I'm just back from a committee meeting at which the subject of grades and grade distributions came up, and it became clear to me that academics (even at the same institution, even in the same field) have wildly different philosophies about just what grades ought to mean.
There are the normal-distribution folks, who think grades ought to convey how you are doing relative to the other people taking the class. The average grade is a C, no matter whether that average corresponds to demonstrating coompetence on 40% of the content or 90% of the content. The grade you get is dependent on how many standard deviations above or below the mean you are. (It should be mentioned that there are universities -- including some with very high tuitions -- where the mean is more like a B than a C, but where the general approach is still a normal-distribution approach.)
Then there are the grading-on-mastery folks, who use grades to identify how well you have mastered the material. An "A" paper will be one where you've mastered almost all of the material, while an "F" paper is one where you show little to no mastery of the material. Folks who approach grading this way often have nice rubrics that will spell out the virtues an "A" paper must instantiate, those that a "B" paper must instantiate, and so on.
Fans of the normal-distribution approach point out that it gives you a sound basis for assigning "average" grades (since C=average) and helps avoid inflated grades (except in cases where the mean is set at something higher than C).
Fans of the grading-on-mastery approach point out that a measure of how much of the course material you have learned is more meaningful and useful than a measure of how well or poorly the other people taking the same course did relative to your performance.
What if a majority of people in a class master the material really well? What if a majority of people in a class master very little of the material?
If the class average on a midterm was 88%, does that necessarily mean the exam was too easy, or could it mean that the class really knew its stuff? If the class average on a midterm was 33%, does that necessarily mean the exam was too hard, or could it mean the class hasn't mastered the material?
(My own first response to situations like this has been to re-examine whether I'm making the material clear enough if the test scores are really low, and to challenge my students more on the next exam if the test scores are really high.)
Should a grading-on-mastery person worry if the grades she awards on mastery don't end up falling in something like a C-centered normal distribution?
Should a normal-distribution person have secondary measures in place to ensure that students who haven't mastered the material don't get high grades,or that students who have mastered the material don't get low grades? (You wouldn't have to worry about both of these problems in a single set of exams or assignments, of course.)
The big question underlying the different approaches is what precisely do you want grades to communicate to you? And, I suppose, that might be connected to the question of who the intended audience of this communication is (the student, the parents, the admissions committee, the potential employer, etc.).
What say you?
Wouldn't the size of the class also come into play?
The normal distribution probably works really well in an undergraduate class of 300. The people who fail that course really deserve to have failed and the A grades probably do have a mastery of the material as you've presented it. The normal distribution works well in large samples that have a substantial variability in them. Then again in my limited experience of grading... grading on mastery in a large class set of essays (300+) created a normal distribution anyway.
However, in a graduate seminar class of about 15-20 a normal distribution is going to cause signficant problems. Hard working grad students putting in some serious effort and with a good understanding of the material are going to fail because of the distribution. This is particularly unfair when you happen by chance to have a cohort of particularly good students come through in that year- some of these will look worse than previous/future students who they are superior to. Furthermore, distribution grading in a grad setting is a bad idea as it encourages the small class to stop communicating/collaborating with each other. They know that the bottom 4 of them are going to fail so it's not a matter of mastery but a matter of doing better than the others. Hardly conducive to a collegial atmosphere. In a small tightly selected class the normal distribution is not going to work well for you. Better to grade on mastery I would have thought.
As a victim of a misapplied distribution curve, I'm all for level of mastery. I also agree with Nat that class size and who is in that class are also important discriminators for deciding which grading approach to take.
Here's my personal example: At a small New England liberal arts college which has consistently rated in the top ten for decades, I was in a Physiological Psychology (That probably dates me) class of 17 students, all of whom were either pre-med, biology, or psych majors. All testing for the course was done with multiple-choice questions (no essays at all) on weekly quizzes, two hour exams, and a three-hour final.
If calculated on percentage of correct answers, the highest grade would have been a 99 and the lowest a 91. The prof (a full prof and the current department head) put those percentages on a normal distribution curve and flunked the poor guy with a 91 average. I had a 97 average and got a B.
