An attempt was made to estimate possible bias in terms of Normal Curve Equivalent (NCE) gains of Elementary Secondary Education Act (ESEA) Title I projects. Twenty-two state Title I reports for 1978 were examined. The analyses presented were directed at two major sources of bias: bias due to errors in local implementation of models, and bias due to incomplete responses. Two subsets of five states each were identified where the State Education Agency (SEA) attempted to ameliorate implementation problems and where no such action was taken. Gains for the two subsets were compared. The numerical gains were large across the group of 22 states. The numerical gains were larger than the verbal gains, and lower grade gains were consistently greater than those in upper grades. Bias from errors in implementation was estimated at from four to ten NCEs. States without implementation controls showed consistently higher gains. Three additional sources of bias were non-response by SEAs, correctable errors, and further regression to the mean in properly implemented Model A evaluations. These sources of bias were identified, but not analyzed. (DWH)