To Curve or Not To Curve

[LostAngeles]

That is the question. Whether tis nobler in the mind to give students the grade they got or by adjusting the scores on curve boost the entire classes grade? To pass, to fail...

Week 6 has passed and in its wake lay the students, broken and battered by an onslaught of exams. It is the Midterm season and for five weeks, exams will be hurtled at them from every corner of the campus. Case in point, I had several exams crammed into a three-day period, one of which was a Sunday. As a master of the procrastination arts, I had postponed studying until the week prior.

So it came to pass that in one class, on a scale of 150 points, the low was a 15 and the high was a 139, with a median around 73. In another, no one scored in the high range, but most scores were around a B.

And on the lone Saturday class with a teacher who had a reputation for being difficult and failing many students, the scores ranged from a 36/50 to a 6/50. There would be no scale, the teacher proclaimed. After all he was dropping one test score, was he not? And lo, many students protested, yet they fell on deaf ears.

On that information alone, should there be a grading curve?

So I'm not even in the 30's in that group (but no where near the lows. I was above the mean) and I do not believe he should curve at all for the follwing reasons:

His other class in the same subject had people getting As. Their exam consisted of seven questions. Ours consisted of six, two of which were omitted due to errors on his part.

As he went over the test, my friend, who's score was half of mine, looked at what he put up on the board and commented that it should have been easy. He's correct. We're given a list of problems that the test will be based on and ALL of them came directly from that list. And by "directly" I mean number for number. Three of which had answers in the back of the book.

This is not my first time with this professor. I know how he works, I know not to fear him. I also missed a class due to five hours in a doctor's office. And I didn't get around to studying the material I had missed. I knew better. I damn better have known better. Also, I lost points for stupid mistakes.

In fact several people in the class admitted that, "Gee. I should have studied better."

Frankly, I think we mostly earned our grades here.

That and we can always get one dropped.

[Zep]

Sorry, I thought you were talking about buying a padded brassiere...

[TragicMonkey]

I've always hated curves. Give me the grade I earned, not the grade based off some statistical trick.

I'm bitter because I got a B in a college course because the professor insisted on grading on a strict curve. I had a 98 percent average. One guy had a 99. Since there were only five students, I got the B. Pretty fair, huh? Because the point of the course wasn't to learn the material, but to compete with each other.

[Jeff Corey]

All grading is subjective.
But,I love it when I give a test that asks questions that I think get to the heart of my topic and over half the students get 85% or better. And I don't use multiple choice tests. Ever.
Grading on a curve means that you don't have a clear idea of what your students should know and that you let some formula take the place of your better judgement.

[Jorghnassen]

Ah, the intricacies of curving... I say the prof should use the order statistics and force them into a gaussian curve of his choice.

/but I'm teh evil

[LostAngeles]

Monkey, there are no words. Regardless of my expression of rule 8 violations in Paltalk.

My Econ class was curved. I set the curve. My grades were B and C. I got an A. I learned nothing.

I've got another teacher who's grading scale is five points lower than the usual (85-100% is an A). Not only is that up front and not curved, he likes to give a few tough questions for us to try.

[Zbu]

The curve is silly because it assumes that there will always be someone who fails, hence it insures that by making sure the ones who didn't understand it (or did poorly) get screwed. This is all nice if you're trying to make your class hard so students will be afraid of it, but you're not doing anybody any favors except the teacher and his/her ego.

[cbish]

I don't use a curve per se. I call it a percentage curve but it's really a scale. I take the top couple of scores (depending on the number of students) and I average those top scores. That becomes my 100% that I break down after that. It allows me to ask a few more difficult/challenging questions. I break my courses into 50% exam, 50% assignments. It's worked well for me. I like it.

[Arkan_Wolfshade]

I am against curves for the following reasons

1) If everyone does well, somebody is going to get screwed. You shouldn't be screwing someone who did well just because the prof made the test too easy.

2) If everyone does poorly, somebody will get a boost, but someone else will get screwed. You shouldn't be boosting or screwing someone because the prof made the test too hard.

In short, I see curves as a way for profs to cover up their teaching mistakes, not as a way to help students.

[roger]

I never quite understood curving - it implies that the professor doesn't know how to evaluate the student's skill set.

I went to a technical university, so basically all the first year students took largely the same courses - everyone was required to take physics, chemistry, calculus, etc., in freshman year. It was also standard for the Physics 101 exam to be the first one taken in all the classes. I think the average in my class for that exam, before curves, was 36. It seemed to be planned as some type of wake-up call. I remember everyone walking out of that test shell-shocked. You were abruptly shifted from being the brainiac of your high school to being one of the hopelessly unprepared, and your next weeks were spent desparately studying for your other tests during every waking moment.

OTOH, just about every freshman stumbled out of the classroom and headed downtown to get drunk after that test, so I wouldn't be surprised that it was all a payola scheme between the bar owners and professors.