In cases like that, trying to apply a distribution curve is not only nonsensical, it's a true injustice to some serious students who are mostly aiming for grad school.
As an afterthought, it occurs to me that someone who insists on using a distribution curve could conceivably be making an effort to disguise their own inability to teach well. Certainly there will be no apparent improvement in their overall grades from year to year unless they also increase the point where grades are centered.
And finally, in answer to your main question, I'd want my grades to reflect my mastery of the subject and the time and effort I'd expended on the course, not simply my position relative to all the others who took the course.
You've really got me going here. ;-)
Here's a case you can put to the distribution curve folks. I can remember being in several classes where it was quite obvious that there were two distinct types of students in the class: those who were really interested and involved in the subject, and those who were just trying to fill a distribution requirement. I'm sure a couple of those would have produced a bi-modal curve. So, how does a distribution curve advocate handle those cases?
You are holding a class for exceptionally advanced students, but the only ones who qualify are Janey and Jenny Jones, who are identical twins.
Either of two results may obtain, grading on the curve:
1. Both do equally well and get a C.
2. One does better than the other, even if their final scores are 1711 to 1710. One gets a B, the other gets a D. Or they both get a C. Or one gets an A, the other an F.
Now, when I look at J. Jones' GPA, what am I to make of the number I see?
My point is that if a scheme cannot work right on a small scale, then it cannot work on a larger scale where the law of large numbers puts enough noise into the outcome to drown out the weaknesses.
Grades should be based on mastery, but sometimes a normal-distrbution approach is a good approximation for or definition of how "mastery" is defined in a large class.
In my experience (not saying this is right, it's just what I've encountered), it has been the lower-level (and therefore, usually larger) classes which employ the normal-distribution approach. The professor knows about where the mean should be, and in general, there will be students who can't cut it who do poorly and those who can who do well. Low level classes generally function as "weeder" classes, and so a C-centered distribution with a decent number of F's is acceptable. With a low-level class, it almost makes more sense to set the mean to be a certain grade because it is harder to gauge at what exact level the material should be taught.
Higher-level classes are different. By the time someone is in a higher-level class, they should be able to master the material, and it is more appropriate to grade based on mastery alone. Kind of like Nat said, these classes also tend to be smaller (I'm also thinking upper-level undergraduate classes), so a normal-distribution grading scheme could be more hurtful than in a large class.
On the other hand, looking at the mean of a set of grades can be instructive. I remember one midterm in an upper-level undergraduate quantum mechanics class I took that had a mean about 15 points lower than the professor wanted or was expecting; the lower grades, and not our complaints, led him to realize the exam had simply been too long, so he renormalized everyone's grades by adding 15 points to each score so that his normal grade delineations for the end of the semester would still work. If we had all done better than expected, however, he probably would have decided he had taught the material exceptionally well that year.
[*** start rant *** ]
Both approaches seem foolish to me. They ignore this question: who is the consumer of the service that the university supplies?
If the students are going to graduate school, the grad school doesn't care how well the students in your class did in competition with each other, they care how those students rank against the best university students in the country, or in the world.
An "A" from Harvard or Tokyo University cannot mean the same as an "A" From West Kidney Community College, even if the courses have the same title.
If the students are going to be in management at a Fortune 100 corporation, the executive management cares how they compete with students from Yale School of Business, or Stanford.
If the student will go to Med School, is their Biology and Chemsitry enough to make them a good doctor, or are they going to be a dishonest Intelligent Design blowhard like Dr. Egnor?
Which is more convincing: an "A" from a Science Writing class at East Egg, or a publication in Science, Nature, Proceedings of the National Academy of Science?
What is the ensemble against which the students are evaulated? With whom have they competed, and with whom will they need to compete a few years from now?
Will they be Postdocs? Will they drive trucks? I've met professors whom I learned were total fools. I've had long deep conversations with truck drivers. I do not think that I'm taking an elitist position. I do think that I'm being more pragmatic than your committee.
