38 University of California Publications in Zoology [Vol. 15 



nometer method, measurements made with the hydrometers are ex- 

 ceedingly erratic, errors in specific gravity of two or three in ten 

 thousand being not uncommon while the average was as great as one 

 in ten thousand. 



Altogether, therefore, the sinker method is the most satisfactory. 

 In addition to being as accurate as the pycnometer and as rapid as 

 the hydrometer method, it has the advantage of enabling two inde- 

 pendent determinations to be made without undue loss of time, which 

 greatly reduces the chance of blunders. 



(&) The experience of others. — Nansen (1902), after careful and 

 detailed investigations made during the Norwegian North Polar Ex- 

 pedition of 1893-1896, concluded that the specific gravity of sea- 

 water determined by any ordinary hydrometer of constant or variable 

 weight may be considerably in error. The magnitude of the error he 

 found to be proportional to the diameter of the stem as compared 

 w'ith the volume of the body of the instrument. The source of error 

 he discovered to be due to surface tension of the water, which varies 

 in an erratic manner depending upon contamination and the time 

 during which the surface is exposed. The readings of a Kiichler 

 hydrometer, such as used bj' the Scripps Institution, he found to vary 

 by 0.0003 in consequence of a variation in surface tension from its 

 maximum to its minimum. Even with the greatest possible care in 

 cleaning the hydrometer and testing jar, if the water surface be neg- 

 lected, an occasional variation in surface tension would occur suiBcient 

 to alter the reading by 0.0002. By taking precautions to clean the 

 water surface thoroughly, Nansen was able to obtain the specific 

 gravity to an accuracy of two in a hundred thousand, but he found 

 the method so troublesome that he recommends the u.se of a pycnometer, 

 sinker, or hydrometer of total immersion. 



This experience of Nansen 's is corroborated by that of Schott 

 (1902) and Buchanan (see Nielsen, 1912), the former having neg- 

 lected the error arising from surface tension while the latter avoided 

 it by using distilled water as a means of control and calculating the 

 error involved in each determination. Schott estimated the error to 

 be fifteen in a hundred thousand, while Buchanan found an error in 

 his determinations of only three in a hundred thousand. 



With respect to Mohr's titrimetric method Dickson (1901, p. 74), 

 relative to his own experience, says: "The results seem to show con- 

 clusively that without spending the time nece.ssary for the highest 

 degree of accuracy, but merely taking ordinary care, the chlorine of 



