182 Prof. Spencer U. Pickering on the Calibration 



It would seem, therefore, that this must represent very nearly 

 the limits of accuracy of which the method is capable — a 

 limit which gives, for the maximum probable error of any 

 point in the mean curve deduced from two such calibrations, 

 a quantity corresponding to about 0'08 millim. of the mercurial 

 column. It would, however, seem difficult to reduce this 

 error much further by increasing the number of calibrations 

 performed ; for the numbers given in the third column of the 

 table show that the mean curve deduced from those obtained 

 with a lens still differs from the mean deduced from those 

 with Brown's instrument by as much as 0*126 millim. 



A possible error of 0*08 millim. represents in the first three 

 and the fifth thermometers o, 0029 C, while in the last two 

 it represents only 0°-0005 C. 



In their Report the Committee recommend a means of 

 separating for calibration- purposes a thread of mercury of 

 any required length, consisting in applying a very small 

 flame to the point at which the separation of the column 

 is desired. I find, however, that it is most undesirable to 

 apply such heat to a portion of the divided scale itself. 

 Some alteration in the bore and in the nature of the glass 

 appears to be effected at the point heated ; in five or six 

 cases it was found that the thread of mercury when travelling 

 in one direction would not pass the point which had been thus 

 heated, whereas when travelling in the opposite direction it 

 rushed past it at an ungovernable rate. In one case a \erj 

 noticeable alteration in the size of the bore was observed to 

 have occurred at the heated part. One of Mr. Casella's 

 assistants, a man on whose powers of observation I can rely, 

 assures me that he can at once tell the point at which a fine 

 thermometer- tube has been heated in this manner, by means 

 of the peculiar jerky motion of the mercurial column when 

 passing it. All the tubes to which the present remarks apply 

 were very flat and fine in the bore, probably much more so than 

 in the instrument examined by the Committee, and this no 

 doubt accounts for their not having observed any injurious 

 effects produced by heating. Separation of a thread by means 

 of the so-called vacuum bubble is not possible in these very fine 

 tubes, so the flame method must be adopted, taking care, 

 however, that it be applied to a point either above or below 

 the graduated portion of the stem. Another, but less con- 

 venient, method consists in passing some of the mercury into 

 the top chamber of the instrument, and distilling a certain 

 quantity of it back into the tube. 



Great difficulty is sometimes experienced in passing a short 

 thread of mercury past certain points in a very line tube, even 



