SEA-FISHERIES LABORATORY. 143 
2 im 3 chances, while the probability that the number was 
ereater than the standard was 1 in 3 chances. 
Even if the curve of variability is not an exponential (as 
it is above) but the ordinary curve of organic variability, 
rising from zero to a maximum and falling to zero again, 
we can still apply the foregoing method. The summational 
curve will then tend to flatten out at its extremities and to have 
a change of curvature (a point of inflexion) somewhere. The 
abscissa of this point of inflexion gives the maximum, and 
then we can find the corresponding point on the vertical scale 
and measure off distances on the latter on either side of this 
point giving the various ranges. The rest of the work is as in 
the figure. 
Evidently, therefore, one plate cannot give us a reliable 
measure of the pollution, and the meticulous care with which 
plates are sometimes counted is labour wasted. We must 
have a measure of variability, and even 20 plates, as in the 
above experiment, is sometimes not enough. 
Another method which is now often employed is the 
inoculation of tubes of bile-salt broth by unit quantities of a 
polluted liquid diluted in various degrees. Thus the analysis 
referred to above might have been made in the following 
way :— 
1 c.c. of the liquid in the flask is put into a tube of bile-salt 
broth. If a positive reaction occurs there must have been 
at least one organism in 1/50th mussel ; 
1 c.c. is mixed with 9 c.c. of sterile water, and then 1 c.c. 
of the dilution is inoculated as before. If a positive reaction 
occurs there must have been at least one organism in 1/500th 
mussel ; 
And so on, through a series of decimal dilutions. 
