18 PROCEEDINGS OF THE AMERICAN ACADEMY 



if Sb = 120.00. "We weighed out, with as much accuracy as if we 

 were adjusting a weight, the smaller of these two quantities of metal- 

 lic silver, and after dissolving the pure metal in pure nitric acid, evap- 

 orating the solution to dryness, and redissolving in water, we gradually 

 added the whole of this silver solution to the litre flask containing the 

 solution of bromide of antimony, in the manner described above. It 

 was then found that 12^^^ cubic centimetres of a standard silver solution 

 {one gramme of silver to the litre) were required to complete the pre- 

 cipitation. It will be seen that the weights of the bromide of anti- 

 mony and silver used could be thus determined with the most absolute 

 precision, and we have the greatest confidence in these values to the 

 -jL of a milligramme. Moreover, it will be noticed that the volumet- 

 ric method is only used to estimate the difference in the atomic weight 

 which has been in question, and that, if the method were only accurate 

 to the ^0 of the quantity to be measured, it would give us the value 

 of the atomic weight within -^^ of a unit, while if, as we had reason to 

 believe, the process was accurate within one per cent, it would fix the 

 atomic weight within y^^ of a unit. 



By the method just described, the following results were obtained. 

 The letters a and b indicate different preparations : — 



The extreme variation from the mean in these determinations is 

 less than one ten-thousandth of the quantity directly estimated, and 

 corresponds to less than two ten-thousandths of the total value in 

 the atomic weight of antimony. We have thus reached the extreme 

 limit of accuracy with determinations on this scale. By using very 

 much larger amounts of material, it is possible that we might still fur- 

 Xhev diminish the limits of experimental error ; but when we consider 

 the further causes of error incident to handling so large an amount of 



