290 



stream is 3.58 per cent; from a floating boat, 11.2 per cent.; at 

 intervals of 1 — 7 days for periods of 2 to 5 days in the more 

 stable hydrographic conditions, 14.1 per cent.; and in the 

 stream as a whole for 200 miles of its course, 57 (total 

 catch) or 89 per cent, (plankton estimated). If, however, we 

 break up the 200 miles into four sections representing sub- 

 ordinate units of environment, each dominated by some local 

 factor, the ± departures from the mean are 12, 51, 32, and 76 

 per cent, respectively for estimated plankton (i. e. after silt 

 deduction), or 44, 5, 36, and 34 per cent, for the total catches, 

 the averages for the two methods being ± 43 and ± 29.7 per 

 cent. 



The average departure from the mean catch in two trans- 

 verse series of 10 catches each is ± 27.2 or ± 22.3 on the basis 

 of plankton content per m. 3 If we eliminate the shallow-wa- 

 ter shore collections, the departures fall to ± 21.9 and ±12.1, 

 or on the basis of volumes under 1 sq. m.,to ± 15.4 and ± 20.2. 

 The departure from the mean number of planktonts is only ± 

 7.8 for the whole cross-section. 



These results are in the main within the ± error of distri- 

 bution of the plankton in lakes arrived at by similar methods 

 of computation. The plankton method may therefore be applied 

 to the quantitative investigation of the life of a stream as legitimately 

 as to that of a lake. The laws of the horizontal distribution of the 

 plankton are in this respect essentially the same in both types of 

 aquatic environment. 



Whether or not a fundamental source of error as large as 

 this — probably the greatest of all the errors in the method as 

 we have used it — vitiates the utilization of such data for scien- 

 tific conclusions must be to some extent a matter of opinion. 

 The extent to which it renders conclusions tentative must de- 

 pend upon the distribution of the error, the extent of the data, 

 and the method of their utilization. Personally I may say that 

 close study of the at first sight aberrant data upon which 

 this paper is founded, has led me to attach less significance to 

 this source of error than I was at first inclined to do. Readers 