I was in the bottom half of my high school graduating class in Grade Point Average. But it was one of the most competitive academic high schools in the world, and my SAT and AP test scores got me the interview that got me into Caltech, the most competitive of all undergraduate colleges by admissions standards.
What did my son's midle school grades mean? He skipped high school completely, and wernt straight to university at age 13. What did his university grades mean? he scored over 99th percentile on the LSAT and has already been accepted to a Top 10 Law School.
What did George W. Bush's grades or MBA mean? Does he have what it takes to run the country, let alone the Free World?
With all due respect to your committee meeting, I suspect that they were lulled into accepting a fallacious either-or.
For that matter, I've never had a job interview yet for a corporate managemernt or engineering position where they asked my what my GPA was. They wanted to know what I could do for their bottom line.
If I ever get rich, I'm tempted to endow a scholarship at Caltech which will go, each semester, to the student who passed -- but with the LOWEST grade point average.
Grades are not about the quality of education. "Leave No Child Behind" is cruelest hoax ever perpetuated on our educational infrastructure.
The competition is global. The malignant political idiocy is home-grown.
[*** end rant *** ]
Normal-distribution grading is an evil mechanism used by teachers without the courage or conviction to use and defend their own judgment about how well students are doing. The whole point of a course is to have the students master the material, and that's how they should be graded, not on where they happen to fit into the distribution. How can it possibly make sense to set an a priori criterion that half the students will have a substandard mastery of the material?
Roy said: "My point is that if a scheme cannot work right on a small scale, then it cannot work on a larger scale where the law of large numbers puts enough noise into the outcome to drown out the weaknesses."
I would argue that in fact that's exactly why it does work in large classes. Large numbers don't necessarily mean more noise by the way. That's usually because of poor measurement. What large numbers does mean is that all of the variation that should be seen in a population is much more likley to actually be observed in you sample. Thus your twin example will give two students with the same mark. In a large class they would be extremely unlikely to straddle a grade boundary. Whereas in a small class they are much more likely to straddle a grade boundary and their performance should be compared to an external benchmark (mastery) rather than each other (distribution curve).
When a professor grades on a curve they are making an assumption that the class has a normal distribution. No class has a normal distribution. Statistical tools should not be used for grading, that is not what they are for. Grades deals with peoples lives and assumptions should not be made about peoples lives. Remember, when one assumes one makes an ass of of ones self. Professor's should grade students on their individual merits because they are teaching individual people. Grading on a curve is lazy and can create injustices as pointed out by Roy and chezjake.
Kevin C wrote: "No class has a normal distribution."
I disagree. My first post did in fact mention that I have observed a large class to approximate a normal distribution. And that was by marking according to merit. The normal distribution just happened to fall out of that process.
Because of this experience I didn't necessarily make clear that I disagree with a process of forcing grades into a normal distribution just so that one would exist. I don't really understand why enforcing normality would be regarded as laudable?
KevinC wrote: "Statistical tools should not be used for grading, that is not what they are for. Grades deals with peoples lives and assumptions should not be made about peoples lives. Remember, when one assumes one makes an ass of of ones self."
I'm not sure I understand your reasoning here. Statistical tools are 'for' whatever you need them 'for'. All of them have assumptions behind them- which you should test to make reasonably certain the particular tool you are using is appropriate. And one does in fact make many assumptions about people's lives in all uses of statistics applied to human problems. If one didn't make assumptions one would have no predictive ability.
The old word play about 'assumes' has no bearing on the real world simply because it 'assumes' that everybody operates in English.
However, I agree completely that grading on a curve creates injustices. But I think that this is not necessarily a problem in large introductory classes. In these large intro classes curve grading would help to adjust for year-to-year marking variability caused by the turnover of Teaching Assistants and the necessary rewriting of finals and essay questions which might be harder or easier from year-to-year. These fluctuations might bring about more injustice than the curve grading. Ever had a really tough TA or a tough exam in a paper that had previously been easier?
