As constructed-response items become an integral part of educational assessments, setting student performance standards on constructed-response items has become an important issue. Two standard-setting methods, one used for setting standards on the National Assessment of Educational Progress (NAEP) in reading in grade 8 and the other used to set standards on the 1993 New Jersey Early Warning Test (EWT) in reading for grade 8, are compared. In the paper selection method, used for the NAEP, judges first conceptualize students who are just at the borderline between categories of performance and then select actual papers from a set of all levels of performance to represent what students at the borderline would have produced. For the EWT, the contrasting groups method was used. Judges internalize the concepts to be assessed and then select students above and below the criterion of success. Papers by these students are scored, and a point between the two score distributions is selected as the standard. Comparison suggests that the contrasting groups method may yield results that are more accurate than those of the paper selection method, although results are not definitive. Two figures and three tables illustrate the comparisons. (Contains 14 references.) (SLD)