主动回想 间隔复习 提取练习 遗忘曲线
复习这件事经常被当成一个动作：把笔记打开，从头看一遍，合上，结束。分数却说明这个动作效率很低，同一章节反复看过三遍的学生，成绩往往不如只看了两遍、但第二遍合上书自己默写要点的学生。学习科学把有效复习拆成两个独立维度：主动回想（active recall，文献里更常写作 retrieval practice）管一次复习内部的动作，间隔复习（spaced repetition）管两次复习之间的时间。前者针对"记不牢"，后者针对"记不久"，机制不同，替换不了对方。很多学生以为这是两种可以二选一的方法，或者在考试前同时用上就算加倍努力，实际决定效果的是两者的顺序与配比。下文按考试场景把两个维度分开讲，最后给出一周可以直接照做的排期。
Revision is usually treated as one action: open the notes, read them through, close the folder, done; Marks tell a different story, since a student who reads a chapter twice and then closes the book to write down the key points from memory usually beats one who reads the same chapter three times. Learning research splits effective revision into two independent dimensions; active recall, or retrieval practice, governs what happens inside a single revision session, while spaced repetition governs the gap between sessions; One addresses weak retention, the other short-lived retention, and neither substitutes for the other. Many students treat them as alternatives or assume that doing both the night before an exam counts as double effort, when the order and the mix are what move the mark.
复习的动作与复习的间隔是两个独立变量
把两个维度交叉起来看，复习方式大致落在四种组合里。重读加集中复习是最常见的做法，也是收益最低的一种，它只制造熟悉感；重读加间隔复习好一些，至少让内容见过多次；合书回想加集中复习能把单次记忆压深，但遗忘来得快；合书回想加间隔复习才是文献里反复验证的组合。判断自己属于哪一种很简单，问一句"这次复习有没有出现过想不起来、然后硬想的过程"，答有就是主动回想，答没有就只是重读。理科和人文学科在这个框架里位置不同：化学的有机反应机理、物理的公式推导需要动手写出来才算回想，历史的时间线、英语文学的人物动机用口头复述或写提纲就够。很多学生以为科目不同所以方法不能用同一套，实际情况是同一个框架，只是回想的形式要跟着科目换。
Cross the two dimensions and revision habits fall into four combinations; Rereading in one long block is the most common and the least productive, because it only produces familiarity; rereading across spaced sessions is slightly better, since the material is met more than once; closed-book recall crammed into one block deepens a single session but fades quickly, and closed-book recall spread over days is the combination that the research keeps confirming. It is easy to see which one you are doing: ask whether the session contained a moment of failing to remember and then pushing through, and if it did not, you were rereading; The same framework covers different subjects, with the form of recall changing rather than the method. Organic reaction mechanisms and formula derivations in physics have to be written out by hand, while timelines in history and character motivation in English Literature can be recalled out loud or sketched as a plan.
提取练习：为什么合上书回忆比重读更能提分
Roediger 与 Karpicke 在 2006 年发表于 Psychological Science 的研究把学生分成两组读同一段材料，一组反复阅读，另一组读完后做一次回忆式测验。短期测验里重读组略占优势，一周后的延迟测验里回忆组明显更好，而且重读组对自己的把握反而更高。解释不复杂：把内容读进去用的是识别，回忆用的是重建，重建过程中大脑必须自己把线索拼起来，这个动作本身就是练习；想不起来的时候再看一眼笔记，得到的是即时纠错，比重读时一路顺畅地滑过去有效得多。考试卷面考的是重建：AQA 与 Edexcel 的官方 spec 都把命令词（command words）单列一章，describe 只要求陈述，explain 与 justify 要求因果与判断，学生最常把这两类题当成同一种准备，结果 explain 的题目写得像 describe，分数自然上不去。错得最多的是计算题的中间步骤和论述题的分论点，这两处恰好都是靠重读补不上的。
A 2006 study by Roediger and Karpicke in Psychological Science divided students reading the same passage into a rereading group and a group tested once from memory; The rereading group did slightly better on an immediate test, while the recall group was clearly ahead on a delayed test a week later, and the rereading group was also more confident about material it had already forgotten. The mechanism is straightforward; reading uses recognition, recall uses reconstruction, and rebuilding the links from scratch is itself the practice, while a failed recall followed by a quick check of the notes fixes an error that smooth rereading slides past. Exam papers test reconstruction, which is why AQA and Edexcel both devote a section of their official specifications to command words, and why answers to explain and justify questions read like description when they have only been reread.
