3.8 



8.9 



15.9 



7. 5 



Sample size 120 



1/ Average of two readings. 



without consulting the other readers; (2) any 

 disagreennents between readers are then rec- 

 onciled in a joint meeting of the readers. 

 Errors ranged from 6,7 to 50.0 percent for 

 150 females in ages 3 to 7, and from 30.0 to 

 100.0 percent for females in ages 8 to 1 1 for 

 two readings (table A-24). For seals < 7 years 

 old and > 8 years old 91 percent and 82 per- 

 cent of the assigned ages, respectively, were 

 within plus or minus 1 year of the correct 

 age. For seals > 8 years old, 96 percent of 

 the assigned ages were within plus or minus 

 2 years of the correct age. In general, a con- 

 ference reading between readers A, B, and 

 C (the most experienced readers) gave lower 

 errors than the individual readings (table 41). 



A binomial test for bias in the nunnber of 

 errors above and below the correct age 

 showed that Reader A had no significant bias 

 for whole teeth and Reader D had no significant 

 bias for sectioned teeth in ages 1 to 6 years 

 (P>.05). For whole teeth from 7-year-olds, 

 all readers tended to assign ages too low 

 (P<.01). Only Reader B had no significant 

 bias for females 8 to 1 1 years old. Among 

 observed errors, there was a tendency for 

 readers to call 4- and 5-year-old seals 1 year 

 younger, and 6-year-old and older seals 1 or 

 more years younger than they actually were. 



An analysis of variance test of the percent 

 errors transfornned by square root showed 

 that the readers, ages of seals, whether the 

 tooth was whole or sectioned, and the age -type 

 of tooth interaction were highly significant 

 (P<.01). The reader-age and reader-type of 

 tooth interactions, and the mean square for 

 duplicate readings were not significant 

 (P>.05). The reader and method with the 

 lowest error should give the best results. 



We conclude that Reader A should determine 

 ages of female seals < 6 years old from a 

 single reading of whole teeth. The error for 

 Reader A is higher than the conference read- 

 ing from sectioned teeth for 3-year-olds, but 

 the difference is too small to warrant section- 

 ing the teeth of 3-year-olds and miaking addi- 

 tional readings. Corrected numbers of females 

 killed on the Pribilof Islands were not com- 

 puted because no females were killed there in 

 1966. A method sinnilar to that used for males, 

 and using the errors for Reader A from whole 



Table 41. --Errors in assigning ages of female fur seals from sectioned canine teeth 



Reader 



Error for age 



10 



11 



-Percent 



28. 3 21. 7 44. 3 



38.4 33.4 48.4 



30.0 31.7 51.7 



31.7 36.7 53.4 



Conference 

 reading 3. 3 



6. 7 



20. 



16. 7 



23. 3 



26.7 



50. 



36.7 



66.7 



Sample 

 size 30 



30 



30 



30 



30 



30 



30 



30 



^-^30 



1/ Average of two readings. 

 7/ Sample size=24 for reader B and conference reading. 



40 



I 



