跳至正文

学习风格「因材施教真的有用吗」

目录

机制裁决第 137 篇 · 对称双向第 132 篇 · section E 元尺(与教育测量/CVSS/计量学同族)· 全库第 195 篇 本篇审的是「学习风格」这个名号背后承重的一件核心事:按学生的风格匹配教学,到底有没有用先读三句红线:

  1. 本篇不裁决任何学生、教师或学校的教学法对错,不给「该不该因材施教」「该不该买某某测评」的建议,也不对任何单一学习风格理论做道德评判——只审「按风格匹配教学(meshing hypothesis)」这句声称从证据里读出来时承重承不承得住。
  2. 「学生有偏好」与「匹配教学有效」是两件不同的事:前者是心理学事实(人们确实会报告偏好),后者是需要特定形式证据(交叉交互)的科学假说。本篇对称审计「匹配教学是科学」与「学习风格是纯骗局」两个方向。
  3. 全篇承重句均给出可点击来源;英文逐字引用一律标注「一手逐字」(全文取回)或「摘要逐字」。本篇的核心判据(Pashler 2008 的 crossover interaction)与核心统计(2024 年元分析 g=0.31 与 26% 交叉率)均有可复现的方法,写在第五节与第八节。

零、一句话裁决

「学生有学习偏好」是真实心理学事实(Pashler 2008 逐字「ample evidence」;学生、教师与公众相信率全球稳定在约 76%–98%〔Newton & Salvi 2020 系统综述加权 89.1%〕)——真的不是「按风格匹配教学有用」(把决定性判据写进文献十八年,完全对齐 Pashler 判据的实验截至 2018 年只有 Rogowsky 2015 一个〔Aslaksen & Lorås 2018 逐字「the only study … totally aligned with Pashler’s criteria」〕,2024 年元分析纳入 21 研究 101 效应量、总体 g=0.31 但仅 26% 的测量出现匹配假说唯一合法的 crossover 交互、作者自己逐字「too small and too infrequent to warrant widespread adoption」)和「学习风格是纯骗局」(真实的偏好差异、多模态教学对所有人的真实益处、认知风格研究的真实成果、有证据的学习策略教学——都是真的,见第九节反虚无账)这两层被声称的定论。 把「一个学生说自己喜欢看图」读成「这位学生必须看图才能学好」,是把「偏好」读成「类别」,把「自报」读成「能力」,把「均值」读成「匹配」。 灵魂句:学习风格测的是「我喜欢什么」——被读成的是「我该被怎么教」。前者几乎人人能答,后者十八年来几乎没人测过。

一、本篇测什么:三跳升格与七件可判定的事

本篇审的不是「有没有人真的喜欢看图/听讲/动手」(有,这是锚),而是「『匹配教学』这个名号从被测量到被使用,中间经历了哪些跳」。拆成七件可判定的事:

  1. 判据真不真:Pashler 2008 为「匹配教学有效」写下的决定性验证判据是什么(按风格分组、组内随机分配教学法、全体同一测试、须观察到 crossover 交互)——这是守真锚。
  2. 锚真不真:「学生有偏好」这个心理学事实有多稳(Pashler 逐字承认;偏好自报与行为选择的对应)——这是守真锚的存在侧。
  3. 存在跳:从「偏好」到「稳定风格类别」这一步——分类稳定吗(同一人跨工具 44.6% 一致〔Krätzig & Arbuthnott 2006〕、VARK 重测严格标准仅 17% 完全一致〔Ortega-Torres 2020〕)?
  4. 判据执行史:crossover 判据写进文献后有没有人照做(截至 2018 年完全对齐者仅 1 个〔Aslaksen & Lorås 2018〕;Newton 2015 的「89% 背书」其实没做方法学筛选——本库纠正)——这是匹配跳。
  5. 正反证据账:2024 年元分析(g=0.31)是否颠覆了「无证据」定论(没有——crossover 仅 26%、WWC 达标仅 5/21、作者自判「不采用」);2025 年 Hattie & O’Leary 的 d=0.04 vs r=0.24 分裂——这是匹配跳的证据层。
  6. 消费账:教育者相信率与制度采纳(Ofsted 官方逐字「没有证据表明学生具有可区分、可辨识的学习风格」;但教师培训与商业测评仍在普及)——这是应用跳。
  7. 反向红跳:学习风格不是纯骗局——真实偏好、多媒体教学对所有人的益处、认知风格研究的真实成果、有证据的学习策略——这是反虚无账。

结构胎记(本库「判据」元尺家族第三个变体):

  • 判据在场而口径缺席(碳核算/SBTi,2026-08-02,第 176 篇):判据写在纸上,口径由承诺人自选。
  • 判据写在模型里、验证缺席(地球工程/SRM,2026-08-08,第 194 篇):判据(有效性)写在模型里,实测缺席。
  • 本篇:判据在场、执行缺席(学习风格,2026-08-08,第 195 篇):Pashler 2008 把决定性实验判据(crossover interaction)写进了同行评审文献,十八年几乎没人照做——Newton 2015 统计 103 篇里仅 1 篇够格测(注:此数字经本库核实实际出处是 Aslaksen & Lorås 2018 的「1215 条扫描仅 1 个完全对齐」,Newton 2015 本身是背书度分析,见第四节)。

共用动作:把一句关于「学生自报的偏好」的话,读成一句关于「学生该被怎么教」的话。

三跳升格

  • 存在跳:测出「偏好」→存在「风格」→存在可操作分类的人群。「偏好」真(人们确实报告喜欢看图/听讲/动手);「风格类别」不稳(同一人跨工具仅 44.6% 一致、VARK 严格重测仅 17% 完全一致);「可操作分类」最脆。
  • 匹配跳:风格+教学方式匹配→学得更好。这是 Pashler 判据要证的,也是 89.1% 教育者相信的。证据:截至 2018 年完全对齐判据的实验仅 1 个(阴性);2024 元分析均值 g=0.31 但 crossover 仅 26%、作者自判「不采用」。
  • 应用跳:教育实践采纳风格匹配→有收益或至少无害。制度层面分裂(Ofsted 官方否定 vs 教师培训与商业测评仍在普及);代价是机会成本(占用了有证据的策略)与贴标签的本质化后果。
  • 反向红跳:无证据≠骗局。真实的偏好差异、多媒体教学对全体学生的真实益处(Mayer 多媒体效应)、认知风格研究的真实成果(Witkin 场独立)、有证据的学习策略(测试练习/间隔练习)——都在第九节。

与库内相关篇的分界:

