i6 



DISCOVERY REPORTS 



Prof. Hardy found it necessary to work on fractions as small as one three-thousandth 

 of the total sample in the case of his larger hauls, and even then it was impracticable to 

 count more than a few squares in the case of the smaller colonial forms. The largest 

 sample he examined occupied a settled volume of 290 cc, while many of those examined 

 here were of twice or four times this bulk. Even with the smaller samples the probable 

 range of error is large (see Hardy, in press), and these facts, coupled with the known 

 ' ' patchiness " of the phytoplankton as it occurs in the sea, made it evident that, practically 

 speaking, a comparative estimation, as distinct from any attempt at a true quantitative 

 analysis, was all that could be usefully aimed at. 



It was found that in practice a settled volume of 20 cc. or so of plankton distributed 

 through 150 cc. of liquid, was the greatest amount that could be used if counts were to 

 be readily obtained after the ultimate fractionization with a 0-5 cc. Stempel pipette. 

 Accordingly, all the larger samples were preliminarily fractionized to about this extent. 



The mathematical errors involved in any such estimation must be enormous, but it 

 may be pointed out that in practice differences of 100 per cent and over are the smallest 

 that can be regarded as of much significance, owing to the unequal distribution of the 

 phytoplankton in nature. The effect of this last factor may be judged from Table i, 

 which is based on the analyses of the material from St. 461. It will be seen that at 

 seven hauls taken within a few miles of each other in the course of 24 hours, the esti- 

 mated numbers of phytoplankton organisms deviated from the mean to the extent of 

 — 547 and + 99-0 per cent, and here the samples were sufficiently small and uniform 

 in character to warrant the assumption that the errors involved by the methods of 

 estimation were not excessive. This may be seen from Table XXVI, p. 242, where the 

 full analyses are to be found. 



Table i 



In extenuation of the rough methods adopted it may be urged that if any high degree 

 of mathematical accuracy had been attempted, the analysis of the material would have 

 occupied many years, as among other factors, the eye-strain involved in the counting is 

 very considerable. 



These counts, then, are to be taken as roughly comparable, not strictly quantitative, 

 and the numbers in the case of the larger samples are to be regarded merely as a slightly 



