Seven prey species (ntotal > 2,700) were fed to seven captive male Pacific harbor seals (Phoca vitulina richardii) in 177 experimental meals to quantify biases associated with scat analysis and current consumption models. Hard parts from an individual meal were recovered in an average of 3.8 ± 1.8 scats (range 1–10; mean ± SD). Overall, 57.7 ± 33.2% of otoliths and 89.5 ± 15.5% of squid beaks were recovered. Recovery rates varied, and prey with smaller, fragile otoliths were recovered in lesser quantities than prey with larger, robust otoliths. Recovery rates of all prey except pink salmon were improved by a mean of 31.7% when all diagnostic structures were included in estimates. Estimated recovery of pink salmon was 9.5 times that fed seals based on the all-structure technique. Mean length reduction of recovered otoliths was 20.4 ± 10.1%. Correction factors calculated from average length reduction improved length estimates for all fish species. Grade-specific length correction factors (gLCFs) reduced variability in all of the estimates and significantly improved estimates of prey with highly eroded otoliths including Pacific hake and shortbelly rockfish. The Biomass Reconstruction (BR) model accurately predicted biomass consumption within 4% of known consumption, whereas estimates based on frequency of occurrence were inaccurate.