Is grading on a scale a flawed method for evaluating student performance?

[lisab]

Is this the sort of program where only the top X% of the class graduated?

Yes! The curve was always centered around 2.6 to 2.9, I think (it's been a while), so some always failed.

In my senior year, it wasn't unusual for the failing students to be first-year grad students from chemistry, geophysics, engineering...there were *lots* of them in the senior classes. Poor folks didn't know they were walking into a buzz saw.

 

[Drakkith]

Yes! The curve was always centered around 2.6 to 2.9, I think (it's been a while), so some always failed.

In my senior year, it wasn't unusual for the failing students to be first-year grad students from chemistry, geophysics, engineering...there were *lots* of them in the senior classes. Poor folks didn't know they were walking into a buzz saw.

Wow, that's nothing like the grading on a curve that I've ever heard of.

 

[Drakkith]

Sure, I know it sounds weird. But believe, designing good exams is difficult. The very first exams I designed were much too difficult. They took way too long to solve and required some ingenious techniques which are very much obvious to me, but not to students of that level. I think every teacher should grade on a curve the first few years until he knows the abilities of the students more.

Thanks for your explanations, Micro. I actually wondered about this last year when my Chemistry teacher graded on a curve. I thought it was silly and that if I'm supposed to learn about chemistry, what good does a curve do? If I have to learn X amount about chemistry, what's the curve for?! Surely if the tests themselves showed that I did/didn't know my stuff then that was good enough, right?

Now I know better.

 

[OrangeDog]

I agree to a point. I think grading on a scale is good for very difficult exams, but useless for easy exams. It also depends on the number of students being graded. But I think testing in itself isn't very useful. In the real world you are almost never in a situation where you have to answer a series of technical questions in a fixed time frame with little to no resources. I think projects or long homeworks are the better way to go.

 

[Choppy]

In the real world you are almost never in a situation where you have to answer a series of technical questions in a fixed time frame with little to no resources.

I think this can depend a lot on your position. You can very quickly lose a lot of professional respect if every time someone asks you a question your response is "hold on while I go look that up." (And it can get even worse if you don't say that when you have to and end up providing wrong answers.)

I think projects or long homeworks are the better way to go.

I agree to a certain extent.

One of the main reasons that tests are timed is because of cost. You have to pay someone to proctor the exam. You have to get everyone together to take the exam in the first place, pay for the heat in the room, etc.

With extended homework problems or projects, you have a lot more time, but because there is less supervision, there is also a lot more opportunity to cheat.

 

[micromass]

I think projects or long homeworks are the better way to go.

Oh come on. Think realistically. With every homework or project I had to do, there was always one person doing everything and the rest doing nothing. You have no control over whether a person truly knows their stuff.

 

[collinsmark]

A bit of humor from Yesterday's xkcd:

Estimating Time

[Source: http://www.xkcd.com/1658/]

With mouseover:

"Corollary to Hofstadter's Law: Every minute you spend thinking about Hofstadter's Law is a minute you're NOT WORKING AND WILL NEVER FINISH! PAAAAAANIIIIIIC!"​

For reference:

Hofstadter's Law: It always takes longer than you expect, even when you take into account Hofstadter's Law.​

[Source: https://en.wikipedia.org/wiki/Hofstadter's_law]

 

[bluemoonKY]

From your first post in this thread. Your reasoning here is flawed, IMO.

My reasoning is not flawed.

The students who know the material being tested on will likely do better than average, and the ones who don't know the material will likely do worse than average.

Agreed. I never said otherwise.

How can you conclude that students who understand the material will fail, and the ones who don't know the material will surpass the others?

The answer is: I didn't.

I'm saying that on a test graded on a curve, it's possible for students who understand the material to fail, and it's possible for students who don't know the material to pass the class (or even get an A!).

Imagine a calculus class in which everyone in the class is a primadonna at calculus. We will call this calculus class: Class #1. In class #1 there are 20 students. All 20 students in class #1 know how to correctly solve any calculus problem in any calculus textbook. There are 4 tests in class #1 , and then a fifth test at the end of the semester that is their final exam. Everyone in the class scores a grade of 100% on the first 4 tests. There are 100 calculus problems on the final exam. Everyone except Johnny gets all the questions correct on the final exam. Johnny is one of the students in class #1, and Johnny gets 99 of the calculus problems correct on the final exam, and Johnny gets one calculus problem on the final exam wrong due to a typo. Since everyone else in the class got 100% of the calculus problems right on the 5 tests in the course, Johnny has the lowest grade in the entire class due to his typo on one question on the final exam. Therefore, since Johnny finished dead last, when the teacher grades on a curve, the teacher fails Johnny, even though Johnny understands the calculus better than 99.9% of calculus students in the average calculus class.

