326 



BULLETIN OF THE BUREAU OF FISHERIES 



for corresponding years of life is less in the young than in the old fish, however. This 

 is indicated more clearly in Table 20, in wliich is shown the amount of correction 

 required to bring the computations of the two series into agreement. 



Table 21. — Amount of deviation of comptited length values (calculated K) from actual (K) length values 

 according to length of time that intervenes between the age of the fish at death and age for which 

 lenqth is calculated 



1 This deviation is not included in the average. The discrepancy is large on account of the small number of nonrepresentative 

 2-year fish available for comparison. 



The question now arises. Which series contains the most accurate computations? 

 To provide an answer to this question Tables 17, 18, and 21 have been constructed. 

 In Table 18 the computed lengths are compared with the measui'ed lengths for certain 

 years of life, whUe in Table 21 the differences between these computed and measured 

 lengths are compared according to the number of years that intervene between the 

 age of the fish at death and the age for which the computations are made. Unfor- 

 tunately no reliable actual values for the early years of life are available. It is dur- 

 ing these years that the most significant dift'erences occur. To indicate the trend of 

 the discrepancies, however, the general averages of the measured lengths of all available 

 herring are compared with similar general averages of the computed lengths as shown 

 in Table 17. 



An analysis of Table 18 reveals that for the fifth year of life the computations of 

 both series deviate from the measured values to the same extent and are too high, and 

 that for the fourth and third years the calculated values are, on the whole, too low, those 

 based on the anterior radius being shghtly lower than those based on the diameter. 

 Further, the discrepancy is greater for the third than for the fourth or fifth year. 

 The calculated values of Table 17 show the same characteristics — they are too high 

 in the fifth and sixth years and too low in earlier years, the discrepancies in the latter 

 years being the greater. 



Further, the figures of Table 21 show that when one year intervenes between 

 the age of the fish at death and the age for which computations are made (that is, 

 when the length is computed for the preceding winter) either dimension may be em- 

 ployed, as the calculations of both series vary from the measured values in the same 

 degree, but that the discrepancy increases considerably more in the computations 