[Soapy Sam]

I have no idea what is meant here. At least I hope I don't.
A result is a result.
A massage is a back rub.
A hybrid of the two is pointless.

[wunky]

With the push towards everything being standards based maybe that means that "curving" will go away.
In some of my grad school classes were K-6 educators who taught in DC (Washington, DC) public schools. They always were talking about how "curving" was such a good thing.

[Jimbo07]

A result is a result.


It almost never is.

I have seen 3 directions:

i) The school competitively ranks students (U of A engineering program) - this is an explicit curve.
ii) Poor results drives professor to say, "Too bad, so sad." (barely scraped that class, despite performing much better on final than midterm)
iii) Did poorly on a midterm. I was allowed to bet it all on the final. Crushed the final and aced the class.

Based on a professor's experience, the abilities of the class, etc. all grading is subjective. Even if straight percentages are awarded, the professor still chooses the questions, assigns the student marker and so on.

In fact, it bugs me when people say, "a result is a result." If the test is too easy, do I really deserve 100? :confused:

[pgwenthold]

It almost never is.

I have seen 3 directions:

i) The school competitively ranks students (U of A engineering program) - this is an explicit curve.
ii) Poor results drives professor to say, "Too bad, so sad." (barely scraped that class, despite performing much better on final than midterm)
iii) Did poorly on a midterm. I was allowed to bet it all on the final. Crushed the final and aced the class.

Based on a professor's experience, the abilities of the class, etc. all grading is subjective. Even if straight percentages are awarded, the professor still chooses the questions, assigns the student marker and so on.

In fact, it bugs me when people say, "a result is a result." If the test is too easy, do I really deserve 100? :confused:

I always tell people, I could write an exam where the average is 75/100, or where the average is 25/100. You tell me what you want the average to be, I can write the exam to accomodate it, for the most part. If I use the combination of writing the exam and grading the exam, I can come even closer.

It never made any sense for me to write an exam where the average is 35 and then just curve the grading scale. Why not just write an exam where the average is 75 and then use the old 90/80/70/60 or other such hard cutoffs?

[drkitten]

It never made any sense for me to write an exam where the average is 35 and then just curve the grading scale. Why not just write an exam where the average is 75 and then use the old 90/80/70/60 or other such hard cutoffs?

Because you don't get the same amount of diagnostic information. For example, if you write an exam where the average is high (and I consider a 75 to be a 'high' average), you can't tell the difference between the very good students and the excellent students, because of the ceiling effect (you can't score more than 100% of the points). Similarly, a test with a low average will lose information that lets you distinguish the truly wretched from the merely underperforming, because there aren't degrees of difference on a blank paper.

[Overman]

If he said he is alrealdy dropping the lowest test than thats fine. It just means you have to drink about 11 beers less the night before the next test.

Big Whoop.

:rolleyes:

[LostAngeles]

If he said he is alrealdy dropping the lowest test than thats fine. It just means you have to drink about 11 beers less the night before the next test.

Big Whoop.

:rolleyes:

Ha!

The class I had with him last semester met every day (meaning four days at my school because an old president was on crack). He tested every half-chapter which amounted to about 9, and dropped the lowest two.

My Physics professor during that semester seemed disappointed that he was behind this particular teacher in handing out tests.:p

[strathmeyer]

Well, obviously there's only one way to get an answer to this questions: randomly assign people into two groups: those who get curved, and those who don't, and then wait 20 years and see which group is the happiest.

[Jorghnassen]

Because you don't get the same amount of diagnostic information. For example, if you write an exam where the average is high (and I consider a 75 to be a 'high' average), you can't tell the difference between the very good students and the excellent students, because of the ceiling effect (you can't score more than 100% of the points). Similarly, a test with a low average will lose information that lets you distinguish the truly wretched from the merely underperforming, because there aren't degrees of difference on a blank paper.

You are infering about the scale parameter with only a location one. It's feasible to design a test with a 75% mean and be able to weight questions properly to distinguish between the very good, the excellent, the underperforming and the plain bad.

[Soapy Sam]

But what does "curved" mean?

It's obviously a well known term to the educationalist types here. I never heard of it.
It seems to me the point of an exam is to tell if the student has acquired an adequate understanding of the course work. If he has, he will achieve a certain score. If he fails to achieve the score, it is assumed he has not done the work or understood it.

What more do you need to know?

[LostAngeles]

But what does "curved" mean?

It's obviously a well known term to the educationalist types here. I never heard of it.
It seems to me the point of an exam is to tell if the student has acquired an adequate understanding of the course work. If he has, he will achieve a certain score. If he fails to achieve the score, it is assumed he has not done the work or understood it.

What more do you need to know?

Assume that a test is taken and the highest percentage is an 84%. This becomes the "A" and all other test scores are then redistributed thusly. During my high school and junior high years, I broke a few of them. I would have, say, a 94% and the next score would be an 82% or something. As I said before, they've been to my benefit at times, but I still don't like them.

So yes, so far as I stand, you are correct sir.

