Worked examples are the most misused study resource: they're meant to be faded, and they're mostly copied. The student writes out the complete solution, watches each line appear, and ends with a page that looks like understanding. It isn't — it's transcription, the worked example passed through a pen without passing through the brain. The solution is on the page; the problem-solving is still in the book.

The fade is the fix. Cover the solution after the setup — the given, the diagram, the first line — and generate the next step yourself. Then uncover one line and compare. Match means the step was right; mismatch means you name the illegal move you almost made — the wrong formula, the skipped assumption, the sign error the worked example would have hidden. Repeat line by line until the example is uncovered and you produced it, then do the same on a cousin problem with one assumption changed. That's using the example — the copy was just having it.

Why copying the solution fails

Copying the completed solution is the worked example's version of rereading — it exposes you to the answer without making you produce it. Each line is written while visible, so the writing is transcription: your hand moves, your eyes track, and nothing is retrieved. The finished page looks like work; the skill it represents is penmanship.

The fade shifts the work onto you — the mechanism worked examples are for (Renkl's fading, Sweller's cognitive load). Each covered line is a retrieval: you generate the next step from the setup, not from the visible answer. The uncover is the feedback — one line at a time, so you know exactly where your version diverged from the correct one. Generating the next step is retrieval; copying the finished one is transcription — and the exam tests the generation.

The fade workflow

Cover everything after the setup. The given, the diagram, the problem statement — visible. Everything below — the equations, the steps, the answer — covered. You generate from the setup, not from the solution's momentum. The cover is the retrieval condition; a partial cover gives you the answer's scaffolding back.

Generate the next step before you look. Write it — the equation, the move, the transformation — before uncovering. The attempt is the learning; the check is the feedback. Generating first means your version exists to compare; looking first means there was never a version, just a copy.

Uncover one line at a time. Not the whole solution — one line, the next step. Compare your generation to it: match means keep going; mismatch means stop. The line-by-line reveal keeps the feedback local — you know exactly which step diverged, not just that the answers differ.

Name the illegal move on a mismatch. "I got it wrong" isn't enough — name what you almost did: "I wrote T = mg, which assumes no acceleration." The named error is the learning — it's the move the worked example would have silently corrected, now explicit so you don't make it again.

End on a cousin, not the same problem. Once the example is produced, do a variant — one assumption changed, one number different, one step removed. The cousin is the transfer test: if the method works on a new surface, you learned it; if it only works on the identical problem, you memorized it.

A 40-minute session on the clock

The weak answer is the illegal move the fade exists to catch — T = mg is the static intuition, the formula that looks right because it's familiar. The strong answer names why it's wrong: T = mg implies a = 0, and the Atwood machine exists to be accelerating. The fade forced the equation choice — the copy would have written the correct equations while never noticing the tempting wrong one.

Why the copy feels like work

A copied solution looks identical to a produced one — same equations, same steps, same answer on the page. The difference is invisible because the page is the product and the process was the learning. The copy's process was reading-and-writing; the fade's process was generate-and-check. Only one of those builds the skill the exam tests.

This is why the line-for-line match is the tell: if your paper duplicates the solution, the solution did the work. The fade's paper has your version first — sometimes right, sometimes wrong, always generated — and the uncover marks where it diverged. That divergence record is the learning; the clean copy is the illusion of it.