 

[vela]

I agree to a point. I think grading on a scale is good for very difficult exams, but useless for easy exams. It also depends on the number of students being graded. But I think testing in itself isn't very useful. In the real world you are almost never in a situation where you have to answer a series of technical questions in a fixed time frame with little to no resources. I think projects or long homeworks are the better way to go.

Many aspects of school don't reflect what the real world is like. So what? The purpose of testing is to assess student performance so that both the instructor and students can see how they're doing. It doesn't have to be the only method of assessment. You can give students a combination of tests, homework, projects, etc.

Grading on a curve should really only be used when you, as the instructor, don't know what to expect. As others have noted, the first few exams can easily become disasters when you misjudge the abilities of the students, but as you become more experienced as an instructor, you should be able to write exams that match a fixed grading scale.

As far as timed testing goes, I agree with Choppy that it's really about scheduling. You only have the room for a certain amount of time. Ideally, time shouldn't be a limiting factor for competent students. Of course, you could eliminate time as a factor by giving take-home exams, but unfortunately, I find you can't trust some students who apparently have no sense of integrity and will cheat.

 

[Mark44]

Imagine a calculus class in which everyone in the class is a primadonna at calculus. We will call this calculus class: Class #1. In class #1 there are 20 students. All 20 students in class #1 know how to correctly solve any calculus problem in any calculus textbook. There are 4 tests in class #1 , and then a fifth test at the end of the semester that is their final exam. Everyone in the class scores a grade of 100% on the first 4 tests. There are 100 calculus problems on the final exam. Everyone except Johnny gets all the questions correct on the final exam. Johnny is one of the students in class #1, and Johnny gets 99 of the calculus problems correct on the final exam, and Johnny gets one calculus problem on the final exam wrong due to a typo. Since everyone else in the class got 100% of the calculus problems right on the 5 tests in the course, Johnny has the lowest grade in the entire class due to his typo on one question on the final exam. Therefore, since Johnny finished dead last, when the teacher grades on a curve, the teacher fails Johnny, even though Johnny understands the calculus better than 99.9% of calculus students in the average calculus class.

The probability of this occurring is unlikely in the extreme -- 1) that in a class of 20 students, all of them would be identically matched, and 2) that if this were to happen, that the instructor would be so robotic and so foolish as to fail a student who had gotten all of the problems correct on the first four exams, and 99 of the 100 on the final exam.

The basic assumption behind grading on the curve is that the abilities of the students are normally distributed. A scenario like the one you describe would be strong indication that the sample of students in this class is NOT normally distributed.

 

[Pepper Mint]

I'm saying that on a test graded on a curve, it's possible for students who understand the material to fail, and it's possible for students who don't know the material to pass the class (or even get an A!).

Imagine a calculus class in which everyone in the class is a primadonna at calculus. We will call this calculus class: Class #1. In class #1 there are 20 students. All 20 students in class #1 know how to correctly solve any calculus problem in any calculus textbook. There are 4 tests in class #1 , and then a fifth test at the end of the semester that is their final exam. Everyone in the class scores a grade of 100% on the first 4 tests. There are 100 calculus problems on the final exam. Everyone except Johnny gets all the questions correct on the final exam. Johnny is one of the students in class #1, and Johnny gets 99 of the calculus problems correct on the final exam, and Johnny gets one calculus problem on the final exam wrong due to a typo. Since everyone else in the class got 100% of the calculus problems right on the 5 tests in the course, Johnny has the lowest grade in the entire class due to his typo on one question on the final exam. Therefore, since Johnny finished dead last, when the teacher grades on a curve, the teacher fails Johnny, even though Johnny understands the calculus better than 99.9% of calculus students in the average calculus class.

This I think is probably an example of spotlight fallacy and/or part of a gambler fallacy too (i.e the conclusion is drawn based solely on previous positive test results whereas that some current or next faulty outcomes are simply ignored or "sympathized"; these tests should only be used as approximate methods to measure system performance). Why could 19 students do it right but only Johnny wrong? We can also consider how serious this exam is (e.g a training one for doctors-to-be in surgical operations). If I am a teacher in that class, I won't likely fail Johnny, though. I personally made many of such mistakes myself during my whole life and sure I will be making more new ones in the future but lessons learned always remind me not to make any similar old ones again.

