The Maryland School Performance Assessment Program (MSPAP) is a relatively new, statewide performance assessment of students in grades 3, 5, and 8. When first administered in May of 1991, the MSPAP included a battery of performance assessment tasks designed to generate written or drawn responses to reading texts. This study evaluated selected reading activities, responses, and scoring apparatuses to improve performance task development and scoring strategies for future MSPAP editions and for other assessments using constructed responses. Focus was on identifying discrepancies between the range of responses anticipated when activities were developed and the actual response behaviors triggered by some activities, problems inherent in the activities themselves, and problems resulting from a mismatch between responses and the scoring apparatus that had been developed to characterize the range of performance on particular activities. Two major test forms in reading/writing/language usage administered in grades 3, 5, and 8 were studied. Each test form contained approximately 26 brief open-ended questions and 1 extended question that yielded a more developed response to be scored for reading, writing, and language usage. Questionnaire data on these 2 test forms were collected from all 12 teams of trained Maryland teachers serving as readers (N=123). Across the test forms examined, contextual errors (those triggered by the discrepancies between classroom and testing contexts) were rare. A seven-item list of references is included. The readers' questionnaire and scoring rubrics are appended. (RLC)