102 . TRANSACTIONS LIVERPOOL BIOLOGICAL SOCIETY. 
Working out all these series (really a very simple matter) 
we get values of the coefficient a, thus enabling us to draw smooth 
probable growth-curves. The coefficient a is really a measure 
of the nutritive condition of the fish and describes the whole 
sample. Fimding mean a’s for the two years 1916-17 we get :— 
May a= 0-00585; weight of herring = 23-5cms.= 76 grms. 
June a= 0-00705; weight of herring = 23-5cems.= 92 germs. 
July a= 0-00825; weight of herring = 23-5ems.= 107 grms. 
Augt. a = 0:0092 ; weight of herring = 23-5ems.= 119 grms. 
Sept. a= 0-0087 ; weight of herring = 23-5ems.= 113 grms. 
These variations in the numerical value of a make really 
quite considerable differences in the mean weights. The curves 
need not be given as they can very easily be reproduced from 
the equation W = al’. We see, however, that a increases from 
May to August, and then decreases because the “ condition ” 
of the fish is falling off during the latter month and some of them 
are beginning to spawn. 
Thus there is a gradual increase in the food value of each 
herring during the progress of the season because the fish is be- 
coming plumper and its roes or milts are enlarging. There — 
is also a change in the chemical composition of the flesh such 
that the fish is, all the while, becoming a better article of food. 
This change consists of (1) an increase in the amount of oil con- 
tained in the flesh, and (2) a decrease in the amount of proteid. 
The former change has a far larger effect than the latter so that 
the net effect is to increase the food value of herrings. These 
changes are recorded in detail in Tables I-III, but they will be 
better seen by making summaries of Tables II and II. In the 
following tables the numbers are percentages of the net weight: 
of fresh herring flesh. 
