In New York City in the '60s all grades were numbers. The lower grades were by 5s, then there was an 88, and above 90 the increment was 1. MIT had a 5 as A, other places I’ve been had 4s. I don’t think I’ve ever been anywhere that used letters.
I believe every class I ever took at MIT graded on a curve - I don’t remember any of this preordained number stuff. Some professors gave really hard tests. After the test, the class average was announced, and grades depended on where you stood relative to that.
At Illinois I TAed for a very large second CS class, that was used to filter out the kids who had it from those who didn’t. We got a lot of drops. We graded on a curve but a censored one, since most kids who were going to occupy the bottom part of it were gone. (Some didn’t get the message.) We summed up the grades at the end of the term, and looked for natural breakpoints to assign the grades. We usually saw clusters, not nice bell curves.
When I moved down south, I taught for one term. In this school, pretty much everyone used the 90% = A formula. I didn’t, because, to put it nicely, the students weren’t quite the level I had experience with, and I didn’t like giving tests that were all multiple choice. I discovered the absolute system when I gave my first test, and everyone freaked out about getting grades that would be Ds in most classes. They were very relieved about a curve there.
But this term did convince me to go into industry.
Technically, you normalize the data, and the median score is a C. Then your students kill you for giving out so many Ds.
In common usage a curve is anytime the scores are adjusted based on the performance of the class. Making the top score 100 is one way to do it. Looking at groupings is another. (I had one prof who would tell you to keep tallying up your scores, after each test he’d write everyone’s cumulative point totals on the board (without names, of course) and say “if I were grading today, these would be As, these Bs…”). Giving X # of As is yet another (I had a great ‘Science for English majors’ type of course - 25% of the class got As, 50% Bs, 25% Cs - you could only fail the course if your test results were worse than guessing - but he’d let you retake the test - they were multiple choice 4 answer tests, you had to get 30% to pass - with a C)
I had one prof who used a 90/80/70 scale - but threw out test questions that had a high rate of people getting them wrong. He figured if half the class missed the question, it was a bad question or he did a bad job of covering the material.
I think the best system depends on the material - and the experience of the instructor.
I agree. If everyone in the class really really knows their stuff—which ought to be the teacher’s goal, though whether it’s realizable will depend on the ability and motivation of the students—then everyone should get A’s. If everyone in the class is a moronic slacker who refuses to do or learn much of anything, the whole class should get F’s.
The argument for grading on a curve is that it is a way of making up for tests/standards that are too hard or too easy (or perhaps poorly designed). A student’s raw score is going to depend, not only on how much that student knows, but also how easy the test was. The solution to this, of course, is to give well-designed, well-calibrated tests and other means of evaluation.
When I was in grade school in the CPS in the 60s-70s, grades on report cards were:
E-excellent
G-good
F-fair
U-unsatisfactory.
Letter grades were qualified with “effort” ratings of 1, 2, 3, meaning something like:
1-excellent effort
2-could try harder
3-unsatisfactory effort.
I always used to get a kick out of E2 grades - excellent work, but could do better! 
Recently got a midterm “progress report” for my HS senior in the western burbs. They gave both a letter and a number grade - such as B-85.
I asked my kid about the “adding points” system, which she said is the only one her teachers have ever used if a class scores below the 90/80/70/60 scale. Says they take the median score, subtract it from 75, and add that many points to everyone’s grade. Seems weird to me - a slavish loyalty to the 90/80/70/60 scale, but that’s what she says they do. When questioned further she said she remembered “one or two” occasions where they added a number equal to the difference between 100 and the highest grade.
Another data point - in this HS they used to use a 94/87/78 scale for honors classes, but just this year or the last they went to 90/80/70 for all classes. Not sure why.
The only grading situation I remember clearly was in college, into to accounting. An A required 94. I had a 93.6. The SOB gave me a B, applying a “Nellie Fox” rule - no rounding up allowed.
I showed THEM! Next semester in the 2d accounting class, I got a solid C! The next semester I transferred the hell out of business! 
