Edexcel心理 记忆 工作记忆 重构记忆 图式
认知心理学在 Edexcel 的 A-level 心理学里只讲一个主题，就是记忆，从多存储模型到工作记忆模型，从图式重建到个案 HM，内容看着集中，考法却从 2 分短答一路铺到 8 分长答。9PS0 的 Paper 1 把这一部分放在 Section B，官方 spec 的 Issue 4 在 2026 年 3 月调整了卷面时长，Section B 的情境题多、分值高，只背定义的答法很难拿到评价分。下面按多存储模型、工作记忆模型、情节与语义记忆、重构记忆与图式、经典研究 Baddeley（1966b）、个案 HM、发育与个体差异、方法检验、三种问法的落笔方式展开，每一节都对着评分口径里真正给分的动作写。
Cognitive psychology in Edexcel A-level Psychology covers a single theme, memory, running from the multi-store model through working memory, schematic reconstruction and the case of HM. It sits in Section B of Paper 1 for 9PS0, and Pearson's Issue 4 of the specification changed the length of Papers 1 and 2 in March 2026, which matters when you rehearse under timed conditions. Each section below pairs the content with the action that actually earns marks, drawn from the mark schemes and examiner reports rather than a summary of the textbook.
卷面位置：Section B 的 70 分与 Paper 3 的经典研究题
认知心理学在 Edexcel 的 9PS0 里排在 Paper 1 Foundations in Psychology 的 Section B，整卷 90 分、占 35%，Section A 到 D 合计 70 分，议题与争论的 Section E 另占 20 分。Pearson 在 2026 年 3 月发布的 spec Issue 4 把 Paper 1 与 Paper 2 的考试时长从 2 小时改为 2 小时 15 分，按旧版时长排模拟考的同学要把这条改动折算进训练节奏。Section B 的分值从 2 分短答一直排到 8 分长答，分值越高越要求先把模型或研究的程序写完整，再落到评价。经典研究 Baddeley（1966b）还会在 Paper 3 Psychological Skills 的 Section B 出现，那一卷 2 小时、80 分、占 30%，Section B 占 24 分，其中一道长答题指定围绕 Topics 1 到 5 的经典研究作答。很多学生以为 Paper 1 考过的内容不会再出现，Paper 3 恰恰是要求把同一批研究搬进新情境的 synoptic 卷。常见的错误是只背模型名词，把 8 分题里的评价分整段空着。
Cognitive psychology occupies Section B of Paper 1 Foundations in Psychology in Edexcel 9PS0, a paper worth 90 marks and 35% of the qualification, with Sections A to D carrying 70 marks and the issues and debates section adding 20. Pearson's Issue 4 of the specification, published in March 2026, extended Papers 1 and 2 from two hours to two hours and fifteen minutes, so timed practice built on the older format understates how much reading and checking time is now available. The classic study Baddeley (1966b) returns in Section B of Paper 3 Psychological Skills, an 80-mark, two-hour paper worth 30%; that section carries 24 marks and one extended answer built around the classic studies from Topics 1 to 5.
多存储模型：感觉登记、5 到 9 项与语义编码
Atkinson 和 Shiffrin 在 1968 年提出的多存储模型把记忆拆成三个储存，信息先进入感觉登记，视觉痕迹在图像记忆里停留不到一秒，声觉信息的停留稍长一点。被注意到的信息进入短时储存，Pearson 的评分方案写的是容量 5 到 9 项、时长最多 30 秒、以声觉方式编码，没有经过复述的信息就在这个阶段丢失。复述把信息送进长时储存，那里的容量与保存时间都被视为无限，编码方式是语义，这个差别是三个储存里最容易被考到的一条。多存储模型对自由回忆里首因与近因效应的解释也是常考点，列表开头的词因为复述多而进入长时储存，列表末尾的词还留在短时储存里，中段词两头不靠，记得最少。很多学生以为短时记忆的容量就是 Miller 的 7±2，官方评分方案用的是 5 到 9 项这个区间，写死成 7 个容易被判过窄。常见的错误是把三个储存的编码方式张冠李戴，把长时储存写成声觉编码。
Atkinson and Shiffrin proposed the multi-store model in 1968, splitting memory into a sensory register where visual traces fade within about a second, a short-term store, and a long-term store. Pearson's mark scheme describes the short-term store as holding five to nine items for up to thirty seconds with acoustic encoding, while the long-term store is unlimited in both capacity and duration and encodes semantically; rehearsal is the mechanism that moves material between the two stores. The model also predicts the serial position curve, because early words collect more rehearsal and reach long-term memory, the final few survive in short-term memory, and middle items benefit from neither.
