that skewness may be introduced by selective availability of the fish 

 or by fisherman selaotion in favor of largest m'smbers of young year- 

 classes, or, in certain localities, converse selection of favored small 

 fisho It may be necessary to include a skew factor to some extento 

 There also are indications of multimodality in the length-frequencies 

 of individual year classes. Whether this is sufficiently pronounced 

 to require statistical recognition remains to be seeno 



De te rmi ne growth rateo A growth curve or curves is needed to give 

 another parameter, tHe mean or modal size, of the frequency distributions 

 of year classes. Having age readings made in any one year it is simple 

 to establish a growth curve, but such a curve would be a true one for 

 only that year. There is evidence that growth differs from year to year, 

 from year-class to year-class, and, more disturbing still, it appears 

 to differ by fishing areas « For instance, a given year"=class has a lower 

 modal length in the San PedrOj than in the Monterey area. We are faced 

 by the dilemma of using either a general growth curve fitting no specific 

 situation, or a variety of curves to fit situations that are largely un- 

 knowTio It remains to be seen whether there is sufficient regularity in 

 geographical or annual variations to permit age determinations of recent 

 years to be applied to the length data of former years, 



Conve r t length-composition to age-c omposition , - The translation 

 of a length-f re quency curve to an age-frequency curve has proved unex- 

 pectedly difficult. As may have been anticipated from the previous dis- 

 cussion of parameters, the only method so far found to be applicable is 

 that of discovering what combination of year=class length-frequencies 

 will fit a season's length-frequency. The fixing of two of the parameters, 

 standard deviation and mean, leaves only the third parameter, relative 

 area, to be varied in the curve fitting. Even so, it is a laborious pro- 

 cess by the trial and error method, but as yet we know no other. Further- 

 more, error in either of the two fixed parameters very seriously affects 

 the results. In other words, the problem should be approached with extreme 

 caution. 



An expedient that greatly facilitates the process may be employed 

 in certain situations v/here one is justified in the assumptions! (1) That 

 the year-classes were of uniform initial numerical strength and (2) that 

 they experienced uniform annual mortality since becoming of commercial 

 size. These assumptions are approximately justified in dealing with a 

 length-frequency distribution of fish pooled from a considerable nijmber 

 (enough to "average out*' inequalities in initial year-class strength) 

 of contiguous seasons during which the fishing intensity was nearly enough 

 constant to have produced nearly uniform mortality. This severely limits 

 the combinations of frequency-curve areas and a unique fit is soon dis- 

 covered. The goodness of fit serves as an indication of how nearly the 

 assumptions were fulfilled, except that uniform trends through the series, 

 either upward or downward and in either year=class strength or intensity 

 of fishing, would not be revealed by poorness of fit. 



IS 



