114 



REPORTS ON THE STATE OF SCIENCE. 



a difference readily determined from the shunts employed. With a 

 galvanometer resistance of 3 ohms the sensitiveness is proportional to 

 0'20iA\/-P. For coils without potential leads, in which case the method 

 of comparison is simplified, the sensitiveness is 0'25iAN/P, the same as for 

 the Carey-Foster bridge employing equal coils and a galvanometer re- 

 sistance of 2 ohms. The latter method is, however, inapplicable to coils 

 with potential leads, necessitates a calibration and standardisation of the 

 bridge wire, is more troublesome in practice, and the accuracy is limited, 

 not by the general arrangement of the bridge arms but by the openness 

 of the bridge wire and the accuracy of the scale and vernier. 



The following table gives the difference in values of four coils with 

 potential leads, every possible combination being taken. The differences 

 in the first column result from the exchanging of the coils in the bridge 

 arms ; the differences in the second and third columns are deduced fi'ora 

 observations of the two coils with a common standai-d. Thus, from the 

 first and second recorded observations, the difference 2206-2205 is 

 1 30 X lO"' ohm. The probable error is of the order of 1 part in 10,000,000. 

 The temperature coefficients of these four coils are not very different, and 

 average 001 per cent, per 1° C. The bath used for the comparison is 

 that described in the Phil. Trans., A 373, p. 87, 1 904. 



The differences recorded above indicate that at least three of the four 

 coils changed between the dates of the observations. In a similar 

 manner, very small changes have been observed in a few coils in an 

 interval of twenty-four hours. Such changes are very interesting, but 

 cannot be discussed here. 



Ten, 100, and 1000 ohm Coils and Resistances of a Higher Value. — 

 By the bridge method the probable error in the evaluation of a resistance 

 of 10" ohm is n times the error of the 10 ohms built up from the unit. 

 This latter error must, therefore, be made as small as possible. The 

 ' build-up ' should contain no variable contacts, and the lines of flow in 

 the coils when these latter are evaluated singly should be practically 

 identical with the lines of flow when the coils are in series. At the 

 National Physical Laboratory three special build-up boxes have been 



