152 EEPORT— 1882. 



' The original calibration was so accurate that the second approxima- 

 tion of Bessel's method was unnecessary in two cases, and was only 

 partially carried out in the third. 



' The maximum positive and negative corrections were in the case of 



Th. 561 + 0°-004 C. and - 0°-004 C. 

 „ 562 + 0°002 C. and - 0°-005 C. 

 „ 563 + 0°-008 C. and - 0°-011 C. 



' As will be seen from the above description of the thermometers, the 

 larger of these quantities are about equal to the limit of certainty in 

 reading. 



' In no case would the calibration error in the determination of a 

 difference of temperature have amounted to 0°02 C 



(9) The remainder of the thermometers, viz. A, B, C, and Ri, R,, R3, 

 had been calibrated by the maker, and were corrected by Bessel's 

 method in the Physical Laboratory of the Torkshire College. 



The instrument used was copied from one devised by and kindly lent 

 for the purpose by Mr. T. Brown, which he has lately described to the 

 Physical Society.' It consists of a wooden base to which the thermometer 

 is attached parallel to a slot in which a small brass plate slides, carrying 

 with it a vertical microscope furnished with a spider line. The microscope 

 is moved by rackwork on the brass plate parallel to the thermometer, and 

 the distance travelled is determined by a millimetre scale. Readings can 

 be made by a vernier to O'l m.ni., and with the help of a lens estimations 

 can be made correct to 0"02 or 0'03 m.m. In the form described by Mr. 

 Brown a split object-glass is used to bring the mercury in the bore and 

 the scale divisions into focus together. In that employed in this in- 

 vestigation this was omitted, and the microscope was moved vertically 

 by rackwork to bring the mercury and scale into view in turn. This 

 worked well. Any error due to want of verticality produced in the 

 measurement at one end of the thread would be compensated at the 

 other, if only the direction of the tube remained the same. That it did so 

 is proved by the fact that two measui^es of the same thread length rarely 

 differed by more than 0'03 m.m., which was well within the limits of the 

 sum of the estimation errors possible at both ends. 



This insti'ument was employed because it was believed that, with 

 proper precautions, and with so accurate a method of calibration as 

 Bessel's, it would serve to correct the thermometers to the required 

 degree of accuracy. This anticipation has been fulfilled, and as the 

 instrument is extremely convenient and relatively inexpensive, it may be 

 well to state precisely the conditions under which it was used. Inasmuch 

 as the thread lengths were required in terms of scale divisions, the instru- 

 ment was never used to measure the entire length of a thread, but only 

 the amount by which it was distant from a neighbouring scale division. 



(10) The position of the end of the thread was always determined by 

 measurinsr the distance from the nearest division which was clear of the 

 thread. Thus, if a and b be the extremities of the thread, and aa', be' 



a b 



the nearest scale divisions, the distances A a and b b were measured, even 

 if they were much greater than A' a and b' b. 



• Phil. Mag. [5] vol. xiv. p. 59. 