Jonathan makes a good point that there's a big difference between Harvard /Caltech /Yale and a typical community college. So whilst have a fail rate of 20% at a large community college course may not be unreasonable the same may not be true in very heavily selected places such as Harvard where all of the students may indeed deserve to pass. My example comes from a University in small country where the selection process for entry to courses typically takes place at the end of the first year of study. I understand that this is not so in many of the places that you work in where students are much more carefully selected before they enter the institution. My sample might be considerably more heterogenious.
Well, for software engineering positions, companies that I have worked for didn't care at all about grades regardless of school. All top companies do heavily technical interviews that leave you tired at the end of the day. Sometimes you need to mail in answers to a small written test to be considered for an interview.
As Jonathan says, we're hiring from the global talent pool, and we would perhaps benefit from something like GRE technical to do some of the filtering for us so that we could concentrate on the non-technical stuff more during the interviews.
I've had good grades, but never better than I thought I needed. You're not always interested in the entire contents of any course, and having to study the less interesting bits because of grades is the perfect way to make you hate the subject altogether sometimes.
"I do not think that I'm taking an elitist position."
I think anyone who has read a comment anywhere by J.V. Post laughed long and hard at that remark.
I think a professor should start out with the assumption of grading on mastery, and tweak things with a distribution as they see fit. For example, my small graduate classical mechanics class is supposed to be graded on mastery, but on our last quiz, the average score was 10+-7 out of 50! Clearly we're not all morons, and the quiz wasn't a good measure of our ability. So you have to step in and fix things when that happens.
In fact, to add to that, my experience is that's how it's almost always done. Every time a friend in college would start whining about how hard a class or test was and how everyone was failing, I'd just say, "do you really think they're going to flunk everyone?", and inevitably, things would turn out fine for the student.
As a math instructor at a small community college, I teach for mastery. It's somewhat easier for me because with math there is an objective standard of perfection: is the work correct? Does it lead to the correct conclusion or answer?
Teaching on a curve for small classes makes no sense to me.
Statistics are tools for evaluating information. It is the fact that they have assumptions that make them poor tools for grading. The problem here is the normal distribution, a large class probably will come close, but is it fair to harm a good student with a bad grade because they happened to join a class that was skewed to the high end or reward an undeserving student because the class was skewed to the low end. Statistics do make a great tool for a teacher to evaluate their teaching and test writing skills.
In jeffk case I would hope the teacher would weight the quiz or just throw it out. There is a big difference between adjusting scores from a test that was to hard or to easy and just taking whatever scores the students made and throwing them onto a normal distribution curve.
In principle, I'm 100% in the mastery camp. We too often mistake competitiveness for competence. The problem is, of course, assessing mastery; there's going to be variation in performance, and on just about any test of a sufficiently large population, you'll probably see that old bell curve somewhere.
Because GPA is used as a primary discriminant for graduate and professional school admissions, choosing a method for assigning grades that differs from common conventions can have terrible consequences for the students. If some professors want to assign grades based on some personal beliefs they really should make sure a letter is sent with all the students' transcripts explaining how the grade was assigned. Harvey Mudd does this already.
N. Johnson's skepticism of my statement:
"I do not think that I'm taking an elitist position"
may be well founded. Just because I don't think I'm being elitist does not gurantee that I am not elitist. My introspection is not the compelling indicator.
In general, although Freud was wrong about most of his predictive hypotheses, I think that he was right, as was Jung, that there is very much more going on in our minds than that portion of which we are consciously aware.
Modern Cognitive Sciences show numerous phenomena in which our brain has already initiated an action substantially before we are aware of what we believe to be our intention to initiate that action.
The Science Section of ther New York Times 2 days ago was a special issue on Desire. Desire differs greatly between people, mentally, physiologically, and between the two. And what we think we desire bears only minor relationships to what our behavior indicates.
So I accept N. Johnson's critique. Consciously, I believe in earned pride in accomplishment. I take a democratic stance against inherited merit, by racial, financial, gender, or other sources. I consciously believe that we are responsible for what we accomplish, that it takes 10,000 hours of concentrated study and practice in any field to achieve minimum professional acapabilities.
But unconsciously, who knows what I know. I don't, despite a lidetime of introspection and feedback from family and colleagues.
Which brings us around again to grades.