工作记忆模型：中央执行、语音回路与情节缓冲
Baddeley 和 Hitch 在 1974 年把短时记忆换成一个由多个成分组成的工作系统，中央执行负责分配注意与协调子系统，它本身不做储存。语音回路分成语音储存与发音控制过程两部分，声音信息先进入储存，再由默读式复述把它维持住；视觉与空间信息走视觉空间模板，Smith 和 Jonides 在 1997 年用 PET 扫描发现视觉任务与空间任务激活的脑区不同，支持这个模板不是单一储存。情节缓冲在 2001 年由 Baddeley 补进模型，用来说明不同通道的信息怎样整合成一段有情境的片段。个案 KF 是支持这个模型的证据，他术后对言语材料的短时记忆明显受损，视觉材料的表现却大致保留，这种分离很难用单一短时储存解释。很多学生以为工作记忆就是短时记忆换个名字，两个模型的差别在于工作记忆强调同时加工与储存。学生最常把中央执行写成负责储存的仓库，评分点上它只负责控制与分配。
Baddeley and Hitch replaced the single short-term store in 1974 with a system of components, headed by a central executive that allocates attention and coordinates the slave systems without storing anything itself. The phonological loop holds verbal material through a phonological store plus an articulatory control process, the visuo-spatial sketchpad handles visual and spatial information, and Smith and Jonides showed in 1997 that PET scans taken during visual and spatial tasks activate different regions, evidence that the sketchpad is not one store. Baddeley added the episodic buffer in 2001 to explain how information from different modalities is bound into coherent episodes, and the case of KF supports the model because his verbal short-term memory was severely impaired while visual performance survived largely intact.
情节记忆与语义记忆：Tulving 1972 的两分法
Tulving 在 1972 年把长时记忆拆成情节记忆与语义记忆，情节记忆保存带时间与地点标记的个人经历，提取时伴随重新体验的感觉；语义记忆保存不依赖情境的事实与概念，比如单词含义与常识。判别的抓手是提取时要不要回到事情发生的那一刻，能回到的是情节记忆，只能陈述事实的是语义记忆，题目给出的情境通常就卡在这条线上。这个两分法在 HM 这类病例上最有解释力，他的手术切除了双侧内侧颞叶之后，新的个人经历几乎进不了长时记忆，早年获得的语言与常识却大体可用。Schmolck 等人在 2002 年比较 HM 与其他双侧颞叶损伤患者的语义知识，发现手术之后才固定下来的词汇他基本解释不了，早年学会的概念相对保留。很多学生以为情节与语义是两套互不相干的系统，新情节其实要靠语义框架才组织得起来，两者的边界在发育早期比成年后模糊。
Tulving split long-term memory in 1972 into episodic memory, which stores personally experienced events tagged with time and place, and semantic memory, which stores facts and concepts that stand independent of any single episode. The practical test is retrieval: if you can travel back to the moment something happened you are drawing on episodic memory, whereas stating what a word means relies on semantic memory, and most scenario questions sit exactly on that line. Patient HM illustrates the split, because after bilateral medial temporal surgery he could not lay down new personal experiences while his earlier language and general knowledge stayed largely usable, and Schmolck et al. reported in 2002 that vocabulary whose meaning became fixed only after his operation was essentially inaccessible to him.
重构记忆与图式：Bartlett 1932 的重复复述
Bartlett 在 1932 年的《Remembering》里提出记忆不是把材料原样取回，而是用图式重建的过程，图式是过去经验形成的组织框架，它决定新信息被吸收进哪个结构。他用「鬼的战争」这个故事做重复复述与系列复述，参与复述的多是剑桥的学生，陌生文化的细节被换成英格兰读者熟悉的说法，故事长度普遍变短，结构逐渐贴近本土叙事习惯。他由此提出 effort after meaning 这个概念，人会把材料往自己已有的框架里塞，越努力解释越偏离原文。这套结论后来常被用来讨论目击证词，重新被问话、与他人讨论、看到媒体报道都会改写原本的记忆，这也是警方要求证人单独陈述的原因。评价这一节要写清程序上的弱点，故事复述没有统一的编码方式，被试知道自己在复述而不是在回忆，结论更多说明重建的倾向，给不出定量的误差。很多学生以为记忆像录像，Bartlett 的实验里没有人复述出原文细节。
Bartlett argued in Remembering (1932) that recall is a reconstruction guided by schemas rather than a playback of stored material. Using repeated and serial reproduction of a folk tale, largely with Cambridge students, he watched unfamiliar cultural details turn into familiar English equivalents, the story shorten, and its shape drift towards local narrative habits, a pattern he summarised as an effort after meaning. The findings carry into eyewitness work, where re-questioning, discussing an event with others or seeing media coverage all reshape what a witness later reports, though the procedure lacked standardised coding and asked people to reproduce rather than recall, so the size of the distortion cannot be quantified.
