What does an A mean?

As a former college instructor, I’ve had this discussion with students before and it’s something to think about. I gave a test one year and the highest numeric grade was 87%, with many grades at or around 75% - 65% and some lower. As one would assume, the percent represents the fraction of questions that were answered correctly, with partial credit given on some answers. When discussing the overall grades with my students, they asked for a ‘curve’. The infamous ‘curve’. I asked if they knew what a curve was. The best answer was along the lines of, ‘some of us should get an A on this test’. Reasonable answer. They said I should look for ‘breaks’ in the scores and give letter grades for scores within certain breaks. That wasn’t a bad idea. So if there were clusters at 87% - 82%, then at 77% - 74%, then at 71% - 55%, those groupings could correspond to the A group, the B group and the C group respectively. I asked where we should make the first break: how many students should be in the top group and get an A on the exam. There were many responses to this, with most students thinking that about 40% to 60% of the class should get an A. The students liked the idea of assigning grades based on where their test scores ranked in relation to others in the class. What makes that mythical 90% so special to receive an A? For this test, grades could be distributed based on “class rank”. The best test scores would correlate to an A, the second group a B and the lowest test scores could get assigned a C grade. There were very few complaints. 

Weeks later, the same class performed much better on the next test, with many grades above 90% and the lowest around 75%. The test wasn’t necessarily easier; the students were simply better prepared. Overall, the class was much happier when I told them the scores. I then suggested we be consistent and apply a curve to this test as well. We could make clusters of grades as follows: 100% - 95% for an A, then 93% - 88% for a B and 86% - 75% would receive a C grade. This still gave about 40% of the students an A on the test, similar to the first exam. Silence followed, then outrage. Nobody had thought that curves work both ways. 

Should teachers provide letter grades based on the long-standing idea that 100% - 90% is and A, 89% - 80% is a B, etc., realizing that there will be times when nobody will receive an A grade for a test (or perhaps a semester) while other times perhaps everyone will get an A? Or should the top students (say 40% - 50%) receive an A because they have the best grades in the class?