145 



of the h.eooii'd order only, so that it could be deterniiued accurately from three 

 iiccurately located points, provided one of these is near the zero end, a second 

 near the center, and a third near tlie upper end of the scale. The distribution 

 ■of the points is therefore of more importance than the actual number. Conse- 

 •((uontly it may be advisable to make a special effort to get a fevr points corre- 

 sponding to salinities as low as 29, and a few as high as 37. It is presumed 

 that there will be available a considerable number of samples having salinities 

 in the range from 30 to 35. 



The general procedure in making these measurements should be identically 

 tlie same as that outlined above for the use of the appai'atus, except that points 

 ■on the curve corresponding to known salinities and observed readings of the 

 apparatus are determined instead of the salinities from a curve or table and 

 the observed readings of the apparatus. 



Any systematic error in the titrations of the sea water to be used in this 

 Avork will of course be carried directly to the curve or table. An estimate 

 of the magnitude of the errors from this source can only be made by the 

 persons who will have the supervision of this part of the work. The indications 

 are that errors from other sources need not exceed 0.01 in salinity. The curve 

 or table determined wdth any one of the cells used as the X cell will not apply 

 in the case of the use of any other cell, and, it will not apply in case of 

 other apparatus even of the same type, unless the ratio of the resistance of a 

 division of the slide wire to the resistance of the slide wire plus the end 

 coils is the same for this apparatus, that is, unless the constants m and n re- 

 ferred to in Appendix I have the same values. 



In case the curve i.s plotted to a scale sufficiently large so that it can be 

 read somewhat better than hundredths in salinity it may be too large for con- 

 venient use on ship board, therefore it may be advisable to construct a table 

 for this purpose and similar to the table given above. The data for such a 

 table might be taken directly from the curve plotted to the large scale. A 

 record should be kept of all readings taken w^ith the apparatus for the purpose 

 of making the curve so that in case difficulties are encountered in the con- 

 struction of the curve by this means, its equation can be determined by an 

 analytical method and the table calculated from the equation. 



It was decided to calibrate the apparatus according to the method 

 just described in Appendix II, which possesses the advantage of 

 eliminating previous computations. The slide-wire readings were 

 plotted on a large scale graph directly against salinities. This work 

 was performed during the winter of 1923-24 at the museum of 

 zoology. Harvard University, there being a large supply of known 

 samples' available from the ice season, all of which had previously 

 been titrated at Cambridge. The procedure in detail was as follows : 



1. A solution of known salinity 30.91 '^/^o was placed in A' cell. 



2. A solution of about 33.00 Voo was placed in Y cell, or of such 

 salinity that for a small setting of Q the reading on the slide wire 

 Avould be about 200. 



3. By varying the salinity of this solution slightly we finally ob- 

 tained a balance of the slide at 235 when Q read 243. This number 

 of 235, therefore, was the first point in the construction of the curve. 



4. KnoAvn solutions of increasing salinity were successively placed 

 in the A' cell and the readings of the slide gave points on the graph, 