[Melendwyr]

In some of my grad school classes were K-6 educators who taught in DC (Washington, DC) public schools. They always were talking about how "curving" was such a good thing. We call those people "the idiots".

[Dogdoctor]

If the classes are static and the information doesn't change much over time such as mathematics then no curve should be necessary. Part of the reason for a curve however is that teachers are not all equal in writing tests. In the third case it sounds like a teacher who is either not good at teaching or not good at writing tests or most likely both of them. If you curve a class it should make up for problems in test design other than trick questions. If you look at test results and compare questions answered incorrectly by high test scorers versus low test scorers then you will find problem questions which can be thrown out. If you want however to test if students have learned certain subject material then a standardized test and no curve is the way to do it.

[athon]

Curving only benefits lazy assessing techniques. If you're confident in the demographic of your students and of their educational backgrounds, and in the integrity of the material you're teaching, and thirdly in the meaning and usefulness of the grade itself, then there is no need to base the results on a curve.

It's long been a gripe of mine. Luckily, as Wunky said, it will die out in most institutions and education systems with time.

Athon

[Hindmost]

I will scale or curve a test...however...only if the students actually do well on the test. With experience, I can determine if the students are putting forth an effort. I only curve the test if a fair number of students do reasonably well. If they all goof off, no curve. (I also typically toss the curve breaker grade)

glenn:boggled:

[bpesta22]

It's easy to write a difficult test. What's hard is writing a difficult but valid one. We have profs here who argue how high their standards are, but it makes me wonder how invalid their tests might be...I'd rather my students just learn the material; if they do that, they get the grade.

I focus solely on content validity-- does the student know the material as presented in lecture. This makes the exams highly objective, but somewhat mickey mouse-ish (in the sense that if you memorize the class notes you will do extremely well on my exams). But, I can tell students exactly why they got an 83 or a 92.

In fact, on a couse evaluation once, a student wrote as a "weakness" for me: If you just memorized the course material you got an A.

Duh.

Can college teachers here defend testing for anything else but content validity?

Since validity is an empirical question, ranging from 0 to 1.0 and teachers have a sense of how valid their average exams are???

[bpesta22]

Ah, I'd also argue that forced distributions are unfair (i.e., only 10% can get a's, etc). If you have that philosophy, why not just give an IQ test and administer grades appropriately....

[Dylab]

I focus solely on content validity-- does the student know the material as presented in lecture. This makes the exams highly objective, but somewhat mickey mouse-ish (in the sense that if you memorize the class notes you will do extremely well on my exams). But, I can tell students exactly why they got an 83 or a 92.

In fact, on a couse evaluation once, a student wrote as a "weakness" for me: If you just memorized the course material you got an A.


What about tests that involve more than knowing content. In many of my physics courses the tests are open book. How do you feel about curves then?

[bpesta22]

Dy-- good point.

I guess it's knowledge of a process (how to apply a formula to a physics problem) versus just knowledge of facts and concepts (as us softies in management teach).

Either way, I think you could structure a content valid test for physics.

I'm not against curves; I just dislike bad / invalid tests.

I think all profs have the right to be very strict graders, if that's what they want to do. I just hope the class gpa's aren't low because the teacher sucked or his exams were invalid. In my experience, "tough" teachers are tough because they're not the best lecturers and their exams suck. There are of course exceptions...

I "curve" by awarding points equally to all students til the class mean is at a level I'm comfortable with. I'd never use forced distributions though.

[Walter Wayne]

The major problem with scaling is it assumes a certain distribution of the population ... in a class of a two hundred, fifty or twenty? Then comes the numerous ways profs set the distribution, based on a median, a max or even forcing it to the bell curve.

TM getting curved in a class of 5 is ridiculous. What are the odds of two students in a class of 5 being in the top twenty percent of a general population. Pretty high.

I was in a class of about twelve, when an exam got weighted for less when people did badly. With the easy assignments almost eveyone got an A+. I had top mark on the exam, and I work with the person who was second. He ain't bright.

Curving did a service to nobody.

Walt

[Jimbo07]

With the easy assignments almost eveyone got an A+.

Don't most peoples' assignments go better than their exams?

[Jorghnassen]

Don't most peoples' assignments go better than their exams?

Except in heavily programming oriented computer science classes. Stupid extra long programs that won't compile... Exams with pseudocodes are much easier.

[pgwenthold]

Because you don't get the same amount of diagnostic information. For example, if you write an exam where the average is high (and I consider a 75 to be a 'high' average), you can't tell the difference between the very good students and the excellent students, because of the ceiling effect (you can't score more than 100% of the points).

Do you need to be able to distinguish between very good and excellent?

Very few places give A+, so why is it important to distinguish among the high As?

If you don't like a 75, you can always lower it if you want. In my current course, I have a little lower average (currently 68 or so) with a wider range of grade cutoffs, accordingly. In this course, it's bloody trivial to distinguish the outstanding and excellent students from the very goods.

