MR. II. X. DICKSOX OX THE CIRCULATIOX OF THE 
zn 
These give the relation between and Cl 
^Si5 = 1-399 Cl - r045 (1), 
a straight line which satisfies excellently the determinations of fifty-two samples in 
which Cl was less than 17'00 (28). 
The samples used as standards in j^revious work (29) give for the same relation 
= 1-389 Cl - 0-805 (2), 
an equation which gives values agreeing fairly well with (l) for values of Cl met with 
in open water, but diftering to a serious amount for very fresh waters, e.g .— 
4815 
481 .^ 
Cl. 
from ( 1 ). 
from ( 2 ). 
Ditf. 
20 
26-94 
26-97 
-03 
17 
22-74 
22-80 
-06 
5 
5-95 
6-14 
-19 
The discre})ancy indicates that in the waters of low salinity the chlorine forms a 
smaller percentage of the total salts in my samples than in those used as standards ; 
the difference may possibly be accounted for by the fact that while in the low-salinity 
standards the salinity has been reduced Ijy admixture of land water from the Baltic, 
my samples liave been freshened by water from melting ice (30). It seems not 
impossible that tlie relative deficiency in chlorides known to exist in surface waters 
near melting ice may make itself ap])arent in this way in the North Atlantic, but the 
differences are so small as t(j re(pnre confirmation by other obser\-ers, and it would be 
quite inq^ossible to deal with differences in valuations quantitatively by any known 
methods. This point will be referred to again in discussing the salinities. 
Partly with the oljject of checking the purity of the water-samples, and partly with 
a lino'erinn hope that some variation inio-ht be found suflicientlv laro'e to be useful in 
studying the circulation, 1 made determinations of the total sul})hates present in a 
number of the samples received during the first year. The method described liy 
Dittmar (31) was followed absolutely all through, about 40 cub. centims. (all that 
could be spared) being taken for each determination. The results are given hi the 
last column of Table I. (p. 117, et 6'cay.). The chief interest of the.se determinations 
is in the ratio of the total sulphate ])er kilogramme of water to the chlorine, usually 
. lOU SO,, 
expres.setl as ~ 
It may be tloubted if, under ordinary conditions, the .sulphate estimations in sea¬ 
water are really trustworthy to much less than 1 per cent. As the chlorines are taken 
as correct to fir ’03, we may neglect their errors, and treat the errors of observation in 
. .. . 100 Sty inn -,11 1 • • 
the fraction —^ as wholly due to eii'ors m the sulphate deternimations; we 
may take 1 pe-r cent, as the probable error of ;i value of the fraction. To make this 
