the total effect of the errors is to reduce the final count. For 

 example, in transferrinc the catch from trap to bottle there is an 

 opportunity to lose organisms but none to §;ain them. Possible errors 

 in making the catch up to volume tend to cancel, but in actual 

 enumeration, there is a greater chance of overlooking an organism 

 than there is of counting it twice. Moreover, in taking a saiaple with 

 the piston pipette, some organisms adhere to the sides and neck of 

 the bottle, and thus are not taken. Without doubt the most important 

 source of error is the failure of the piston pipette to take an 

 'absolutely representative sample of the contents of the bottle. This 

 error probably depends largely on the fact that the organisms cannot 

 be distributed with exact uniformity in the bottle. The accuracy of 

 this part of the method has not been determined. A few preliminary 

 experimental counts with knoim concentrations of organisms indicated 

 that the accuracy varies with the concentration, and since it probably 

 is different \rith different organisms, the number of counts necessary 

 to solve the problem would be very great. 



An idea of the precision of the method, that is, the degree 

 of similarity of duplicate samples from the same bottle, can be gained 

 by examination of the data in Tables 6U, 65, and 66. Table 6U shows 

 counts of duplicate samples from each of four bottles containing 

 plankton from four depths at Station l58. Tables 65 and 66 show a 

 number of duplicate counts on composite samples from Station 117 and 

 13h. At these stations the individual samples of a vertical series 

 were combined in one bottle for counting. 



Inspection of the many pairs of counts shows that, in general, 

 the absolute difference between the two counts of a pair increases 

 with an increase in the number of organisms in the pipette sample, but 

 that the percentage deviation from the mean decreases. It is advan- 

 tageous, then, to have the catch highly concentrated in the sample 

 bottle. However, to bring the rare forms to the proper concentration 

 would result in such high concentration of the abundant forms as to 

 lengthen unduly the time required in making the count. Most of the 

 pairs of counts in the tables show close agreement. The principal ex- 

 ception is in the first two counts at Station 117 on August 30. In 

 these samples the lack of agreement for Cyclops and Bosmina was so 

 striking that a third count was made. It appears that, whatever the 

 accuracy of the method may be, it gives fairly consistent results. In 

 the opinion of the writers, the agreement shown by the pairs of 

 counts given, and by many others at hand, is sufficiently close to 

 validate the routine procedure of making two counts, particularly since 



205 



