Page:Proceedings of the Royal Society of London Vol 69.djvu/96

88

It is believed that in the majority of cases the mean error does not greatly exceed one part per thousand. In every case the determinations of the concentration of the samples were conducted either immediately before or after those of the densities or refractive indices, as the most concentrated acid is as hygroscopic a substance as sulphuric acid.

In order to show the degree of concordance obtainable from various methods the following cases may be cited.

The value of a certain soda solution was determined at constant temperature as against a certain hydrochloric-acid solution; using variable quantities five values were obtained with a mean error of 1 part in 2844 parts, calculated by the method of least squares. The hydrochloric acid solution had previously been standardised on two dif- ferent occasions ; at one time as against recrystallised silver nitrate and purified sodium carbonate, and at the other against another sample of silver nitrate of presumably different past history ; the mean error by the one method was 1 part in 1850 parts, and the total mean error by the two methods was 1 part in 1298 parts.

The same solution was determined twice as against a certain sul- phuric acid solution, the value of which was ascertained by determining its density at 15/4, ascertaining its percentage strength by interpola- tion from Pickering's tables, and hence the amount of acid in 1 c.c.

The results of all the determinations may be summarised as follows, and expressed as nitric acid equivalent to soda in 1 c.c.

(i.) By hydrochloric acid method = 0*06136 gramme (ii.) By sulphuric = 0-06144

Mean 0-06140

Namely, a mean error of 1 part in 1533 parts.

By soda it is desired to imply alkalinity only ; the metallic sodium used for its preparation might have and probably did contain small quantities of other alkaline metals, but as such soda solutions merely served as go-betweens to express the one acid in terms of the other two the possible effect of such impurities is thereby eliminated.

Attention has recently been drawn to errors in burettes, not only by the Eeichsanstalt of Berlin but also by individual writers, especially Wagner ;* these errors may be summed up as those of (i) Parallax, (ii) temperature, and (iii) drainage. The last of these varies, of course, with the length drained, the viscosity of the solution, and the often wholly unknown condition of the cleanliness of the glass.


 * 'Zeits. Physikal. Che:n.,' vol. 29 (1839), pp. 193-219.