This paper discusses the application of Hierarchical Linear Modeling (HLM) to the evaluation of the effectiveness of Basic Skills programs in the Newark (New Jersey) school district. Whether HLM is able to handle regression effects in Basic Skills examination data is studied. The analysis uses data from the Stanford Achievement Tests for mathematics for grade 8 for 3 years (7,738 students). A comparison between HLM and a conventional pre- versus posttest design indicates that the size of the correlation between parameter estimates is considerably smaller in HLM, and not consistently in the predicted direction. It is concluded that while HLM is better able to contain regression effects, those effects remain a source of concern in the interpretation of basic skills evaluation data. (Contains one table and five references.) (SLD)