[drkitten]

Do you need to be able to distinguish between very good and excellent?

Yes, because those are the students for whom I will be likely to need to write intelligible (and compelling) letters of recommendation.



Very few places give A+, so why is it important to distinguish among the high As?


I don't think you understand. My "excellent" are the A's. My merely "very good" are the B's. (Down with grade inflation!)

[drkitten]

You are infering about the scale parameter with only a location one.

I am indeed. Because the scale "parameter" isn't.

It's feasible to design a test with a 75% mean and be able to weight questions properly to distinguish between the very good, the excellent, the underperforming and the plain bad.

Experientially, no, it isn't. In practical terms, you want the greatest possible scale because students will have a greater understanding of their relative positions the wider you make the scale -- every student will intuitively recognize the idea that a 95 is a high mark and that a 5 is a low mark, irrespective of where the mean lies.

Trying to explain to students that a 70 is a C, a 75 a B, and an 80 an A merely encourages them to quibble over individual points in grading because they think (correctly or not) that they can argue about misgrading, partial credit, and badly phrased test questions. and eke out the extra few points they need. If the cutoffs are 50, 70, and 90 respectively, students will recognize the diagnostic signifciance of the larger gap.

[pgwenthold]

I don't think you understand. My "excellent" are the A's. My merely "very good" are the B's. (Down with grade inflation!)

Then your objection that the grading scale fails because it doesn't discriminate enough at the high end fails.

A is 90 - 100
B is 80 - 90

How does the fact that they can't get higher than 100 affect the A/B distinction? Of course, there are border line cases, but that is always the case.

[drkitten]

A is 90 - 100
B is 80 - 90

How does the fact that they can't get higher than 100 affect the A/B distinction? Of course, there are border line cases, but that is always the case.

But it's precisely the borderline cases that are at issue here (see the issues of scaling, discussed earlier). What do I do about an 89, especially an 89 that 'obviously' understands the material, but who made a number of minor and careless errors that reduced his score from a 91? Alternatively, how do I make the distinction between 89 and 91 appear to the students to reflect a genuine difference, enough to make the letter grade legitimate, instead of simply a random variation within the margin of error?

Using a larger scale means that the relative effects of the 'margin of error' are lessened. If we assume, for example, that your actual grade represents your level of understanding plus or minus five points, then there's a legitimate difference between an 80 and a 95, but not between an 89 and a 91. A more typical curve for me might center on 50 points, and the A/B cutoff might be as low as a 70 or 75 -- but I've set the questions such that very few people will score in the 70-75 point range; if you understand the material well enough to get 70 points, you almost certainly understand it well enough to get 80.

[pgwenthold]

Kitten, something else is bothering me.

Do you really need to distinguish between an A and an A+ in order to write letters of recommendation?

I would never agree to write a letter for anyone for whom I didn't know a lot more about them than their grade, and if I know that much about them and they were in my class, I can provide a lot more insight than if they had a high A or a high B.

I read a lot of app files, with lots of recommendation letters, and I don't learn much from comments in letters that say "he took my class and got a high A" unless there is something more to it.

[drkitten]

Kitten, something else is bothering me.

Do you really need to distinguish between an A and an A+ in order to write letters of recommendation?

No. I need to know their strengths and weaknesses, which is a much harder task and requires much more diagnostic information than merely being able to distinguish between an A and an A+.

[Soapy Sam]

So a "Straight A Student" in a class of complete numpties might be averaging 20% in his tests?
How does this relate to the 1-4 grade point average scores I hear about?

I'm not a teacher and it's 30+ years since I was a student. We tended to deal with straight numeric scores in my day. The big debate was whether 50% constituted a "pass". Clearly it did to the pupil and it didn't to the teacher or potential employer. Relativity at work.

[drkitten]

So a "Straight A Student" in a class of complete numpties might be averaging 20% in his tests?
How does this relate to the 1-4 grade point average scores I hear about?

By tradition, in US-schools, grades are given on a letter basis (A-F, with A representing exceptionally good performance, C representing "average," and F representing failure or unacceptable performance).

For convenience in comparing students, one typically computes a numeric "grade point average" or GPA by giving 4.0 "grade points" per A per credit, 3.0 per B, 2.0 per C, and so on. So a student with a 4.00 GPA has gotten an A in every credit they have attempted -- a student with a 3.5 GPA has probably gotten A's in half of their credits and B's in the other half. A student with a 2.0 GPA has probably gotten either all C's or a mixture of B's, C's, and D's, centered around C's.

Et cetera.

In theory, the task of the instructor is to assess at what level the student is performing. There's an inherent tension here -- if C means "average," then average of what? The average student at a high-end school like Bronx Science or Harvard is probably substantially better than the top student at St. Agnes' School for the Disciplinarily Disadvantaged. Does an 'A' mean that one is performing substantially better than one's peers, or does it mean that one has an objective mastery of course content? (That's the curving question in a nutshell.)

