766 MB J. D. HAMILTON DICKSON ON 



Notwithstanding the uncertainty attaching to the observations above 10° C, it was 

 thought right that those to be used in the equations of least squares should be fairly 

 distributed ; those selected are indicated by arrow-heads on the sheared A-curve, 

 and will show that this idea was carried out. The calculation led to the equation 



( T - 8^7! -131-6952,18) 2 = 102-722 6) 04(H + ^^ -88-1434, 8 ) . . (54) 



Hence is derived, in a form more suitable for calculation, the equation 



E=48691-329 + 131-287480-5115-O152 N /0 .... (55) 

 where 



# = £ + 276-90308 ..... (56) 



From this we get the equation of the Tait-line 



f = 131-28748 - 2557 ™ 6 (57) 



at *Jv 



E being, as usual, in C.G.S. absolute units, and t in degrees centigrade. 



By means of these equations the E.M.F. and sheared B-curves were drawn ; and 

 on these curves the observations were plotted — for the sheared curve by small crosses, 

 for the E.M.F. curve by large points. The sum of the errors between the observed 

 and calculated values of E/20 was 28*1 — 28*3, which, being practically zero, verifies 

 the correctness of the calculation. Hence the probable error was found to be ±'76 

 micro- volt, over a range of 126 micro- volts. Omitting the erratic observations above 

 10° C, and one obviously inaccurate at — 51-25°, the remaining seventeen observations 

 are seen to lie remarkably close to the calculated curve. 



The vertex of the E.M.F. parabola is at the point t = 85-56°, E = - 1 103-6. Through 

 this point on the diagram the axis of the curve is drawn. This metal allows both 

 branches of the Tait-line and its two asymptotes to be shown clearly ; the lower branch 

 of the curve is that appertaining to the present question, the upper branch has at present 

 no physical meaning. The horizontal asymptote is at E= 131 '28, the vertical one is 

 at t = -276-9°. 



Gold. 



As this gold wire was in a state of very great purity, it was one of the earliest 

 metals to be examined. The observations were plotted and a curve drawn among them 

 in the manner already described, the A-curve in fig. 14. From this curve ten values of 

 dE/dt were measured by means of the stretched thread, which, on being plotted, 

 appeared to lie so nearly on a straight line (the line C) that this was assumed to be the 

 case, and consequently that the axis of the associated parabola was without slope. The 

 work at this point was laid aside, and on resuming it two years later, fresh measure- 

 ments of dE/dt, eleven in number, were made on the same curve, A, which on being 

 plotted gave again apparently a straight line, D. The measured values of dE/dt are 

 shown on both these lines, the D line being placed a little above the C line for the sake 



