326 



Mr. J. C. McConnel. 



[Mar. 12, 



Detailed Account of the Experiments. 



The first two experiments have been sufficiently described already. 



The place of experiment was a north room in the Buol Hotel, 

 Davos. A box without a lid was placed on a wooden table, and 

 across the top of this box were laid two pieces of wood, which 

 served to support the ends of the bar of ice. From the middle of 

 the bar was suspended a weight with a loop of thick string. In the 

 bottom of the box, but at the other end, i.e., about a foot from the idr 

 and 6 inches below it, was placed a registering thermometer of the Sii 

 pattern. Over the whole was put a thick wooden cover. As there 

 was nothing inside the cover of great capacity for heat, I believe 

 that any variation of the temperature of the ice was nearly simul- 

 taneously felt by the thermometer. This thermometer, which was 

 used throughout, was divided into Fahrenheit degrees ; its correction 

 at freezing-point was tested both before and after the experiments. 

 The error did not exceed $ F. At 6 F. I compared it with a spirit 

 thermometer which had been verified at Kew ; it read | F. too high; j 

 These errors are negligible in the present work. 



Exp. 3. A bucket of water left in the ice room over night wai 

 found in the morning covered with ice about 15 mm. thick, consisting 

 of several crystals. From this I sawed out a bar and planed it smooth 

 and straight. The breadth was 10 mm. ; the depth, 9 mm. 



The bar contained many long bubbles in a vertical position, 

 the middle of it was one crystal with the axis nearly vertical. T 

 two ends of the bar were composed of many crystals. A weight 

 1'29 kilograms was applied from 11.20 A.M. to 8.30 P.M. on Decem 

 14. During this time the maximum temperature was 2*"8 C. 

 the minimum, 5'6 C. ; and the mean, about 3'6 C. The bar 

 taken the shape of the diagram, fig. 1, which is copied from a t: 



Fio. 1. 



made soon after the experiment. The bends at the points indicated 

 by a and 6 were more decided in the bar than in the trace. The exact 

 position of the supports was not noted at the time, but they certainly 

 did not extend right up to the bends at a and 6. The fact that the 

 two end pieces are still nearly in line suggests that the end surfaces 



