Cobra Effect · Learning and memory
The generation effect
Trying to produce an answer first makes it stick, even when you are wrong.
6 cards, read aloud in 2:13, with a test and sources.
Two groups get the same word pairs. One group gets half of each pair.
One group simply reads pairs like rapid fast. The other reads rapid f, and has to produce the missing word itself. Same pairs, same time, one letter of difference in the task. The difference in what they remember is not small.
Norman Slamecka and Peter Graf, 1978.
They ran the comparison several ways and the pattern held every time. The people who produced the word remembered more of them than the people who read it. They called it the generation effect, and it has been replicated for decades since.
Making the answer costs something, and that cost is the point.
Reading is receiving. Producing is searching, and the search leaves a trail behind it. The word you built has your own route back to it. The word you read has none. Effort at the moment of learning buys retrieval later.
The odd part is that a wrong guess works too.
Nate Kornell and colleagues had people guess an answer before being told it, on pairs they could not possibly know. The guesses were wrong. Almost all of them. Those pairs were remembered better than pairs that were simply studied. The failed reach prepares the ground for the answer.
Which makes most teaching backwards.
The clean explanation first and the problems afterwards is comfortable and weak. A minute of guessing beforehand changes what the explanation lands on. Even a bad guess cuts the slot that the right answer drops into.
So guess first, every time, especially when you are sure you cannot.
Before you read the answer, say what you think it is. Out loud or on paper. Before the meeting reveals the number, write down the number you expect. Being wrong is not the cost of this. Being wrong is most of the benefit.
Sources
- The Generation Effect: Delineation of a Phenomenon, Norman J. Slamecka and Peter Graf, 1978. The original experiments, in the Journal of Experimental Psychology. Search the title to find the paper. This page summarises the finding and the work that followed it.
- Understanding How We Learn, Yana Weinstein, Megan Sumeracki and Oliver Caviglioli, 2018. A short illustrated tour of what the research says about memory, by the people behind the Learning Scientists project. This is their site.
- Retrieval Practice, Pooja K. Agarwal and colleagues. A free library of guides on producing answers rather than reviewing them, written for teachers and full of examples you can steal.
Nearby ideas
- The testing effect. Pulling an answer from memory teaches more than reading it again.
- Desirable difficulties. The practice that feels hardest is often the practice that sticks.
- The spacing effect. Spread your study out over time and it lasts far longer.
- The learning styles myth. Matching lessons to a preferred style does not improve learning.
- Sleep and consolidation. Sleep is when the day’s learning gets filed away.
- Interleaving. Mixing problem types feels worse and teaches better.
- The Feynman technique. If you cannot explain it simply, you have found a gap.
- Chunking. Experts see patterns where beginners see pieces.