  • 与教育测量篇(2026-08-03,第 181 篇)分界写死:那篇审「两把尺被读成什么」(PISA 排名、增值归因、NAEP 名号——尺子被当客观数用·教育域,section E 元尺);本篇审「一个名号(学习风格)从偏好被读到教学」——那篇是「尺子读数被误读」,本篇是「尺子本身没被造好还当客观数用」。同族(E 元尺)不同对象。
  • 与 CVSS 篇(2026-08-03,第 179 篇)分界:那篇是「纯造尺人问题」(软件漏洞评分被全行业当客观数);本篇是「造出来的尺子本身信效度存疑、还被当成必须因人定制的客观数」——CVSS 尺子稳定、效标清楚;学习风格尺子连重测一致性都不过。
  • 与认知科学相关篇分界:自由意志/镜像神经元等篇审的是「神经科学被神化」;本篇审的是「教育心理学概念被神化」——学习风格本质上是心理学神话而非神经科学主张(它不依赖脑科学证据,靠的是直觉合理性)。
  • 与记忆/学习策略相关篇分界:本库未单做「学习策略」专篇;本篇只在反虚无账里用 Dunlosky 2013 的高效用学习策略(测试练习/间隔练习)作为「机会成本的正面替代物」,不展开做策略综述。

编号落位:机制裁决第 137 篇·对称双向第 132 篇·section E 元尺·全库第 195 篇。去重实测(python 全库 194 篇 6,977,613 字符):学习风格/learning style(s)/learning-style/VAK/VARK/Kolb/Pashler/神经神话/neuromyth/meshing 全库零命中VAK 4 处全为 McIntosh-Vakorin 人名词尾、meshing 1 处为摩擦篇「啮合」——专篇零命中,处女地

二、守真锚(一):Pashler 2008 的决定性判据——先看清「什么才算证据」

本库原则:审「声称」必须回到声称的原件。先看清 Pashler 2008 到底写了什么。

2.1 Pashler 2008:写进同行评审文献的决定性判据

Pashler, McDaniel, Rohrer & Bjork (2008)《Learning Styles: Concepts and Evidence》发表于 Psychological Science in the Public Interest 9(3), 105–119(DOI: 10.1111/j.1539-6053.2009.01038.x;印刷 2008-12,在线 2009-12,Crossref/OpenAlex/PubMed 三源交叉)[一手逐字]。这篇是学习风格领域的判据文本,被反方综述链(Willingham 2015/2018、Kirschner 2017、Aslaksen & Lorås 2018、Hattie & O’Leary 2025)全部引用为基线。

判据四要件(正文逐字) [一手逐字,SAGE 全文 pashler2008-sage.txt]:

“To provide evidence for the learning-styles hypothesis—whether it incorporates the meshing hypothesis or not—a study must satisfy several criteria. First, on the basis of some measure or measures of learning style, learners must be divided into two or more groups (e.g., putative visual learners and auditory learners). Second, subjects within each learning-style group must be randomly assigned to one of at least two different learning methods (e.g., visual versus auditory presentation of some material). Third, all subjects must be given the same test of achievement (if the tests are different, no support can be provided for the learning-styles hypothesis). Fourth, the results need to show that the learning method that optimizes test performance of one learning-style group is different than the learning method that optimizes the test performance of a second learning-style group.”

归纳:① 按学习风格分两组及以上;② 组内随机分配至少两种教学方式;③ 全体使用同一成就测试;④ 须观察到风格×方式的交叉交互(crossover interaction)。

crossover 交互判据(正文逐字)

“Thus, the learning-styles hypothesis (and particular instructional interventions based on learning styles) receives support if and only if an experiment reveals what is commonly known as a crossover interaction between learning style and method when learning style is plotted on the horizontal axis.”

这是关键:匹配教学有效的唯一合法证据形式是「交叉交互」——视觉学习者看图更好、听觉学习者听讲更好,方向相反。不是「视觉组平均分更高」(那可能只是视觉材料对所有学生都好),也不是「匹配组比不匹配组高 0.31 个效应量」(那混入了「某种教学法本身更好」的可能)。Pashler 特别强调(正文逐字):

“One important thing to notice about such a crossover interaction is that it can be obtained even if every subject within one learning-style group outscores every subject within the other learning-style group (see Fig. 1B).”

即:即使一组学生整体上都比另一组强,仍可能观察到交叉交互——所以「风格与能力相关」不构成障碍。

最宽松口径(正文逐字)

“To provide the most liberal criterion in our search for evidence supporting the learning-styles hypothesis, we cast the hypothesis so that it requires only the style-by-method crossover interaction described previously.”

Pashler 自己把判据放宽到只要 crossover 交互就行(连「最优方法须匹配偏好」都不要求)。即便如此,结论仍是:

“We concluded that any credible validation of learning-styles-based instruction requires robust documentation of a very particular type of experimental finding … [four criteria above].” [一手逐字,PubMed 官方摘要]

2.2 Pashler 2008 的结论:四句逐字

  • 结论一(SUMMARY):”We conclude therefore, that at present, there is no adequate evidence base to justify incorporating learning-styles assessments into general educational practice. Thus, limited education resources would better be devoted to adopting other educational practices that have a strong evidence base, of which there are an increasing number.”
  • 结论二:”But at present, these negative results, in conjunction with the virtual absence of positive findings, lead us to conclude that any application of learning styles in classrooms is unwarranted.”
  • 结论三:”At present, however, such validation is lacking, and therefore, we feel that the widespread use of learning-style measures in educational settings is unwise and a wasteful use of limited resources.”
  • 结论四:”In summary, our efforts revealed at most one arguable piece of evidence for the learning-styles hypothesis in general.”

[以上均一手逐字,SAGE 全文 pashler2008-sage.txt / pashler_sage.txt]

关键点:Pashler 的结论措辞是「no adequate evidence base」「virtually no evidence」「at most one arguable piece of evidence」——他没有说「匹配教学被证伪了」,说的是「没有足够证据支持采纳」。这个区分很重要:学理上「证据不足」与「已证伪」是两回事,但教育实践采纳需要的是「支持证据」——证据不在场,就是「不应采用」,无论是否已证伪。

2.3 APA 官方科普页的转述交叉

APA 官方 Article Spotlight(”Learning Styles: Out of Fashion”,2014)对 Pashler 判据的转述逐字 [一手逐字,apa-learning-styles-outoffashion.html]:

“In a critical review of the literature, Pashler, McDaniel, Rohrer, and Bjork (2008) concluded that there has been little empirical scientific research to support or negate the learning styles theory. They detailed the specific experimental design needed to investigate this theory empirically.”

