Educators' current enthusiasm for alternative assessment has roots in political and intellectual concerns. American science students do not do well on international comparisons or on tests of science literacy. Blame for these shortcomings is laid on multiple-choice tests and their focus on isolated facts. But if alternative assessment is to fulfill its promise, research and development need to be done in a coordinated way. This article lays out an agenda for developing assessment tools and researching their effectiveness. A series of design tasks is proposed, followed by related research questions for each task. Needed theoretical research includes studies of the effects of alternative assessment on policy and practice as well as the development of new psychometric techniques. Influences of new assessment forms on learning and motivation theory are also discussed.