OF ARTS AND SCIENCES. 181 



sources. This conclusion still remains in force, even supposing there 

 be a constant error in the process, for the constant error must affect all 

 the results equally, and could not possibly equalize unlike results. 



A third conclusion pointed out by the determinations is that the 

 argentic nitrate was the normal compound, and quite pure; for it will 

 be remembered that two entirely different samples had been used in 

 the course of the work. 



There is but one point which remains to be considered, and that is 

 the existence or non-existence of a constant error in the reaction. 

 That this is by far the most important point in the whole discussion, it 

 is unnecessary to state. Whether there be such a constant error, 

 future investigations may show ; for the present, it is sufficient to say 

 that it is extremely difficult to see where such an error might creep in. 

 The whole reaction is so simple and so sharp, that the probability of 

 error is reduced to a minimum, and in every case any possible cause 

 of error has been guarded against. 



Professor Cooke, under whose direction the whole investigation has 

 been conducted, suggested that similar experiments be made, using 

 argentic sulphate instead of the nitrate ; but after a large number of 

 trials this was found to be impracticable : first, because the solution 

 has a much higher freezing point than that of the nitrate; and sec- 

 ondly, because the solution was necessarily so dilute, on account of 

 the slight solubility of argentic sulphate, that the complete precipita- 

 tion of the silver required a much longer time, giving more opportu- 

 nity for secondary reactions. The silver was always accompanied by 

 a very slight admixture of some basic cupric sulphate ; and hence 

 this method, which, if successful, would have been able to throw much 

 light on the question of a constant error in the previous results, had 

 to be abandoned. 



Cambridge, December 15, 1887. 



