154? Royal Society : — 



SPT' S'Q 

 reduce the error to the fraction . ■ ^~— 1, of the small resist- 



ance R*. But further, when, as in experiments on short thick bars 

 like those of Mr. Calvert, S'T cannot by auy management be got to 

 be small in comparison with TT', the use of the secondary testing- 

 conductor becomes essential, and the most accurate possible fulfilment 



of the condition SPT' ^Q 



S'QT ' SP p * 



must be aimed at. This is to be done by dividing the secondary 

 testing-conductor at Q, in very exactly the same ratio as the primary 

 at P, and taking care that the resistances in the connexions S'K, LT 

 are very small in comparison with KQ and QL. 



Part II. Suggestions for carrying out these principles in practice. 



When high accuracy is not required, the two " testing-conductors" 

 may be made of wires stretched straight in parallel lines, and the 

 connexions for the galvanometer electrodes may be applied to them 

 by means of a slide on a graduated scale — as in one of the common 

 forms of Wheatstone's balance, with sliding contact on single test- 

 ing-conductor. This form is very objectionable, however, whether 

 for Wheatstone's balance or the method I now propose : (1) because 

 it is impossible to secure that the different parts of each testing-con- 

 ductor shall be accurately at the same temperature ; (2) because the 

 resistances at the ends of the fine stretched wire or wires are always 

 sensible in comparison with the smallest measured differences pro- 

 duced by the slide ; (3) because the stretched wire itself is never of 

 absolutely equal gauge throughout, and, even if sensibly so when first 

 put into the instrument, soon ceases to be so in consequence of the 

 friction of the sliding contact which it experiences in uset ; (4) be- 

 cause, in even the hastiest experiments, provided a rationally 

 planned galvanometer is used, a far higher proportional degree of 

 accuracy is easily attained in measuring electrical resistances against 

 a standard of resistance than can be at all attained, without very 

 extraordinary precautions and the assistance of a microscope, in 

 measuring lengths under a yard or two against a standard of length. 



Yv^hen the highest accuracy is required, I always use for primary 

 testing-conductor the bisected conductor which, I described to the 

 British Association at its Glasgow meeting in 1855. This consists 

 of a fine, very perfectly insulated wire, doubled on itself and wound 

 on a bobbin, with very stout terminals soldered to its ends, and an 

 electrode soldered to its middle, for joining to the galvanometer elec- 

 trode. The two terminal and the middle electrodes thus attached 

 to the testing-conductor, I have generally hitherto made flexible, 

 either of thick wire, or strand of wires like the conductor of a sub- 

 marine cable ; but, for many applications, it is more convenient 



* In such cases R will, according to equation (1) above, be nearly equal to 

 S'BCT, but somewhat less, 



f This defect I have remedied by frequently putting in a new wire for 

 testing-conductor in working with a sliding-scale Wheatstone's balance. 