注意 APA 官方也写「little empirical scientific research to support or negate」——再次确认这是「证据不足」判词,不是「已证伪」判词。

三、守真锚(二):偏好是真实的——先承认心理学事实再谈承重

本库原则:审「声称」先清点「事实」。学习风格话题最容易被两极化,而真实的锚是:「人们确实会报告偏好」。

3.1 Pashler 2008 对偏好的承认(逐字)

“Our review of the literature disclosed ample evidence that children and adults will, if asked, express preferences about how they prefer information to be presented to them. There is also plentiful evidence arguing that people differ in the degree to which they have some fairly specific aptitudes for different kinds of thinking and for processing different types of information.” [一手逐字,pashler_sage.txt]

Pashler 同时明确:偏好存在 ≠ 偏好对教学有含义

“Having noted the reality of these preferences, we emphasize that the implications of such preferences for educational practices and policies are minimal. The existence of preferences says nothing about what these preferences might mean or imply for anything else, much less whether it is sensible for educators to take account of these preferences.” [一手逐字]

3.2 偏好自报与行为选择的对应(Massa & Mayer 2006)

Massa & Mayer (2006) 用三实验考察偏好:自报偏好的学生在实际点选帮助屏时会选择与偏好一致的呈现方式——偏好自报与行为有对应 [一手逐字,经 Pashler 转述+原文核验]。偏好是行为层面真实的。

3.3 分类学:什么是「学习风格」

术语谱系(每个模型一句话+一手出处):

  • VAK(Visual–Auditory–Kinesthetic):按信息接收通道分三类。一手出处转述级(Barbe & Milone 1981 原书未抓取;经 Felder & Silverman 1988 全文脚注转述「The ways people receive information may be divided into three categories, sometimes referred to as modalities」)。
  • VARK(Visual–Aural–Read/write–Kinesthetic):Fleming & Mills 1992 提出,四模态自评问卷。官方逐字(vark-modalities-jina.txt):”The acronym VARK stands for the four VARK modalities – Visual, Aural, Read/write, and Kinesthetic – sensory modalities that are used for learning information.” [一手逐字,官方页面]
  • Kolb 体验学习四类(divergers/assimilators/convergers/accommodators):Pashler 转述逐字 [一手逐字,pashler2008-sage.txt];Kolb 1984 原书未抓取(转述级)。
  • Felder–Silverman ILS:工程教育四维(Sensing/Intuiting、Visual/Verbal、Active/Reflective、Sequential/Global)。NCSU 官方全文 [一手逐字,felder-silverman1988.pdf]。
  • Dunn & Dunn:五类偏好。Pashler 转述官网定义逐字 [一手逐字,pashler2008-sage.txt]。

3.4 学术分层:偏好 vs 学习风格假设 vs 特定能力

Pashler 原文明确的三层区分(逐字):

  • 偏好:”The existence of preferences … amounts simply to the fact that people will, if asked, volunteer preferences about their preferred mode of taking in new information and studying.”
  • 学习风格假设:偏好对教学有含义的断言(需要 crossover 证据)。
  • 特定能力(specific aptitude):”evidence for specific aptitudes does not, by itself, validate the learning-styles hypothesis.”(能力强不意味着需要不同的教学方式。)

[一手逐字,pashler2008-sage.txt]

四、存在跳:从「偏好」到「稳定风格类别」——分类到底有多稳

存在跳要回答:即使偏好真实,从「偏好」到「可操作分类的人群」这一步立不立得住?证据直指:分类不稳定,量表信效度存疑。

4.1 同一人跨工具只 44.6% 分到同一类(Krätzig & Arbuthnott 2006)

Krätzig & Arbuthnott (2006) Journal of Educational Psychology 98(1), 238–246 [一手逐字,开放版全文]:

“The results from these two assessments did not show good correspondence: Only 29 of the 65 participants (44.6%) were classified as having the same learning style using both assessment procedures. This impression of poor correspondence was verified with a correlational analysis, which indicated a nonsignificant relationship between the two measures (Spearman’s ρ = .057).”

同一批人,两个不同的学习风格量表,只有 44.6% 被分到同一类,两个测量的相关 ρ=.057(不显著)。如果「学习风格」是稳定的个人特质,两个量表应该高度一致——它们几乎不相关。

Krätzig 还发现 [一手逐字]:

“learning style assessed either through individuals’ self-assessment or with a commonly used questionnaire did not correlate with objective memory performance using the various stimulus types.”

风格(无论自评还是量表)与客观记忆表现零相关。以及一个耐人寻味的发现:

“The findings indicate that participants answered the inventory using general memories and beliefs rather than specific examples of learning in different modalities.”

——人填量表时用的是泛泛的记忆与信念,不是具体的学习经历。

4.2 VARK 重测:严格标准仅 17% 完全一致(Ortega-Torres 2020)

Ortega-Torres, Solaz-Portolés & Sanjosé (2020) Periódico Tchê Química 17(36) 440–456,N=582 中学生、128 人重测、间隔数周 [一手逐字,期刊开放 PDF]:

“According to the stricter criterion, only 17 percent of participants obtained precisely the same assignation from both VARK administrations. Nonetheless, when the permissive criterion was adopted, 89 percent of students shared at least one pure mode in both administrations.”

严格标准(完全相同分类)只有 17%;宽松标准(只要共享任一通道)才有 89%。且自评一致性 Kappa 仅 0.30 [一手逐字]:

“The obtained Cohen’s Kappa was K= 0.30. Therefore, the students’ selfassignations appeared to have low reliability, at least when periods of few months were considered.”

关键判读:分类稳定性高度依赖判读口径——宽松口径(共享任一通道)Kappa 0.78,严格口径(完全一致)只有 17%,学生自评只有 0.30。一个需要精确「视觉型/听觉型/动觉型」标签的工具,却给不出稳定的标签。

4.3 两个主流量表对同一批人零相关(Martin 2010,经 Rogowsky 2020 逐字转述)

“Martin administered two learning style inventories to 394 secondary students: Kolb’s Learning Style Inventory-2 and Honey and Mumford’s Learning Styles Questionnaire. Despite the similar descriptors for the classifications between the two measures, no correlation between the two measures was found and there was a lack of construct validity for both inventories as well. Martin concluded, ‘Teachers would have as much information if they assigned the learning styles randomly to students rather than using the Kolb test’ (p. 1586).”

