Well, I should be studying for my last final, so of course, I’ve been reading blogs and watching the snow fall while searching for the right Christmas music station on Pandora. (The Vince Guarladi Trio, of Charlie Brown fame, should work nicely for studying, while not bringing up any Bing Crosby or Perry Como to drive me nuts.)
So, since I’m being so productive anyway, I thought I’d share a few more thoughts on grading and curving that have come to mind since reading the comments on the last post…
Most of the grading schemes I addressed work best when applied to large classes, where you might reasonably expect something approximating a normal distribution. And of course, it assumes that at least some of the students have demonstrated knowledge at a satisfactory level, worthy of an A. I would argue (and I think most teachers would agree) that a student at the top of a very weak class does not deserve an A just because he or she outperforms everyone else, just as the student at the bottom of a strong class does not necessarily deserve an F just because other students were more successful. However, the sheer number of students in classes like introductory physics at a huge university will almost guarantee that there are students at both ends of the spectrum.
Things get trickier, however, in smaller classes, or upper level classes, where most of the students are expected to do a good job of learning the material. For example, what sort of grade distribution should there in a senior-level course with seven students, like the quantum mechanics class I took in undergrad? There is no a priori answer to this question, which I’m sure makes it more difficult to grade. In addition, the ability level of classes can fluctuate wildly from year to year with such small sample sizes. Maybe one year you’ll have a class where everyone deserves an A, and the next year, no one does.
It is nearly impossible to make two exams of the same difficulty level, and accounting for those differences in difficulty is feasible only for tests with huge numbers of test-takers, such as national standardized tests. So if a professor notices that the average test scores are much lower than last year’s, he must decide whether he made a more difficult test this time around, or if his students this year are just not quite as good.
Recycling questions from old exams can help in this regard if you’ve been teaching the class for awhile, but the sample sizes are still not large enough to be truly useful. This method also runs the risk of having students finding copies of old exams floating around, and having an unfair advantage because of it.
This is of particular importance in classes for upper level majors, as grad schools often look at grades in particular classes to see how well a student learned that particular subject. An A should mean that the student understands quantum mechanics, not “well, he’s better than the rest of this group of screw-ups.”
Of course, I have no experience doing this myself, so I would welcome comments from those who have dealt with this issue. How did you handle it?
Well, I guess I’d better brave the subzero temperatures and head down to the coffee shop down the block in hopes of making myself study for my exam by removing some distractions.
Oh, and of course, if you’re a professor reading this who happens to be grading one of my exams, the best curving system is the one where I get an A. Just a suggestion.