26 



DISCOVERY REPORTS 



covered. A ^-in. objective conveniently covers the width of the 2 mm. squares; but 

 a l-'m. objective has continually to be brought into use. If a large number of very small 

 forms are present it will be necessary to travel backwards and forwards over each 

 individual square with the /.-in. objective; this can be done without covering the same 

 ground twice by using a grid form of micrometer eyepiece and counting the number of 

 organisms viewed through the grid, and then moving the slide until another portion of 

 the square comes below the grid. A particular organism or mark on the slide lying under 

 the right side of the grid is kept under observation and the slide moved until it comes to 

 lie on the left. If small organisms are very numerous it will only be possible to count 

 those present in a limited number of squares taken at random over the slide. By taking 

 the average in these squares and multiplying by the appropriate factor an estimate of 

 the total number on the slide is obtained. 



Accuracy of N 50 V analysis methods 



Usually only one fraction of a sample collected has been analysed. A number of such 

 fractions might have been analysed for each sample and their average taken ; but we do 

 not believe that the time taken in gaining a more accurate estimate of the contents of 

 the sample is justified when we know that the methods of collection are necessarily only 

 giving us an approximate indication of the plankton in the vicinity of the station. 



In order to estimate the degree of reliance which can be placed on the analysis of 

 single fractions, six such fractions were taken from a single sample collected at St. 133 

 and analysed for five different diatoms which varied considerably in size and shape. 

 The fraction of the sample examined in each instance was 1/600. 



The results are tabulated as follows : 



Thus we see that the result of an analysis of a single fraction is usually 18 per cent 

 larger or smaller than the true content of the samples and it may in exceptional instances 

 be as much as nearly 50 per cent larger or smaller. These facts must be kept in mind 

 in interpreting the results which follow and will be referred to from time to time in the 

 later text. Thus, when studying the distribution of any form, should we see 30,000 of 



