1915] Michad.fi al.: Hydrographic lircords of Hcrippslnsiitution 37 



This table shows that the average ditferenee between the mean of 

 the two determinations made by the sinker and that made by the 

 pycnometer method is within 0.00002, while the extreme difference 

 does not exceed 0.00004. Since these differences involve the errors 

 inherent in both methods and since the average error due to the pyc- 

 nometer (see p. 30) is approximately 0.00002, it is obvious that 

 that due to the sinker can exceed 0.00001 but slightly, which checks 

 the estimate heretofore given (see p. 36). Of course blunders are 

 occasionally made which result in an error exceeding 0.00004, but 

 the second te-st which is always made brings it to light immediately, 

 whereupon a third te.st determines which of. the two is correct Since, 

 in addition to its great inherent accuracy, the observations and reduc- 

 tions involved in two independent determinations of .S' - require onlv 



4 



five minutes, the method is by far the most satisfactory of any thus 

 far tried. 



3. COMPARISON OF OUR EXPERIENCE RELATIVE TO ACCURACY AND 



EFFICIENCY IN WATER ANALYSIS WITH THE EXPERIENCE 



OF OTHER INVESTIGATORS 



(a) Brief review of our own experience. — The experience with the 

 methods used by the Scripps Institution has shown that measurements 

 of specific gravity made with the pycnometer are more satisfactory 

 than titrations of chlorine content. Although, after some practice, 

 we obtained fairly consistent results with the titration method, we 

 failed to reduce the time of a single test below that required by the 

 pycnometer method. The latter also gave much more accurate re- 

 sults, the error in specific gravity seldom exceeding 0.00003 and aver- 

 aging less than 0.00002; the error in the titration method averaged 

 0.00007, frequently being as great as 0.0001, and occasionally as much 

 as 0.0002. 



The hydrometer and sinker methods proved to be the most rapid, 

 the former requiring five minutes for a single determination, including 



the necessary reductions to S-, while two independent determinations 



and reductions can be made by the sinker method in the same time. 

 However, while this method has proved to be as accurate as the pyc- 



