( 202 ) 



conducting pieces of glass, which partij projected out of the batii. 

 The scale (comp. Comm. N°. 85) was wrapped round with a 

 thick layer of wool enclosed in card-board of which the seams had 

 been pasted together as much as possible. The temperature of the 

 room was kept as constant as possible by artificial heating and cooling 

 with melting ice, so that the temperatures of the scale vary only 

 slightly. 



They w^ere read on three thermometers at the bottom, in the 

 middle and at the top. 



The scale and the points of the glass rods were illuminated by 

 mirrors reflecting daylight or arc-light, wiiich had been reflected by 

 paper and thus rendered diffuse. 



The vacuum tube (comp. Comm. N°. 85) has been replaced by a 

 new one during the measurements. The evacuation with the latter 

 had succeeded better. So much liquid gas Avas economized. For the 

 measurement with liquid oxygen we required with the 

 first tube V/^ liter per hour and 74 liter with the 

 second. Of N,0 we used with the first only ^^ liter 

 per IY4 hour. 



In order to prevent as much as possible irregularities 

 in the mean temperature the bath has been filled as 

 high as possible, while dry air was continually blown 

 against the projecting points. The}' were just kept free 

 from ice. In two extreme cases which had been chosen 

 on purpose — the bath replenished with oxygen as high 

 as possible and the points covered with ice, and the 

 bath with the float at its lowest point and the point 

 entirely free from ice — the difference of the mean 

 temperature of the ends was 10 degrees, corresponding 

 to a difference in length of 4 microns. The greatest 

 difference Avhich has occurred in the observations has 

 certainly been smaller and hence the entire uncertainty 

 of the length cannot have surpassed 2 microns. 



At the lower extremities the difference is still smaller. 

 All this holds with regard to oxygen, in nitrous 

 oxide such variations in the distribution of the tem- 

 perature can be entirely neglected. 



With some measurements we have observed that the 



length of the rods, when they had regained their 



ordinary temperature after cooling, first exceeded the 



original length, but aftei' two days it decreased again 



Fig. 1. to that value. 



