Thus in three tests involving parallel readings of 2,6U3 specimens, 

 there was disagreement on U5U of the cases — "disagreements" — and agreement 

 in 2,1;^3 — "agreements." ViThen the former were re-examined by the two par- 

 ticipating workers, the final decision agreed with one or the other of 

 the two original readings in i;08 cases and disagreed with both in only U6. 



In close to 90 percent of the cases, disagreement involved a choice, 

 of two adjacent ages, and, therefore, uncertainty as to the existence of 

 only one of the rings ; and only 10 percent involved uncertainty as to more 

 than one ring. Save for the first of the three tests, no one annulus was 

 the subject of disagreement more frequently than any of the others. For 

 the first test, disagreement over the marginal annulus, which was somewhat 

 in excess of expectation, was later corrected as a result of a special study 

 (see page 9 ). 



A reader's decision as to which of two adjacent ages a "disagreement" 

 should be referred, might possibly be determined by chance. Assuming this 

 to be true, half the specimens should be assigned to the younger age, half 

 to the older. Consequently the size composition of fish comprising the 

 two halves should be statistically identical, and should agree equally well 

 or badly with the size composition of the "agreements" .of the two adjacent 

 ages in question. To examine this hypothesis, a series of chi-square com- 

 parisons was made, the results of which are summarized in table 2. The 

 tabulated figures give the probability of obtaining, in a second series 

 of samples, differences in size composition as great or greater, by random 

 error . 



Table 2.- P values obtained on chi-square comparison oetween size 

 composition of agreements and revised disagreements. 



* Final determination on revised disagreements 



—' Figures in parenthesis indicate degrees of freedom. 



102 



