58 University of California Publications in Zoology [ Vou. 14 
in San Francisco Bay, which had been saved for this purpose, were 
titrated with the ‘‘standard’”’ outfit and solutions. Finally, in August, 
1913, seven of the recently taken water samples were titrated accord- 
ing to both the older and the newer methods. 
These comparisons indicate a mean difference of 2.3 per cent be- 
tween the results obtained by the two methods. The magnitude of this 
difference agrees fairly closely in the various cases, and is invariably 
of the same sign, the figures obtained by the ‘‘standard’’ method 
being smaller. Since the latter is probably more accurate than that 
earlier employed by us, we have corrected all of the first year’s 
figures by the amount of this mean difference, i.e., we have deducted 
2.3 per cent of their value. (In reality, they were multiphed by 0.977). 
Dr. G. F. McEwen, of the Scripps Institution of the University of 
California, has been so good as to determine the specific gravity of 
eight of our samples from San Francisco Bay, by means of an appa- 
ratus of high precision. For seven of these samples, the mean 
difference between Dr. McEwen’s figures and ours (with the 
““standard’’ titration outfit) was 0.7 per cent, our figures being the 
lower. In other words, our figures were, on the average, less than 
Dr. McEwen’s by seven-thousandths of the latter. In the eighth 
case, it is obvious that an error of some sort was made, for the recorded 
difference is far greater than could be accounted for by the difference 
of method. 
We are disposed to question whether the accuracy of the titration 
method of determining the salinity of sea-water has not been some- 
what overestimated. A considerable source of error is to be found 
in the reading of the burette scale, and an even greater one in 
determining the point at which the color change takes place in the 
solution. The mere fact that an observer can with practice learn to 
obtain almost identically the same figure upon repeating the test of 
a given sample, or even the fact that two experienced observers may 
obtain readings which differ by a very minute quantity, does not 
prove that their errors are infinitesimal. It may merely prove that 
their errors are equal. As a result of adopting uniform methods of 
procedure, our two observers obtained, after some months of practice, 
salinity figures which agreed with one another very closely. And yet 
it has been shown that throughout this phase of the work there was 
a nearly constant error of more than two per cent in their figures, 
standard’’ method to be absolutely accurate. 
We believe, none the less, that the titration method, even as em- 
ee 
assuming the 
