METHODS EMPLOYED IN CALIBRATION OF MERCDRIAL THERMOMETERS. 153 



In what follows, the nearest division therefore means the nearest 

 division beyond and clear of the thread. 



The word reading means reading on the scale of the calibrating instrn- 

 ment, and in terms of millimetres. 



In a particular thermometer graduated from 55° to 107° C, the mean 

 length of a degree between 55° and 75° was found to be 10'50 m.m. 



Hence, to convert readings between 55° and 75° into degrees, we have 

 to multiply by jJ^„= -095. 



The corresponding factors for the whole scale were as follows : — 



Between 55° and 75° the factor = -095. 

 75° and 85° „ „ = -097. 

 85° and 90° „ „ = -100. 

 90° and 107° „ „ = -104. 



The initial points chosen were 55°, 58°, and 61°, &c. The method of 

 exhibiting the results of the measurement is shown in the accompanying 

 Table II. which is made more cumbrous than in actual practice by a full 

 explanation of the meaning of the figures in eacb column. 



The results of two sets of measures only are given. 



The scales of the thermometers were carefully tested in various parts, 

 and it was generally found that a group of consecutive divisions, all of 

 which were equal in magnitude, was bounded at either end by other 

 groups, in which the length of a division difiered more or less considerably 

 from that of those included in it. 



This was no doubt due to the fact that the scales were already 

 adjusted by the maker to the varying dimensions of the bores, though 

 not with the necessary accuracy. It was therefore possible to divide the 

 scale into parts, throughout each of which the same factor could be used 

 for converting the readings of the calibrating instrument (given in milli- 

 metres) into degrees. An example of such a division is given above. 



It may here be stated once for all that in the case of all the thread 

 lengths required for the observations detailed in Part II. the corrections 

 for the varying lengths of the scale divisions were thus made. The tem- 

 perature, too, was frequently observed by means of a thermometer placed 

 close to that operated on, and the thread lengths were corrected by sub- 

 tracting 000016 of their length for every degree Centigrade above the 

 standard temperature for the set of observations in which they were made. 



The apparent mean error, taken without regard to sign of the 

 thread lengths measured by this instrument is about double that of the 

 more accurate instruments employed at Kew and the Owens College. In 

 the case of Thermometer C, which was carefully calibrated by Bessel's 

 method for the purposes of this Report, it amounts to 0°'0024, or about 

 0'033 m.m. As each thread was measured only once, and required four 

 readings of the vernier, this result is satisfactory. The error in the 

 estimation figures obtained by the vernier is indeed much greater 

 than the error of setting the cross-wire to the point to be measured. 

 Additional accuracy could, of course, readily be obtained by fitting 

 the eye-piece with a micrometer screw to read over 2 or 3 m.m. only. 

 Inasmuch also as in the method of measurement described the sum of the 

 two small quantities Aa, Bb (see p. 152), is alone required, it would be 

 convenient, if a uniform scale were used, to transfer the mici-oscope 

 from a to 6 by a screw or rackwork, without altering the position of the 