 

[bluemoonKY]

The probability of this occurring is unlikely in the extreme -- 1) that in a class of 20 students, all of them would be identically matched, and 2) that if this were to happen, that the instructor would be so robotic and so foolish as to fail a student who had gotten all of the problems correct on the first four exams, and 99 of the 100 on the final exam.

If everyone in the class could correctly solve every calculus problem in any calculus textbook, then the class would be identically matched as far as calculus goes. As for #2, if the teacher graded the class on a curve, and the grading was completely objective, then the teacher would fail a student who had gotten all of the problems correct on the first four exams and 99 of the 100 on the final exam. By the way, my example was somewhat an exaggeration to make a point. You asked me how those scenarios could happen, and I answered you.

The basic assumption behind grading on the curve is that the abilities of the students are normally distributed. A scenario like the one you describe would be strong indication that the sample of students in this class is NOT normally distributed.

Two points which I will label A and B so as not to confuse them with your #1 and #2. A: The basic assumption could be wrong. B: Even if the abilities of the class were distributed normally, what if all the students were crummy students?

Regarding B, it does happen, especially at community colleges. My father once told me about an electrical engineer who used to teach calculus at a community college. This electrical engineer taught calculus at a community college as a second job. Anyway, this electrical engineer said that all or almost all of the students in his calculus class had very crummy calculus skills. The electrical engineer said that most of the students barely paid any attention in class. Almost all or maybe all of the students in the calculus class would do a miserable job on any calculus test the engineer issued. They wouldn't do homework. The engineer failed almost all or all of the students in his class. The Dean told him that he was grading too strictly. So eventually the engineer started issuing the tests to the students before the test, and the engineer even gave his students the answers to all the test questions before the test. The only stipulation was that the students could not use the cheat sheets during the test itself. All or almost all of the students still got all or almost all the calculus problems wrong on the test! Therefore, the engineer failed all or almost all of the students. Finally, the community college fired the engineer for failing so many students. The engineer thought he should fail all or almost all the students since all or almost all the students had such poor calculus skills. I agree with the engineer. If nobody in a class can competently solve the calculus problems, all the students should fail.

 

[OrangeDog]

Many aspects of school don't reflect what the real world is like. So what? The purpose of testing is to assess student performance so that both the instructor and students can see how they're doing. It doesn't have to be the only method of assessment. You can give students a combination of tests, homework, projects, etc.

Grading on a curve should really only be used when you, as the instructor, don't know what to expect. As others have noted, the first few exams can easily become disasters when you misjudge the abilities of the students, but as you become more experienced as an instructor, you should be able to write exams that match a fixed grading scale.

As far as timed testing goes, I agree with Choppy that it's really about scheduling. You only have the room for a certain amount of time. Ideally, time shouldn't be a limiting factor for competent students. Of course, you could eliminate time as a factor by giving take-home exams, but unfortunately, I find you can't trust some students who apparently have no sense of integrity and will cheat.

I completely disagree. I feel the purpose of schools should be to prepare students to work, unless they go a more advanced route such as graduate school. As an engineer in industry, I feel if my only goal was to get a job and make money that my education could have been shortened to two years and was a complete waste of time. I have other goals though and enjoy learning so this is not the case. I think the purpose of any class should be to educate, which at the end of the day will always be up to the student. I know plenty of people who would cram or exams or study for the test and are terrible engineers because they can't remember any of their rigorous course material. I like projects (not necessarily group projects) or extended homeworks because I feel it gives a student the opportunity to exercise their creativity and maybe incorporate stuff they find interesting into their education, which will ultimately help them learn more. I do agree that the fundamentals like calculus and physics need to be a in a test style, but as the material gets tougher the standard college exam becomes less useful. There will always be kids that cheat, cram, are lazy, etc. I have found in my limited experience that making the class fun and engaging students in their work minimizes these obstacles and promotes learning.

