让阿忆以为自己都会了的，是“眼熟”。有实验把单词印成大字和小字，大家都觉得大字的更好记，真考起来几乎没有差别，事先提醒过也照样这么觉得。2009 年一份调查里，177 个大学生八成多靠重读复习，把合上书考自己当头号办法的，只有 2 个。

1879 年起，赫尔曼·艾宾浩斯拿自己做实验：背没有意思的音节表，隔一段时间再背同一张，看能比第一次省多少时间。二十分钟后省 58%，一天后 34%，一个月后还有 21%。2015 年有人照原样重做，一个月那一点低得多（4% 左右），但也不是零。忘了，不等于没了。

加州大学洛杉矶分校的 Robert Bjork 和 Elizabeth Bjork 是一对夫妻，结婚二十三年后才第一次合写论文。1992 年，他们提出：每样记住的东西都有两个数。提取强度是现在想不想得起来，储存强度是在脑子里扎得多深，理论假定后者只攒不丢；“我记住了”的感觉，只量得到前一个。

这个理论最反直觉的一条：每想起一次，储存强度都会涨，而当时越难想起来，涨得越多；自己想起来，也比照着再看一遍涨得多。前提是最后想起来了，或者实在想不起来时马上看到了答案。Robert Bjork 在 1994 年给这类“学的时候费劲、记得更牢”的安排起了个名字，叫 desirable difficulties，可以译成“值得的难”。

2006 年，Henry Roediger 和 Jeffrey Karpicke 让大学生读一篇短文：一组读四遍，一组读一遍、再合上书写三次。五分钟后读四遍的领先，83% 对 71%；一周后倒过来，61% 对 40%。读四遍的那组，反而对自己更有信心。

越有把握的错，看过正确答案以后反而记得最牢（Butterfield 和 Metcalfe，2001）。先猜也有用：先猜一个答案、猜错了再看，比直接看答案记得牢，哪怕题目是编出来的、根本猜不对；不过猜的东西得和答案沾点边，毫不相干的瞎猜帮不上忙。长城那题：杨利伟 2003 年从太空回来就说过，他没看到长城。

艾宾浩斯 1885 年就发现，38 遍摊在三天里，顶得上前一天一口气背 68 遍。1993 年，心理学家 Harry Bahrick 发表了一项九年的研究，四位受试者据报道就是他们一家四口：隔八周复习 13 次，和隔两周复习 26 次记得差不多。隔多久合适？一项一千三百多人的实验给的粗略说法是：想记一周，隔一两天复习；想记一年，隔三四个星期。

2008 年，Nate Kornell 和 Robert Bjork 让人认十二位画家的画风，混着看的人认得更准，可多数人还是觉得一位一位看好；2014 年有人找 123 人重做，结果一样，认对 75% 对 66%，而 71% 的人站错了队。2020 年佛罗里达 787 名七年级学生的课堂试验里，多做打乱题的一组一个月后得分 61%，老办法 38%。不过混着学不是万能的：分辨长得像的东西最有用，背一串单词，反而分开背更好。

最后，把五种学法画成替身面板。重读和划重点，当时舒服、看着眼熟、当场见效，偏偏记不牢；考自己、隔开、混着练，正好反过来。在学习这件事上，感觉总是站在另一边。

最后一条有个名字，叫“执行意图”。Peter Gollwitzer 和 Paschal Sheeran 汇总过 94 项研究、八千多人：写成“如果……就……”的计划，比只下决心更容易做到。

做这些事的时候，你会难受，会发现自己忘了好多。这种难受是对的。它不说明你笨，它说明你在长。

艾宾浩斯的曲线常被拿来吓人：看，忘得多快。我更愿意记住它的另一半：它没有掉到零。你学过的东西，比你以为的留下得多，等的只是一个线索。

所以下次卡在那儿、怎么也想不起来的时候，别骂自己。深呼吸，再想几秒。

“想不起来的那几秒，
不是白费，
是记忆在长。

What convinced Ayi that it was all learned was the feeling that it “looks familiar.” One experiment printed words in large or small type. People felt the large words would be easier to remember, but on the real test there was almost no difference, and they kept feeling that way even when warned in advance. In a 2009 survey of 177 college students, more than 80% studied by rereading, and only 2 named closing the book and testing themselves as their top strategy.

