126 DONALD J. MATTHEWS. 
together, when it was filtered on a small washed paper and washed twice: 
with hot water. The precipitate on the filter and adhering to the beaker 
was dissolved in warm dilute nitric acid and evaporated to dryness on 
the water-bath to remove silica. Seven cubic centimetres of 25% HNO; 
and 20 cem. of water were added, the dish covered, and the solution 
heated for twenty minutes, when it was transferred to a 50 ccm. gradu- 
ated flask. If there is much insoluble residue the solution should be 
filtered. The bulk was then made up to about 47 ccm., leaving space for 
2ccm. of reagent. The standards were prepared by making up the re- 
quisite amounts of P,O,; to about 47 ccm. with 7 cem. of 25%, HNO, 
and water; the writer has generally diluted the ;,-molecular phosphate 
solutions used in determining hydrogen-ion concentrations by the Soren- 
sen method. Two cubic centimetres of the strychnine-molybdate re- 
agent were then added to each flask, the bulk completed to the mark, and 
the whole well shaken. In twenty minutes the solutions are ready for 
comparison. 
The writer has never been able to secure perfect equality of illumina- 
tion in the two halves of the field of the colorimeter owing to the shape 
and setting of the window in the Laboratory, so the precaution was always 
taken of reversing the position of the tubes after six readings and then 
taking another six. The accuracy with which the readings could be 
made varied very much. On some days a set of six have been obtained 
which did not differ by more than 0-2 mm. on 40 mm., while at other 
times the uncertainty was from five to ten times as great. A large sheet 
of ground glass between the colorimeter and the window was often of 
great assistance. Comparisons by artificial light were very difficult and 
fatiguing, but the results were satisfactory. 
To test the accuracy of the method the following experiments were 
made :— 
Part of a standard, containing 0-0237 mg., was analysed against itself. 
In two experiments the results were too low by 0:0005 mg. and 0-0004 mg. 
Three lots of 500 cem. each of distilled water, to which 0-0237 mg. had 
been added, were analysed ; the errors were, +-0-0010 mg., +0-0041 mg. 
and +-0:0001 mg. From another 500 ccm., to which no phosphate had 
been added, 0-0036 mg. was obtained. The filtrate from this was acidi- 
fied and analysed again without adding any more ammonium chloride. 
The amount found was again 0-0036 mg. This value was taken as the 
blank instead of the mean, 0-0025 mg. The result in which an excess of 
0-0001 mg. was found was probably erroneous owing to the evaporation 
having been carried out on the sand-bath, which might give rise to over- 
heating and formation of pyrophosphates. An artificial sea-water was 
