MEAN DENSITY OF THE EARTH. 645 



on the lever. The guides on the sides were reuewcd and made stronger 

 so as to counteract all swinging motions of the masses. 



While the changes and improvements just referred to were being 

 made to secure a more accurate and ready determination of the con- 

 stants, protection against the ill effects of sudden and irregular changes 

 in temperature in the neighborhood of the api)aratus was so complete 

 as to obtain a closer agreement in the results of the experiments. The 

 door and window were also tightened and the wooden coverings coated 

 with paint. The box containing the pendulum was covered with tin- 

 foil and white-lead, and the attracting masses were likewise subjected 

 to a covering of tin-foil so as to free thdm from the action of the lamp 

 which illuminated the scale. To still further secure the apparatus from 

 sudden changes of temperature during a series of observations all oper- 

 ations were conducted without entering the room, einplo^'ing electricity 

 to mark the correct height of the cylinders when their positions were 

 changed. 



With the precautions just referred to, it was ])ossible not only to keep 

 the temperature of the room constant during a series of observations, 

 but also to eliminate, so far as the thermometer would show, the effect 

 of the warmth of the lamp, which was apparent during the tirst set of 

 experiments. 



A disturbance was however noticeable, caused by a temporary hori- 

 zontal stratification of temperature, which in the neighborhood of the 

 apparatus was interrupted by the changing of the masses whose tem- 

 peratures were unequal. I sought to compensate the operation of this 

 interruption by so attaching lead disks of the same diameter as the 

 cylinders that they remained in every position opposite the masses, in 

 this way securing at each change the same temperature throughout any 

 horizontal cross-section of the pendulum. 



The order of observing is essentially the same as during the former 

 series; however, instead of having a pause in the middle of the set, with 

 four readings on each side, or before and after, the eight or ten readings 

 of the elongations of the pendulum were consecutive, in this way 

 making it possible to discuss the observations of a day as equably dis- 

 tributed over the period of observation. In the determination of the 

 time of oscillation of the pendulum after the removal of the extra weight 

 on its upper end a better method was adoi)ted by comi)uting the desired 

 quantity directly from the means of two series of observations separated 

 from one another by a considerable interval. 



In the first experiments 1 computed numerically the attraction of the 

 masses upon the pendulum rod. The result was important, es|)ecially 

 as a control or check upon the experiments, but on account of the un- 

 certainty in the computation of certain constants which entered into 

 the work, it was not to be compared with the latter results which were 

 obtained by an empirical elimination. In these computations this con- 

 trol could be dispensed with, thus relieving one of the numerical work 