Starting in 1879, Hermann Ebbinghaus experimented on himself. He memorized lists of nonsense syllables, memorized the same list again after a while, and measured how much time he saved compared with the first time. After twenty minutes he saved 58%; after a day, 34%; after a month, still 21%. A 2015 replication followed the original procedure. Its one-month figure was much lower (about 4%), but it was not zero either. Forgotten does not mean gone.

Robert and Elizabeth Bjork of UCLA are husband and wife, and they wrote their first paper together twenty-three years into their marriage. In 1992 they proposed that everything you remember has two numbers. Retrieval strength is whether you can recall it right now; storage strength is how deeply it is rooted in your mind, and the theory assumes this one only builds up and is never lost. The feeling “I’ve got it” measures only the first.

The theory’s most counterintuitive claim: every time you recall something, its storage strength goes up, and the harder it was to recall at the time, the more it goes up. Recalling it yourself also raises it more than reading it over again. This holds as long as you do recall it in the end, or see the answer right away when you really can’t. In 1994 Robert Bjork gave such arrangements, where learning takes more effort and the memory holds better, a name: desirable difficulties.

In 2006 Henry Roediger and Jeffrey Karpicke had college students study a short passage. One group read it four times; the other read it once, then closed the book and wrote down what they could recall, three times. After five minutes the four-reads group was ahead, 83% to 71%. After a week it was the other way around, 61% to 40%. And it was the four-reads group that felt more confident.

The more confident you were in a wrong answer, the better you remember the right one once you have seen it (Butterfield and Metcalfe, 2001). Guessing first helps too. Guess an answer, get it wrong, then look, and you remember better than if you had simply been shown the answer, even when the questions are made up and impossible to guess correctly. But the guess has to be somewhere near the answer; a wild guess that has nothing to do with it does not help. As for the Great Wall: when Yang Liwei came back from space in 2003, he said right away that he had not seen it.

Ebbinghaus found as early as 1885 that 38 repetitions spread over three days did as much as 68 in one go the day before. In 1993 the psychologist Harry Bahrick published a nine-year study whose four subjects were reportedly his own family of four: 13 reviews eight weeks apart worked about as well as 26 reviews two weeks apart. How far apart should reviews be? An experiment with more than 1,300 people gives a rough rule: to remember something for a week, leave a day or two before reviewing it; to remember it for a year, leave three or four weeks.

In 2008 Nate Kornell and Robert Bjork had people learn the styles of twelve painters. Those who saw the paintings mixed together identified the styles more accurately, yet most people still felt one painter at a time worked better. In 2014 a replication with 123 people found the same thing: 75% correct versus 66%, and 71% of people backed the wrong side. In a 2020 classroom trial with 787 seventh graders in Florida, the group that did more shuffled problems scored 61% a month later, against 38% for the old way. Mixing is not a cure-all, though. It helps most with telling apart things that look alike; for memorizing a list of words, learning them separately actually works better.

Finally, here are five ways of studying drawn as Stand stat cards. Rereading and highlighting feel comfortable at the time, look familiar and pay off on the spot, yet they don’t stick. Self-testing, spaced review and interleaved practice are the exact opposite. When it comes to learning, your feelings always take the other side.

The last one has a name: “implementation intentions.” Peter Gollwitzer and Paschal Sheeran pooled 94 studies covering more than 8,000 people: plans written as “if…, then…” are carried out more often than plain resolutions.

Doing these things will feel uncomfortable, and you will find you have forgotten a lot. That discomfort is as it should be. It doesn’t mean you are stupid. It means you are growing.

Ebbinghaus’s curve is often used to scare people: look how fast we forget. I would rather remember the other half of it: it never drops to zero. More of what you have learned stays with you than you think. It is only waiting for a cue.

So the next time you are stuck and just can’t remember, don’t scold yourself. Take a deep breath and think for a few more seconds.

“Those seconds when you can’t remember
are not wasted.
They are your memory growing.