This study was designed to evaluate biases associated with four methods of estimating stage-specific parasitism and one method of estimating time-specific parasitism for two parasitoids attacking gypsy moth,Lymantria dispar(L.), larvae. Estimates for all five methods were calculated using both observed death rates and marginal attack rates. Calculating percentage parasitism by pooling all hosts collected during the season resulted in an underestimation of stage-specific parasitism by bothCotesia melanoscela(Ratzeburg) andParasetigena silvestris(Robineau-Desvoidy). Percentage parasitism byC. melanoscelapeaked after completion of recruitment of hosts and parasitoids and before substantial losses (death or emergence) of either hosts or parasitoids. Consequently, peak sample percentage parasitism was an unbiased estimate of stage-specific parasitism byC. melanoscela. Oviposition and emergence ofP. silvestrisoverlapped considerably with each other and also with losses of hosts not parasitized byP. silvestris. As a result, the peak value of percentage parasitism was not appropriate for use with this species. The release of laboratory-reared larvae (trap-hosts) to directly measure the number of parasitoids recruited into the susceptible stage of the host did not provide accurate estimates of percentage parasitism by either species, possibly because of different sizes and behaviors of trap-host larvae. The graphical method of estimating numbers entering a stage used by Southwood&Jepson in 1962 is unbiased only if minimal mortality occurs during sampling. This was true forC. melanoscela, but not forP. silvestris. Mortality of parasitized hosts can also affect estimates of time-specific parasitism of hosts collected in the field. Because mortality of gypsy moths parasitized byP. silvestriswas high, this method is not recommended forP. silvestris, and an alternative technique is suggested.