经典研究 Baddeley（1966b）：四个词表与编码结论
Baddeley 在 1966 年发表的第三个实验用四个独立组做词表学习，四份十词词表按声似、声不似、义似、义不似分成四类，被试来自应用心理学研究单位的被试名册。每人按顺序学习自己的词表四遍，第五次复测安排在 20 分钟之后，中间的间隔用分散任务填满，目的是清掉短时记忆里的痕迹。结果分两层，学习过程中的顺序回忆在声似词表上最差，20 分钟之后的长时回忆在义似词表上最差。这个结果支持短时记忆用声觉编码、长时记忆用语义编码，也让 Baddeley 开始怀疑单一短时储存的假设，几年之后他与 Hitch 合作提出了工作记忆模型。方法上，20 分钟的延迟是控制变量，四个词表之间的差异是自变量，同一被试的第 4 次与第 5 次成绩构成重复测量设计。评价时写清词表都是互不关联的单词、缺少句法线索，结论推广到句子与篇章需要打折。
Baddeley's third experiment in 1966 ran four independent groups, each learning a ten-word list drawn from a pool that varied acoustic similarity and semantic similarity. Every participant practised their list in order four times, and the fifth recall trial came after a twenty-minute delay filled with a distractor task to clear short-term traces; immediate order recall suffered most on the acoustically similar lists, while delayed recall suffered most on the semantically similar lists. That pattern supported acoustic encoding in short-term memory and semantic encoding in long-term memory, pushed Baddeley away from a single short-term store, and led to the working memory model, though the unrelated single words used here say little about sentences and longer prose.
个案 HM 与案例研究这种方法的操作
HM 是认知这一节里唯一被点名的个案，Scoville 在 1953 年为控制他的癫痫切除了双侧内侧颞叶的一部分，术后他出现严重的顺行性遗忘，新的个人经历留不住，智力与语言没有出现明显下降。他能学会镜子描画这类操作技能，却记不住自己练过这些任务，程序性学习与陈述性记忆因此被放到两个抽屉里讨论。Edexcel 要求的不只是背症状，还要写清案例研究作为方法的操作方式，包括用纵向观察与访谈收集定性数据、用记忆测验收集定量数据，再讨论这种组合的证据等级。这类研究的强项是能接触到无法用实验复制的罕见损伤，一名患者在几十年里被反复测量，数据的细节密度远高于实验室样本。弱点是样本只有一个人、无法重复，切除范围与术前病史同时变化，健康组织被牵连的程度也不完全一致。很多学生以为个案研究能直接推出因果，单个案例加上手术本身的复杂性，通常只支持方向性的结论。
HM is the one named individual in the cognitive topic: Scoville's 1953 operation to relieve his epilepsy removed parts of both medial temporal lobes, leaving severe anterograde amnesia in which new personal experiences could not be retained while intelligence and language stayed largely intact. He could acquire motor skills such as mirror drawing yet had no memory of having practised them, a separation that helped mark the boundary between procedural and declarative memory; Edexcel asks for the method as well as the symptoms, including longitudinal observation and interviews for qualitative data and memory tests for quantitative data. The obvious price is that a single unrepeatable case cannot establish cause, since surgical extent and pre-existing history shift together and the surrounding healthy tissue is never spared to the same degree twice.
