I had a colleague opine yesterday that if you have to curve a test, it’s not a good test. I would say that’s only true if your goal is to write a test that doesn’t need curving. If you have experience with a course and a very homogeneous class, then I agree that you might not need to curve tests. Absent either of those conditions, it’s very difficult to make a really good test (ie, one that doesn’t aim for the lowest common denominator) and not scale the scores.
I think of college tests: the reason they’re scaled is so that the best students can be challenged without overwhelming the weakest. Such is always my goal as well.
I think the practice of aiming for tests that “don’t need curving” has several bad effects in practice (though I don’t attribute these bad effects as motives to the people writing them, of course!):
- it implies that we’re aiming for a collection of a standard body of skills (which by definition have to be algorithmic or not everyone will have mastered them equally at the same time, even in principle) or body of knowledge rather than aiming at developing people’s ability to think creatively
- it suggests that students should be fungible units, where each one is capable of doing exactly the same thing in response to the same prompt as every other one
- it further implies that if a student *can’t* perform just like every other student on a given prompt the student failing to do so is inferior in some measurable, and presumably valuable, way
I acknowledge this post may be controversial, and I apologize if you are offended by it, but that’s how I see it.