194 
About five liters of liquid air were required for an experiment. Thea 
tube was left in the bath for about one hour before the glass rod was 
sounded. Corrections were applied for scale errors and for the expansion 
of the tube prior to measurement. ‘The coefficient of expansion” for glass 
at liquid air temperature was taken as 0.00000738. 
One complete set of average wave length measurements for an experi- 
ment is given in table III, and the assembled data together with the caicu- 
k 
lated values of —1 are givep in table IV. The subscripts relate to liquid 
k 
air temperatures. All temperatures are referred to —273°.04 as the abso- 
lute zero’. 
: ky . : ; 
Any change in T will alter Jeon. inversely in about the same ratio. 
k 
Temperatures were probably accurate to one-fifth of one per cent. An un- 
rae ; k ; 
favorable combination of errors might invest ~* with an error of one- 
k 
half of one per cent. 
From the results in table IV it would appear that k for liquid air 
temperature is something more than four per cent greater than for ordi- 
nary temperatures, about 22° C. 
12 Phil. Mag. 49, June, 1899. 
13 Ann. der Physik 9, p. 1149. 
Purdue University, Dee., 1910. 
