A team of university and public school mathematics educators designed performance-based mathematics assessment tasks designed to align with the Texas Assessment of Academic Skills for 93 students who had been identified as at-risk in mathematics. Scenarios were developed based on four contexts: (1) familiar activity; (2) social issue; (3) hands-on; and (4) technology. Each context was administered in three settings: individually, aided by a proctor, and small group. The data analysis consisted of two repeated measures analyses of variance with context and setting, and content and setting as the main factors. The repeated measures were operations, concepts, or problem-solving scores. The results indicated that context was not significant, but content was significant. Setting was significant in both analyses. Generalizability studies (G-studies) were conducted to measure dependability of raters and students. The G-studies indicated that the six raters were dependable when assigning scores. The problem-solving domain was the most dependable knowledge domain rated and the concept domain was least dependable. An appendix provides scoring rubrics and sample questions. Thirteen tables and four figures. (Contains 48 references.) (SLD)