While substantial energy has gone into understanding the mechanics of assessment in higher education, little attention has been paid to developing means by which university lecturers can monitor the quality of the assessment tasks they develop. This paper introduces a research project, funded by the Australian Learning and Teaching Council, which was designed to (a) enhance the quality of assessment in the social sciences in the tertiary sector and (b) refine and evaluate a model for analysing and improving the quality of assessment tasks in the social sciences, primarily in first year courses. The research has its foundation in the NSW Quality Teaching model, with the major focus on the link between task quality and student performance. Initial findings are reported in the paper. We discuss the validity of the Quality Teaching model for the tertiary setting. We report a strong positive correlation between task quality and student work as measured by our instruments. We share some of the tasks before and after refinement to illustrate the kinds of gains to be made when diagnosing and redesigning tasks with reference to the detailed specifications provided by the Quality Teaching model.