Double reads indicated that reader uncertainty was predictive of reader error rates.
Maturity asymptotes of 1.0, 0.96 and 0.81 suggest differences in skipped spawning.
Models with multiple covariates (e.g. length, age, date) improve maturity estimates.
Variation among readers analysis was largely due to slide quality and experience.
Readers with little training can be used for accurate analysis, if uncertainty is noted.