I've had students in my classes tell me: "I know the material when I do the homework, but it just flies out of my head when I take your exam."
It may be true to them when they tell me, but it has been analogized to "I can swim, until I get into a pool or the ocean."
As a teacher, I give grades as a side-effect of my trying very VERY hard to figure out what is going in in my students' minds. I don't care if they get a problem right or wrong. I care HOW they find their way to an answer, or where they go wrong if they go off track.
Hence I tell them: if you don't know the formula, draw me a picture. A good picture gets half-credit. If you don't know the formula, nor want to draw me a picture, write me a narrative of what you think and why.
It takes me MUCH longer to grade exams than most of the other teachers I know. But by about the middle of each semester, I know each student's cognitive style and learning style rather well, and can guide people towards what works, playing on their strengths.
I have succeeded nearly 100% of the time in getting students who've failed the course 2, even 3 times before, to finally "get it."
It is routine to teach people what they don't know and know that they don't know. It is VERY hard to teach people what they don't know and don't know that they don't know, or what they incorrectly think that they know.
My wife has published papers on how bad many High School Physics textbooks and teachers are, and what stratregies do and don't appear to work in getting peole to UNLEARN the bad models and false assumptions that they bring with them into colege or university.
There are deep problems of enormous importance in education. Superficial grading theories do not help.
Some unorganized thoughts:
I thought it was a sad day a few years back when UC Santa Cruz got rid of its credit/no credit system and moved to giving letter grades -- at the urging of the students no less.
Two professors teaching the same course at the same school do it differently, giving different tests, and grading with different approaches. How do you compare the students' results?
Jonathan Vos Post: An "A" from Harvard or Tokyo University cannot mean the same as an "A" From West Kidney Community College, even if the courses have the same title.
Why not, what basis do you have to compare the two at all? Could well be the community college kid with an "C" has a greater mastery of the material than the Harvard student with an "A".
Show me the quality control applied to university teaching and to test creation. When I was in the system (long ago and on both sides, student and teacher) there didn't seem to be any. It seems just a free-for-all on the part of the instructors (or their TAs) who generally have only their own personal experience to guide them. It's a strange way to produce a metric in what we call "institutions of higher learning."
Kevin W. Parker says,
Normal-distribution grading is an evil mechanism used by teachers without the courage or conviction to use and defend their own judgment about how well students are doing. The whole point of a course is to have the students master the material, and that's how they should be graded, not on where they happen to fit into the distribution. How can it possibly make sense to set an a priori criterion that half the students will have a substandard mastery of the material?
We just had an exam in my biochemistry course where the class average was 62%. There are 115 students in the course and they are all honors students. They needed to have a GPA of >73% just to get into the course.
The Professors did their very best to set an exam that would test for mastery of the material but some of the material was new to the course. 25% of the students got a failing grade.
There are two ways we could handle this situation. We could assume that Professors are perfect human beings and they always make precisely the right judgements about what students should, and should not, know. In this case we would flunk 30 honors students.
Or, we could assume that we goofed and made the exam too hard based on twenty years of experience in grading this course where students generally did much better. In that case, we would add 10-15% to everbody's mark.
What would you recommend? If we take your words at face value then our "courageous" Professors will be arrogant enough to assume they can never make a mistake and lots of lives will be ruined. That ain't gonna happen in my course. I can't wait until you try it in one of your own courses.
Now, let's assume that it's reasonable to raise grades when you recognize that you've set an exam that's too hard. Most students don't have a problem with this concept. What happens when the average grade comes out to be 88%? Does the same reasoning apply or do students somehow feel that a low average is unfair but a high one is merited?
You can probably guess the answer. Last Fall we had an exam where the class average was 81%. In the previous twenty years the class average had always been closer to 70%. Naturally, the students assumed that this year's class was much smarter than every other class had ever been. They deserved to get higher marks.
We pointed out that their grades in all other courses were not significantly different than the grades obtained by students in other years. That didn't seem to matter. (That, by the way, was quite a relavation to us. We're supposed to be teaching critical thinking and these are science students. Where have we gone wrong?)
