For the first few years I graded, I took what I thought was a pretty standard approach. I had a stack of exams, and I went through them one by one, grading each exam fully based on the rubric I had developed.
Last year, however, a colleague persuaded me to try a different approach - grading by the question, rather than by the exam. Under this approach, if you have three questions on your exam, you go through your stack three full times -- the first time, grading all answers to question 1; then all answers to question 2; then finally all answers to question 3.
I was very skeptical when I tried this. I thought that grading one entire exam would give me the ability to better assess a student's output as a whole. I could take notes if a student was particularly good or, probably more importantly, needed help.
However, my experiment with by-the-question grading was a huge success, and now I wouldn't do it any other way. I found two big benefits to grading by the question rather than the exam. First, I could stay focused on the answers to that question rather than having my thoughts jump around from subject area to subject area. I think this sped up the grading for me. Second, I avoided any unintentional bias I might have developed while grading an exam. For instance, as much as I tried to avoid it, would I be biased toward a student while grading question 3 if she answered questions 1 and 2 very well? Or, the flipside, would I be harsher on a student while grading question 3 if I had grown frustrated with his answers for questions 1 and 2? Grading by the question avoids this for the most part.
I'm now a convert and don't think I'll go back to the other system. But, I welcome others' thoughts in the comments on this important question of how to grade exams.
Here's a post I did on the subject over at PrawfsBlawg:
http://prawfsblawg.blogs.com/prawfsblawg/2010/09/how-do-you-grade-exams-exam-by-exam-or-question-by-question-ie-exam-hopping.html
Posted by: Colin Miller | December 13, 2010 at 01:22 PM
Thanks Colin. I hadn't seen your post before writing this one, but it's remarkable how similar they are in content....
Posted by: David S. Cohen | December 13, 2010 at 01:34 PM
The answer to the specific question that is the title to this post is, of course, "by stair."
Posted by: Eric Muller | December 13, 2010 at 02:02 PM
There's a good empirical study for you. Do the stair method, record the results. Then grade by whichever is your preferred "serious" grading methodology, record the results. Then compare the results.
Posted by: David S. Cohen | December 13, 2010 at 02:04 PM
I have always graded my exams on a question-by-question basis. In addition to the two advantages you note (increased efficiency and avoidance of halo or reverse-halo effects), this approach is more equitable because it allows for more consistency in scoring across exams.
Posted by: Barry L. Johnson | December 13, 2010 at 02:31 PM
I would not have exam questions that overlap if using the question by question method, or at least warn students to repeat themselves in each question, if the questions do overlap.
For example, if question A and question B BOTH have a similar issue, a student may not think to explain the "rule" in-depth in question B, having just explained it in excruciating detail in A. If you grade question by question you may mistakenly assume that the student did not know the rule, or did not explain it in enough depth, when grading question B.
Posted by: Jimmy | December 13, 2010 at 05:04 PM
I also grade question by question. In addition to the increased consistency as Barry observes, it also allows me to reverse the order of the stack each time through, which means that if I have any bias toward the beginning or the end, it also tends to even out. Also more efficient if I discover there's something in the question I missed and the students are picking up (lets me amend the rubric appropriately).
I publish my exam instructions well before the exam itself, and I make it clear that I grade on a question by question basis, so that cross-referencing answers is not a good idea.
And, indeed, it does give each student, for better or worse, a fresh start with each question.
Posted by: Jeff Lipshaw | December 13, 2010 at 05:30 PM
I grade my exams "on the can." This is a cutting edge technique, but I don't think I will explain it at this time.
Posted by: anon | December 13, 2010 at 08:28 PM
I also grade question by question, mainly for the reason that Jeff mentioned. Not only is it faster, but I find I can be more consistent about partial credit. And, I always worry that the 15th student to make a mistake will get less (or more) generous treatment than the first. His/her place in the exam pile is not a student's fault, and I want to make sure that nobody gets penalized unfairly.
Posted by: rebecca bratspies | December 14, 2010 at 03:11 PM
I have always graded question by question, also to make sure that the answer I expected to get would surface in the answers. If it didn't, or did not surface enough, I'd reassess and revise my answer key.
Posted by: Darren Rosenblum | December 15, 2010 at 09:01 PM
I've always graded question by question. My exams tell the students that each question will be graded in isolation.
Posted by: Steve Ware | December 16, 2010 at 04:52 PM