To learn then whether such a difference laight be centered about fish 

 of certain size categories, a chi^square comparison of age composition 

 V\ras made for each size class. 



For this purpose, size classes vrere made as small as possible — usually 

 two millimeters — but lar<];e enough to contain, for each member, a minimum of 

 ten specimens. The results of this analysis indicated no significant dif- 

 ferences betvreen the two readers — all P values virere well above .05 — except 

 for eight classes in the size range 211-226 millimeters. For each of these 

 was indicated a probability of less than 0.001 that a second pair of samples 

 from the sajiie population could differ as much or more by chance. Accordingly, 

 scales of the 1,131 fish in this size range vrere reexamined, the two readers 

 exchanging their respective portions so as to check each other's first inter- 

 pretations. It transpired from the subsequent joint readings of the disagree- 

 ments, that the chief cause of disagreement had been confusion as to the 

 criteria for distinguishing betvreen old and neiJ- rings .at. the margin, par- 

 ticularly during January and February, v/hen accelerated grovrth is exposing 

 the nev;-ly formed annulus. 



This difficulty vfas at length resolved by a special. study as described 

 on pages 9 to 11, which provided an objective method of making the neces- 

 sary distinction. After the method had been applied, the two original por- 

 tions, Vfelford's and Mosher's, >vere again compared as to age composition, 

 by the chi-square test. On this occasion the probability v^as O.kh that the 

 two distributions would differ as much or more by chance in a second pair 

 of samples drawn from tho same population. 



Thus an equivalent of parallel reading has been devised, v/hich saves 

 considerable labor, yet serves to detect and correct errors in the inter- 

 pretation of tho critv.ria of age marks. The method was applied during the 

 19U1-U2 season, when three persons were employed. The average P value ob- 

 tained on first readings was .U8,. indicating no significant difTerence in 

 results among the three readers^;/. 



RECOGiJITION'OF THE NST Alff^LUS 



In a previous paper (L'alford and Mosher, 19h3), it was shown that the 

 nevf annulus appears on the scales of young pilchards, up to three years of 

 age, from late fall to early spring. Judging from the present data, it ap- 

 pears at the same season in the scales of adults. Because commercial fishing 



1/ It has been suggested by those critically reading the manuscript of this 

 paper that the above discussion tends to over-simplify age determination of 

 pilchard scales. It must be emphasized that a considerable degree of' judgment 

 must be exercised in recognizing age marks. This reauires skill obtained 

 from experience illuminated by Imowledge of the biology of groirth. It also 

 requires much patience and a certain sympathetic attitude tbvrard scale reading. 

 Nevertheless, the evidence described on pages llff v/as obtained despite the 

 subjective nature of these imponderables. It appears, however, that age 

 reading cannot be practiced equally successfully by all investigators. 



106 