Anyway, we solved the problem by making the next test harder and that's why the class average was only 62%.
The distributional approach is better for assessing the instruction and testing than the students. Even in large classes, there can be large differences in the population the students are selected from and who select them. Grades being public indications are most appropriate across cohorts where they will be considered, rather than schools, classes, or individuals. Many schools don't even credit grades below B towards degrees which make such comparisons almost impossible though.
Quoth J vos P:
I consciously believe that we are responsible for what we accomplish, that it takes 10,000 hours of concentrated study and practice in any field to achieve minimum professional acapabilities.
Thats most of us buggered then, given that we change jobs more often than every five years. I'm sure that ties into my rant on the decline of British manufacturing somewhere.
Those who will be judging the students need to know at least 3 things:
1) What is their general intelligence?
2) What is their work ethic like?
3) How much do they know about some subject?
De facto IQ tests like the SAT and GRE should be used for #1. Since IQ is normally distributed, these scores can reflect percentile rank.
Grades reflecting mastery would answer #3.
And a separate grade should be given for work ethic. Basically, you norm their mastery to their pre-existing IQ and knowledge base. Say an MIT freshman didn't take calculus in high school. He likely scored above 750 on both the SAT Math and SAT II Math 2C subject test. Accordingly, he should be mastering just about everything in his calc course, assuming he puts in the effort. If he got a B (for mastery), then his effort-based grade would be a C or lower -- he should've gotten an A+.
Conversely, the kid who barely squeaked into MIT with a 600 on the same two SAT tests, he deserves an A for his effort-based grade if he gets a B for his mastery-based grade in calculus. He must have worked his ass off to get the B.
So, that's it: SAT / GRE to measure IQ, mastery-based grades to measure what the student knows, and effort-based grades to measure the student's work ethic.
Trinifar has already commented on Jonathan's statement, but I have my own two cents' worth. I went to school with Jonathan in Pasadena, having transferred to Caltech from a community college. I took Halliday-Resnick physics at my cc, followed by Feynman physics at Tech. I noticed that I did better than most of my classmates because my problem-solving and computation skills had been honed in the cc course. My A in physics at the cc lead to another A in physics at Tech. (And in those days Tech was offering a Halliday-Resnick track as an alternative for those not majoring in physics, but I took the Feynman track anyway.)
I won't claim that my junior college managed to hit the mark in every instance and I definitely believe that Tech's classes were pretty uniformly tough, but "West Kidney Community College" may have nothing to apologize for. I am quite certain that many of the classes at my current college are every bit as good as the equivalent courses offered at CSU or UC campuses.
P.S.: At the risk of taken the edge off my paean to my home town cc, I do recount in To curve or not to curve, my cc physics prof was not the cream of the crop. He thought he could curve a class with three students in it.
I think that the more fundamental question is: what is education for? Is the goal to teach people, or to weed out the talented from the untalented? That sounds like a rhetorical question, but it really isn't. A lot of the tougher classes in many colleges are specifically for the purpose of weeding out those who don't have what it takes to succeed (in medical school, or in engineering, or whatever).
Just for comparison, I've known two other grading systems that differ from the two "Dr. Free-ride" mentioned:
1) In South Africa, grades are given based on % mastery of the SUBJECT matter, not the material in the course. You could master 100% of the subject presented in the course (via lecture, readings, etc.), but still only get a 50% because you only know 50% of "microbiology" or whatever the topic was
2) In Poland, grades are given subjectively by teachers, who will dock letter grades for unrelated things such as misbehavior in class. However, the highest grade any student can get is a "B". To get an "A", a student has to go above and beyond the course work, doing an extra project or something that the other students didn't do, to deserve that special letter.
I'm not advocating either of these, but there is a wealth of different ways to address grading. Personally, I lean towards mastery for assigning grades, and using the distribution of grades to evaluate teaching and compare grades across years.
I grade on mastery, which sometimes surprises the students who are used on being graded on a curve. After the first half of the semester I often have more Cs than if I had used the curve (I work at one of those institutions where the average is a B), but once I point that out to the kids they start working harder. I also try to change letter grades when there is a gap in the numerical scores of the students, as opposed to turning 90.0 into A- and 89.9 into B+.