I had one professor who had a great style of giving assignments: rather than tests he assigned 6 extended homework assignments every 2 weeks. Once you got your grade back you had the opportunity to do "corrections", where by incorporating the professors comments and fixing erroneous answers you could earn back some points. I remember on one of his assignments I got a very high mark because I opted to include an additional analysis of solar technology on a battery powered UAV vs a standard 2-stroke design. I felt I learned the most from his class because the work he assigned required lots of outside the box thinking and extensive time trying to produce a realistic answer.

 

[Mark44]

The probability of this occurring is unlikely in the extreme -- 1) that in a class of 20 students, all of them would be identically matched, and 2) that if this were to happen, that the instructor would be so robotic and so foolish as to fail a student who had gotten all of the problems correct on the first four exams, and 99 of the 100 on the final exam.

If everyone in the class could correctly solve every calculus problem in any calculus textbook, then the class would be identically matched as far as calculus goes. As for #2, if the teacher graded the class on a curve, and the grading was completely objective, then the teacher would fail a student who had gotten all of the problems correct on the first four exams and 99 of the 100 on the final exam. By the way, my example was somewhat an exaggeration to make a point. You asked me how those scenarios could happen, and I answered you.

My intent was "how could those scenarios happen in the real world?" How often do you suppose a scenario like the one you describe actually occurs?

The basic assumption behind grading on the curve is that the abilities of the students are normally distributed. A scenario like the one you describe would be strong indication that the sample of students in this class is NOT normally distributed.

Two points which I will label A and B so as not to confuse them with your #1 and #2. A: The basic assumption could be wrong. B: Even if the abilities of the class were distributed normally, what if all the students were crummy students?

Then they wouldn't be normally distributed. In either situation, with all students at essentially the same level, whether all aces or all duds, you don't have a normal distribution.

Regarding B, it does happen, especially at community colleges. My father once told me about an electrical engineer who used to teach calculus at a community college. This electrical engineer taught calculus at a community college as a second job. Anyway, this electrical engineer said that all or almost all of the students in his calculus class had very crummy calculus skills. The electrical engineer said that most of the students barely paid any attention in class. Almost all or maybe all of the students in the calculus class would do a miserable job on any calculus test the engineer issued. They wouldn't do homework. The engineer failed almost all or all of the students in his class. The Dean told him that he was grading too strictly. So eventually the engineer started issuing the tests to the students before the test, and the engineer even gave his students the answers to all the test questions before the test. The only stipulation was that the students could not use the cheat sheets during the test itself. All or almost all of the students still got all or almost all the calculus problems wrong on the test! Therefore, the engineer failed all or almost all of the students. Finally, the community college fired the engineer for failing so many students. The engineer thought he should fail all or almost all the students since all or almost all the students had such poor calculus skills. I agree with the engineer. If nobody in a class can competently solve the calculus problems, all the students should fail.

I agree.

 

[vela]

If everyone in the class could correctly solve every calculus problem in any calculus textbook, then the class would be identically matched as far as calculus goes. As for #2, if the teacher graded the class on a curve, and the grading was completely objective, then the teacher would fail a student who had gotten all of the problems correct on the first four exams and 99 of the 100 on the final exam. By the way, my example was somewhat an exaggeration to make a point. You asked me how those scenarios could happen, and I answered you.

That's like saying every time you drive your car, you run the risk of dying in a car accident, so you should never ever drive anywhere. The probability of an event happening matters.

Regarding B, it does happen, especially at community colleges. My father once told me about an electrical engineer who used to teach calculus at a community college. This electrical engineer taught calculus at a community college as a second job. Anyway, this electrical engineer said that all or almost all of the students in his calculus class had very crummy calculus skills. The electrical engineer said that most of the students barely paid any attention in class. Almost all or maybe all of the students in the calculus class would do a miserable job on any calculus test the engineer issued. They wouldn't do homework. The engineer failed almost all or all of the students in his class. The Dean told him that he was grading too strictly. So eventually the engineer started issuing the tests to the students before the test, and the engineer even gave his students the answers to all the test questions before the test. The only stipulation was that the students could not use the cheat sheets during the test itself. All or almost all of the students still got all or almost all the calculus problems wrong on the test! Therefore, the engineer failed all or almost all of the students. Finally, the community college fired the engineer for failing so many students. The engineer thought he should fail all or almost all the students since all or almost all the students had such poor calculus skills. I agree with the engineer. If nobody in a class can competently solve the calculus problems, all the students should fail.

To me, this sounds like the engineer was a really bad instructor, if the anecdote is even remotely true.