Martin 的原话「教师随机给学生分配学习风格,得到的信息量也和用 Kolb 量表一样多」——这是存在跳最锐利的一击(注意:Martin 2010 原文经 Rogowsky 2020 全文逐字转述,含页码,标转述级)。

4.4 量表信效度全景:Coffield 2004 的 13 工具评审(经 Kirschner 2017 逐字引用)

Coffield et al. (2004) 评审 13 种最常用学习风格量表的四项最低心理测量标准(内部一致性、重测信度、构念效度、预测效度),经 Kirschner (2017) 全文逐字转述 [一手逐字,kirschner2017.txt]:

“three of the models did not meet any of the four criteria (Jackson; Riding; Sternberg’s multiple intelligences), four met only one criterion (Dunn & Dunn; Gregorc; Honey & Mumford; Kolb), three met two criteria (Entwistle; Herrmann; Myers-Briggs personality type indicator), two met three criteria (Apter, Vermunt) and only one met all four criteria (Allinson & Hayes cognitive style indicator), but this instrument measured cognitive styles and not learning styles.”

13 种量表没有一种完全通过四项最低标准;唯一四项全过的(Allinson & Hayes)测的是认知风格而非学习风格。

4.5 VARK 的独立因子验证(Leite et al. 2010)

Leite, Svinicki & Shi (2010) Educational and Psychological Measurement 70(2) 323–339 [一手摘要逐字]:CFA 找到四因子结构、信度 .85/.82/.84/.77「可接受」——但作者同时警告「item wording 与 scoring algorithm 有潜在问题」「对在研究中使用 VARK 要谨慎」。

存在跳小结:偏好真实(§3),但「稳定风格类别」不成立——跨工具 44.6%、严格重测 17%、跨工具零相关、13 工具无一全过最低标准。从「我喜欢看图」到「我是视觉型学习者」这一步,测量学上站不住。

五、匹配跳(判据执行史):判据在场,执行缺席

5.1 2008 年前:Pashler 认定的「够格实验」

Pashler 2008 原文评估的、方法学合格(或接近合格)的实验:

  • Massa & Mayer 2006(三实验,电子课件 verbalizer/visualizer 匹配):Pashler 评价「particularly informative and well-designed」「methodologically strong」,结果阴性——”Massa and Mayer found no support for any of these interactions despite exhaustive analysis of nearly 20 individual-difference measures” [一手逐字,经 Pashler 转述+原文核验]。Massa & Mayer 结论逐字:”the idea that different instructional methods should be used for visualizers and verbalizers” 无支持(该句经原文全文核验)。
  • Constantinidou & Baker 2002(VVQ×视觉/听觉自由回忆):阴性——”no relationship between a visual learning style and the actual learning of verbal items that are presented visually or auditorily” [一手逐字,经 Pashler 转述]。
  • Sternberg et al. 1999(324 资优生,三成分能力):Pashler 判定的「唯一可争论证据」,但有严重方法学问题——”the interaction was found only with highly derived measures … only about one third of the subjects were classified into the groups that produced the interaction … the interaction was achieved only after the outliers were excluded for unspecified reasons” [一手逐字]。
  • Cook et al. 2009(123 内科住院医,医学教育):阴性 [一手逐字,经 Pashler 转述]。

Pashler 总结:「at most one arguable piece of evidence」[一手逐字]。

5.2 判据执行史的量化:1215 条记录、10 篇、1 个完全对齐(Aslaksen & Lorås 2018)

Aslaksen & Lorås (2018) Frontiers in Psychology 9, 1538 做系统检索 [一手逐字,PMC 全文]:

“A total of 1215 records were initially scanned, and 10 studies … were found that had applied the appropriate methodology according to the criteria by Pashler et al. (2008).”

“The only study located through the systematic literature search across six different databases and the screening of more than a 1000 records that was totally aligned with Pashler’s criteria was Rogowsky et al. (2015). These authors report no statistically significant relationship or crossover-interaction effect between modality-specific learning styles and modes of instruction.”

1215 条记录 → 10 篇「按判据应用了适当方法」→ 其中严格完全对齐者仅 Rogowsky 2015 一个。 而这唯一完全对齐的研究结果是阴性(§5.3)。

Aslaksen 的效应量汇总 [一手逐字]:

“This resulted in a small and non-significant effect size for visual matching (g = -0.09, 95% CI [-0.74-0.58], p = 0.80, n = 484) as well as for auditory matching (g = -0.27, 95% CI [-0.87-0.32], p = 0.37, n = 356).”

视觉匹配 g=-0.09、听觉匹配 g=-0.27,均为非显著,甚至方向为负(匹配条件表现还略差)。Aslaksen 逐字结论:

“there is still no replicable statistical evidence for enhanced learning outcome by aligning instruction to modality-specific learning styles.”

5.3 Rogowsky 2015:唯一完全对齐判据的实验——阴性

Rogowsky, Calhoun & Tallal (2015) Journal of Educational Psychology 107(1) 64–78 [一手逐字,作者自传全文 PDF]:

方法:成人 121 人随机分四组;先测听觉/视觉文字风格(标准化量表);随后随机分配到数字有声书或电子文本;同一书面理解测试(即时+两周后);显式采用 Pashler crossover 判据。

结果:

“Results demonstrated no statistically significant relationship between learning style preference (auditory, visual word) and instructional method (audiobook, e-text) for either immediate or delayed comprehension tests.”

“As shown in Figure 1C, contrary to the crossover pattern that would be expected to support the meshing hypothesis, the auditory and visual word learning style preference groups both scored higher on the same method, and hence there is no crossover.”

唯一完全对齐 Pashler 判据的实验,结果没有 crossover。

5.4 Newton 2015:本库纠正一处流行误传

传播中常出现「Newton 2015 统计了 103 篇学习风格文献,只有 1 篇满足方法学标准」。此说经本库核实是张冠李戴

  • Newton (2015) 实际做的是背书度分析——统计 ERIC/PubMed 里新近论文有多少篇「背书」学习风格使用:“The overwhelming majority (89%) of recent research papers, listed in the ERIC and PubMed research databases, implicitly or directly endorse the use of Learning Styles in Higher Education” [一手逐字,PMC 全文]。纳入标准是「是否关于学习风格」「参与者是否高教师生」「是否英文」,没有做 Pashler 判据筛选
  • 「仅 1 篇完全对齐判据」的真实出处是 Aslaksen & Lorås 2018(§5.2,1215 条扫描仅 Rogowsky 2015 一个)。

这个纠正很重要:Newton 2015 的价值在「背书率 89%」(学习风格在文献里被广泛背书),Aslaksen 2018 的价值在「完全对齐判据的只有 1 个」——两件事,两个数字,都要写对。