What I like about grading on mastery is that it encourages students to challenge themselves as opposed to hiding behind the curve to justify their grade. ("I only got a C because the course was full of geniuses the year I took it." or "Nobody bothered studying for the test but since we all did poorly she will be forced to curve.") Grades do matter for the students' job prospects; most companies which recruit in my department have a minimum GPA requirement that students must exceed if they want to apply for the position. I will admit, though, that those thresholds are rarely that difficult to clear. (Put differently, the students whom you feel confident would do well if they were offered a job at that company always have GPAs much higher than the threshold.) For graduate school, I feel recommendation letters count a lot more than GPAs and students considering that career path should invest their time doing research with a professor rather than obsessing about their grades. Some college superstars have a very difficult time in graduate school as they have trained themselves to perform well on clear, well-defined questions with all the facts in front of them - some struggle with the exploration and the inherent vagueness associated with delving into a research topic. In that sense, undergraduate GPAs aren't always good predictors of success in graduate school anyway. At the undergraduate level, it would be interesting to discuss the students' GPAs in the major - what GPA means that the student knows what he is talking about? If all professors in a major use the same system, the grades they give to the same student should balance each other after the kids have taken enough courses.
I feel many of the rules, especially those about curving the grades, are used as a shield to escape the difficulty of evaluating student performance. I can understand why it is done at the freshman level (who wants 300 freshmen storming into one's office contesting a grade? and it is also an excellent way to fight grade inflation), but many of the senior electives have much smaller enrollments and that makes curving more arbitrary. Maybe the solution is to keep curving for underclassmen and non-major courses and grade on mastery for the courses students take in their major.
What you really need to look at is how the two different systems alter success strategies for the students. If you grade on a curve, then the worse my classmates do relative to me, the better grade I will receive. If you grade based on mastery, than my only competition is the test, not my fellows.
Thus in the class graded on a curve, it is against my interests to help my classmates. No lending notes, no study partners, nothing. Indeed were I dishonest then things that actively injure my classmates (taking the reference books needed for research and hiding them in the library after I've finished to take an example that actually happened in an undergrad class I took) would begin to look very appealing.
In the class graded on mastery, anything that helped me pass would be in my best interests, so co-studying with classmates and other social behavior would be encouraged.
Do you want a class where the students are actively set against one another by the system, or one where they can and should work together to learn the material? The choice seems blindingly obvious.
So you throw 2 or 3 tests and maybe one make-or-break paper at the students. How accurately does the final grade reflect anything? Does that fit a normal curve? Should grades have error bounds?
I think what's important is not the grade in the individual course, but the grade point average of an entire academic career. While individual grading methodologies might vary, over the course of 30 to 35 courses with 20-25 different teachers measuring the same student, you can get a fairly good idea of the talent, mastery, and work ethic of an individual student.
I'm a returning student finishing my undergraduate degree after 15 years away, and, in retrospect, I think the thing that is most important for students to learn, and thus, the thing teachers should measure them on, is grit, work ethic, consistency, etc. If someone is talented enough to get 100% of the material right, but only shows up to class 50% of the time and does not engage others, I think the most that student should get is a C.
In the real world, people of average intelligence can make good salaries if they are grinders. People of exceptional intelligence will be fired if they are flaky. Universities should help students understand that fact well before it can affect their income.
David
Grading should be done on mastery. Period. In a large class it will likely fall into a Gaussian distribution anyways, but that's not the point. Basically, take the following few assumptions:
1) You go to school to gain knowledge.
2) Grades are a measure of success in whatever endeavor you undertake.
It directly follows from 1 and 2, then, that your grades should be based on what you know. I do not attend school to have some kind of work ethic forced upon me, I do so to gain a better understanding of whatever the subject of the class is.
If Student A knows 70% of the material, Student A should receive a 70%, period. Nothing else matters, except for what the student's ability. Period.
People severely overestimate the concept of "work ethic". Results are what matters, not "effort".