 

[bluemoonKY]

My intent was "how could those scenarios happen in the real world?" How often do you suppose a scenario like the one you describe actually occurs?

I think the hypothetical scenario of all students being primadonnas (sp?) where most of the class gets every single answer correct for every test the whole semester is extremely rare. However, like I said, it was somewhat of an exaggeration to make a point. It is not rare for the midpoint of the Bell Curve distribution to be higher for some some classes than others even at the same course at the same school.

I think the scenario of all the students being extremely low ability is more common than the scenario where all the students are primadonnas. What do I base that on? Just my experiences and observations from about 15 years of education in my life at public schools and a university and a community college, and I base it on what other teachers and professors I've known have told me and what my father told me that that electrical engineer told him.

Then they wouldn't be normally distributed. In either situation, with all students at essentially the same level, whether all aces or all duds, you don't have a normal distribution.

That's part of the reason it's stupid to grade on a curve.

 

[bluemoonKY]

That's like saying every time you drive your car, you run the risk of dying in a car accident, so you should never ever drive anywhere. The probability of an event happening matters.

My hypothetical scenario where all the students except one get every single test questions correct for the whole semester is ridiculously unlikely. However, the midpoint of the Bell Curve distribution being significantly different for one class than another at the same course at the same university is not highly unlikely.

To me, this sounds like the engineer was a really bad instructor, if the anecdote is even remotely true.

If the anecdote is true, how would it make the engineer a bad instructor? If it's true, the students did not pay attention in class and did not do homework, and the students were of extremely low ability.

 

[Borek]

Not that I wish you such a situation, but I wonder if you would still say "its stupid" if you were of victim of a too hard exam, in which nobody passes - but the results, despite being below the passing minimum, are still nicely curved.

 

[Drakkith]

If the anecdote is true, how would it make the engineer a bad instructor? If it's true, the students did not pay attention in class and did not do homework, and the students were of extremely low ability.

The problem is that this anecdote is potentially highly biased. It's entirely possible the teacher was actually a very poor teacher and their students didn't pay attention in class because they were confused, lost, or bored out of their minds from the teacher's teaching style, and didn't do their homework because they did not learn enough in class to be able to do it. I see it happen to some of the folks I tutor all the time.

 

[Mark44]

My intent was "how could those scenarios happen in the real world?" How often do you suppose a scenario like the one you describe actually occurs?

I think the hypothetical scenario of all students being primadonnas (sp?) where most of the class gets every single answer correct for every test the whole semester is extremely rare.

But this was the basis for your example of good students getting a failing grade, and poor students receiving a passing grade.

However, like I said, it was somewhat of an exaggeration to make a point. It is not rare for the midpoint of the Bell Curve distribution to be higher for some some classes than others even at the same course at the same school.

I think the scenario of all the students being extremely low ability is more common than the scenario where all the students are primadonnas. What do I base that on? Just my experiences and observations from about 15 years of education in my life at public schools and a university and a community college, and I base it on what other teachers and professors I've known have told me and what my father told me that that electrical engineer told him.

Then they wouldn't be normally distributed. In either situation, with all students at essentially the same level, whether all aces or all duds, you don't have a normal distribution.

That's part of the reason it's stupid to grade on a curve.

As I mentioned a couple of times, the primary assumption in grading on the curve is that the students are normally distributed; i.e., that they are a representative sample of the population at large. In a situation where all of the students are perfoming at the same level, use of the Bell curve would not be appropriate.

I base what I'm saying on my time as a student (about 18 years), plus the time I spent teaching (about 21 years), and including the courses I took on statistics (a year-long upper division course in mathematical statistics, plus a couple of other stats courses).

 

[Drakkith]

I'm apparently not familiar with this type of curve. In every curve I've ever heard of, until this thread, everyone moves up and no one moves down.

 

[bluemoonKY]

Not that I wish you such a situation, but I wonder if you would still say "its stupid" if you were of victim of a too hard exam, in which nobody passes - but the results, despite being below the passing minimum, are still nicely curved.

If the test questions were all things that anyone who passes the course should know, and I and everyone else in the class failed the test, I would still think, in the privacy of my own mind, that it would be stupid for the teacher/professor to grade it on a curve so that some people pass, even if the results of the test were in a Bell Curve distribution. I admit that I would not complain if the professor graded it on a curve and I passed because it was graded on a curve.

In my mind, there is just no getting around my idea that anyone who does not competently know the course content should fail.

