526 Marine Microbiology 



of four trips, reduced the error to ± 13 per cent (± 36%), of 

 which ahnost half is due to handHng. 



Apparently the membrane filter technique achieves greater 

 precision with marine sediments than it displayed with lake sedi- 

 ments. This may reflect a greater inherent stability in the marine 

 population, but it must be borne in mind that the present ob- 

 servations cover only one season, whereas those of Hayes & 

 Anthony extended over several years. 



In view of the practice of averaging trip means via their 

 logarithms (see Table 1), it would be more appropriate to ex- 

 press the standard error as 2^ 1.13 instead of ± 13 per cent, but 

 for sake of direct comparison with Hayes & Anthony, the latter 

 method of expression has been retained. 



TABLE 1 

 Bacteria in Top 5 cm of Marine Sediment 

 "Station" does not mean a precisely located spot, but one of the sites 

 chosen at random upon each visit to a region. Each station provided one sediment 

 sample from which four membrane filters were prepared. A mean colony count 

 per filter was determined for each station, the station means were averaged 

 arithmetically to provide a trip mean colony count, and the logarithms of the 

 trip means were averaged to obtain the final mean counts shown in this table 

 and in Figure 2. Column 6 records, as percent of the arithmetic mean, the 

 reduction in recorded numbers resulting from averaging trip means logarithmically. 

 For example, the Fundy count, had it been averaged arithmetically, would have 



been = 130. A small difference between the two means reflects a high 



100-4 ^ 



degree of agreement between replicates. The rationale of averaging by logs is 

 discussed by Robertson (5) and Hayes and Anthony (4). 



Strictly speaking, only one cruise was made to this region. Explanation in text. 



