24 EXPERIMENTS WITH THE DISPLACEMENT INTERFEROMETER. 



These results are quite regular. The extremely slow cooling shows how 

 near the temperature is to that of the environment, the temperature excess 

 being nearly negligible. It shows also how difficult it is to obtain rigorous 

 temperature constancy in the metallic truss exposed to the surrounding at- 

 mosphere. Thus it would have taken considerably over two hours to dissipate 

 the negligible difference of temperature in the last case. 



Again, a Bunsen flame placed about 30 cm. from the brace, about at e in 

 fig. 15, gave the following result: 



9 h i5 m x=i.$ cm. Cold. Burner placed as stated. 



20 3.0 Gradually heating. Burner now removed. 



28 2.2 Cooling. 



The radiation of the Bunsen burner at a distance of about i foot is thus 

 quite perceptible, in spite of the fact that the brace and standard are to some 

 extent affected differentially. 



With this experiment, therefore, the mysterious temperature variation has 

 been cleared away, and it provides definite specifications for the installation 

 of such an apparatus. I will only add that similar experiments tried on the 

 scaffolding of the permanent mirrors produced only negligible effects. Thus 

 a Bunsen flame run rapidly along any of the horizontal braces changes the 

 deflection x only a few millimeters. 



14. Inferences. If we abstract from discrepancies introduced by the pendu- 

 lum truss, which are to be separately treated, it may be assumed that the data 

 obtained with the partially floating pendulum represent the actual tilting of 

 the concreted subfloor of the laboratory. With a reasonably constant tem- 

 perature in a cellar room, in the absence of artificial heat, temperature dis- 

 crepancies should no longer be seriously menacing. To account for the differ- 

 ence between the small variations of a obtained in the absence of a float and 

 the large variations on addition of the float to the same pendulum, it is suffi- 

 cient to admit that the friction at the vertical pivots in the former case (float 

 absent) was excessive, and that the full deflection can not appear, unless the 

 weight is taken off the pivots as in the latter case (float present). This is 

 particularly the case, since the inclination resulting from the expanding brace 

 presently to be mentioned should in any case have been present. Capillary 

 forces at the float, mounted axially as above, have produced no appreciable 

 distortion, as they did when the water damper was mounted at the end of the 

 beam. Finally, the float is itself a sufficient damper, and in the absence of 

 air-currents the front of the case may actually be kept open, as was done in 

 most of the later experiments, in a room free from artificial heat. Moreover, 

 it does not seem necessary to construct the floating horizontal pendulum on so 

 large a model as was done in the above paper, so that a smaller portable model 

 may be a serviceable instrument for many laboratory purposes, seeing that the 

 constants are determinable by the direct method indicated. In how far the 

 sensitiveness may be increased by applying the buoyant force at the center 



