923 
it is often done, in order to construct the formula such that it 
identically equals 1 at the freezing-point — it would have been 
impossible to obtain the close agreement at low temperatures. 
Besides I give the constants for formula (4), which may be useful 
in comparing the results for other (wires with the data used here. 
These constants can evidently be found from 5, c, and / by division 
by a: 
a B Y 
0.1536 0.1255 0.0958 
A survey of the course of W, as it follows from our formula, is 
given in Table [V, in which we find WW for every ten degrees, 
with first and second differences. The same numbers which served 
for this computation, were also used for the ten times smaller 
temperatures of Table V. 
It will be seen from the first numbers in that table, that even at 
helium-temperatures the formula is in a sense in accordance with 
experiment, yielding a resistance invariable within the errors of 
observation. However it may very well be that more accurate 
measurements in that region will show rather large deviations 
from the formula. Indeed our formula is based on the assumed 
behaviour of the resistance of pure metals at the absolute zero, 
which appeared probable partly on theoretical grounds, and may 
also be expressed thus: the resistance and its first three derivatives 
with respect to 7’ vanish at 7’=0O. Now I only dropped the 
first of these four conditions in the application to a slightly 
impure metal, whilst it seems possible, in view of the results 
for constantin and manganin, that at least also the second one, 
dW/jrp=0, is no longer exactly satisfied as soon as impurities are 
present. Therefore the accuracy of the four assumptions can only 
be tested experimentally by comparing wires of different degree 
of purity. 
Further Table V gives a part of a table for every degree, as it 
would be calculated from the formula for practical use. The numbers 
given have been obtained from Table IV by interpolation with third 
differences. Even this did not yield the exact value for the first five 
temperatures, 55° to 60°. Therefore they have been computed 
downright from the formula. On the other hand the interpolation 
becomes much simpler above 90°. 
7. Application of the formula to other metals. 
In what is mentioned above the suitability of the new formula 
has been tested for the most important case in practice, that of the 
