NO. 13 SALINITY OF PACIFIC SURFACE WATER — CLARK 



In the cases in which the reading from the circulating pump was 

 greater than that from the surface water (excluding No. 9 in which 

 the surface reading, from rain or some other cause, is obviously 

 abnormally low ; and No. 99 in which the reading from the circu- 

 lating pump, probably because of evaporation, is obviously too high) 

 we find a variation between 0.00092 (No. 97) and 0.00003 (Nos. 19, 

 29, 51 and 81), with a mean of 0.00018. 



In the cases in which the reading from the circulating pump was 

 less than that from the surface water (excluding Nos. 5 and 12 in 

 which the surface reading is obviously too high) we find a variation 

 between 0.00065 (No. 11) and 0.00003 (No. 50), with a mean of 

 0.00022. 



The fact that the variation was plus in twenty-four instances and 

 minus in twenty-three with no variation in three, that there is no cor- 

 relation between the variation and the latitude, and that the average 

 of all the variation (+0.00018 — 0.00022) is —0.00004, a quantity 

 altogether too small to be detected by the instruments used, shows 

 that in reality the difference in the water at the surface and a few 

 feet below it is entirely negligible, not falling within the scope of pos- 

 sible detection by the instruments used. 



This being so we have a personal and fortuitous error of 

 0.00018 + 0.00022 = 0.00040 to take into consideration so that, 

 under the best of conditions we must admit of a possible variation 

 of ±0.00020 from the truth. This shows graphically the absolute 

 insignificance of the figures in the fifth decimal place and their entire 

 dependence upon chance. 



THE REJECTION OF INACCURATE READINGS. 



Assuming that the readings from the surface and from the circu- 

 lating pump are, so far as lies within the power of our instruments 

 to detect, the same, we are able to compare the two readings and i:i 