5.5 2008 年后的其他执行

  • Rogowsky 2020(五年级儿童版):结果三连否定 [一手逐字,PMC 全文]:”(1) no significant positive relationship between auditory learning style and listening comprehension, (2) no significant positive relationship between visual learning style and reading comprehension, and (3) no differential effect of learning style on performance on a listening as compared to a reading comprehension test.” ⚠️ 注意:本篇风格在第 3 天才测(教学活动之后),违反 Pashler「先分类再随机」判据——2024 元分析 Table 1 标注「completed after learning activity」,故不是严格判据完全满足者(与 2015 版不同)。
  • Knoll et al. 2017(VVQ×图片/词对):结果区分主客观 [一手逐字,PubMed 摘要]:”learning style was associated with subjective aspects of learning but not objective aspects of learning.”——风格影响「你觉得学得怎么样」,不影响「实际学得怎么样」。
  • Aslaksen & Lorås 2019(随机实验 N=22):阴性,且听觉匹配 g=-1.25(听觉学习者在匹配条件下反而明显更差,反向结果)[一手逐字,Crossref 官方摘要]。
  • Melzner & Kappes 2024(Instructional Science,N=222 准教师):阴性 [一手逐字,官方摘要]:”There was no significant interaction between presentation mode and learning styles on learning outcome.”

5.6 2024 元分析:判据执行的最新合计——均值小正,crossover 稀少

Clinton-Lisell & Litzinger (2024)《Is it really a neuromyth? A meta-analysis of the learning styles matching hypothesis》Frontiers in Psychology 15, 1428732(2024-07-10)[一手逐字,Frontiers+PubMed 双通道]:

纳入与结果

“A systematic search of the research findings yielded 21 eligible studies with 101 effect sizes and 1,712 participants for the meta-analysis. Based on robust variance estimation, there was an overall benefit of matching instruction to learning styles, g = 0.31, SE = 0.12, 95% CI = [0.05, 0.57], p = 0.02.”

但关键限定

“However, only 26% of learning outcome measures indicated matched instruction benefits for at least two styles, indicating a crossover interaction supportive of the matching hypothesis.”

作者自己的结论

“Given the time and financial expenses of implementation coupled with low study quality, the benefits of matching instruction to learning styles are interpreted as too small and too infrequent to warrant widespread adoption.”

判据放宽与质量 [一手逐字]:

“the criteria from Pashler et al. (2008) are generally followed. One exception is that non-randomized quasi-experiments are included”

“Based on these standards, a study must be a randomized experiment to meet standards … As can be seen in Appendix Table B2, five studies were determined to WWC standards and the remainder did not meet WWC standards.”

两个数字必须分清(本篇最核心的统计区分):

  • g = 0.31:所有 101 个效应量中「匹配条件比不匹配条件平均好」的量。它不是「按风格匹配有效」的证据——它混入了「某种教学法本身对全体学生更好」的可能(视觉材料可能对所有学生都好,与风格无关)。
  • crossover 仅 26%:匹配假说唯一合法的证据形式(视觉学习者看图更好+听觉学习者听讲更好,双向交叉)只在 11/42(26.19%)学习结果测量中出现 [一手逐字:“11 learning outcome measures in which matched instruction benefited at least 2 learning styles … out of a total of 42 learning outcome measures … 26.19%”]。

2024 元分析没有颠覆「无证据」定论——而且作者本人就是最大刹车。 它找到的是「均值小正、交叉稀少、达标仅 5/21、作者自判不采用」。

5.7 2025 年 Hattie & O’Leary:d=0.04 vs r=0.24 的分裂(方法学澄清)

Hattie & O’Leary (2025)《Learning Styles, Preferences, or Strategies?》Educational Psychology Review 37:31(2025-04-01,开放获取)[一手逐字,HTML+PDF 双版本交叉验证]:

“Drawing on 17 meta-analyses, we distinguish between studies testing the matching hypothesis (effect size d = .04) and correlational studies (average correlation r = .24), revealing that the latter often conflates learning styles with learning strategies.”

这是理解「为什么不同人读元分析得出相反结论」的关键

  • 匹配假说严格测试(d-研究):四个匹配元分析、143 研究、401 效应量、d = 0.04(近零)。逐字:”There is NO evidence that matching learning styles to teaching styles makes much, if any, positive impact on learning.”
  • 相关研究(r-研究):r = 0.24(d = 0.50)——但这是「学习风格量表得分与成就的相关」,不是匹配证据。逐字:”Causality and directionality are lost in these correlations.”

Hattie & O’Leary 的结论逐字:

“There is still no evidence for matching, and as many previous reviewers have noted, we should not perpetuate myths, unsupported claims, and wishful beliefs. The recent surge of interest in promoting the correlates of learning styles is misleading, of little value, and should be resisted.”

(注:原文摘要写「eight confounding factors」但正文编号到 7,为本库抓取的原文内部计数不严,如实注记。)

匹配跳小结:判据写死十八年,完全对齐的实验截至 2018 年仅 1 个(阴性);2024 元分析纳入 21 研究但 crossover 仅 26%、WWC 达标仅 5/21、作者自判「不采用」;2025 年 d=0.04 vs r=0.24 的方法学分裂把「相关被误读为匹配」彻底切开。执行缺席,不是证据缺失。

六、反向红跳(一):2024 元分析的效应量真的「正向」吗

元分析 g=0.31 常被宣传为「匹配有效的新证据」。本节拆解这个数字:

第一层:crossover 才是唯一合法形式。 匹配假说的判据是交叉交互(§2.1)。g=0.31 是「匹配 vs 不匹配」的均值差,其中「匹配」条件往往是「视觉学习者看视觉材料」——但视觉材料可能对所有学生都更好(多媒体效应),这时「匹配组平均高」不代表「风格×方式交互」存在。crossover 出现率 26% 才是真正检验假说的指标 [一手逐字,§5.6]。

第二层:作者自己的效用量级参考。 Aslaksen & Lorås (2018) 引实践相关效应量标准 [一手逐字]:

“Ferguson (2009) recommended that a minimum effect size representing a ‘practically’ significant effect amounts to g ≥ 0.41, and Hattie (2009) has advanced the view that effect sizes ≥ 0.40 represent a ‘hinge-point’ at which deliberate interventions provide relevant outcomes for teaching and learning.”

g=0.31 低于实践相关阈值 0.40–0.41。 且 Aslaksen 指出 95% CI 跨零([0.05, 0.57] 不跨零但接近;Aslaksen 引的是其综述内各单项研究的跨零 CI)。

第三层:与其他教育干预的对比。 2024 元分析作者自己引用多媒体效应 g=0.70(Noetel et al. 2022,约为匹配效应 0.31 的两倍,且不需评估风格、成本更低)[一手逐字,经 meta2024-frontiers.txt]。同一批学生,投入多媒体教学比投入风格匹配收益大一倍多。

第四层:假风格实验。 Moser & Zumbach (2015) 给学生随机分配假的「视觉型/言语型」标签,结果学生「匹配」假风格时得分更高——按真实风格匹配则无益 [一手逐字,经 meta2024 引用]。「贴标签」本身就有安慰剂式效果,这正是存在跳与应用跳之间的通道。

七、应用跳:消费账——相信率与制度采纳

7.1 相信率:全球教育者约 89.1% 相信匹配教学有效

Newton & Salvi (2020) 系统综述 37 研究、15,405 名教育者、18 国 [一手逐字,Frontiers 全文]:

“Self-reported belief in matching instruction to Learning Styles was high, with a weighted percentage of 89.1%, ranging from 58 to 97.6%. There was no evidence that this belief has declined in recent years; 95.4% of trainee (pre-service) teachers agreed that matching instruction to Learning Styles is effective.”

