Assessing students' knowledge of statistics with "objective" test questions

In this talk I will give some principles and many examples of "objective" questions that test statistical concepts. In some of our large classes we must of necessity use machine-gradable tests. Is it still possible to test concepts (and not just computations) in such a setting? I will try to argue (mostly by example) that we can test concepts this way. The most important principle for any testing is: Test what you believe is important!