 

[bluemoonKY]

But this was the basis for your example of good students getting a failing grade, and poor students receiving a passing grade.

Not just that. Also this: It is not rare for the midpoint of the Bell Curve distribution to be higher for some some classes than others even at the same course at the same school.

As I mentioned a couple of times, the primary assumption in grading on the curve is that the students are normally distributed; i.e., that they are a representative sample of the population at large. In a situation where all of the students are perfoming at the same level, use of the Bell curve would not be appropriate.

Use of the Bell Curve can also not be appropriate where the students are not performing at the same level if the weight of the distribution is too high of an ability or too low of an ability.

 

[bluemoonKY]

I'm apparently not familiar with this type of curve. In every curve I've ever heard of, until this thread, everyone moves up and no one moves down.

That's not the curves that I'm familiar with. Probably your teachers only graded on a curve if everyone in a class did extremely poorly. The type of curves I'm talking about are not unheard of. Just look at lisab's example on page 2.

 

[bluemoonKY]

Mark44, I think that I would not violently disagree with anything in your method of grading. I'm saying that grading on a curve is stupid largely because of two possible things that could happen: #1 everyone in the class is competent or #2 everyone in the class is highly incompetent. You're saying that a teacher should not grade on a curve if either of those two examples happen, but it's okay otherwise. I don't like when teachers grade on a curve period, but I don't think I would try to avoid taking your classes due to your grading since if either of those two examples happened, you would not grade on a curve.

 

[Mark44]

But this was the basis for your example of good students getting a failing grade, and poor students receiving a passing grade.

Not just that. Also this: It is not rare for the midpoint of the Bell Curve distribution to be higher for some some classes than others even at the same course at the same school.

Well, of course. The students in each class are just samples from the overall population of students. It's entirely possible that one group of students will have a mean and standard deviation somewhat different from another.

As I mentioned a couple of times, the primary assumption in grading on the curve is that the students are normally distributed; i.e., that they are a representative sample of the population at large. In a situation where all of the students are perfoming at the same level, use of the Bell curve would not be appropriate.

Use of the Bell Curve can also not be appropriate where the students are not performing at the same level if the weight of the distribution is too high of an ability or too low of an ability.

What does "the weight of the distribution is too high of an ability or too low of an ability" mean?

The Bell curve for a given class is based on the statistics (mean and standard deviation) for that class. That's what "grading on the curve" means.

 

[Mark44]

I'm apparently not familiar with this type of curve. In every curve I've ever heard of, until this thread, everyone moves up and no one moves down.

That's definitely not "grading on the curve." "The curve" is based on the distribution for that class. A grading system where everyone moves up and no one moves down is different from grading on the curve.

 

[Mark44]

Mark44, I think that I would not violently disagree with anything in your method of grading. I'm saying that grading on a curve is stupid largely because of two possible things that could happen: #1 everyone in the class is competent or #2 everyone in the class is highly incompetent. You're saying that a teacher should not grade on a curve if either of those two examples happen, but it's okay otherwise. I don't like when teachers grade on a curve period, but I don't think I would try to avoid taking your classes due to your grading since if either of those two examples happened, you would not grade on a curve.

The reasons you're giving for saying that this system is "stupid" are either extremely unlikely (everyone competent) by your own admission, or not appropriate (everyone incompetent). Either way, these are pretty rare, notwithstanding the anecdotal evidence from your father.

The alternative to grading on a curve is a system where getting 90% - 100% of the available points in the class earns you an A (or 4.0, whatever), 80% - 89% gets you a B, and so on. What typically happens is that out of a class of 30 students, 3 or 4 get an A, 10 more get a B, 15 get a C, and so on. It can happen that no one gets over 79%, so no B's and no A's.

One other thing. Calling something that you don't like "stupid" is pretty childish, if you can't give solid reasons for that opinion.

 

[bluemoonKY]

Well, of course. The students in each class are just samples from the overall population of students. It's entirely possible that one group of students will have a mean and standard deviation somewhat different from another.

