Beach seines are widely used to estimate the density and species richness of fishes in estuaries. We evaluated the causes and extent of bias in estimates from seines using a series of field experiments in small estuaries in southern California, USA. Seining in spatially paired areas that were either enclosed by block nets or not, revealed that seines used without block nets underestimated density by more than 4-fold and species richness by more than 2-fold relative to blocked areas. Seining in paired blocked areas with seines of two lengths revealed that net length affected estimates of density, but not species richness; a 7.6-m long seine produced 1.6-fold higher estimates of total density than did a 15.2-m long seine due to increased catches of demersal fishes, but not midwater species. Paired sampling in blocked areas also revealed that many fishes initially evaded capture by the seine. Estimates of density but not species richness were significantly higher in areas through which a seine was swept 5 times compared to once. This was due to higher catches of demersal fishes but not midwater fishes in areas seined 5 times. Repeated seining through blocked areas revealed that the vast majority (90% or more) of species and individuals of midwater fishes were captured within the first 5 sweeps, compared to only about 50% of the individuals of demersal species. A mark-recapture study in blocked areas revealed lower probabilities of capture for demersal species relative to midwater species.