I9I4-] 



ATMOSPHERIC PRESSURE. 



71 



Considering the results indicated in Figs. 7 and 8, it appears that 

 the slopes of the various straight lines are nearly proportional to the 

 atmospheric pressure. This means that, at least to a first approxi- 

 mation : 



{ PA^ ,,, , /abwatts per cm. \2 



where pk' = k- 



Consequently, when in Professor Morris's diagram, the atmospheric 

 pressure p changes, the ordinates are increased in like proportion. 

 We have then 



p = evpk'iv + v^) 



abwatts per cm. 

 dee. C. 



(8) 



so that the linear convection is nearly proportional not only to the 

 square root of the velocity, but also to the square root of the atmos- 

 pheric pressure. This agrees with Boussinesq's formula (o) when 

 it is remembered that the air-density <t of the medium is proportional 

 to the pressure p. The remaining constant k' involves, among other 

 things, the diameter of the wire. According to Boussinesq's formula 

 (o), the constant k' should be proportional to the square root of the 

 wire diameter. A special investigation should be directed to this 

 question : but the measurements recorded in our earlier paper of 

 1909 seem to indicate a higher ratio than the square root. 



TABLE III. 



Values of k' in the Expression Pc/0^ Vk'p(v^-vo) derived from the 

 series of observations at different atmospheric pressures p, at the velocity 

 (v-\-z'o) =1,000 cm. per sec. 



