This paper reports the problems involved in measuring the extent to which the practice in program sites matches model theory formulated in Head Start Planned Variation. The main focus in studying program implementation is the question of internal validity, whether the children in the HSPV programs actually receive the treatment being evaluated. Several attempts were made to develop an appropriate measure, though none of the measures was found to be completely adequate. Three types of information needed to develop a good measure of implementation were determined: a definitive model description, clear-cut standards of implementation, and judgments concerning comparability within and between models. The problems connected with obtaining this information were described. It is concluded that precise and accurate evaluation of program implementation is difficult and some suggestions are made to alleviate evaluation problems. (SDH)