5D Library › Books › 5DL-B-023
学习的85%最优错误率定律、髓鞘化动力学与刻意练习
从随机梯度下降、维果茨基最近发展区到间隔重复的认知神经科学及古典学思观并读
The 85% Rule for Optimal Learning, Myelination Kinetics and Deliberate Practice: From Stochastic Gradient Descent and the Zone of Proximal Development to Spaced Repetition, Read Beside Classical Chinese Learning Theory
DOI10.5281/zenodo.23761339摘要
太简单学不会，太难会崩溃：最有效的练习落在「刚好够难」的地方。本书用学习科学、神经科学与可运行的模型考察这个「之间」在哪里：威尔逊等人 2019 年的「85% 法则」（用梯度下降学习的二分类学习者在高斯噪声下，训练错误率为 ½ erfc(1/√2) ≈ 15.87% 时每次试次的学习增益最大；最优区平坦，噪声不同则最优值不同，拉普拉斯约 18.4%、柯西 25%），它的前提，以及它与维果茨基最近发展区只是类比；刻意练习与「一万小时」，元分析中刻意练习平均解释约 12% 的个体差异（游戏 26%、音乐 21%、体育 18%、教育 4%、职业不到 1%），1993 年研究的重复与定义之争；赫布可塑性、长时程增强与活动依赖性髓鞘化，弥散成像指标不等于髓鞘，电缆理论中的 g 比与传导速度；遗忘曲线、间隔效应、检索练习（一周后 61% 对 40%）、ACT-R 激活模型与 SM-2 调度器；交错练习与「必要难度」及线索竞争模型；反馈的指导假说与三分之一以上使表现变差的反馈干预。
八章与《论语》「学而不思则罔」「不愤不启，不悱不发」、《礼记·学记》与朱熹「循序而渐进，熟读而精思」并读，并说明古人讲的是修身成德，不是认知心理学实验。经核对修正：最优错误率公式应为 ½ erfc(1/√2)，而非 ½(1 − 1/√2)（约 0.146）；「一万小时」不是 1993 年论文提出的规则；0.77 是理论而非实测的 g 比；「朱子读书法」六条由张洪、齐熙在辅广旧辑的基础上釐定，而非辅广归纳。最后给出四步刻意练习核查表与一项预注册随机对照试验方案：「约 85% 自适应难度 + 交错提取练习」对「约 95% 正确率 + 集中练习」，成年人学习人工符号推理规则，主要结局为 14 天后的迁移测验，每组 198 人可分析、招募 233 人，共 466 人；写明是组合干预并给出 2 × 2 析因变体的样本量。书中以标注框区分证据、可运行的 Python 计算（15 段程序）、原典与作者的假说。本书不是针对学习障碍、ADHD 或任何状况的诊断或治疗建议；「85%」是特定模型下的最优值，不是普适常数。
关键词：85% 法则；最优错误率；随机梯度下降；最近发展区；刻意练习；全距限制；赫布可塑性；长时程增强；活动依赖性髓鞘化；g 比；电缆理论；遗忘曲线；间隔效应；检索练习；ACT-R；间隔重复；交错练习；必要难度；指导假说；论语；学记
Abstract
Practice that is too easy teaches little; practice that is too hard breaks down. This book uses the learning sciences, neuroscience and runnable models to ask where "just hard enough" lies: Wilson and colleagues' 2019 "Eighty Five Percent Rule" (for binary-classification learners that learn by gradient descent under Gaussian noise, learning per trial is fastest at a training error rate of ½ erfc(1/√2) ≈ 15.87 %; the optimum is flat and moves with the noise, about 18.4 % for Laplacian and 25 % for Cauchy noise), its assumptions, and why Vygotsky's zone of proximal development is an analogy rather than the same theorem; deliberate practice and "10,000 hours", meta-analyses in which deliberate practice explains on average about 12 % of individual differences (games 26 %, music 21 %, sports 18 %, education 4 %, professions under 1 %), the replication of the 1993 study and the dispute over definitions; Hebbian plasticity, long-term potentiation and activity-dependent myelination, why diffusion-imaging measures are not myelin, and the cable-theory relation between g-ratio and conduction velocity; forgetting curves, the spacing effect, retrieval practice (61 % against 40 % after a week), the ACT-R activation model and an SM-2 scheduler; interleaving, desirable difficulties and a cue-competition model; and the guidance hypothesis of feedback and the more than one third of feedback interventions that make performance worse.
