Define: Grading On The Curve?

Two-part question:

A) Are there two meanings to “grading on the curve”? Using TomCat’s thread as one example at this link: http://boards.straightdope.com/sdmb/showthread.php?t=377241, I have heard this simplified explanation.

But, technically, grading on the curve refers to the Bell Curve and requires a statistical analysis to assure an even distribution by a statistical definition. You know, std deviation and such.

B) Can the latter method really hurt as many as it helps? One Senior High math teacher claimed he wouldn’t curve for this reason. Is that really true? Has anyone personally experienced this? Or, is it a theorhetical result, but unlikely if the class needed the help of curve in the first place, I WAG?

What’s the SD? (Please don’t get bogged down in the statistical nitty-gritty).

  • Jinx

I am not sure I am following the question. However, I have seen teachers use the dirty method and make say, an 80, become the new 100 if that is the highest score. A cheater could obviously screw stuff up for everyone by scoring way above the legitimate scores. They call that grading on the curve but it obviously isn’t really a curve,

My psychology professors did grade freshman level classes on a normal curve sometimes. The classes were 150 people or more and a large sample is needed to form a good curve. Grading on a normal curve wouldn’t be very good in a class of 5 for example. True normal curve grading can be brutal however. Following it completely means that half the class is going to be in the C range or lower. You could skew the curve so that it is piled more at the high end but you still need to find a way to generate failures. That can be more complicated.

Grading software can handle all of this stuff easily today however.

Grading on a curve is designed to correct for poorly designed examinations. If the test were done properly, you could say that 100% - 90% was an A, 90% - 80% was a B, and so on. But what if the highest grade was a 47%? Or the lowest grade an 88%?

You can’t say that the 47% on the first exam wasn’t exceptional, it’s the highest mark in the class. You can’t say that the 88% on the second is better than average, it’s manifestly far worse than average. Most tests I have taken have been graded on some sort of a curve.

With a large sample there are usually no real problems, but a small sample can have curve breakers who can throw things off. Most of the small classes I’ve been in (organic chemistry, physical chemistry, and the like) have had a bi-modal distribution of grades, with a normal distribution of grades in a small high scoring group and another normal distribution in a larger lower scoring group.

Usually the entire high scoring group ends up with an A, with the rest of the grades distributed normally among the other group.

I took a physiological psychology course once where anyone who didn’t score at least 65% of the highest score got an F. This caused no end of agony, especially when the highest score was a 93%, and the rest averaged around 60% or so. Having 1 “A”, 2 “C’s” and 8 “D’s” and 5 “F’s” makes for a funny looking curve.

In most of the classes I’ve graded, large (150 or so) and small alike, and at levels ranging from freshman to grad school, this is typical. The high bell represents those students who made an effort to learn, and the low one (which is also typically much wider) represents those who didn’t. Some of the profs here don’t even have a rigid grading scale at all; they look for the natural cutoffs in the distribution and assign the grade ranges accordingly.

This is simply wrong. The “curve” referred to is not the bell curve. In the first place, there are numerous other ways to represent a population statistically. Bell curves are applicable only to certain, well-specified, populations, few of which can be found in a given classroom.

Second, “curve” is an idiomatic term in this usage, not a mathematical one.

This entire part of the argument is irrelevant.

Whether or not grading on a curve is justifiable depends entirely on the situation. Is the teacher creating the tests or are standardized tests being given? Is the class a mixed group of students of all levels or a group of honor students or a collection of ESL students? One size fits all rules could easily harm specialized populations. But you can’t determine this without knowing the actual situation.

That you’ve completely misunderstood the question, the process, and the terminology. Again.

Exapno, you missed one part of Jinx’s OP - “Please don’t get bogged down in the statistical nitty-gritty.”

I think it’s clear Jinx was looking for information about the general principles of grading on a curve rather than specific information about whether or not a bell curve was the most applicable model.

Previous thread: ‘Curved’ grades

Glad you know all. Have you taken any math courses? Grading on the curve is most certainly not a figure of speech. You’re confused with a teacher throwing you a curve. If I must, I will find a cite…

Please do check for a cite. Although I would like to see your casting about off the Board for information, you might want to check the thread that Thudlow Boink linked to. It talks a great deal about the variation between the ideal bell curve that a regrading might be intended to achieve and the real world grading that is the true practice.

But you may not want to use your particular contribution to that thread:

Sweetheart, Exapno’s right. I’ve taken a million tests that were “graded on a curve” and not one was actually transformed into a normal distribution. “Grading on a curve” refers to any method whereby a student’s grades on a test depend on how their classmates did; while I don’t doubt that sometimes instructors actually try to massage the scores onto a normal curve, it’s already been explained why that’s not really an ideal methodology in a classroom and it’s certainly not the only - or even a common - way to grade on a curve.

