Variability of Bacteria in North Atlantic Sediments 525 



differences between observed and calculated values in the other 

 two parts of tlie same figure. The points exhibit those charac- 

 teristics expected of handling error and are thought to represent 

 it. The line drawn through the points is an abscissa at ± 7 per 

 cent and tliis may be compared with a handling error of ± 9 

 per cent found by Hayes & Anthony (4) for lake sediment counts. 



The point bearing an arrow in Figure lA & C calls attention 

 to tlie fact that above a certain count (say 500 colonies per filter) 

 agreement with the theoretical distribution begins to break down, 

 even when allowance is made for handling error. The D^ ( Fisher's 

 X^) values for these high counts also indicate that they cannot be 

 accepted as part of a Poisson series (2 and 3), hence they were 

 excluded from present treatment. 



Estimation of bacterial populations involves further errors 

 that have also been examined by Hayes & Anthony (4), whose 

 paper gives methods of calculation. Their results are shown in 

 brackets for comparison with similar analysis of marine sediment 

 counts. The numbers are standard percentage errors of the mean. 

 The average enor of colony counts plotted in Figure lA is ± 17 

 per cent ( ± 23%). Measurement of mud volume is assumed to be 

 subject to the same error, namely ± 10 per cent, and the error 

 involved in converting the volume to weight was found to be 

 the same, namely ± 9.6 per cent, hence the error combined in 

 counting four sediment samples is ± 11 per cent (=ti 13%). The 

 variability between four samples from one station at one time 

 averaged ± 33 per cent (± 31%). If four samples collected at 

 the same time were each taken from a different station, the 

 error was found to be ± 21 per cent (± 35%). If, on the other 

 hand, the four samples were collected one at a time on each of 

 four trips, tiie variability was found to be ^ 27 per cent ( ± 71%). 

 Finally, sixteen samples, one from each of four stations upon each 



error is: 



X 1.2.: 



4 



C shows the result of subtracting the theoretical values indicated by B 

 from the observed points in A. The line is an abscissa passing through ± 7 



percent. 



