RICHARDS AND MERIGOLD. ATOMIC WEIGHT OF URANIUM. 393 



The so-called " probable error " of the average atomic weight computed 

 from the six analyses numbered 7 to 12 inclusive, if each is given the 

 same weight, is 0.015. That is, according to the theory of least squares, 

 the atomic weisht of uranium should be between 238.515 and 238.545. 



The magnitude of the maximum deviations in these two final series is, 

 moreover, about as large as would have been expected from known ana- 

 lytical uncertainty. The observed variation in the amount of sodic 

 bromide, for which a constant correction had to be applied, would account 

 for three quarters of it, and the rest, corresponding to less than the 

 tenth of a milligram in the weighings, might easily be due to unavoidable 

 errors of weighing or manipulation. 



Further evidence of the trustworthiness of the figures is to be found in 

 the comparison of the amounts of silver used in analyses 10, 11, and 12, 

 with the corresponding amounts of argentic bromide, found in analyses 

 7, 8, and 9. This comparison is given in the following table, which 

 gives the weights of silver corresponding to 100.000 parts of argentic 

 bromide. 



The result not only verifies the mechanical work, but affords evidence 

 that the precipitate must have been pure argentic bromide. Clearly, 

 then, the analysis is as accurate as need be. Further repetition of the 

 process might reduce the so-called " probable error," but could not 

 change the average by a significant amount. In the present state of the 

 question, the method seems to have been carried as far as expediency 

 demands. 



It is worth while to inquire whether or not the method may conceal 

 some source of constant error beyond the reach of the experimental 

 precautions detailed above. Such an error could hardly have occurred 



