for some Pure Metals. 223 



The column headed " mean error " shows the mean error 

 of each set of observations, expressed in absolute units of 

 E.M.F., and also in fractions of a degree. The last column 

 is the calculated E.M.F. of a pair one of whose junctions is 

 at 0° 0., the other at 100° C. The discrepancies between the 

 values determined from different sets of observations are very 

 much greater than the mean error of a single set, which 

 appears to indicate, as already mentioned, that the thermo- 

 electric " constants" are not absolutely constant. In deter- 

 mining these constants, it is an open question whether they 

 should be found from a short range of low temperatures, or 

 whether a long range of temperatures should be observed. 

 In the former case the most accurate observations can be 

 taken, and are practically uninfluenced by the irregular effects 

 mentioned above ; but in this case the Thomson effect comes 

 in as a quantity of a much smaller order of magnitude than 

 the absolute height. If the range of temperature be greater, 

 the Thomson effect is a much more important quantity, and 

 can consequently be determined more accurately. In this 

 case, however, the accuracy of the observations is very 

 seriously interfered with by the irregularity of the effect 

 referred to above. Different ranges of temperature with 

 different junctions might be advisable, but I considered that 

 a range of 85° or 90° would give sufficiently accurate results. 

 I was disappointed in some of them, e. g. antimony and cad- 

 mium, but in these the mean result is probably not very far 

 from the truth, especially in calculating electromotive forces ; 

 the error in the constants considered apart from one another 

 is of course greater. In writing down the thermoelectric 

 height of each relatively to lead, some metals will have two 

 values ; that is, in cases of three metals where the three pairs 

 were examined separately, as for instance in lead, copper, 

 and zinc, the height of the zinc line might be taken as the 

 observed height above lead, or the sum (or difference) of the 

 lead-copper and copper-zinc heights. In this case, where the 

 zinc line lies between the copper and the lead lines, the 

 former of the two values will be the more accurate, and in 

 estimating the position I have given the two values weight in 

 the ratio of three to one. In the case of tin, which lies very 

 close to lead compared with copper, I have taken the directly 

 observed height as the true one, as the other is the difference 

 of two large and nearly equal quantities — a very inaccurate 

 method of observing the value. The following table is the 

 result of all the observations, which is also shown graphically 

 on the diagram with the exception of the antimony line : — 



