122 PROCEEDINGS OP THE AMERICAN ACADEMY. 



heating of the bulb to a high temperature, — an operation inevitably 

 causing an increase of volume which only slowly readjusts itself, as in a 

 thermometer thus heated. Regnault continually adopted this practice, 

 but it cannot but be considered as dangerous when great accuracy is 

 desired. These objections are less applicable to the use of mercury in 

 the case of the smaller supplementary bulb, for this was so stout as to 

 be scarcely susceptible of deformation by the slight pressure of the small 

 column of mercury which it contained, and could easily be freed from air 

 by exhaustion. 



In the next place, the conditions of measurement also must as nearly 

 as possible approach those of the actual expansion-experiment. Espe- 

 cially, because the gas volume is always in contact with mercury, the 

 meniscus used to determine the volume must in each case assume the 

 same shape, and therefore be measured at the same temperature and be 

 set in the same fashion. This condition decides that mercury must 

 be used in measuring the volume of the supplementary bulb, and de- 

 mands that the temperature at the upper setting should be zero, and that 

 at the lower setting 32.88°, the actual temperatures used in the final 

 experiments. Fortunately, because the volumes of mercury are small, 

 this demand is not inconsistent with adequate accuracy as to the other 

 attendant conditions. 



One other point proved to be at least equally important ; namely, 

 the increase of volume of the large bulb upon warming from J to 

 32.38°. This increase, of course, adds to the volume of the auxiliary 

 bulb, constituting the total gain in volume with change of temperature. 

 Two alternatives were open, — either the coefficient of expansion of the 

 glass could be calculated from the weight of water filling the bulb at 

 32.38°, or else the linear coefficient of "expansion might be measured 

 directly. 



After a number of time-consuming preliminary experiments it was 

 decided to adopt the latter course, because the expansion of water over 

 the range in question is six times as great as that of glass, and hence 

 any error in the assumed value of the former or in our measurements 

 would cause a sixfold percentage error in the coefficient of expansion 

 sought. 



In view of these considerations, the following means of measurement 

 were adopted ■ The bulb and supplementary bulb were cut off from the 

 rest of the apparatus somewhat below the lower pointed platinum wire (b), 

 and a glass stopcock was attached at the junction. The total volume 

 between the stopcocks was now found by weighing the bulb full of water 



