482 HENRY A. ROWLAND 



readings which one can make on an air thermometer will vary several 

 hundredths of a degree. 



Hence we can never use with accuracy the direct comparison with the 

 air thermometer but must express the difference of the two instruments 

 by some formula of the form: 



J = a + bt + ci 1 + &c. 



Should we take an infinite number of terms this formula would ex- 

 press all the irregularities of our observations. But by limiting the 

 number of terms the curve of differences becomes smoother and 

 smoother and the formula expresses less and less the irregularities of 

 the experiment. The number of terms to be used is a matter of judg- 

 ment, and this point I sought to determine by the use of the observa- 

 tions of Eegnault and others. The rejection of the higher powers of t 

 is more or less of an assumption founded on the fact that we are 

 reasonably certain that the curve of differences between the mercurial 

 and the air thermometer is a smooth curve. It is evident that the 

 less the correction to be introduced the less the rejection of the higher 

 powers of t will affect our results. 



We now come to my criticism of the Geissler thermometer for not 

 having a reservoir at the top. Dr. Thiesen has in some way misunder- 

 stood my principal reason for its presence. My reason was not that 

 " es vermindert die Schadlichkeit der im Quecksilber zuriickgebliebenen 

 Spuren von Luft " but that only by its use can the mercury in the bulb 

 be entirely free from air. Take a thermometer and turn it with the 

 bulb on top. If the thermometer is large, in nine cases out of ten the 

 mercury will separate and fall down: allow it to remain and observe the 

 bubble-like vacuum in the bulb. Turn the bulb in various directions so 

 as to wash the whole interior of the bulb, as it were, and then bring 

 the thermometer into a vertical position, keeping the bubble in sight. 

 As the mercury flows back, the bubble diminishes and finally, in a good 

 thermometer, almost disappears: but in most thermometers a good 

 sized bubble of air, in some cases as large as the wire of a pin, remains. 

 It is the most important function of a reservoir at the top to permit 

 such manipulations as to drive all such air into the top reservoir and to 

 make the mercury and the glass assume such perfect contact that the 

 bulb can be turned uppermost without the mercury separating, even in 

 thermometers of large size and with good generous bulbs. In many 

 Geissler thermometers such a test might succeed, not on account of the 

 freedom from air, but because the capillary tube and bulb are so small 



