REPORT ON THE COMPOSITION OF OCEAN-WATER. 
37 
total salts, the value 1'692; and 0*029 as the probable deviation (r) of the individual 
result from this mean. Referring from chlorine = 55'42 to chlorine =100, we have — 
Mean, 100- = 3 ‘05 3 
% 
r= ±0-0523. 
The corresponding values brought out by the 12 analyses of the three mixtures are — 
Mean, 100- = 3‘0261 
% 
r= ±0-0046. 
The method used for the 77 previous determinations must have been infected with an in- 
herent source of inconstancy not observed at the time. Exact analyses of two different 
mixtures of crude lime precipitates obtained in these had, by accident, given very nearly 
the same value of about 0"91 for L 0 -v-L. Our new analyses give us the individual values 
of L 0 and L for 16 cases, and show that the quotient L 0 -=-L, even under apparently 
constant conditions, is subject to considerable variation. And this casts a doubt upon 
the important conclusion which I (somewhat diffidently) drew from those analyses in 
regard to the relation of L 0 -^x to the depth from which the water is taken. Let us 
examine the new analyses with respect to this point. Taking s, m, cl as symbols for 
lime per hundred of chlorine in water-mixture I., II., and III. respectively, we have — 
(A) If we take the three mean values as our basis. 
m — s = 0-0124 = 4-8 x (0-0012 + 0-0014). 
^-5 = 0-0132 = 5-8 x (0-0012 + 0-0011). 
The differences m — s and d — s, as we see, have positive values, and to explain them as 
accidental we must assume that, supposing even in both cases the two actual errors had 
opposite signs, their sum was in the case of m, 4 - 8 times ; in the case of cl, 5 "8 times the 
probable error r 0 of the mean. But the corresponding probabilities are no more than about 
O'OOOZ — i.e., even disregarding the fact that the original assumption as to the signs of 
the errors has only the probability ^ on its side, the probability is 9993 to 7 that the 
lime per unit of chlorine was greater in the deep-sea and medium waters (II. and III.) 
than in the surface waters (I.), for some definite and ascertainable reason. 
This is exactly the interpretation we got from the 77 complete analyses ; except that 
we then considerably over-estimated the difference. 
