I view examinations as a necessary evil. We need them to assess students and the primary goal is to come up with numbers that are strongly correlated with their understanding of the course material. They are evil because they put students unnecessarily under stressful conditions and hence lead to situations where the primary goal is missed. I keep trying out various methods of evaluation in an attempt to come up numbers that do measure the student's understanding. Here is a sample of things I have tried over the years.

**Assignments**that are graded do help in this regard. However, given that I have to grade them on my own (IITM doesn't give me TA's), I can do this when the class size is small. The second requirement is that the students are mature enough to understand that they cannot hand in assignments that are carbon copies of someone else's assignment. I have done this successfully in my Classical Field Theory course where the student gets to do around 40 non-trivial problems that complement/supplement my lectures.**Open notes examinations**do reduce the stress of memorizing formulae which I consider a waste of time.- In lieu of open notes examinations, I have asked students to bring one A4 sheet of handwritten material containing anything they consider pertinent to the examination. This is wonderful as it forces the student to summarize the whole course material and in the process create some order in the chaos that is the human brain. This works for big classes as well as small ones.
**Open book examinations**are similar to open notes but I find that students waste a lot of time searching through books looking for the answer.

For the GR and Cosmology course during Spring 2009, for Quiz I students brought in one handwritten A4 sheet, Quiz II was open notes. Inspired by what Mehran Kardar did at MIT, I tried out the following for the final examination. I split the examination into two parts. Part A (about 40%) consisted of short questions that test concepts and simple application of formulae. Part B (about 60%) consisted of longer questions that required good understanding as well as a reasonable amount of algebra. Questions in Part B was made available to the student 24 hours in advance. They were expected to solve it during this period. However, unlike in a take home examination, there were *not* expected to provide written solutions. The actual examination was a conventional closed book examination where they were expected to reproduce their solutions to Part B in addition to solving Part A. I am yet to grade the examination.

**Update:** I finished grading the examination. I would rate the experiment a partial success. In an attempt to avoid doing messy calculations, some students memorized the Christoffel connection and Riemann curvature that they computed using symbolic manipulation programs in their room. Some others forgot to compute certain terms which (eventually) ended up not contributing to the quantity of interest but should have been computed if you didn't know the final answer. There was a ‘phase separation’ among the students splitting them into the *have learnts* and *have not learnts* making the experiment a success. Students also indicated that the pressure of writing an examination was immensely reduced. I also think I need to learn a lot more about how to set questions suitable for this format.

**Update (Nov 2009):** I introduced surprise quizzes in the course, *Classical Physics*, during Fall 2009. There were three such quizzes and the best two of three was chosen. The attempt here was to make the student brush up course material on a regular basis instead of a late-night cram session.