Curve

[woodwardiocom]

Prompted by this article on scholarships (courtesy mzrowan), I ask the teachers, professors, and other academic professionals on my list: When is it actually appropriate to grade on a curve? I've always seen it as saying, "I don't care how good you are, somebody has to get an F."

(For those not familiar with the term, it means that, for example, in a class of 10, for each assignment, the teacher gives one A, two Bs, four Cs, two Ds, and one F, regardless of absolute quality of the work. (Well, in reality, Fs are usually not on the curve. but still.))

Comments

[tamidon.livejournal.com]

I always saw grading on the curve as something that happens when the teacher misjudges his test. Say he thought it was fine, but everyone,including his best students,got below 70%. Clearly if it's all the students then something was wrong with the test and needed to be adjusted. Redoing the test could screw up the class schedule too much.

At CIA we had teachers who had the set grade scale, hated that lmost as much as the ones who said the whole class should work as a team and would get the lowest students grade. These sometimes alternated

[moonshadow.livejournal.com]

I agree with tamidon. If everyone in the class gets an F, then the problem is likely with the teacher, not the students, and grading on a curve might be one of the appropriate ways to deal with that situation.

[spoothbrush.livejournal.com]

If everyone in class gets an F, there are ways of adjusting means that don't involve fitting the results to a normal curve -- which are probably better and more fair.

[moonshadow.livejournal.com]

I'm not an education professional - could you give me a fr'instance?

[spoothbrush.livejournal.com]

There are a few ways of scaling grades that don't require making the grades fit a normal curve.

One of them is the classic "add x to every score". I don't like that one much.

One of them is to increase every score by the same percentage (like, multiply everything by 1.1).

One of them is to use a scaling factor and do something like 100 - .3(100 - Original Score).

[archangelbeth]

I think this happened at UT once, in the Biology lab segment. The grade distribution was so weighted towards Ds that they wound up going on a curve instead. (Which, as a usually-decent student, pleased me; it's nice when you don't have to outrun the bear, but only have to outrun the other people sufficiently... >_> )

[feste-sylvain.livejournal.com]

Speaking as a former student (not as, per your request, a teacher, professor, or other academic professional), we had one professor in a course required of all engineers ("Thermodynamics") aim for the average grade on any given test to be 50%. His thinking was that the sample size for this course was sufficiently large that he needed the bell curve not to bunch up on either end.

He made a grave mistake in creating one test, and the average come out around 30%. That was easily the most grueling test I had at college. I got a 45 on it, which worked out to a high 'B'. The professor apologized to the entire class.

However, the only grade that truly mattered in that course was the final one; the final exam included several questions which were scattered among the others, and anyone who got all of those questions correct could not get an 'F', even if they punted the rest of the exam. By creating a baseline, he kept the tail of the curve from forcing artificial failures.

[woodwardiocom.livejournal.com]

His thinking was that the sample size for this course was sufficiently large that he needed the bell curve not to bunch up on either end.

Why did he think that? And, was he actually grading on a curve, or just making the tests hard enough that the average percentage correct was 50%?

[feste-sylvain.livejournal.com]

He thought that if the results were bunched up at either end, there would be no meaningful distinction between grades (the difference between 'C' and 'A' could be a momentary brain-fart).

Yes, he was actually grading on a curve.

And yes, he was making the tests hard enough that, over the course of the course, the average percentage correct was very close to 50%.

Do remember, this was an engineering course; there was partial credit given for demonstration of understanding the underlying concepts, but most test questions were trying to determine whether we students could apply the knowledge.

Oh, right: the other important thing about that was that he could tell whether the student body was missing out on any basic concepts; sometimes, he'd refresh us on any such concepts that we hadn't managed to absorb yet, and include some refresher work on the homework to make sure we'd absorb it before the final exam.

[ectropy.livejournal.com]

Hmm. I took an Thermodynamics course which was also graded on a modified curve - general knowledge was that the unadjusted binomial distribution had 20% getting A, 30% getting C, 40% getting D, and 10% getting F, so the teacher promised to adjust the lower group to spread into B. I thought it was weird but it didn't bother me one whit, since I consistently got the highest grade in the class.

[spoothbrush.livejournal.com]

There are times when I would love to grade on a curve. I wouldn't have to evaluate how much every individual response is worth, or how well it answers the question. I could just make 5 piles. In reality, my students are not normally distributed. If anything, the distribution is typically very bimodal, and I can't think of a classroom situation when it would be appropriate to transform a bimodal distribution to a normal one.

Amen to that

[drwex]

While I agree with the earlier commenters who said that if all the students did poorly on the exam then the fault likely lies with me (the teacher) that doesn't change the fact that a level of work is a level of work and if everyone happens to make the level I've set then I give them all As. (And then I revise the level for the next year.)

I think that any system that forces students' grades downward such that some are guaranteed to fail no matter how proficient they are is an abomination.

[gwendolynclare.livejournal.com]

I've never heard of a professor grading on a curve in such a way that the people at the low end of the curve get a worse grade than their numeric score would indicate.

I think curving is appropriate when 1) the class size is large, 2) each exam has something close to a normal distribution to begin with, and 3) the means and/or variances of the exams are substantially different. If all these conditions are true, it is a relatively fair way to correct for the unintentional variability in how hard the exams are (ie the professor's inability to write consistent exams). If any of these conditions are violated, it doesn't make statistical sense to grade on a curve.