The chapters are read beside the Analects, the Record of Learning and Zhu Xi's "advance step by step, read until familiar and reflect with care", making clear that the classical authors were concerned with moral self-cultivation, not experiments. Corrections after checking: the optimal error rate is ½ erfc(1/√2), not ½(1 − 1/√2) (about 0.146); "10,000 hours" is not a rule from the 1993 paper; 0.77 is a theoretical, not measured, g-ratio; the "six principles" of Zhu Xi's method of reading were set out by Zhang Hong and Qi Xi on the basis of Fu Guang's earlier compilation, not by Fu Guang himself. The book ends with a four-step deliberate-practice checklist and a pre-registered randomised trial comparing "adaptive difficulty at about 85 % plus interleaved retrieval practice" with "about 95 % accuracy plus blocked practice" (adults learning an artificial symbolic rule system; primary outcome a transfer test after 14 days; 198 analysable and 233 recruited per arm, 466 in all; stated to be a combined intervention, with a 2 × 2 factorial variant). Marked boxes separate evidence, runnable Python calculations (15 programs), classical sources and the author's hypotheses. The book is not a guide to diagnosing or treating learning disabilities, ADHD or any condition; "85 %" is an optimum within a particular model, not a universal constant.
Keywords: Eighty Five Percent Rule; optimal error rate; stochastic gradient descent; zone of proximal development; deliberate practice; range restriction; Hebbian plasticity; long-term potentiation; activity-dependent myelination; g-ratio; cable theory; forgetting curve; spacing effect; retrieval practice; ACT-R; spaced repetition; interleaving; desirable difficulties; guidance hypothesis; Analects; Record of Learning
目录 · Contents
- 序：在「刚好够难」的地方练习English 中文Preface: Practising Where It Is Just Hard Enough
- 第一章 为什么太简单学不会、太难会崩溃：85% 法则English 中文Chapter 1 Too Easy to Learn, Too Hard to Cope: The 85 % Rule
- 第二章 刻意练习与「一万小时」English 中文Chapter 2 Deliberate Practice and "10,000 Hours"
- 第三章 赫布可塑性、长时程增强与活动依赖性髓鞘化English 中文Chapter 3 Hebbian Plasticity, Long-Term Potentiation and Activity-Dependent Myelination
- 第四章 遗忘曲线、间隔效应与提取练习English 中文Chapter 4 Forgetting Curves, the Spacing Effect and Retrieval Practice
- 第五章 交错与集中：必要难度English 中文Chapter 5 Interleaved versus Blocked: Desirable Difficulties
- 第六章 反馈的时机与粒度English 中文Chapter 6 The Timing and Grain of Feedback
- 第七章 原典并读：学思、愤悱、学记与朱子读书法English 中文Chapter 7 Reading the Classics: Learning and Thinking, Striving, the *Record of Learning* and Zhu Xi's Method of Reading
- 第八章 实践与预注册试验English 中文Chapter 8 Practice and a Pre-Registered Trial
How to cite · 引用
Fan, Y. (2026). The 85% Rule for Optimal Learning, Myelination Kinetics and Deliberate Practice: From Stochastic Gradient Descent and the Zone of Proximal Development to Spaced Repetition, Read Beside Classical Chinese Learning Theory [学习的85%最优错误率定律、髓鞘化动力学与刻意练习] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23761339
BibTeX
@book{fan2026eighty,
title = {{The 85% Rule for Optimal Learning, Myelination Kinetics and Deliberate Practice}: From Stochastic Gradient Descent and the Zone of Proximal Development to Spaced Repetition, Read Beside Classical Chinese Learning Theory},
author = {Fan, Yafu},
year = {2026},
publisher = {5D Library},
edition = {Version 1.0.0},
doi = {10.5281/zenodo.23761339},
url = {https://doi.org/10.5281/zenodo.23761339},
note = {Chinese title: 学习的85%最优错误率定律、髓鞘化动力学与刻意练习}
}
| DOI | https://doi.org/10.5281/zenodo.23761339 |
|---|---|
| DOI（全部版本 all versions） | https://doi.org/10.5281/zenodo.23761338 |
| ORCID | https://orcid.org/0009-0004-3519-1203 |
| 5D Library ID | 5DL-B-023 |
| 版本 Version | 1.0.0 · 中英双语版 Bilingual edition |
| 出版日期 Date | 2026-10-29 |
| 网址 Online | https://5dlibrary.com/books/eighty-five-percent-rule-myelination/ |
| SHA-256 · eighty-five-percent-rule-myelination_bilingual_v1.0.0.pdf | 89fe1f175d89d600fd30fbffee340a018bc7460d8c853e256f7e194c6b2ffcd0 |
| SHA-256 · eighty-five-percent-rule-myelination_en_v1.0.0.pdf | 98b6912211dad7859e3264fcb55190fff65997fc6241226f7cb5f847a8caae02 |
| SHA-256 · eighty-five-percent-rule-myelination_zh_v1.0.0.pdf | d351353e668d8f78dc3bb2cadb0a303994f3a8c787b89958be9765264464c70e |
| Licence · 许可 | Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0). https://creativecommons.org/licenses/by-nc-nd/4.0/ |
Disclosure. The author uses the works of 5D Library, including this book, as the methodological basis of his community courses, the 5D Ascension Cards (Dimensional Shift Card Decks) and the Yafu AI knowledge base, and therefore has a commercial interest in the subject of this book. The book promotes no course, card deck or service; the Chapter 8 trial is designed to be run and analysed by researchers independent of the author's business and does not recruit from the author's community, students or customers; all the code and the study protocol are published openly for anyone to check.