The other question is how the instructor is to assess the students. In particular, I can pose a set of test questions in which the "average" student is not expected to be able to answer any at all. (The infamous Putnam examination in math is a good example; the median score among those who take it is typically a zero.) I can also pose a set of test questions in which the average student is expected to be able to score perfectly (or near-perfectly). I prefer (for reasons stated upthread) to pose a set of question such that the "average" question scores at 50% or a bit above, because that lets me make the greatest possible use of the standard 0-100 scale.

Since I'm aware of the difficult of the exam when I write it, it's no problem for me to mentally convert a raw score to an overall assessment. ("Gee, I expected a baseline "competent" student
to score a 65, and this student scored a 68, so they are pretty much performing at baseline." vs. "This students scored an 80, so they're doing substantially better than baseline.") On the other hand, a student who comes from a prior educational background where the pass mark was defined as a 75 might be appalled to get an 80.

The more fool they, put bluntly.

A straight-A student might have a test average of 20% for two reasons, depending upon the teaching philosophy of the instructor. First, the instructor might have written difficult tests for some reason (perhaps to scare the students into studying harder.... nothing says "you don't know as much as you think you do" like a test covered in red ink), but is "objectively" demonstrating a very good understanding of the material. Alternatively, he might be doing rather poorly -- but substantially better than the rest of his colleagues. Again, the question of which is the whole curving question in a nutshell. Do I deserve high marks simply for being the best student in the room?

[Complexity]

I taught computer science at the university level for several years.

The only way that I could decide where grade breaks should be was by scoring the exams, estimating where I thought grade breaks should be, and then reviewing several exams to see if the letter grade they would receive under that scheme reflected what I thought of each exam. It usually took two or three iterations of this before I was comfortable with the grade breaks.

Another practice that I found helpful: Rather than assigning points to problems based on their difficulty, I assigned minutes to each problem - the number of minutes that I expected students to spend on the problem. Since most exams were for a 50-minute class, each problem was worth twice the number of minutes in points. I found that few students left early and few left much unanswered. They were happy with this scheme, and it helped them use their exam time wisely.

[bpesta22]

There is a definate "right answer" to the questions raised here, but it depends on what the goal of the test is.

Testing is science-- are you guys familiar with item response theory?

http://ourworld.cs.com/Bpesta22/image332.gif

On the x axis is ability, or how much of the latent trait the test taker has (i.e., how much of the course content he knows). It's measured in z scores, so 0 is an average level of ability; +1.0 would be one standard deviation above average, etc.

On the y axis is the probability of getting the test question right.

This creates a set of curves-- one for each item on the test.

The item curves are informative. They tell you if the item's any good (i.e., does having more or less of the latent trait predict if you will get the item correct); how difficult the item is for people of different ability levels, and where in the range of ability the item best discriminates (e.g., real easy items might still be good if they discriminate trully poor students from slightly less poor ones...).

The "a" parameter is shown in the graphs above. It tells how good the item is at discriminating people who have more ability from those who have less (steeper is better)

If this were an IQ test, for example, A1 would be a very good item for telling the difference between an IQ of 100 versus, say, 102, but it would be a worthless item for anyone else.

a5 however is not as good an item-- it takes a relatively big difference in ability level to get a big difference in the probability the test taker will get the item right (due to it's shallow slope).

Both a1 and a5, though, have difficulty levels of p = .50.

So, just making a test so that the "average question is 50%,"' doesn't tell you whether you have a test more like A5 or A1.

What's needed is steep slopes for items, but also curves that span the range of ability you want to test for.

http://ourworld.cs.com/Bpesta22/image338.gif

These curves all have the same discrimination ability, but they assess different levels of the ability.

The far left curve would represent a good but easy item. It discriminates dummies from slightly less dumb people. Notice that you can have an IQ of only 68 (2 sd's below average) but still have a 60% chance of getting this item right. If your iq is 116, you will basically never get this item wrong.

The far right curve would do just the opposite (nicely discriminate high IQs from higher IQs, but give no information on people below average in IQ).

If you're at an elite school, you probably don't need test items at the far left of the graph. They're a waste of time, as everyone would get them right.

The danger with having many or too many items at the right end of the curve is that although you will tell the difference between the 2 very best students in the class, you do so at the expense of not spreading the rest of the class out fairly.

So, depending on what you want the test to do, you should structure the items accordingly.

[athon]

I think what a lot of people are forgetting is that a grade is only relevant to those who utilise it as a system for discriminating out of a spectrum of students. A curve might well be useful in relation to this if it is necessary to have individuals from a select group, as many universities opt for (a secondary school system, for instance, where universities are out to fill seats). In other words, the knowledge a student brings is secondary to the numbers. No university is going to have reduced numbers if a spectrum of results is not fulfilled. Therefore a curve is preferred.

However, where an accurate description of how much a student has learned is needed (which is where most educators are concerned), a curved assessment is pointless.

Athon

[clarsct]

Hmmmmmmmmm.