间隔效应与遗忘曲线：复习该隔多久
Ebbinghaus 在 1885 年用无意义音节做的自我实验画出了最早的一条保持曲线，遗忘先快后慢的形态此后被反复复现。Cepeda 等人在 2006 年的元分析（Psychological Bulletin）汇总了上百个实验，结论是同样的总复习时间，分散到多次的长期保持优于集中在一次，这条结论在各个学科和年龄段都成立。间隔取多长有经验刻度：Cepeda 等人在 2008 年的后续研究里给出的做法是把复习间隔定在距离考试的 10% 到 20% 之间，一个月后的考试对应大约三到六天的间隔，半年后的考试对应两到四周。这条刻度解释了一个常见现象，把公式和定义集中在考前三天反复看，考完当天还觉得清楚，两周后的模考就只剩模糊印象。人文学科需要背的量更大，同样时长下间隔可以拉长；理科题目要动手，间隔短一些但每次都要重新算一遍。
Ebbinghaus drew the first retention curve from his own experiments with nonsense syllables in 1885, and its shape, a fast early drop followed by a flatter tail, has been reproduced many times since; The meta-analysis by Cepeda and colleagues in Psychological Bulletin in 2006 pooled over a hundred experiments and concluded that the same total study time produces better long-term retention when it is spread across sessions instead of gathered into one, and the pattern holds across subjects and age groups. Their follow-up work in 2008 put a rough number on the gap, around 10% to 20% of the interval between the last revision session and the exam, which means roughly three to six days before an exam a month away and two to four weeks before one six months away. That scale explains why definitions drilled for three days in a row feel clear on the night and fuzzy two weeks later in a mock paper.
命令词与答题结构：把回想对准卷面要求
复习方法要和卷面要求对上，否则努力落在了不考的地方。AQA 的 GCSE 数学（8300）从 2015 年 9 月起施教、2017 年首考，它的官方 spec 把评估目标分成使用与运用、推理与论证、解题三类，非计算器卷占了三分之一左右的分值；Edexcel 的 iGCSE 与 A Level 数学同样在 spec 里写清 AO1 到 AO3 的权重。这意味着纯计算的熟练度只能拿一部分分，推理与判断要靠能复述思路来支撑。阅卷时看得出来谁在做回想式复习，因为这类答案会写出为什么选这个公式或者为什么排除了另一种解释，而不是只给一个结果。OCR 与 CAIE 的 mark scheme 把分点拆得很细，用回忆的方式把历年 mark scheme 的常见分点默写一遍，比再读一遍教材更能提高得分密度。考试局官网的 past paper 与 mark scheme 都可以免费下载，用它做自测是零成本的提取练习。
Revision only pays off where it meets the paper, and the assessment objectives show where that is; AQA's GCSE Mathematics specification, taught from September 2015 with first exams in 2017, divides its objectives into using and applying, reasoning and arguing and solving problems, with the non-calculator paper carrying a large share of the marks, and Edexcel's IGCSE and A Level specifications state the same kind of weighting between AO1 and AO3. Fluency with routine calculation therefore only reaches part of the marks, while the reasoning and judgement marks need an answer a student can explain out loud. A script reveals which sort of revision was done, since recall-based preparation produces sentences explaining why a formula was chosen or why an alternative was ruled out, and the free past papers and mark schemes on the exam board websites are the cheapest source of self-testing material available.
把间隔排成时间表：从当天到考前四周
可执行的排期不需要复杂工具，一个章节按四次接触就够：学会当天合书回忆一次，第三天只做回忆不做重读，第十天把前两次没答出来的部分再做一遍，第三十天整章重测。中间两次的目的不是重新学，而是让"差点想起来"的状态出现，这种状态对长期保持的作用比一次流畅的重读更大。学生最常把排期理解成"多复习几遍"，于是把同一个章节排进连续四天的晚上，四天都在重读，提取动作一次都没发生。考试前四周是排期的分界点，这时的复习密度要往上走，间隔缩到两三天，但每次仍然要先合书再对照。不同科目混排还有一个好处：数学和物理的推导与历史的论述在脑子里走的是不同的路，交替进行时每一科都像被间隔过一遍。
A workable schedule needs no software, since a topic only has to be met four times: a closed-book recall on the day it is learned, recall only on day three, a pass on day ten over whatever was missed the first two times, and a full retest on day thirty; The middle sessions are not there to relearn the content; they are there to produce the state of almost remembering, which does more for long-term retention than a smooth reread. What most students do instead is read the same chapter on four consecutive evenings, which counts as four rereads and no recall at all; The last four weeks before an exam are where the schedule tightens, with gaps of two or three days rather than weeks, and every session still starting from a closed book. Alternating subjects helps as well, because a derivation in maths and an essay plan in history use different routes through memory, and each subject gets spaced by the other.
