1032 
PROFESSOR J. W. MALLET ON A REVISION 
Third series. 
A. 
B. 
Experiment. 
Riff, from mean. 
Experiment. 
Diff. from mean 
I. 
Al = 27-012 
+ •007 
VII. 
Al = 26-995 
+ •005 
II. 
„ 27-005 
•ooo 
VIII. 
„ 26-999 
+ •009 
III. 
„ 27-022 
+ •017 
IX. 
„ 26-977 
— •013 
IV. 
„ 26-988 
-•017 
V. 
„ 27-006 
+ •001 
VI. 
„ 26-996 
-•009 
Mean 
,, 27’005 
Mean 
„ 26-990 
Probable 
error of mean result + "0033 
Probable 
error of mean result 
+ •0046 
In view of the gradual loss of water which, as has been shown, crystallised ammo¬ 
nium alum undergoes on exposure to the atmosphere, I feel that of these various 
sets of experiments, B of the first series is entitled to least confidence, and the con¬ 
siderable difference between its results and the others leads me to favour its rejection. 
On the other hand, I am inclined to attach most weight to series 3, A, since the 
method used was very simple in principle, the determination of one of the two 
quantities concerned was rendered very exact by the grea/t volume occupied by the 
hydrogen, the comparison was made directly with the standard element in our system 
of atomic weights and not through the intervention of any other substance whose 
atomic weight must be assumed, and the agreement of the results among themselves 
is particularly good, as shown by the probable error of the mean being the smallest 
reached. 
General mean of results .—The general mean from all of the thirty experiments, if 
all be included in the calculation, is Al = 27‘032, with a probable error for this mean of 
±•0045. 
If series 1, B, be excluded, the mean of all the remaining twenty-five experiments is 
Al = 27 , 019, with a probable error of ±'0030. The third decimal having no positive 
value, we may take Al=27'02. If integer numbers be used for O, N, C, Na, &c., 
Al=27. 
All the experiments which I have made are reported, except three or four in which 
there was manifest failure, as by accidental loss of material to a visible extent, and 
which were on that account not completed. 
The general result adds, I trust, aluminum to the, unfortunately still limited, list of 
those elementary substances whose atomic weights have been determined within the 
limits of precision attainable with our present means of experiment. 