I had a Chem Prof that subscribed to the following theory:

I do not give partial credit nor do I curve, which is merely a way to give partial credit without directly doing so, and thre reason why is that you either know the material, or you don't. What if your doctor got through Medical school on partial credit. What if the part that was wrong with you was the part he didn't know?


As a student, I found such a philosophy sensible, but a little scary, too. I knew I was going to have to study and do well, because the instructor had no pity and made no bones about his lack of pity. Of course, Analytical Chemistry is a hard science and the numbers rarely lie. Easy to do.

How do you grade a philosophy test using that method?

I also took Physical Chemistry with a different prof. The tests were designed to take as much as 8 hours. You had 24 hours to complete the test, using anything you wished aside from the Internet, each other, or a Professional Scientist(Profs were in this catagory). Ordering pizza was allowed, so long as you didn't leave the Chemistry floor.

He did curve, based on scores. I didn't study as muich for his exams. Then again, studying did you little good as the tests were based on applications of the material, not memorization.

So was he measuring IQ?


Overall, I would say that not curving is better and that more students ought to be failed if they deserve such.

[drkitten]

I also took Physical Chemistry with a different prof. The tests were designed to take as much as 8 hours. You had 24 hours to complete the test, using anything you wished aside from the Internet, each other, or a Professional Scientist(Profs were in this catagory). Ordering pizza was allowed, so long as you didn't leave the Chemistry floor.

He did curve, based on scores. I didn't study as muich for his exams. Then again, studying did you little good as the tests were based on applications of the material, not memorization.

So was he measuring IQ?

Rather obviously not. IQ is in theory knowledge-independent, and there's little or no way that one could pass a test such as you describe without subject knowledge in P Chem.

However, what he was measuring was your ability to apply the material (subject knowledge) to novel situations -- essentially a test of generalization performance of the subject knowledge.

[clarsct]

Ah.

A concise and informative post, thanks.

So, should he have curved the grade in that instance, do you think?

[exarch]

Assume that a test is taken and the highest percentage is an 84%. This becomes the "A" and all other test scores are then redistributed thusly. During my high school and junior high years, I broke a few of them. I would have, say, a 94% and the next score would be an 82% or something.
So THAT's how American students get to be straight A students. It's not about getting consistent 80 or 90% plus scores, just make sure you're close enough to the highest score to be awarded the highest mark. I think in elementary school even I may have qualified as an A student that way, and I was a real lazy-ass bum back then.

[drkitten]

Ah.

A concise and informative post, thanks.

So, should he have curved the grade in that instance, do you think?

It depends on the purpose to which the grade was being applied.

In some schools, P Chem is one of the classes used as a weed-out class, with the intention (implicit or explicit) of discouraging the poorly prepared or insufficiently motivated. For example, if we only have lab space for a hundred majors to do their senior projects, we need to reduce the cadre size down to 100 students by the beginning of the senior year. If there are three hundred sophomores, then we need to get rid of nearly 70% of them, either to other majors or other schools altogether.

Curving the grades would be a very easy way to do this. The top 30% of the class get A's or B's, and the rest are "encouraged" to change majors.

On the other hand, if this is supposed to prepare students to take some sort of professional certification exam, then the grades should reflect expected exam performance.

[glsunder]

I got a B in a college course because the professor insisted on grading on a strict curve. I had a 98 percent average. One guy had a 99. Since there were only five students, I got the B.

It wasn't a statistics class was it?

I can see using curves for big classes where there are different instructors every semester. But if a class is always taught by the same prof and it's typically small, I doubt it'd be needed.

[TragicMonkey]

It wasn't a statistics class was it?

Anthropology. That professor wound up in a mental institution three years after I graduated.

[Jorghnassen]

Anthropology. That professor wound up in a mental institution three years after I graduated.

I thought only set theorists went crazy.

Especially since the advent of category theory...

[DevilsAdvocate]

So THAT's how American students get to be straight A students. It's not about getting consistent 80 or 90% plus scores, just make sure you're close enough to the highest score to be awarded the highest mark. I think in elementary school even I may have qualified as an A student that way, and I was a real lazy-ass bum back then.Sort of. I was a good student. In high school my GPA was a little under 4. There were valedictorians with a 4.0 GPA. Of course they completed school within the absolute minimum requirements for graduation and filled their courses as much as possible with courses in home economics, typing, social studies, art, phys ed, etc. Meanwhile I was taking a full “college prep” course load consisting of courses in foreign language, ancient history, advanced calculus, chemistry, physics, etc.

[DevilsAdvocate]

I've always thought the curve was bad, and actually only needed because of lazy teachers. I was always the one that scored high and threw the curve against everyone else. I never understood the curve. Any course should have certain objectives. Any test should test whether the student can meet those objectives. If you need to scale the results, the scale should be set BEFORE the test is given, not based on the results of the test. The scale should be set based on expectations of meeting or exceeed the objectives of the class. Curving is just a lame excuse for a teacher to not now the objectives or be able to give a test that can determine wether a student met or exceeded the objectives.

