Of the second typo ar^ those cases in vfhich a reader simply mis-counts 

 the rings on a clearly marked scale; or those in which he overlookc one or 

 more unusually obscure marks j or those in wluch, by faulty judgment, he 

 miscalls an adventitious mark a true annulus. Th'jse errors are, in gen- 

 eral, the fault of the reader, and are relatively ■'insusceptible to dupli- 

 cation in a second reading, particularly by a diff'^rent reader. Therefore, 

 the results of parallel readings may be expected to provide a measure of 

 the personal error — the reader fault — in interpreting our criteria of year 

 marks and in counting them. 



Scales of 1,187 specimens from the San Francisco fall fishery of 19hO 

 were read by Walford and by Ifoshcr, each \vorking separately and independently 

 Subsequent comparison sho^-ved that l,Oli? or" 88. percent of the pairs of read- 

 ings were identical. The 138 specimens for vj-hich readings were in disagree- 

 ment vrere then subjected to a third examination, this time by the two opera- 

 tors working jointly, and without reference to either of the two previous 

 readings. On this occasion, k specimens v/ere discarded as unreadablei/ ; 

 123 of the joint readings agreed with one or the other of the tvro previous 

 readings ("single disagreements'") J ^^^ 1^ were different from either 

 ("double disagreements"), Jn 31 instances the "single disagreements" 

 concerned the innermost ring, in h9, the marginal ring, and in U3, one 

 of the intermediate rings. In t^^.rms of percentage these quantities are 

 25, ho, and 35, respectively. Since the average number of rings present 

 in specimens of this group is 11.2, those percentages should be close to 

 25, 25 and 50 respectively, if all rings WGro equally difficult to recog- 

 nize. Obviously, the marginal ring was more obscure than the others. In- 

 deed, during the joint readings, it v^as found that uncertainty about the 

 criteria for recognizing a newly formed ring at the margin during January 

 and February was the cause of several of the disagreements, and a special 

 study v/as indicated (see page 9 ). 



A similar test was made by three workers, Tfelford and Mosher; and Phillips 

 of the Calif oniia St9.te Fisheries Laboratory. Scales from li83 specimens, 

 read first by Phillips, Vrcre subsequently divided into t\fo lots, of xvhich 

 one vras read by ITalford, the other by Mosher. The two sets of readings, 

 i.e., by Phillips and by Walford-Hosher, wore identical for U31 or 39 per- 

 cent of the scales. The 52 specimens for which readiogs were in disagree- 

 ment wore then subjected to a third reading, this time by the three workers 

 together. IVhen these joint readings vrerc compared with the original read- 

 ings, it v>ras found that throe disagreed vj-ith both Phillips and with IValford- 

 Mosher ("double disagreements"), and I4.9 agreed v;ith one or the other ("single 

 disagreements"). Disagreements concerned the first annulus in 13 of the cases 

 the ■■marginal annulus in 15, and an intermediate ' one in 21: or, expressed as 

 percentages: 26, 31, and U3, respectively. Thus there was no marked tendency 

 for any one annulus to cause more disagreement than iuiy other. 



A third test v/^ith parallel independent readings was made by Walford 

 and Mosher on 973 specimens taken from lYashington. This sample contained 

 a higher percentagd of fish over throe years old than those described above. 

 For that reason, and also because the scales had not been viell cleaned vfhen 

 collected, this test v;as more severe than the others. Nevertheless the two 

 readers agreed in 709 or 73 percent of the cases. Of the remainder, 235 

 were "single disagreemerfts", and 29 were "double disagreements." 



±/ Less than 0.2 of 1 percent of the scales had been discarded as unreadable 

 on first reading. These were specimens with regenerated centers. 



101 