“Calculating the weighted average revealed that 79.7% of participants said they used, or intended to use, the matching of instruction to Learning Styles.”

相信率 89.1%、使用/计划使用率 79.7%、且「没有证据表明随时间下降」——对照判据执行史(完全对齐实验 1 个),这是本库最爱的「相信率×证据率」剪刀差。

其他相信率数据点(人群/方法/年份全标注)[一手逐字,各抓取件]:

研究 年份 人群 N 相信率
Dekker et al. 2012 英/荷中小学教师 242 英国 93% / 荷兰 96%
Ferrero et al. 2016 西班牙教师 284 91.1%;跨文化 85.8–97.1%
Horvath et al. 2018 获奖教师 vs 非获奖 50 84% vs 93%
Macdonald et al. 2017 美国公众/教师/高神经科学接触 3,045/598/234 93% / 76% / 78%
Hughes et al. 2020 澳大利亚教师 228 >79%
Grospietsch & Mayer 2019 德国职前理科教师 550 93%
Newton & Salvi 2020 系统综述 15,405 加权 89.1%

值得注意:获奖教师(84%)比非获奖(93%)低——但仍是压倒性多数。「越懂教学的人越不信」这个方向真实存在,但差距不大,不信者仍是少数。

7.2 制度采纳的分裂

英国 Ofsted(官方否定)——Education Inspection Framework Overview of Research (2019) [一手逐字,官方 PDF]:

“it should be clearly stated that there is no evidence that pupils have distinct and identifiable learning styles. Trying to design tasks with this misconception in mind will increase teachers’ workload but is very unlikely to improve learning.”

英国教育标准局官方文件逐字写「没有证据表明学生具有可区分、可辨识的学习风格」,并点名反对按风格设计任务。

但英国教师培训仍有残留——Ofsted ITE Thematic Monitoring Visits 报告 (2025-10) [一手逐字,官方 HTML]:

“In a small number of cases, leaders have not updated the research that underpins their ITE curriculum. Therefore a few trainees continue to study learning styles without critique.”

美国——Nancekivell et al. (2020) J Educ Psych 逐字 [一手逐字]:”Some teacher certification programs incorporate learning styles into training courses, wasting valuable time and resources”;APA 官方新闻稿同句。具体哪些州/项目无公开名录(诚实空位)。

中国——教育心理学教材正面讲授「学习风格」专节,并对接教师资格考试大纲(超星慕课《教育心理学》「第二节 个体的学习风格差异……(1) 视觉型学习者 (2) 听觉型学习者 (3) 动觉型学习者」)[一手逐字,教材页抓取]。诚实空位:美国州级与中国的匹配教学推荐措辞未逐字核实。

制度采纳小结:官方机构(Ofsted)已明文否定,但教师培训教材、认证课程、教材仍广泛正面讲授——制度层面「否定在官方、传播在民间」的分裂,正是应用跳的现场。

7.3 商业市场:三大测评工具的产品化

  • VARK(vark-learn.com):订阅制 B 端产品 $50/年+30 人、增购 $0.40–1.00/人、报告 $1.20–10.57/份 [一手逐字,官方价格页]。
  • Honey & Mumford LSQ:Peter Honey Publications 出版、Pearson 分销。
  • Kolb LSI:经 Korn Ferry/Hay Group 分销(2008 年 10 册 $100)。
  • Dunn & Dunn:认证项目 $1,225/人+研究级 $1,000 [一手逐字,经 Pashler 转述]。

7.4 消费的代价

  • 机会成本:匹配教学占用了有证据的策略(§9.4 的测试练习/间隔练习)。2024 元分析逐字:「有限资源应投向有证据基础的做法」。
  • 贴标签的本质化后果:Sun et al. (2023) 三项研究(经 Hattie & O’Leary 2025 逐字转述):
    > “parents, teachers, and children judged children described as visual learners as more intelligent than children described as hands-on learners” (p. 46)

「视觉型学习者」标签被判定比「动手型学习者」更聪明——贴标签不是中性行为,它携带社会评判。

  • 自我归因:Nancekivell et al. (2020) 发现人们把学习风格本质化(essentialism)——把它当作生物性的、稳定的、可以预测一切的特质。

八、正反证据账的对账:三套统计口径不混

本库纪律:多套统计语言不混。 学习风格话题里有三套数字,分别属于不同口径:

口径 数字 说什么 承重什么
判据执行 截至 2018 完全对齐 Pashler 判据的实验 = 1 个(Rogowsky 2015,阴性) 判据在场、执行缺席 「匹配教学有效」无合格证据
元分析均值 2024: g=0.31 (p=0.02);2025: d=0.04 匹配 vs 不匹配均值差(混入「某教学法本身更好」);严格匹配假说近零 「均值正」不等于「匹配有效」
crossover 率 2024: 26% (11/42);2025: d=0.04 匹配假说唯一合法形式的出现率 「匹配有效」的承重指标

对账结论

  1. g=0.31 与 d=0.04 不矛盾——前者是 21 研究 101 效应量的单研究元分析(含放宽随机判据、含 2 个非随机准实验、crossover 仅 26%);后者是 4 个匹配元分析的再汇总(143 研究 401 效应量)。两者都指向「不采用」。
  2. crossover 26% 是承重判据:即便按最宽松口径(含低质研究),匹配假说唯一的合法证据形式也只出现在约 1/4 的测量里。
  3. 「无证据」≠「已证伪」:Pashler 措辞是「no adequate evidence base」;但教育采纳需要「支持证据」——证据不在场即不应采用。这两层都要写。

九、反向红跳(二):学习风格不是纯骗局——反虚无账

防「学习风格是彻头彻尾骗局、毫无价值」的虚无化。真实的东西有四件:

9.1 真实偏好差异(守真锚)