I think curves were used because the teachers just bought books with pre-set lessons and tests and really had no idea what their objectives or expectations were. So, the teachers just figure the class represents a normal distribution and curve grades.

Curving makes no sense. It should be possible that everyone in the class gets an "A", or everyone gets a "C", or a "D", or a "B". Otherwise the results (rade) mean very little. For example, you could have two identical geometry classes: one class better than another. A student with a grade "C" in the "smarter" class could well have a better knowlegde of geometry than any student with an "A" in the other class. So what do the grades mean?

Grqades should be a representation of a student's ability to fulfill the pre-define objectives of the class (which is depdent only on the objectives), not a representation of a comparisson of a student with other students in a particular class (which is dependent on the students in the class).

[Jon the Geek]

I've always thought the curve was bad, and actually only needed because of lazy teachers. I was always the one that scored high and threw the curve against everyone else. I never understood the curve. Any course should have certain objectives. Any test should test whether the student can meet those objectives. If you need to scale the results, the scale should be set BEFORE the test is given, not based on the results of the test. The scale should be set based on expectations of meeting or exceeed the objectives of the class. Curving is just a lame excuse for a teacher to not now the objectives or be able to give a test that can determine wether a student met or exceeded the objectives.
That sounds more like scaling (set 100% = whatever the top person got), not curving (fit the students onto a distribution such as the bell curve, since statistically you supposedly should have students spread out into such a distribution). I think curving is ridiculous, because it assumes any given group of students fits the bell curve in ability. But scaling can make sense.

My high school chemistry teacher's objective was to teach us as much chemistry as we could learn. He loaded everything he thought we might have understood onto each test, then used scaling as a check to determine how well he did. He also had a minimum score that he'd scale to, so it wasn't possible to "game" the system and conspire together to all get 10 out of 100 or something.

Was his method the best solution? Probably not. But I learned more in that chemistry course than I did in pretty much any course since (through high school, undergrad, and grad school).

[NobbyNobbs]

I will scale or curve a test...however...only if the students actually do well on the test. With experience, I can determine if the students are putting forth an effort. I only curve the test if a fair number of students do reasonably well. If they all goof off, no curve. (I also typically toss the curve breaker grade)

glenn:boggled:

Interesting. I do the complete opposite. If everyone uniformly does poorly, I assume that it was either an unfair test or I did not teach the material properly. I will then curve the test so the the students' grades do not suffer from my faults. I will then re-teach the material. Should at least one student do relatively well, that indicates to me that I did teach the material, and had they studied the test is fair, and so I don't curve.

The method of curving I use, by the way, is to take the square root of the grade and multiply by 10. 100% stays 100, 81 moves up to 90, 36 becomes a 60.

[LW]

Didn't notice this thread before.

I don't like curving in normal situations, but there is one specific application area where I think they are useful: in nation-wide tests that thousands of students take every year.

Here in Finland we have a matriculation examination at the end of high school. Every high school student takes it. The exam results are one important entrance criterion for universities. So it is important that the results from different years are directly comparable. Curving is the only way to do this.

I wouldn't even dream of grading on curve if a course had less than 100 participants. I could think about using curve on a larger course but I wouldn't in practice. My experience in grading theoretic computer science exams is that the exam points form a nice bell curve that has two extra upwards bumps near the edges (the students who don't know and those who do). The midpoint of the main curve is in the area where the students know the most important constructs and algorithms and can apply them but who show weak understanding of more theoretical stuff. If we used a strict curve, we would end up passing a lot of students who can't really apply the algorithms in practice. On the other hand, if we issued a hard point limit for passing and curved those who passed the line, we would be giving too high grade for some borderline students.

So we don't use a curve.

[Hindmost]

Interesting. I do the complete opposite. If everyone uniformly does poorly, I assume that it was either an unfair test or I did not teach the material properly. I will then curve the test so the the students' grades do not suffer from my faults. I will then re-teach the material. Should at least one student do relatively well, that indicates to me that I did teach the material, and had they studied the test is fair, and so I don't curve.

The method of curving I use, by the way, is to take the square root of the grade and multiply by 10. 100% stays 100, 81 moves up to 90, 36 becomes a 60.

In my first year or two teaching, I would make adjustments for tests or questions that were too hard or unfair. Now that I have a few years under my blackboard, I can make up a test that is reasonable. I will always try and stretch the limits of what I taught to see how many students are pushing themselves. I always put in some questions that will be hard to answer. It helps me know what the students are learning and prepare them for college when everything steps up. I won't compromise on what I teach either...they get a full class with essentially no time off...unless watching an episode of "Mythbusters" is considered time off.

If someone gets a 98 on a test, then I know it was fair. If the next highest grade is 90-92, then I will scale it up about 6 points for the class...however, no passing anyone for "trying hard." If the whole class bombs what I know to be a fair test, it's tough luck, I don't want to give out grades for free...so, no curve or scale.

