EXPERIMENTS WITH THE DISPLACEMENT INTERFEROMETER. 29 



to do this during the second half, where temperature in general rises and the 

 a curve falls. The curves may be real and indicate a gradual settling of the 

 ground during the whole period, modified by rains, etc. ; or there may have 

 been a gradual viscous yield of the support of the pendulum in its concrete base. 

 Whatever be its nature, the pendulum fully recovers from this apparently 

 continuous subsidence during the second period of observation (BB), between 

 August 23 and September 28, omitting the gaps at a and b. The temperature 

 observations (not drawn) show no relation to the a curve whatever. The 

 curve being undulatory, it can not be referred to any persistent yielding or 



FIG. 18. 



other similar discrepancy. It is probable, in fact, that both curves AA and 

 BB show the actual tilting of the concreted subfloor of the laboratory. 



On September 28 the steam heat was turned on (scale of a, on the right) and 

 the totally new character of both the a curve CC and the temperature curve 

 furnish abundant evidence of the importance of this disturbance. In fact, the 

 pendulum behaves at first like an extremely sensitive thermometer. Observing 

 that the temperature scale is enormously smaller, rise and fall of temperature, 

 i.e., depression and elevation of both curves, may in general be coordinated 

 throughout. But there is no quantitative relation between the two curves. 

 Thus the marked rise of temperature from about 16 to nearly 30 at the end 



