ACOUSTICS AND GRAVITATION. 123 



differences of m and M must be less than this, since there is some accommo- 

 dation, all of which shows the minute temperature discrepancies which are 

 quite adequate to annul the effect of gravitation. 



Next day, August 21, with the water-bath removed and a free case in air, 

 the alternations throughout the day given in figure 156 e were obtained, show- 

 ing that the full deflection should have been (deducting of course the 7/11.3 

 part due to the presence of water) : &y = 4.45 a. m., and 4.30 p. m. The be- 

 havior of apparatus /, however, announced this to have been a day of large 

 excursions; but at least one-half of the Ay should have been exceeded. 



The drift observed must be ascribed to the very slightly rising temperature 

 of the water-bath, of which the case partakes. It is an increasing pull toward 

 the front of the case or, better, a decreasing push toward the rear, and agrees 

 in sense with the pull of the weight M when in front. There is also marked 

 drift in the air values in figure 156 e. 



I made these experiments with scrupulous care, as they furnish the only 

 quantitative estimate of the thermal relations involved. It is astonishing that 

 this standard and effective method of obtaining temperature constancy here 

 utterly breaks down, notwithstanding the 30 pounds of water used. While the 

 mere exposure in a medium of air (fig. 156 e) gives results which are fluctu- 

 ating indeed, but not abnormal. In speculating as to a cause, it seems prob- 

 able that the surface temperatures, on radiation through air, tend to become 

 equal to a greater extent than happens in the other cases. 



97. Attraction in vacuo. The remarkable result embodied in figure 149, in 

 which the gravitational attraction decreases in a high vacuum to about one- 

 third of its value in a plenum, induced me to give the subject further attention 

 and ultimately to construct a metal apparatus for the further study of this 

 strange behavior. Unfortunately it was not possible to get apparatus II 

 (case of wood, impregnated when hot with melted soft sealing-wax) quite 

 tight. The leak was originally about 7 cm. per hour at the highest exhaustions. 

 After many improvements and much labor I reduced this to 2 cm. per hour. 

 I then found, however, that a supply of air was absorbed in the wood. This 

 came out very gradually on reducing the internal pressure, but was reabsorbed 

 on increasing it. Thus one meets the curious result of a gradually increasing or 

 a decreasing vacuum, according to the direction of the change of the internal 

 air-pressure. An increase by as much as 10 cm. or more was recorded at mean 

 pressures. These annoyances, as they must be accompanied by slight temper- 

 ature variations in the wood complicate the results. They can not possibly 

 be smooth. I hoped, however, to get a mean result by operating both with 

 increasing and with decreasing internal pressure. 



There is the additional difficulty of the change of atmospheric temperature 

 during these essentially long-period observations, which operates with a lag. 

 This induces a change of Ay, as explained above (figs. 119, 120, and 121). My 

 first endeavor to eliminate it consisted in comparing the Ay 2 of apparatus II 

 with A^t of apparatus/, placed on the ad joining face of the pier; but although 



