1915] Miclincl. (i al.: H ijdrogrdpJiie Records of Scripps InstUution 39 



large miniluTs of sjiiiiples can ho (Iptcriiiined with eoinparativcly little 

 trouble to within the admissible limit of error, ±0.03 Voo-" Since 

 this corresponds to an error in specific gravity of five in a hundred 

 thousand, his results are more accurate than ours. 



Regarding the accuracy and rapidity with which this method may 

 be carried out. Ilelland-Hansen (1911-12, p. 39) states that in some 

 Norwegian investigations, "one series of 260 water samples was 

 titrated twice, with a mean difference between two determinations of 

 the same sample equal to 0.016 "/„„ in salinity," which corresponds to 

 about 15 in 1,000,000 in specific gravity. Relative to the speed he 

 says: "With some practice the whole process of titration can be gone 

 through in about four and one-half minutes." 



To ascertain the accuracy of the same method Nansen (1902) had 

 a number of water samples independently titrated by several investi- 

 gators. By comparing the results he concluded that, under favorable 

 conditions, when the investigator had had special practice and the de- 

 terminations were controlled after everj^ tenth or fifteenth test by the 

 titration of "normal sea-water," the specific gravity could be deter- 

 mined within 15 or 25 in 1,000,000. However, he says (1902, p. 231) : 

 "Considering that even the results obtained by the methods of gravi- 

 metric analysis of the chlorine (halogens) may differ as much as 

 0.037 %„ [which corresponds to 3 in 100,000 in specific gravity] with 

 such skillful investigators as Pettersson and his assistants, I think the 

 probability is that the accuracy attained by titration will in many 

 eases be lowered." This is corroborated by Bjerrum (1904) who, 

 after making a thorough study of all the sources of error inherent in 

 the method, concluded that their combined effect could not be reduced 

 below a mean error in a single specific gravit.y determination of 22 in 

 1,000,000, and, since this does not take into consideration errors which 

 ought not to be committed but which are not always avoided in prac- 

 tice, the actual error is apt to be somewhat greater. 



Finally, there are errors in titration peculiar to the observer him- 

 self. Perhaps the best illustration of this is afforded by the follow- 

 ing measurements, where the results obtained by two observers (A and 

 B) in the titration of the same samples are compared. While the 

 results were originally published by Homen (1904, p. 39) in terms of 

 chlorine content, we have reduced them to specific gravity (Sf ) by 

 means of Knudsen's tables, in order to make them comparable to those 

 so far discussed. 