I am open to other ideas however. In general I don't get complaints about fairness, but do get some about how hard stuff is.

glenn

I didn't pruff read this post.

[Godmode]

I don't understand the curve. I suppose it depends on whether you think it's important to make sure children have learned a specific amount of information (I do) or whether you think it's more important children learn to compete amongst themselves.
The problem I see with the curve is if you are the smartest in the class, it doesn't follow you are smart. You could just be in a class full of idiots. The reverse is also true.
I think we need to test on knowledge. We should keep track of how we compare to others, but more as a tool of information. (If everybody is getting an A+, I think its safe to say you could be adding more challenging studies)

[Almo]

Curves are necessary IF the scores warrant it.

As an example, when I took the PhD qualifier at U of Iowa, we had scores like this out of 120:

105
90
44
35
33
32
31
25
15

They flunked everyone below 90. These are physics grad students we're talking about here. I knew everyone here, and the top 2 were the usual "grinders" who could do problems all day, but were worthless in terms of creativity in research; their scores are clearly outliers. Mine was the 33. This score distribution happened two years running, as a certain unnamed prof was head of the question selection committee those years. Both years they flunked 7 out of 9 students. Hence I have a master's in physics and work at a video game company programming games for the PSP. I even had all the course work for the PhD and only about 6 months of research to do to finish.

Several of the students that were flunked on that exam left physics, even though they would have been an asset to the community.

With regard to the OP, failure to look at the distribution to assess performance cost the academic world several really good people.

Having said all that, using a curve to enforce 10%-F 20%-D 40%-C 20% B-10%-A is massively stupid.

[Jorghnassen]

Well, a programming job in Montreal beats any job in Iowa if you ask me ;)

/or any physics job anywhere, but I'm no physicist...
//well maybe not all physics jobs, some sound kind of cool

[lenny]

shouldn't the answer to "to curve or not to curve?" depend on the decision that the results of the exam are intended to inform?
Curves are necessary IF the scores warrant it.
As an example, when I took the PhD qualifier at U of Iowa, we had scores like this out of 120:
105
90
44
35
33
32
31
25
15
They flunked everyone below 90.
How could one possibly "curve" this distribution? given only a small number of students and a test that changes from year to year, there is simply not enough information in these 9 numbers to say much about anything, certainly not regarding the candidates ability to be research physics.

even simple rank ordering is rather dubious with scores of 31,32,32,35 (out of 120).

it seems to reflect a general attempt to avoid the responsibility of evaluation, which might be better made subjectively (even with all the shortcoming there), rather than basing it on these numbers.
Having said all that, using a curve to enforce 10%-F 20%-D 40%-C 20% B-10%-A is massively stupid.agreed. and this is the case even with thousands of students taking an exam north of 60°N. the exam changes from year to year, we need to normalise to take this into account, not merely "curve" to get the same distribution as if we were sampling from the same population (distribution) in different years and merely accounting for sampling fluctuations.

even then, if the students only take the test once their individual score will not reflect the talent of many, due it illness, personal issues (and short-comings due to what skills the exam quantifies, as noted by Almo).

from an individual student's point of view: what is the aim of a statistically objective evaluation of a set of uncertian numbers?

[curt]

Trying to explain to students that a 70 is a C, a 75 a B, and an 80 an A merely encourages them to quibble over individual points in grading because they think (correctly or not) that they can argue about misgrading, partial credit, and badly phrased test questions. and eke out the extra few points they need.

I have on my syllabus that students are welcome to come to office hours and discuss why their grade should have been higher (tests or papers). However, part of this deal is if I decide that I graded too high, I have the right to lower their grade. When I started this policy, the number of students who wanted to debate their grade dropped steeply!

[Beleth]

Sounds like it's a case of different methods to achieve different goals.

Curving is good when it's a huge class, and you are trying to weed out students (reply 49) or if the teacher is suboptimal. Other methods are better if you have different goals or quality of teacher.

So, LA, to answer your original question ("On that information alone, should there be a grading curve?"), I have to give a definite "There is no way to give a good answer with that information alone."

[LostAngeles]

Sounds like it's a case of different methods to achieve different goals.

Curving is good when it's a huge class, and you are trying to weed out students (reply 49) or if the teacher is suboptimal. Other methods are better if you have different goals or quality of teacher.

So, LA, to answer your original question ("On that information alone, should there be a grading curve?"), I have to give a definite "There is no way to give a good answer with that information alone."

Oh Beleth...

It was a class of 8, the minimum number needed to keep an upper-division (in this school) running.

Also, I got a B.

:p

[Rob Lister]

I haven't read every post but I read most of most.

By all means...CURVE!

It ain't really fair, but neither is life.

In life, we're [almost] all graded on an equally weighted curve. Learn it in college.

ETA: a curve doesn't require you do great to pass, just that you do better than that percentage of your classmates who did not protect their notes from your opportunistic theft. :)