case about one-eighth of the first-order curve in ampli- tude, but the third-order curve is not always entirely negligible. At Kew its amplitude is only one-thirtieth of the second-order curve amplitude, and at Falmouth it is only one-seventh, so that in these
