Should people give final exams?

The Boston Globe has a piece up about the declining use of final exams, and In Socrates' Wake responds with more general thoughts about frequent, low-stakes testing vs. infrequent high-stakes testing, as a part of their ongoing discussion of the issue. I have long been on the frequent, low-stakes side, which mostly means I spend a lot of time shuffling paper. ISW comes down in favor of retaining some role for the final exam, but everything I'm seeing here reinforces the idea that frequent low-stakes testing is the way to go.

The Globe article was prompted by a change in Harvard's policy about scheduling finals. Rather than assuming that every class will have a final, and them booking a room for it which may go unused, Harvard now requires instructors to request a room for a final. This decision in tern was prompted by the discovery that only 23% of the classes at Harvard actually have finals. (This, by the way, is a change that LCCC might want to consider as well. Our inefficient room scheduling has been cited by outside consultants as a place where we might save money, and our finals scheduling has often been a mess, with rooms double booked.)

I know there is research out there backing the frequent, low-stakes side, but I've never really delved into it. The Globe article one cites on empirical study, by M. Vali Siadat, which showed that algebra students who were given small weekly quizzes did better overall, and on the final exams, than those who were given less rigorous weekly assignments. (This makes me feel good about the structure of my logic class.) This is the original study. A commenter at ISW also mentions the spacing effect, which shows that periodic reinforcement over a long period of time leads to better memory than short cramming sessions. This, by the way, has been known since 1885.

Michael Cholbi at ISW suggests that the frequent low-stakes vs. infrequent high-stakes may a be a false dichotomy. Indeed, the Siadat study seems to have really been about two different versions of the mixed approach, with the winning strategy leaning more to the frequent low-stakes side. My logic class follows roughly that structure, with 10 short quizzes (down from 15 in a 15 week course.) In place of a final, you are allowed to re-take 3 of your quizzes. I think this definitely works for math-like subjects. The issue is more complicated for the humanities.