What does grading on a curve actually do?
Grading on a curve means assigning grades according to a distribution of the cohort's performance, rather than against fixed criteria. It converts an absolute measure into a relative one.
The strict version, sometimes called norm-referenced grading, allocates fixed proportions of each grade — the top 10% get an A regardless of their actual scores, the bottom 10% fail regardless of theirs. Looser versions simply scale marks upward if a paper proved harder than intended.
What it actually does:
It measures rank, not attainment. A student's grade depends on who else sat the exam. The same performance in a strong cohort and a weak one produces different grades. That is the intended feature for selection purposes and the core objection for everything else.
It compensates for unpredictable exam difficulty. This is the strongest argument for it. Setting an exam of precisely calibrated difficulty is genuinely hard, and without adjustment a harsh paper punishes a cohort for the examiner's misjudgement.
It caps the number of good grades. In the strict form, not everyone can do well even if everyone learned the material — which is why it is widely criticised as pedagogically perverse.
It creates competition rather than cooperation. If grades are zero-sum, helping a classmate lowers your own position. Research and long-standing complaint in competitive programmes both point at this.
The alternative is criterion-referenced grading, where grades are defined by what a student can demonstrate against published standards. Everyone can, in principle, achieve the top grade. This is the direction most school qualifications have moved.
In practice many systems are hybrid. UK GCSE and A-level grade boundaries are set after marking, using statistical information about the cohort alongside examiner judgement about the standard of work — which is neither purely norm-referenced nor purely criterion-referenced.