That’s typically the way it’s done in large enough classes in most math departments (and we’re the ones who know how to curve right). You get down to a 20-30 person class and it’s just easier to recognize A work from C work.

Is this method of marking very common in America? I’ve never come across it in the UK, especially not at degree level - if you get 70% + you’re awarded a first, no matter how many people got that mark. More to the point, how fair is it compared to the non-curved marking scheme?

I didn’t start seeing it until my upper-level high school classes, and it was fairly common in college. As far as the fairness goes, it really depends on how well the test was designed, how well the instructor taught, and the population of the class.

In one of my freshman calculus courses (college), I scored 40 out of 80 on the first midterm and 40 out of 100 on the second. According to the 90-80-70-60 scale (90 and above is an A, 80-89 a B, 70-79 a C, 60-69 a D, below 60 an F), I would’ve failed. But because pretty much everyone scored poorly, I got a BC (my college gave grades for those on the border between A and B and between B and C) on the first exam and a B on the second. And boy, were most of the class shooting daggers at the girl who was questioning the professor about why she only scored a 93 on the second exam!

Conversely, my AP biology class in high school was packed with overachievers. After the first class, our teacher said he had decided to grade on the curve and started writing what the scores would be for each letter grade. As the class had done well, you had to score higher than 90 to get an A and higher than 80 to get a B. People started to complain that the curve hurt their score. To which he replied, “That’s exactly what grading on a curve will do in this class. Never ask me to curve test in the future.”

I really have to object to the insinuation here. While not a strict curve (see my comment below), in our calculus classes here we have moving grade cutoffs, and the tests are far from poorly designed. Let’s see you write a single exam for 300 students in a dozen sections, each taught by a different instructor, and see how fair you can make it for all the sections at once without a little massaging.

Besides which, unless you use the same exam over and over again, every test is a little different, and to hit the sweet spots you point to every single time is all but impossible even in a small class.

We manage to write exams that average around 70% with a 10%-20%[1] standard deviation three times a semester for each of three classes. At the end of the course, we come up with each student’s weighted course score and find good cutoffs with relative ease. If you think this is a poor design, you come here and fix it.

[1] 20% goes for calculus 1, which has such a wide variety of students coming in with so many different backgrounds it’s impossible not to have a large spread.

So can an outlier who legitimately got that score. I’ve been on both sides in classes before, and it’s not fun in either case. On the one side you make no friends being the guy who lowered every other person’s score by a letter grade or more. On the other, you feel cheated of hard work studying and testing because there’s someone in the class who’s truly gifted and is always going to outscore the class on a sufficiently hard test.

In one of my senior courses as an undergrad, I had one particular professor who graded on a curve that I felt hurt my grade unfairly. He graded such that a C was -1/2 – +1/2 SD from the mean, a B was 1/2 - 1 SD, and A was greater. My final grade was about a 92-93 (which is an A by the old fashioned method, and I think I was about 5th or so in the class), but because I was just less than 1 SD above the mean, I got a B. Nevermind that the mean was a high 60-something (yes, there were plenty of grades in the single digits and teens, that’s why the SD was so high). He should have seen he was only giving out about 2 or 3 As (out of quite a large class), and WAY too many Bs and Cs and adjusted. That, and he was a horrible teach

The lesson is, curving is good, as it obviously helped most of the class, but its STILL only a guideline. As others have already mentioned, a bell curve (especially one so strictly and arbitrarily defining the boundaries) is generally a poor feel for how hard the class worked and how much they learned. I’ve noticed as I’ve gotten farther in grad school that most professors tend to do what was also mentioned, they still get the scores the same way, but they cluster them naturally. In graduate and especially PhD courses, its unlikely that you’ll have a lot of C or lower quality work, so its entirely possible a professor may give out As and Bs and nothing else, even the number grades alone aren’t as impressive.

When I taught physics at university, at the end of the quarter the professor and all the teaching assistants (of which I was one) would have a grade meeting. The prof had a list of all the final scores and would set the point levels for A, A-, B+, etc. The TAs could pipe in and suggest small changes up or down. Borderline cases were often decided by the student’s participation level and other non-quantifiables.

Overall, this system worked quite well. The difficulty of exams didn’t need to be precisely set. The professor could choose approximately how students would get each grade (the department would review grade distributions and scold profs who gave too many high or low grades). The TAs contributed a human element–students who had steadily improved, but ended a few points shy would get bumped up. Slackers who aced a few tests early on but then bombed later tests and never participated in class would get bumped down. It seemed more fair than using only the raw numbers.

I think the onus is on you to find a cite supporting your position (namely, that the ORIGIN of “grading on a curve” did not refer to the bell curve. It’s true that the term is often used today to describe any type of adjustment, but that does not mean that your point is correct.

Does that mean you have a cite that shows that the origin of grading on a curve did in fact refer to the bell curve?

Or that the origin rather than the current meaning of the term was asked for anywhere in the OP?