16 M. Humboldt on Isothermal Lmes, and the 
state, that even in 1777, the mean temperature of Toulon was 
estimated by Cotte ^ at 78°.08, though he afterwards found, by 
employing the whole mass of observations, that it was not moi-e 
than 60°.26. 
In order to diminish the errors of the method of annual ex- 
tremes, it was perceived, though very late, that it was necessary 
to subdivide tlie curve which expresses the variation of tempe- 
rature. Twenty-four extremes divided among twelve months 
of the year, give an annual mean more exact than the two ex-, 
tremes of all the observations. The ordinates do not increase 
uniformly and uninterruptedly up to the maximum of the year, 
and there are partial inflexions sufficiently regular. The more 
we subdivide, and the more we know the terms in the series, 
the more will these terms approximate, and the less error will 
there be in the supposition of an arithmetical progression, and 
in that of the equidistance of the different maxima and minima 
of temperature. These considerations enable us to appreciate 
the three methods according to which observations are at present 
made. 1. Observations are made three times a-day, at sunrise 
and sunset, and at two o’clock in the afternoon. This was done 
at Geneva during the three years 1796, 1797 and 1 798. In the 
observations, the hour of mid-day was preferred to that of sunset. 
2. Observations are made twice every day, at the two periods 
which are supposed to give the maximum and the minimum, 
namely, at sunrise, and at two o’clock in the afternoon. 3. Ob- 
servations are made once a-day, at an hour which, in different 
seasons, has been found to represent the mean temperature of 
the day. It is thus that M. Raymond, by a judicious induc- 
tion, has proved, that the height of the barometer, at mid-day, 
gives, in our climates, the mean atmospherical pressure, correct- 
ed for the diurnal variation. 
In calculating a great number of observations made be- 
tween the parallels of 46*" and 48°, I have found, that a single 
observation at sunset, gives a mean temperature which differs 
only some tenths of a degree from that which is deduced from 
observations made at sunrise and at two o’clock. The devia- 
Mem. de la Soc. Ror/ale de Med. 1TT7, p. lO-t. f De la Formule Barom. p. 213, 