Pashler 2008 逐字承认偏好「ample evidence」、特定能力差异「plentiful evidence」[一手逐字,§3.1]。「学生有偏好」是真的——它只是不构成「匹配教学有效」。

9.2 多媒体/多模态教学真实有益(但非「匹配」意义)

Mayer 多媒体学习认知理论:多媒体呈现(图文并茂)平均优于单一通道——这是对所有学生,不是「视觉学习者匹配」。Mayer (2001) 多媒体效应 d=1.39(11/11 一致)[一手逐字,经 F 捆抓取;Massa & Mayer 2006 原文确认图示帮助「同时」帮了视觉型与言语型]。

关键区分:「多模态教学有益」与「按风格匹配教学」是两件事——前者提供多种表征让所有学生受益(UDL 通用学习设计也持此立场,CAST 官方逐字「没有任何一种表征对全体最优;提供选项是必要的」,只谈选项不谈匹配);后者要求「视觉型看视觉、听觉型听听觉」的交叉交互。把前者当作后者的证据,是元分析混淆的根源之一(Hattie & O’Leary 2025 Confound 6)。

9.3 认知风格研究的真实成果(但≠学习风格)

Witkin & Moore (1974) 场依赖-场独立是「可识别、显著的个体差异维度」[一手逐字,ERIC 原件]——但 Witkin 自己就承认匹配证据缺失 [一手逐字]:

“First, studies are still needed to determine whether matching for cognitive style does indeed make for better student learning and not only for more positive feelings.”

认知风格研究从来就承认匹配证据缺失——这是 50 年前一手作者的诚实限定,却被后来的商业测评丢弃。

9.4 有证据的学习策略(机会成本的正面替代物)

Dunlosky et al. (2013) Psychological Science in the Public Interest 的「用有效学习技巧」评审 [一手逐字,全文]:

“practice testing and distributed practice received high utility assessments because they benefit learners of different ages and abilities and have been shown to boost students’ performance”

测试练习、间隔练习是高效用学习策略——它们不是「匹配风格」,而是「教所有人更好的学习方式」。这才是教育者时间该投的地方(Pashler 逐字「limited education resources would better be devoted to adopting other educational practices that have a strong evidence base」)。

9.5 反方学者自己的限定(不虚无论原句)

  • Willingham (2018):”Although there’s scant evidence that matching the manner of processing to a student’s preferred style brings any benefit, there’s ample evidence that matching the manner of processing to the task helps a lot.” [一手逐字,American Educator 开放版]
  • Kirschner (2017) 开篇承认偏好真实:”We all differ from each other in a multitude of ways, and as such we also prefer many different things whether it is music, food or learning.” [一手逐字]
  • 2024 元分析作者(防虚无论的主力):”However, the majority of learning outcomes did not indicate a crossover interaction that would validate accommodation to learning styles. However, a non-trivial minority of learning outcomes did indicate the crossover interaction”——作者保留了一个「不可忽视的少数」,没有一棒子打死,但也判「too much of an overreach to insist learning styles should be incorporated」。
  • Pashler (2008) 自己:”it would clearly be a mistake to label these negative results as a conclusive refutation of the learning-styles hypothesis in general”——反方自己都拒绝「已证伪」的表述。

十、母题收口:偏好与判据之间的十八年

学习风格测的是「我喜欢什么」——被读成的是「我该被怎么教」。 从偏好到匹配教学,经历了四跳:

  1. 存在跳:偏好真实(§3),但「稳定风格类别」测量学上不成立(§4)——跨工具 44.6%、严格重测 17%、13 工具无一全过最低标准。「我是视觉型学习者」这句话,比「我喜欢看图」多出来的部分没有测量学支撑。
  2. 匹配跳:判据写死十八年(§2),完全对齐的实验截至 2018 年仅 1 个且阴性(§5.3),2024 元分析 crossover 仅 26%、作者自判「不采用」(§5.6),2025 年 d=0.04 的严格口径近零(§5.7)。「匹配教学有效」没有合格证据。
  3. 应用跳:89.1% 教育者相信、79.7% 使用(§7.1),官方否定 vs 培训/教材/测评普及的分裂(§7.2),商业测评明码标价(§7.3),代价是机会成本与贴标签的本质化后果(§7.4)。消费跑在证据前面。
  4. 反向红跳:真实偏好、多媒体益处、认知风格成果、有证据的学习策略——都是真的(§9)。虚无化与神话化同样没有证据。

灵魂句:学习风格把「一句关于『我喜欢什么』的自报」读成「一句关于『我该被怎么教』的处方」——自报是真的,处方十八年来没有人开得出证据。判据不是没有,是被写进文献后,十八年几乎没人照做。

对称金句:Pashler 2008 写「virtually no evidence for the interaction pattern」——而 Newton & Salvi 2020 统计「89.1% 的教育者相信匹配教学有效」。判据在场而执行缺席,执行缺席而相信在场。 同一篇文献里的判据,十八年后仍是同一套判据;同一批教育者里的信念,十八年后仍是同一个信念。科学与信念之间隔的不是知识更新的时间,是证据标准。

来源清单

一手统筹(✓ 全文取回 / ◐ 摘要级 / ○ 多源交叉 / ⚠ 需读者亲核)

