(221) 
the influence of the surrounding tubes is negligible in view of the 
accuracy obtained. 
We arrive finally at the following values 
for nitrous oxide: 
bs K.P) oy 
14.197 + 3.731 (= + 1.912 =) 
Pert oe == 1,988 
: 5.640 + 3.731 
for oxygen: 
83 5 
10.200 + 3.731 (= EBS =) 
Ko,= ease WGE, 
: 5.843 + 3.731 
As one sees the values of yo are a little more variable than those 
which refer to our condensers when filled with air (see § 4). This may 
be due to variations of temperature or small impurities. When we 
take this into account and assume for the other numbers the accu- 
racy arrived at above, we find that the maximum error (if we assume 
that the errors are additive) in the dielectric-coefficient of nitrous oxide 
is 0.5 °/ and in that of oxygen 0.7°/, while the error probably can 
be smaller. 
The value 1.491 given by Dewar and Fremine for the D-C. of 
oxygen at the normal boiling point, differs from my value by 1.8 °/,, 
an agreement which may be considered as satisfactory, if we take 
into account the deviations in the various values arrived at by 
different workers, even where the experimental substance can be 
more easily produced than liquid gases. 
6. Application of the Cuaustus-Mosorti formula to the results. 
An obvious application of the above results is employing them to 
test the Crausrus-Mosorrr formula. 
This is usually expressed 
K+2 
. d = Const. = D 
ke I 
where K is the dielectric-coefficient, d the density. 
This equation enables us to calculate the D.-C. of a substance in 
the liquid state when we know the D.-C. in the gaseous state and 
the densities of both aggregates. 
This is unfortunately not possible for nitrous oxide as the density 
at the normal boiling point is not accurately known. However as 
it is very interesting to see how my value agrees with the observations 
