Wanting to remember something does almost nothing for your memory.
That sounds wrong. It felt wrong to me too, the first time I read the study. But it's one of the oldest, most replicated findings in memory research, and it explains why you can read the same page of docs three times and still open it again next week.
The experiment where intent didn't matter
In 1969, Thomas Hyde and James Jenkins at the University of Minnesota gave people a list of words and asked each group to do a small job with them. Some rated each word as pleasant or unpleasant. Others checked whether each word contained the letter "e".
Here's the twist. Some people in each group were warned there'd be a memory test afterwards. Others had no idea.
The warning barely mattered. People who rated pleasantness recalled about two-thirds of the list whether they'd been told to learn it or not (Hyde & Jenkins, 1969; figures summarised in this SFU lecture). The letter-checkers recalled far less, around 40 percent, and knowing about the test didn't rescue them.
So the thing that predicted memory wasn't effort or intention. It was what their brains did with each word.
Shallow, rhyme, meaning
A few years later, Fergus Craik and Endel Tulving at the University of Toronto turned that into a cleaner test. People saw a word flashed for a fraction of a second and answered one quick yes/no question about it. The question was the only thing that changed.
"Is the word in capital letters?" That's about shape.
"Does it rhyme with rain?" That's about sound.
"Does it fit the sentence He kept the ___ in his room?" That's about meaning.
Nobody was told to memorise anything. Then came a surprise recognition test.
For words that got a "yes", recognition jumped from 15 percent after the capital-letters question to 81 percent after the sentence question (Craik & Tulving, 1975, Experiment 2). Same people, same session, same flash of a word. A fivefold difference, decided by a question that took less than a second to answer.
Craik and Robert Lockhart had already given the idea a name, "levels of processing": the deeper you work with something's meaning, the longer the trace lasts (Craik & Lockhart, 1972).
You can watch it happen in the brain
In 1998, Anthony Wagner and colleagues at Harvard and Massachusetts General Hospital put people in an fMRI scanner while they judged words as abstract or concrete. Later, outside the scanner, they tested which words stuck.
The words people went on to remember confidently had produced more activity, at the moment of reading, in the left prefrontal cortex and in the left parahippocampal and fusiform regions near the hippocampus (Wagner et al., 1998, Science). The difference held even after they matched trials for how long people spent on each word. Spending more time wasn't the whole story; how hard those regions worked was.
That means part of whether you'll remember a line of documentation is decided in the second you read it.
The catch: "deep" depends on the test
This is where it gets more interesting, and a bit less tidy.
In 1977, Donald Morris, John Bransford and Jeffery Franks at Vanderbilt ran the rhyme-versus-meaning comparison again, with one change. Half the people got the usual test. The other half were asked to spot words that rhymed with the ones they'd studied.
On the normal test, meaning won. On the rhyme test, the rhyme group won (Morris, Bransford & Franks, 1977). They called it transfer-appropriate processing. Memory works best when the way you studied matches the way you'll need it.
Critics, Alan Baddeley among them, also pointed out that "depth" was hard to measure on its own, apart from the memory it was supposed to explain (Baddeley, 1978). Fair. The finding held up anyway: process for meaning, and process the way you'll later use it.
For code, that second part matters. If you'll need a flag at 2 a.m. during an incident, you have to practise pulling it out of your head, without reading it from the page.
The deepest question is about you
One more layer. In 1977, Timothy Rogers, Nicholas Kuiper and William Kirker at the University of Calgary added a fourth question to the Craik and Tulving setup: "Does this word describe you?"
Those words were recalled best of all (Rogers, Kuiper & Kirker, 1977). Twenty years later, a meta-analysis of 129 comparisons by Cynthia Symons and Blair Johnson found the self-reference advantage held up, with an average effect of about half a standard deviation (Symons & Johnson, 1997). Their explanation is plain: your sense of self is the most worked-over structure you own, so tying a new fact to it gives the fact a lot of hooks.
What this looks like at a keyboard
Most of us read docs in "capital letters" mode. Eyes on the syntax, scanning for the snippet, copy, paste, move on. That's the letter-checking group. Being keen to learn it doesn't change the outcome.
A few small swaps push the same reading deeper:
Ask one "why" per page. Why does git rebase rewrite hashes when merge doesn't? Answering it, even badly, is a meaning question.
Tie it to your own code. "Where in my repo would useMemo have saved me last month?" That's the self-reference question, aimed at your work.
Practise the way you'll use it. If you'll need it from memory, close the tab and write the command out cold. Then check. That's Morris, Bransford and Franks, applied to a terminal.
Come back to it. Depth gets a memory started. It still fades without a second visit.
None of this takes longer than reading the page twice. It just swaps which question your brain is answering.
The bit I built for myself
I noticed I was great at saving talks and articles and terrible at asking anything about them afterwards. So in Recl, each thing you save gets a short summary and key takeaways, and then a Recall side that comes back later with a question about it. It's a nudge into the meaning question, days after I'd have otherwise forgotten the thing existed.
Next time you open the docs, skip "I need to remember this." Ask one real question about what you're reading. That question does more for your memory than the wanting does.
I'm Asutosh. I write about how the brain learns and remembers, and I build Recl, a tool for remembering what you watch and read.
Sources
- Hyde, T. S., & Jenkins, J. J. (1969). Differential effects of incidental tasks on the organization of recall of a list of highly associated words. Journal of Experimental Psychology, 82(3), 472–481. https://doi.org/10.1037/h0028372
- Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671–684. https://doi.org/10.1016/S0022-5371%2872%2980001-X
- Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104(3), 268–294. https://www.rotman-baycrest.on.ca/files/publicationmodule/@random45f5724eba2f8/JEP75.pdf
- Wagner, A. D., Schacter, D. L., Rotte, M., et al. (1998). Building memories: Remembering and forgetting of verbal experiences as predicted by brain activity. Science, 281, 1188–1191. https://www.science.org/doi/10.1126/science.281.5380.1188
- Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519–533. https://doi.org/10.1016/S0022-5371%2877%2980016-9
- Baddeley, A. D. (1978). The trouble with levels. Psychological Review, 85(3), 139–152. https://doi.org/10.1037/0033-295X.85.3.139
- Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977). Self-reference and the encoding of personal information. Journal of Personality and Social Psychology, 35(9), 677–688. https://doi.org/10.1037/0022-3514.35.9.677
- Symons, C. S., & Johnson, B. T. (1997). The self-reference effect in memory: A meta-analysis. Psychological Bulletin, 121(3), 371–394. https://digitalcommons.lib.uconn.edu/cgi/viewcontent.cgi?article=1008&context=chip_docs


