There are two main approaches to group testing -- criterion based testing and normative testing.
Criterion based testing establishes a series of standards that are meant to measure competence on a given task or in a given field. To be a good driver, you must successfully complete x number of tasks out of a total of x+y tasks. The number of people who successfully complete the overall test is independent of what any particular subject or group of subjects do. Out of ten test-takers, none may pass, ten may pass, or any number in between. Being at an acceptable level is based on the criteria set by the test designers.
Normative testing compares a test taker with other test takers and reports his or her standing based on the performance of the group. If an exam of 100 questions has half the test-takers score at or above 75 correct answers and half below, 75 correct answers is the 50th percentile (50%ile, also known as the median, a certain type of average). But on the next test, if half the test takers score above 40 correct answers and the rest below, on the second test, the 50%ile is 40 correct.
In many districts in the US, schools (and students) have traditionally been ranked based on their percentile scores that are derived from looking at a very large nationwide pool of test-takers.
How many of the schools in your district are above the 50?%ile in reading?
How many of the children in your class are above the 50?%ile in math?
The problem is that, unlike Lake Wobegone, not all of our children are above average. By definition. Unfortunately, in the past people often equated the median score, 50%ile, with grade level.
But if next year, if we work to improve our schools and succeed, and get many more of our students answering many more questions correctly, they have more overall knowledge and skill, but 50% of them are still below the median score! Or, if all teachers did a lousy job and kids learned much less, the median would drop and 50% of all the kids would still rank above the median score. This would result in a constantly shifting grade level. And that would mean that half our students (out of the whole pool) were always below grade level. This is the ultimate grading on the curve. And it often leads to complaints from the broader community that children are being passed through schools but do not seem to have mastered enough fundamental skills.
Criterion testing seeks to establish a base level of success -- if you get a score of x in a given set of challenges, you are considered acceptably proficient. That doesn't mean perfect, or expert. It means acceptable. Such standard setting is hard, but it is objective, at least as to measurement. And it is no guarantee that the criteria or the acceptable level of success was wisely chosen. But it doesn't change based on who is taking the test and how well they do in a given year.
Most schools now tend to use criterion based measures to describe grade level. There is much controversy, especially as changes in testing lead to reports of achievement that seems to slip, but is often just reporting something different.