两个动作叠加：提取之后再间隔
叠加的顺序是提取在前、间隔在后。一个章节第一次接触时先合书回想，把能写出来的写在一张白纸上，再翻开笔记补漏，补漏时只标记没写出来的部分，不要顺手把整页再读一遍，这一步最容易走回重读的老路；下一次复习只做上次标记的部分，做对了就从列表里划掉。错题本是这个循环里最有价值的一件工具，但它的用法常被写反：有效的错题本记的不是题目和标准答案，而是当时卡住的判断，比如"没有先判断定义域就用求根公式"、"把 marginal 当成 average"。这样一条记录在下一次间隔里只需要三十秒就能重测，而抄一整道题的整理通常只做一次就再也不会翻。物理的电学题、经济的图像题用这种做法收益明显，因为卡点集中在判断而不在计算。
The order that works is retrieval first, spacing second; On the first meeting with a topic, close the book and write what you can on a blank sheet, then open the notes only to mark what was missing, because reading the whole page again at that point quietly turns the session back into rereading; the next session covers only the marked items and crosses off whatever comes back correctly. An error log is the most valuable tool in this loop, although it is usually kept the wrong way round: what earns marks is a record of the judgement that failed, such as applying the quadratic formula before checking the domain or treating marginal cost as average cost, rather than a copy of the question and the official answer. A single line of that kind can be retested in thirty seconds during the next interval, while a full transcription of the question is normally read once and never opened again.
四种常见误用：看着答案说我会了
Dunlosky 等人在 2013 年发表于 Psychological Science in the Public Interest 的综述里，把 practice testing 与 distributed practice 列为高效益技术，把划重点、重读、总结三类列为低效益技术。第一种误用是拿划重点当复习，彩色笔划过的页面在合上书之后什么都不剩。第二种误用是看着答案判断自己会不会，看着答案时用的是识别，识别在考试时用不上，正确的做法是合上答案先写完，再对照分点。第三种误用是只在考前集中刷一轮，把总量堆在一起，遗忘曲线在这种情况下几乎垂直下落。第四种误用是相信熟悉感，很多学生以为"读起来眼熟"等于掌握，实际上一道题从"眼熟"到"能独立写出完整过程"通常还差三次提取。常见的错误是复习完就算完，缺少一次自测来确认到底留下了什么。
A 2013 review by Dunlosky and colleagues in Psychological Science in the Public Interest rated practice testing and distributed practice as high-utility techniques, and put highlighting, rereading and summarising in the low-utility group; Treating highlighting as revision is the first misuse, since a page of colour leaves nothing behind once the book is closed. Judging your own grasp while the answer is visible is the second, because that judgement uses recognition, which the exam hall does not offer; the working method is to write the answer first and mark it against the scheme afterwards; Cramming the whole syllabus into one block in the final week is the third, and it puts the forgetting curve almost vertical. Trusting familiarity is the fourth, because reading something that looks familiar and producing it unaided are usually three retrievals apart, and a revision session that ends without a self-test leaves no evidence of what stayed.
一周执行表与科目差异
把上面的规则排进一周，可以这样安排。周一学新章节，当天用十分钟合书默写要点。周三只做十五分钟回忆，不看笔记，答不出的地方记进错题本。周五用二十分钟重做周三卡住的题，同时对上一章做一次快速回忆。周日留三十分钟，把本周两章各做一次整章自测，只写提纲或关键步骤。理科每天的提取时间里至少一半要动手，数学的推导、物理的计算、化学的方程式都要落在纸上。人文学科的提取以口述加提纲为主，历史按因果链复述，英语文学按论点与证据复述，因为这两科的分数来自论证结构而不仅是事实。这套排期每天占用的时间在四十分钟以内，比每天读两小时更容易维持，也更容易坚持到考前。
Fitted into a week, the rules look like this: learn a new topic on Monday and spend ten minutes writing its key points from memory the same evening; on Wednesday run a fifteen-minute recall session with the notes shut and log whatever does not come back; on Friday redo the Wednesday items for twenty minutes and give the previous topic a short recall; and keep thirty minutes on Sunday for a full self-test on both topics of the week, writing only the plan or the key steps. At least half of the daily retrieval time in science subjects has to be on paper, covering derivations in maths, calculations in physics and equations in chemistry, while arts subjects work better through spoken recall and plans, retelling a causal chain in history or an argument with its evidence in English Literature. The whole schedule takes under forty minutes a day, which is easier to sustain into the exam season than two hours of reading that stops after a fortnight.
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导，物理化学生物数学经济
16621398022 同微信
aleveler.com 预约试听
或截图后用「扫一扫 → 相册」选择本图