发育与个体差异：数字广度从 5 岁到 17 岁
发育这条线在官方 spec 里给了三个可选方向，数字广度随年龄增长的证据来自 Sebastián 和 Hernández-Gil 在 2012 年对西班牙儿童与青少年的测量，5 岁的广度很低，随后逐年上升，到 17 岁前后接近成年水平。第二个方向是阅读障碍，工作记忆与短时记忆的容量偏小会直接拖累解码与阅读理解，语音回路环节的缺陷是最常被提到的机制。第三个方向是阿尔茨海默病，早年记忆相对保留而近期经历快速流失，这个模式与内侧颞叶及海马的病变位置吻合。个体差异部分要求写清两条线索，加工速度慢的学生在同一套材料上表现更差，图式结构不同的学生对同一段材料重建出的版本也不一样。自传体记忆尤其私人化，同一段经历在不同人的回忆里连时间顺序都可能换位。阅卷时看到只写年龄数字、不区分记忆类型的答案，通常拿不到展开分。
The developmental strand of the specification offers three routes, the first being the digit span data reported by Sebastián and Hernández-Gil in 2012 on Spanish children and adolescents, where span is low at five years old and climbs steadily towards adult levels around seventeen. Dyslexia is the second route, since reduced short-term and working memory capacity constrains decoding and reading comprehension, with the phonological loop the mechanism most often implicated, and Alzheimer's disease is the third, where earlier memories hold up while recent experience drains away. Individual differences add two more levers, because slower processing speed depresses performance on the same materials and differing schemas make two people reconstruct one event differently, which is why autobiographical memory resists averaging.
方法题：Mann-Whitney U 与 Wilcoxon 的选择
认知这一节的方法部分围绕实验展开，独立变量与因变量的操作化、实验假设与零假设、方向性与非方向性假设都是短答的常客。设计决定检验方法，独立组设计比较两组人的得分用 Mann-Whitney U，重复测量或配对设计比较同一批人在两种条件下的得分用 Wilcoxon，题目里出现前后对照或者同一批被试的两个条件，就该想到 Wilcoxon 并写清选择理由。显著性水平在试卷里按 p≤0.10、p≤0.05、p≤0.01 三档给出，临界值表随卷附上，判断时先比观测值与临界值，再回到具体研究情境说明是否拒绝零假设。反平衡与随机化用来压住顺序效应，前后测之间安排分散任务是常见的控制手段，被试猜到研究目的带来的需求特征会削弱内部效度。常见的错误是把类型一与类型二错误说反，类型一是本来没有真实差异却拒绝了零假设。阅卷时，只报检验名称而不写选择理由的答案通常只拿到一半的分。
The methods strand of cognitive psychology is built on experiments, so operationalising the independent and dependent variables, writing experimental and null hypotheses, and choosing between directional and non-directional wording all appear as short-answer questions. Design dictates the test: independent groups compared on one score call for the Mann-Whitney U test, while repeated measures or matched pairs comparing the same people under two conditions call for the Wilcoxon test, so a scenario showing a before-and-after measure should trigger Wilcoxon together with a justification. Significance is offered at p ≤ 0.10, p ≤ 0.05 and p ≤ 0.01 with critical value tables supplied in the paper, and counterbalancing, randomisation and distractor tasks are the standard controls against order effects and demand characteristics.
AO1、AO2、AO3 三种问法的落笔差别
Edexcel 的评分口径把 30% 到 35% 的分给 AO1，另外 30% 到 35% 给 AO2，分析与评价占 35% 到 40%，三类问法的落笔方式差别很大。AO1 要求把模型或研究的程序与结果写完整，包括研究设计、被试情况与结果的方向；AO2 要求在题目给出的情境里指认出具体环节，2021 年 11 月的评分方案里有一道场景题要求学生把漏记教室物品解释成图式重建的结果，官方说明写明通用答案给 0 分。AO3 要求给出带推理链的评价，2024 年 6 月的考试报告提到，很多考生描述了工作记忆模型的成分，却没有说明它比多存储模型强在哪里，评价分就是这样丢掉的。备考时给每个模型准备两条支持证据与两条削弱证据，每条证据后面接一句为什么。常见的错误是把研究者名字排成清单当作评价，名字本身不构成论证。
Edexcel allocates 30 to 35% of the marks to AO1, a further 30 to 35% to AO2 and 35 to 40% to analysis and evaluation, and the three command words demand genuinely different writing. AO1 wants the model or the study set out completely, including design, participants and the direction of the results, while AO2 requires naming the relevant component inside the scenario printed on the paper, as in a November 2021 question that expected candidates to explain a student's misremembered classroom contents through schematic reconstruction; the mark scheme states plainly that generic answers score zero. AO3 wants a chain of reasoning, and Pearson's June 2024 examiner report notes that candidates often described the components of the working memory model without saying why it explains more than the multi-store model, which is where the evaluation marks were lost.
牛津高阶英汉双解词典 第10版 商务印书馆¥168📱 去京东购买京东精选
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
16621398022 同微信
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图