This study presents psychometric characteristics of the mathematics problem solving performance assessment used in the Minneapolis Public Schools, focusing on the interrater reliability, scoring reliability, and validity of the assessment. The Minneapolis Math Problem Solving Assessment (MPSA) was established in 1991. Students are asked to solve two complex problems, and written communication is a strong component of the assessment. The problems are scored on strategy, quality of solution, math communication, mathematical reflection and connection, and sentence structure, spelling, and mechanics. This study analyzed the 2,600 papers done by about 1,300 students in grade 5 using the FACETS model. FACETS is an extension of the one-parameter item response theory, or Rasch model, which parameterizes each facet of performance assessment into a common scale and enables one to compare elements within one facet and to compare different facets. Using FACETS with classical test theory allowed evaluators to examine interrater reliability, interrater agreement, and validity. Results support the validity and reliability of the MPSA and provide information that can be used to improve the assessment. (Contains 4 figures, 8 tables, and 14 references.) (SLD)