646 



Fishery Bulletin 91(4). 1993 



between-reader variance may have been atypically high 

 early in the counting period as well. The regression 

 analyses of PD values with time, however, did not pro- 

 vide strong evidence for this because reader counts 



were close to the mean values and no significant tem- 

 poral trends were detected (Fig. 4). Probable cause for 

 the significant effect was that readers were "out of 

 practice" at the start, especially Reader 1. Prior to the 