That is what a curve is, technically. That said, I’ve never met a teacher, in high school or college, who used a normal distribution curve to determine grades. The term “curve” has evolved to colloquially mean any sort of method used to adjust grades. I’ve had teachers use the method where the highest score gets 100, the rest of the students get bumped up by the difference between the highest score and 100. I’ve had other teachers use some mysterious method in which “curve busters” could score well above 100. (I had one series of biology tests in high school where I scored in the 60s [a solid F] on one test, and then in the 120s the next three tests.) Then there were professors (in college mostly) who would look for clumpings in the grading, and then announce in class what the cutoffs for each letter grade was (one chem class I remember a 40% being a B). But they didn’t have a blanket “add x amount of points” to your score policy. That one most closely resembled a classical curve, but the distribution was more like 20% A, 40% B, 30% C, 5% D, 5% F. I assume this was based on how the scores clumped.
I should add, we always had letter grades from grammar school to college (1980-1998). Well, actually, from kindergarten through second grade we only had three grades: + (excels), O (satisfactory), and checkmark (needs improvement.) Until college, 93-100 was an A, 84-92 B, 75-83 C, 70-74 D, 0-69 F (or “U,” unsatisfactory, in grammar school.) In college, grades were much more fluid, but if teachers had numerical scales, they would generally be graded in 10s (90-100 A, 80-89 B, etc.)
I think the phrase is nigh meaningless. I hate when students ask if I will use a curve, because I always ask what they mean by “curve” and they never seem to know.
This is the only reasonable way to grade.
This only works if the tests are standardized. Otherwise, each test will be difficult in different ways, and the best way to figure out if I made a test too hard or too easy is to look at how my students did on it. Also, grading is always a judgment call. For example, I am now teaching an upper-level Maths course with 12 people. Some don’t do any homework. They will fail. Others sort of understand the material. Some understand some of the material pretty well, but fuck up some of the other shit. And some understand most of it. Who gets what grade? 12 is too small to look at any kind of average, of course. I also know that the bar is set a bit lower for a course this hard. Understanding half of the important concepts is probably worth a B-? (Maybe a B?) How does one quantify “understanding half of the important concepts?” It’s not as easy as you think.
I have no idea how Humanities profs do it. At least I have numbers that I can pretend mean things.
+1 to Manda JO. Especially about the “dynamic assessment” bit. Consider even that I will base the content of the second exam in some part on how he students did on the first! Also to the bit about 90% being meaningless. If I make my tests hard, then maybe an A student should get 65% or so. If I make them easy then maybe an A student should get 95%. Also, I don’t neccesarily grade exactly the same each time. Way too many variables.
I think a curve is anything that deliberately alters student’s grades based on total class performance or the “accepted norms” of the system (usually 90-100 A 80-89 B etc), for worse or better. And I’ve heard some horror stories about curves, the worst type are what I think amounts to a bell curve, Engineering classes are the most notorious here. My friend did tell me about the Curve From Hell™ in her 7th grade math though, everybody scored over 95%s consistently, the only people who ever got A’s were the ones who got 100%s or 99.5%s, a “D” on a particular test was often around a 97-95 depending on the difficulty of the material. I heard about one quiz where everyone got 100% so EVERYONE got a “C”, meaning 32 students pretty much got 30% arbitrarily shaved off their score.
But sometimes curves are just a way of acknowledging you may have written a bad test, or that maybe something was taught badly. I have seen 10 points added to everyone’s score, I actually got extra credit on a humanities test because I very nearly 100%d it (about 2 points off) but most people failed or got close (fault of the kids, not the teacher in this case I can assure you, but the general gist was she thought it was partially her fault and didn’t want to trash any GPAs).
I think the most non-evil use of curves is probably weighting by the relative importance of the material, I’ve rarely seen this done, but I’ve known (but never had classes with) teachers that would widen the allowed area for point for an A (meaning in the gradebook it would get converted to somewhere in the 90s) for particularly regurgitative things they admit they themselves usually look up (like Trig Identities was a big one, “you should know a few basic ones, but even when I’m doing stuff if the odd chance comes around that I need it for a paper I’ll look it up”), or on occasion tighten the ranges for really, really important concepts to the material.
For example, if a teacher applies a scale where 90-100%=A, 80-89%=B, 70-79%=C, 60-69%=D, and 59 and below=F - did the teacher apply a curve?
That’s generally the definition of NOT grading on a curve. Any other method that somehow moves the average grade (whatever that is) towards the mean or median score is a form of grading on a curve. AFAIK there is no uniform way in which this “curving” process is done. For instance, the average grade may not be considered a C and there are not necessarily hard cut-offs for the various grades. It’s because of this arbitrariness in grading on a curve that some instructors just use the 90-100=A, 80-89=B, etc. approach for grading in order to avoid grade disputes.