Yeah, that's another reason I don't like grading on a curve. Let me give an example. Let's say there are two calculus I classes at the same university with the same professor teaching both. The professor performs the same in presenting the material to both classes. The professor grades both classes based on how well the students solve calculus problems from the same textbook. In class #1, the median student can solve 85% of the calculus problems in the textbook correctly. In class #2, the median student can only solve 60% of the exact same calculus problems in the same textbook correctly. The professor grades both classes on a curve. In class #1 , student X has the lowest grade in the class. Student X can only solve 70% of the calculus problems in the textbook correctly. In class #2, student Y has the highest grade in the class. Student Y can solve 70% of the calculus problems in the textbook correctly. Because it's grade on a curve, student X fails the calculus class, and student Y gets an A in the same class, even though Student X and Student Y are at the exact same ability level! Is that how it should be graded and is that fair? NO!

What does "the weight of the distribution is too high of an ability or too low of an ability" mean?

If everyone in a calculus class is competent at solving calculus problems, then the weight of the distribution is too high to grade on a curve because you would have to fail some of the (competent) students if you graded on a curve. If nobody in a calculus class is competent at solving calculus problems, then the weight of the distribution is too low to grade on a scale because you would be passing incompetent students.

 

[bluemoonKY]

The alternative to grading on a curve is a system where getting 90% - 100% of the available points in the class earns you an A (or 4.0, whatever), 80% - 89% gets you a B, and so on. What typically happens is that out of a class of 30 students, 3 or 4 get an A, 10 more get a B, 15 get a C, and so on. It can happen that no one gets over 79%, so no B's and no A's.

The alternative you mention here is better. A knowledgeable and experienced and competent teacher/professor should not need to grade on a curve.

 

[bluemoonKY]

One other thing. Calling something that you don't like "stupid" is pretty childish, if you can't give solid reasons for that opinion.

I've been giving solid reasons for my opinion throughout this whole thread.

 

[bluemoonKY]

The reasons you're giving for saying that this system is "stupid" are either extremely unlikely (everyone competent) by your own admission, or not appropriate (everyone incompetent)..

One more thing. When you say examples of situations where it would be inappropriate to grade on a curve, you're basically agreeing with my position.

 

[Mark44]

Yeah, that's another reason I don't like grading on a curve. Let me give an example. Let's say there are two calculus I classes at the same university with the same professor teaching both. The professor performs the same in presenting the material to both classes. The professor grades both classes based on how well the students solve calculus problems from the same textbook. In class #1, the median student can solve 85% of the calculus problems in the textbook correctly. In class #2, the median student can only solve 60% of the exact same calculus problems in the same textbook correctly. The professor grades both classes on a curve. In class #1 , student X has the lowest grade in the class. Student X can only solve 70% of the calculus problems in the textbook correctly. In class #2, student Y has the highest grade in the class. Student Y can solve 70% of the calculus problems in the textbook correctly. Because it's grade on a curve, student X fails the calculus class, and student Y gets an A in the same class, even though Student X and Student Y are at the exact same ability level! Is that how it should be graded and is that fair? NO!

Since in this scenario, the same instructor is teaching both classes, the instructor would likely use one curve and incorporate both classes. Class #1 would have more students with the higher grades, and class #2 would have more students with the lower grades.

What does "the weight of the distribution is too high of an ability or too low of an ability" mean?

If everyone in a calculus class is competent at solving calculus problems, then the weight of the distribution is too high to grade on a curve because you would have to fail some of the (competent) students if you graded on a curve.

If nobody in a calculus class is competent at solving calculus problems, then the weight of the distribution is too low to grade on a scale because you would be passing incompetent students.

"Weight" of a distribution is not a term that is used. And as I've said numerous times, the first situation does not occur often enough to worry about, and the second situation is one in which a curve should not be used.

 

[Mark44]

I've been giving solid reasons for my opinion throughout this whole thread.

Not really. A scenario you've mentioned multiple times with everyone in the class performing at the same level, is so rare that it's not worth discussing.

 

[bluemoonKY]

Since in this scenario, the same instructor is teaching both classes, the instructor would likely use one curve and incorporate both classes. Class #1 would have more students with the higher grades, and class #2 would have more students with the lower grades.

What if the two classes are taught by different professors, but everything else in my scenario is true? There is professor A for class 1 and professor B for class 2, but the results for each classes are the same as in my example with the median of class 1 being 85% correct and the median of class 2 at 60% correct. It's not like professor A and professor B are going to put the results of both of their exams together with each other and incorporate both classes.

"Weight" of a distribution is not a term that is used. And as I've said numerous times, the first situation does not occur often enough to worry about, and the second situation is one in which a curve should not be used.

But some teachers/professors do use a curve for the second situation I mentioned.