The Best Way to Remember Things, According to a Forgetting Curve Replication
Updated: 1 day ago
Most advice about remembering things assumes the problem is storage. Repeat it, visualize it, make a mnemonic. But the thing people actually forget is not a fact. It is an intention: call the pharmacy, move the car, send the form. Psychologists call this prospective memory, and it fails in a specific way that the data below makes clear.
What the forgetting curve really looks like
In 1880, Hermann Ebbinghaus measured his own memory for nonsense syllables and found that most forgetting happens almost immediately. In 2015, Jaap Murre and Joeri Dros ran the experiment again under the original conditions, one subject learning and relearning 70 lists over about 70 hours of work, and published every number in PLOS ONE.
Their measure is "savings": how much less time it takes to relearn a list than it took the first time. If you still remember most of it, relearning is fast and savings are high. If it is gone, savings go toward zero. Here are the savings scores at each delay, replication first, then Ebbinghaus's original from the same paper's Table 3:
20 minutes: 0.472 (Ebbinghaus 0.582)
1 hour: 0.373 (0.442)
9 hours: 0.276 (0.358)
1 day: 0.317 (0.337)
2 days: 0.230 (0.278)
6 days: 0.168 (0.254)
31 days: 0.041 (0.211)
How fast the first hour is
I took the 20-minute score as the starting point and worked out how much of it was gone at each later delay, then compared the rate of loss per hour early versus late. This is simple arithmetic on the published table, not a new experiment, but the papers do not present it this way.
From 20 minutes to 1 hour, the replication lost 21 percent of its savings. Ebbinghaus lost 24 percent. That is a quarter of what you could still recall, gone in 40 minutes.
By one day, the replication was down 33 percent from the 20-minute mark. Ebbinghaus was down 42 percent.
By 31 days, the replication had lost 91 percent. Ebbinghaus had lost 64 percent.
Rate of loss in the first 40 minutes: about 0.15 savings points per hour in the replication, 0.21 for Ebbinghaus.
Rate of loss between day 1 and day 31: about 0.0004 points per hour in the replication, 0.00018 for Ebbinghaus.
Put those two rates side by side and the early period is roughly 400 times faster in the replication and over 1,000 times faster in the original. Both curves also show a small bump at 24 hours, which the authors attribute to sleep. Whatever you still have the next morning, you tend to keep for a while.
The practical lesson is that a memory decays most in the window right after you form it. If you think "I should book the dentist" while making coffee and do nothing with that thought for an hour, the arithmetic above says a good chunk of it is already going.
Forgetting is the main reason intentions fail
This matters for intentions specifically. In a chapter summarizing their 2006 meta-analysis of 94 studies, Peter Gollwitzer and Paschal Sheeran describe the first obstacle to acting on a goal as simply remembering to act, and cite a study in which 70 percent of people who intended to do a health check but did not gave "forgetting" as the reason.
Their meta-analysis tested a fix. An "implementation intention" is an if-then plan: "If it is Tuesday at 7pm, then I will call mom." It attaches the action to a concrete cue, so the cue does the remembering instead of you. Across the 94 studies, forming one had a medium-to-large effect on goal attainment, d = 0.65. For the specific problem of getting started, which includes remembering to act, the effect was d = 0.61. These are standardized effect sizes, not percentages, but 0.65 is large for a behavioral intervention that costs nothing but a sentence.
Putting the two findings together
The forgetting curve tells you when memory is fragile: immediately. The implementation intention literature tells you what form a durable intention takes: a specific when and where, paired with the action. So the best way to remember things is not a better memory technique. It is to move the intention out of your head, in if-then form, inside the window where it is still intact.
In practice that means three things.
Capture the intention at the moment you have it, not later. The first hour is where the loss is steepest.
Write the cue, not just the task. "Dentist" is a goal intention. "Dentist Thursday 9am" is an implementation intention.
Make the cue come to you. A plan written in a notebook still depends on you remembering the notebook. A plan that fires on its own does not.
A text reminder is an implementation intention with the "then" delivered. You text "book dentist tomorrow at 9am" to Reminder Textapp during the coffee, while the thought is at full strength, and the sentence you wrote arrives on your lock screen at the moment you specified. The capture takes less time than the first ten minutes of forgetting.
What this does not fix
The forgetting data comes from two single-subject experiments, Ebbinghaus on himself and Dros on himself. That is the design the replication set out to repeat, and the two curves agree closely, but it is still two people. The implementation intention effect comes from many studies, though the authors note it is smaller when people have no real commitment to the goal. A reminder for something you do not want to do will arrive, and you will still have to do it.
What the evidence does support is narrow and useful: capture early, specify when, and let something other than your memory hold the plan.
Try it by text
Text Hi to (720) 315-7147 to start Reminder Textapp. Then text any reminder in plain English, like "move the car Tuesday at 8am" or "call the pharmacy every Monday at 9". You get a confirmation with the time it understood, and a text when it is time. It works on any phone that can text, with no app to install and no account. Your first 20 messages are free. Details at textapp.co/apps/reminder.
Comments