48 EXPERIMENTS WITH THE DISPLACEMENT INTERFEROMETER. 



place was probably something like this: the access of heat in the room 

 reaches the outer layers of the pier first and only gradually penetrates to the 

 interior parts of the masonry. It has seemed to me, also, that the occurrence 

 is not merely a question of temperature, but rather a case of drying out the 

 parts of the brickwork, from the very damp conditions in the summer to the 

 desiccated condition during the winter months. The lower curve is naturally 

 more sinuous than the upper, but not nearly as much so as would have been 

 expected in comparison with the curve of fig. 14 in 17. In fact, on comparing 

 these two curves after September 28, the initial maximum on October i and 

 the minimum on October 6 correspond. After this the curves diverge and there 

 is no correspondence until the maximum of November i is reached. Thus, 

 for instance, on October 19, fig. 14 finds no counterpart in fig. 9. A compar- 

 ison of the two curves is naturally immensely in favor of the pier, though 

 even the latter is again altogether inadequate for the kind of work contem- 

 plated; i.e., work involving variations in a of hundredths of second of arc. 

 Changes of a of half a second are by no means uncommon, and even in the 

 summer changes of 2 seconds within a month would be a small estimate. 

 The curve shows the nature of the difficulties encountered, for instance, in 

 endeavoring to measure the repulsion of two disks in Chapter II. 



Between September 17 and September 28 the temperature curve (light line) 

 is drawn on a large scale, between September 28 and November 5 on a smaller 

 scale. With regard to the former it is evident that the sinuosity of the curves 

 is about the same, but that the minima and maxima are not cotemporaneous. 

 Thus, for instance, at a the temperature minimum precedes the inclination 

 minimum. The same is true for the maximum at 6, etc. In their details and, 

 in general, quantitatively, the two curves do not coincide in character. Hence, 

 the effect of temperature, if admitted, can at best be indirect; i.e., tempera- 

 ture changes the inclination a by straining or warping the pier. 



If we compare the temperature curves between September 28 and November 

 5 with the inclination curve, there is again a general resemblance. Thus the 

 maxima at/, g, k and the minimum at h occur in both. But there is no detailed 

 resemblance, even when the difference of scale is taken into consideration. 

 Neither is the temperature effect as marked as in the corresponding case of 

 fig. 14, Part I. The temperature maxima tend to precede the inclination 

 maxima, etc. Hence, as before, temperature acts, not upon the pendulum 

 mechanism directly, but rather indirectly through the supports, which become 

 displaced by unequal expansions in the pier and a corresponding tilting 

 from its position. 



Finally, the changes of inclination a shown by the aluminum pendulum are 

 quite as marked as those occurring in the corresponding case of the steel pen- 

 dulum, although the viscosity error of the former is much greater. It does 

 not therefore appear that the effect of changes of magnetic field has produced 

 any error, such as was surmised above in case of the steel pendulum. The 

 latter is, therefore, preferable for work of the present kind. 



