40G 
APPENDIX NO. XI. 
All of these at temperatures above —40° agreed within 1*8°, and 
were selected as the most consistent of nearly thirty spirit thermometers. 
At 9 a.m. of the same day eleven similar thermometers gave under 
like circumstances a mean of 08°, the extreme readings being —56-4° 
and —80°. For the purpose of obtaining the most probable tempera¬ 
ture from these conflicting records, my first impulse was to reject the 
lowest (coldest) extremes, and take the mean of those which accorded 
best; but upon advising with our astronomer, Mr. Son tag, I determined 
to take the mean of all, without rejecting any,—the view which he took 
being simply that those instruments which indicated the extremes in 
the low scale had never in temperatures above —40° shown any 
anomaly which deprived them of an equal claim to confidence with the 
rest, and that there was no reason a priori to consider the results which 
they gave as less probable than those shown by the others. 
In a word, I adopted the views of Professor Airy, as published in 
the 95th number of the American Astronomical Journal. The causes 
which had produced the errors were mostly unknown, and the quantity 
to which these errors might amount was entirely so. 
Our thermometers were made with great care by Taliabue, of New 
York. But, independently of other mechanical sources of error, I 
am obliged to say that I do not regard the contraction of colored 
alcohol at very low temperatures as sufficiently investigated to enable 
us to arrive at the causes or the quantity of error. In most of the 
spirit thermometers the uniform thickness of the tube was tested before 
leaving New York; and the freezing of carefully-distilled mercury 
which I had taken with me for the purpose, gave excellent determina¬ 
tions of absolute temperature. 
But it may not be uninteresting to state that the freezing-point of 
this metal varied between — 38*5° and —11*5°, and that its rate of 
contraction as a solid was so uniform, that in our long and excellent 
36-inch standards it descended after freezing as low as —44°. This 
n result is in accordance with that obtained by Sir Edward Belcher, 
whose experiments go eveu further than my own, — the mercury having 
been observed by him to descend as low as 4G° below zero. 
I may mention the fact as in some degree confirming the propriety 
of not excluding an eccentric result from the computation of means, 
that two or more instruments may agree well together and still differ 
considerably from the most probable temperatures. This was the case 
with two long spirit thermometers, which never, even at the lowest 
temperatures, showed differences amounting to one degree, but which 
at 68° varied 7*7° from the mean of eleven others. The cause was in 
