ON STANDARDS FOR USE IN ELECTRICAL MEASUREMENTS. 61 



temperature is, however, not given in the Report for 1867, p. 483, nor is 

 it marked out in coil itself. Chrystal says that it was used in some of the 

 experiments ' because its variation coefficient was small, but otherwise we 

 have not bestowed much attention on it. ' In his first table he states 

 that the results given for F came from a single experiment, aud he gives 

 as its variation coefficient per 1° Centigrade 28 divisions of his bridge, while 

 Flat and G, also platinum silver coils, have coefficients of 34 and 35 divi- 

 sions. Now the observations of Fleming and ourselves show that without 

 any doubt these coils Flat, F, and G- have practically the same coefficient, 

 viz., -00028 B.A.U. per 1° C. Taking Chrystal's bridge wire as -075 ohm- 

 as stated by him, his value for Flat and G comes to '00026 B.A.U., which 

 is in fair agreement ; while for F we find "00021, a value which is now un- 

 doubtedly too low. We must infer either that the value of F has changed 

 considerably or that there is some accidental error in the one observation 

 given in Chrystal's table. The change necessary to account for the tem- 

 perature difference recorded in Table V. would he an increase in resist- 

 ance of -00067 B.A.U. 



Let us now examine the numbers for A and B. It will be seen at 

 once that they have altered appreciably, having, in fact, just changed 

 places. Their temperature coefficients are nearly the same, and there is 

 no doubt that throughout Chrystal's observations the coil he called B 

 was slightly higher in value than A, while throughout the obser- 

 vations of Fleming and ourselves the reverse has been the case. The 

 question naturally arises, have the coils been interchanged ? Chrystal 

 (Report, 1876, p. 17) states that, though they have no proper labels, they 

 are marked in some way or other so as to be identifiable. At the present 

 time they have brass labels screwed on to the ebonite of the frame bearing 

 the stamp B.A. 76 A and B.A. 76 B respectively. These were placed on 

 at the time of Chrystal's observations, and there seems just the possibility 

 of an accidental interchange. 



The coil H, No. 6, of the original report is marked as correct at 15°'3. 

 It is now correct in the sense used above at 17°*9, and here again we have 

 apparently a large change. The resistance would appear to have gone 

 down by about '00070 B.A.U. in the twenty-two years which have 

 elapsed since it was made. This corresponds closely to the change in G 

 observed by Chrystal between 1867 aud 1876. Now we know that G has 

 not changed relatively to C, D, and E, since 1876 — unfortunately H was 

 not examined by Chrystal— and we are led to ask whether the change 

 was a real one, or due in some way to the observations. The suggestion 

 already made in case of G applies again. The temperature coefficient used 

 by Matthiessen and Hockin is certainly too high, '00032 instead of '00028. 

 If his observations on the platinum silver coils were made at low 

 temperatures, and then the value of the temperature at which the coil is 

 correct were found by the use of the temperature coefficient, the result 

 would be too low. It will be seen shortly that all the platinum silver 

 coils examined, not merely those already mentioned, appear to have fallen 

 appreciably in value relative to the others. 



But we have another method of comparing the results. Chrystal has 

 given a table of the differences at 10° between each of the coils and Flat. 

 Now we have seen reason to believe that there is not much change in 

 C, D, B, and G. Let us find from Chrystal's table the valueof the difference 

 between G and the various coils at 10°, and compare these with our results. 

 In doing this some uncertainty is introduced from the fact that the value 



