220 
BIGELOW. 
century ago he regarded the ether as subject to gravitation, 
but he now treats it as a substance outside the law of gravi¬ 
tation. He developed his famous theory of the atoms con¬ 
sisting of the ether isolated in dynamic vortices, but now 
regards this idea as untenable and has taken up again the 
old Lucretian mass atom as most likely to prove correct. In 
his presidential address of 1893 he gave an example to show 
that the sun and the earth have no causal magnetic connec¬ 
tion, but it is understood that he now thinks that the ob¬ 
served synchronism between the variations of the solar 
faculae, spots, and prominences on the one hand, and the ele¬ 
ments of the terrestial magnetic field on the other, is so per¬ 
sistent as to make it necessary to reverse that conclusion. 
The influence of an apparently valid result of the discussion 
of observations by a scientist of undoubted ability, fortified 
by a powerful mathematical treatment, sometimes turns aside 
the advance of knowledge into a wrong path, and this may 
even stop for a time all further efforts to solve the problem. 
Such failures, of course, should be reckoned as only the profit 
and loss in the book-keeping of research, and such temporary 
checks must not be taken too seriously. 
THE THEORY OF LEAST SQUARES IN METEOROLOGY. 
Professor Schuster has recently urged upon meteorologists 
the importance of submitting their researches to the analysis 
of the Fourier series, and the theory of least squares, in order 
to test properly the periodicities derived from the observa¬ 
tions, and he has illustrated his views by applying his pe- 
riodogram or probability curve to check the various periods 
that have been derived for the solar rotation. Fourier’s 
Theorem has already been widely used by meteorologists 
to express many of the periodic functions observed in the 
atmosphere, and some prefer this method to the numerical 
or the graphic methods, in spite of its great additional labor. 
Astronomers and Physicists have used the probability theo¬ 
rem freely and with valuable results, but only in certain 
restricted classes of observations. This theorem requires 
