When to Curve Grades: A Teacher’s Guide to the Five Methods

The signs a test actually needs a curve

Curve the test, not the students. The legitimate triggers are about the instrument: a question that was ambiguous or mis-taught, a test pitched far above the level it replaced, or a disruption that cost the whole class time. The numbers show it too.

• The top score is far below the maximum (a best score of 72 out of 100 is the classic case).
• The class mean sits well below the same cohort’s usual mean on comparable tests.
• One question was missed by nearly everyone, including the strongest students.
• Scores bunch tightly: the test separated nobody, so small raw gaps are deciding letters.

The five methods, on the same scores

Take five raw scores — 55, 62, 70, 78, 85 — with a class mean of 70. Flat +5 gives 60, 67, 75, 83, 90: everyone gains equally and gaps are untouched. A bell curve set to an average of 75 and a spread of 10 gives 61, 67.6, 75, 82.4, 89: the order is preserved, but the class is rebuilt onto the target distribution, and notice that the strongest score can come down as well as go up.

Top-score-to-100 would add 15 to everyone here (the best raw score is 85), the square-root curve lifts the bottom most (55 becomes about 74), and the target-average method adds exactly the gap between the current mean and your target. Same class, five different gradebooks: the method is the policy decision, not the arithmetic.

When not to curve

Do not curve to reach a quota of As, to match another section’s distribution after the fact, or because individual students lobbied. Do not curve a test that did its job: a mean of 78 with a sensible spread is a working instrument, and moving it teaches students that preparation matters less than pressure. And never curve selectively — a rule that applies to some students is not a curve, it is a favour.

Fairness checks before you post curved grades

Run these four checks on the before-and-after table (the grade curve calculator prints one). No student’s rank should change unless the method says so, as with the square-root curve’s narrowing gaps. The cap decision (at 100, or uncapped) should be made before you see names. Students near a letter boundary should be explainable in one sentence. Finally, announce the method with the grades: “the top score became 100 and everyone gained 15” ends arguments that “grades were adjusted” starts.