  1. ✓ Pashler, McDaniel, Rohrer & Bjork (2008)《Learning Styles: Concepts and Evidence》PSPI 9(3) 105–119,SAGE 全文 PDF:https://journals.sagepub.com/doi/10.1111/j.1539-6053.2009.01038.x(DOI 解析;SAGE 全文经 open 镜像 pashler2008-sage.pdf)
  2. ✓ Pashler 2008 PubMed 官方摘要(PMID 19178297):https://pubmed.ncbi.nlm.nih.gov/19178297/
  3. ✓ Pashler 2008 Crossref 元数据(卷期页三源交叉):https://api.crossref.org/works/10.1111/j.1539-6053.2009.01038.x
  4. ✓ APA 官方 Article Spotlight「Learning Styles: Out of Fashion」(2014):https://www.apa.org/pubs/highlights/spotlight/issue-22
  5. ✓ VARK 官方 The VARK Modalities 页:https://vark-learn.com/introduction-to-vark/the-vark-modalities/
  6. ✓ Fleming & Mills (1992)「Not Another Inventory, Rather a Catalyst for Reflection」To Improve the Academy 11, 137–155:https://vark-learn.com/wp-content/uploads/2014/08/not_another_inventory.pdf(含作者逐字「testing for construct validity and reliability was unnecessary and inappropriate」)
  7. ✓ Felder & Silverman (1988)「Learning and Teaching Styles in Engineering Education」NCSU 全文:https://www.engr.ncsu.edu/wp-content/uploads/drive/1QP6kBI1iQmpQbTXLW08lQePzZjJ7PaDs/1988-LS-plus-note.pdf
  8. ✓ Krätzig & Arbuthnott (2006) J Educ Psych 98(1) 238–246(开放版全文):https://doi.org/10.1037/0022-0663.98.1.238
  9. ✓ Ortega-Torres et al. (2020) VARK 重测 Periódico Tchê Química 17(36) 440–456(瓦伦西亚大学 RODERIC 机构库开放版):https://hdl.handle.net/10550/76595
  10. ✓ Rogowsky, Calhoun & Tallal (2015) J Educ Psych 107(1) 64–78(作者自传全文):https://doi.org/10.1037/a0037478
  11. ✓ Rogowsky et al. (2020) Frontiers in Psychology 11, 164(PMC 全文):https://pmc.ncbi.nlm.nih.gov/articles/PMC7033468/
  12. ✓ Knoll et al. (2017) Br J Psychology 108(3) 544–563(PubMed 摘要):https://pubmed.ncbi.nlm.nih.gov/27620075/
  13. ✓ Aslaksen & Lorås (2018) Frontiers in Psychology 9, 1538(PMC 全文):https://pmc.ncbi.nlm.nih.gov/articles/PMC6113575/
  14. ✓ Newton (2015) Frontiers in Psychology 6, 1908(PMC 全文):https://pmc.ncbi.nlm.nih.gov/articles/PMC4678182/
  15. ✓ Clinton-Lisell & Litzinger (2024) Frontiers in Psychology 15, 1428732(PMC 全文):https://pmc.ncbi.nlm.nih.gov/articles/PMC11270031/
  16. ✓ Clinton-Lisell & Litzinger (2024) PubMed 记录(PMID 39055994):https://pubmed.ncbi.nlm.nih.gov/39055994/
  17. ✓ Willingham, Hughes & Dobolyi (2015)「The Scientific Status of Learning Styles Theories」Teaching of Psychology 42(3) 266–271:https://journals.sagepub.com/doi/10.1177/0098628315589505(摘要逐字)
  18. ✓ Willingham (2018) American Educator 开放版(ERIC EJ1182080):https://files.eric.ed.gov/fulltext/EJ1182080.pdf
  19. ✓ Kirschner (2017)「Stop propagating the learning styles myth」Computers & Education 106, 166–171:https://doi.org/10.1016/j.compedu.2016.12.006(全文 PDF 镜像)
  20. ✓ Hattie & O’Leary (2025) Educational Psychology Review 37:31:https://doi.org/10.1007/s10648-025-10002-w(HTML+PDF 双版本逐字交叉)
  21. ✓ Newton & Salvi (2020) Frontiers in Psychology(系统综述,相信率):https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2020.00820/full
  22. ✓ Dekker et al. (2012)「Neuromyths in education」Frontiers in Psychologyhttps://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2012.00429/full
  23. ✓ Ferrero et al. (2016) 神经神话相信率 Frontiers in Psychologyhttps://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2016.01594/full
  24. ✓ Brown (2023)「The persistence of matching teaching and learning styles…」Frontiers in Education 8:https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2023.1147498/full
  25. ✓ Ofsted《Education Inspection Framework: Overview of Research》(2019, 2021-02-22 更新) 官方 PDF:https://assets.publishing.service.gov.uk/media/6034be17d3bf7f265dbbe2ef/Research_for_EIF_framework_updated_references_22_Feb_2021.pdf
  26. ✓ Ofsted《Initial teacher education thematic monitoring visits: overview report》(2025-10):https://www.gov.uk/government/publications/initial-teacher-education-thematic-monitoring-visits-overview-report
  27. ✓ Nancekivell, Shah & Gelman (2020) J Educ Psych(本质化):https://doi.org/10.1037/edu0000408
  28. ✓ APA 官方新闻稿「Learning styles don’t exist, but they persist」(2019-05-30):https://www.apa.org/news/press/releases/2019/05/learning-styles
  29. ✓ Leite, Svinicki & Shi (2010) Educational and Psychological Measurement 70(2) 323–339:https://doi.org/10.1177/0013164409344507
  30. ✓ Dunlosky et al. (2013)「Improving Students’ Learning With Effective Learning Techniques」PSPI 14(1):https://journals.sagepub.com/doi/10.1177/1529100612453266
  31. ✓ Massa & Mayer (2006)(经 Pashler 转述+原文核验):https://doi.org/10.1037/0022-0663.98.1.238
  32. ✓ Melzner & Kappes (2024)「Testing the meshing hypothesis in prospective teachers」Instructional Science(官方摘要):https://doi.org/10.1007/s11251-024-09689-1
  33. ✓ Martin (2010)《Teachers using learning styles: Torn between research and accountability?》Teaching and Teacher Education 26(8) 1583–1591:https://doi.org/10.1016/j.tate.2010.06.009(DOI 经 Crossref 与 Teesside 研究门户交叉确认;正文经 Rogowsky 2020 逐字转述,含页码 1586)
  34. ⚠ Sun et al. (2023)(经 Hattie & O’Leary 2025 逐字转述):https://doi.org/10.1037/edu0000762
  35. ⚠ Stahl (1999)《Different strokes for different folks? A critique of learning styles》American Educator 23(3) 27–31(正文经 Kirschner 2017 逐字转述;无稳定公开 DOI,不挂链接)
  36. ✓ Witkin & Moore (1974)(ERIC 原件 ED097356):https://files.eric.ed.gov/fulltext/ED097356.pdf
  37. ✓ CAST UDL Guidelines 官方文本:https://udlguidelines.cast.org/
  38. ✓ VARK 官方产品/订阅页:https://vark-learn.com/articles/(文章与产品入口;订阅价格页直连 403,经 B 捆 vark-for-teachers 抓取件逐字核对)

诚实空位(各捆已登记,正文不引为承重):

  • Kolb (1984)《Experiential Learning》原书未抓取(转述级);Dunn & Dunn (1978) 原书未抓取(转述级)。
  • Pashler 全文无「strategy」一词——「风格 vs 策略」区分以 Hattie & O’Leary 2025 为准,不冒认 Pashler。
  • VARK 模态自述依赖 jina 提取文本(vark-learn.com 直连 403)。
  • Martin 2010 正文未取回(Rogowsky 2020 逐字转述含页码,标 ⚠)。
  • 2024 元分析附录表 B2(WWC 达标 5 项清单)在 JS 加载的补充材料,未抓实体表(数字句本身逐字到手)。
  • 美国州级与中国的匹配教学推荐措辞未逐字核实(教材页为正面讲授证据)。