Edexcel心理 多存储模型 工作记忆 图式
认知心理学研究的是看不见的内部过程：注意、记忆、语言与问题解决。Edexcel 9PS0 把认知心理学排在 Topic 2，考在 Paper 1 的 Section B。Paper 1 全卷 90 分，2026 年 3 月更新的 Issue 4 把时长定为 2 小时 15 分，占总成绩 35%。Sections A 到 D 共 70 分，Section E 的议题与争论占 20 分，认知部分在整卷里的分量并不轻。很多学生复习时把所有模型背成一张流程图，动笔却分不清哪一问要描述、哪一问要评价。阅卷时看到的现象是，能背出三个仓的名字但写不出编码、容量、时长三项具体差异的答案，通常停在低分档。这部分内容在 Pearson Edexcel 官网的 specification 里逐条列出，复习前值得先对照一遍。
Cognitive psychology deals with processes you cannot watch directly, which is why the exam leans so heavily on inference from measurable behaviour. In Paper 1 Section B, Pearson Edexcel expects you to describe a model, apply it to a scenario, and then weigh the evidence behind it; a candidate who only memorises the flow diagram rarely reaches the upper bands. Many students lose marks here because they confuse coding with capacity, and the specification is precise about which feature each study was designed to test.
认知心理学的定位与 Paper 1 Section B 的版面
版面先要弄清，否则复习重点会跑偏。Edexcel 9PS0 的 Paper 1 叫 Foundations in psychology，认知心理学是其中的 Topic 2，对应 Section B。这一节的题目既可能给一个情境让你用模型解释，也可能直接问某研究支持了什么。Topic 2 的内容清单在 spec 里写得很死：多存储模型、情节与语义记忆、工作记忆模型、重建记忆与图式、个体记忆差异、一个经典研究、一个当代研究。复习时按清单逐条打勾比按教材章节推进更省时间。分值分布上，Section B 的简答题与短文题混排，短答通常四到六分，评价型长答会要到八分以上。
The layout matters before anything else, because revising by textbook chapter rather than by specification heading wastes time. Topic 2 sits inside Section B of Paper 1, and questions there either hand you a scenario to explain or ask what a named study established. The content list is explicit, covering the multi-store model, episodic and semantic memory, working memory, reconstructive memory and schema theory, individual differences, one classic study and one contemporary study. Short answers usually carry four to six marks, while an evaluation-style response can ask for eight or more.
多存储模型的三个仓：感觉登记、短时记忆与长时记忆
多存储模型由 Atkinson 与 Shiffrin 在 1968 年提出，把记忆写成一条单向的信息流。外界刺激先进入感觉登记，被注意到的部分进入短时记忆，经过复述再转入长时记忆。三个仓的差别不在名字，而在编码、容量、时长三项特征。感觉登记按感觉通道分别编码，短时记忆以声音为主，长时记忆以意义为主。模型还规定了两道控制闸门：注意决定信息能否进入短时记忆，复述决定信息能否留在短时记忆并转入长时记忆。提取是唯一的反向通道，把长时记忆里的内容拉回短时记忆供当下使用。很多学生以为这条线是自动发生的，实际上每一道闸门都可以被打断，这正好解释了分心时为什么记不住刚听到的号码。
Atkinson and Shiffrin (1968) described memory as a one-way flow through three stores, and the appeal of the model lies in how precisely each store can be tested. Attention gates entry into short-term memory, rehearsal keeps material there and moves it onward, while retrieval is the only route back; break any gate and the memory fails. That is why a phone number heard while you are reading survives for seconds only, and why the model is still worth learning as a benchmark rather than as a finished answer.
感觉登记的编码、容量与时长：Sperling 的部分报告法
感觉登记保存的是尚未加工的原始信息，视觉形式叫图像记忆，听觉形式叫回声记忆。Sperling 在 1960 年用部分报告法测它的容量：要求报告整张字母表时只能说出四五个，改为按音调提示只报告一行时，几乎每一行都能报全。这个结果说明图像记忆的容量远大于短时记忆，只是衰减极快，视觉痕迹在不到一秒内就淡掉。回声记忆维持得稍久，大约两到四秒，这解释了没听清一句话时还能在脑子里回放一遍。复习时容易忽略的一点是，感觉登记的编码是通道特异的，它不给信息赋予意义。常见的错误是把它当成短时记忆的入门仓，两者在时长上差着一个数量级。
The sensory register holds raw, unprocessed input, and Sperling's partial-report technique is the cleanest demonstration of how much it can hold. When a tone told participants which row to report, performance rose sharply, showing that the trace was briefly complete before fading in under a second. Echoic memory lasts a little longer, roughly two to four seconds, which is why a half-heard sentence can still be replayed in your head; modality-specific coding is the point to remember, not the store's name.
短时记忆的三个数字：Miller 的七加减二与 Peterson 的十八秒
短时记忆的三项特征各有一个经典数字支撑。容量方面，Miller 在 1956 年给出七加减二，并指出分块可以把有效容量放大。时长方面，Peterson 与 Peterson 在 1959 年让被试看三个辅音字母，用倒数三个数字阻断复述，三秒后回忆率约八成，十八秒后掉到不足百分之五。编码方面，Baddeley 在 1966 年发现声音相似的词表在即时回忆中明显更差，说明短时记忆偏向声码。三个数字要分开记，不能混成一句话。很多学生以为十八秒是短时记忆的绝对上限，其实那是没有复述条件下的衰减速度。
Three classic numbers anchor short-term memory: Miller's seven plus or minus two for capacity, Peterson and Peterson's eighteen seconds for duration without rehearsal, and Baddeley's acoustic similarity effect for coding. The trigram task matters because counting backwards blocks rehearsal, so the fall in recall measures decay rather than a storage limit; after about eighteen seconds almost nothing survives. Each number answers a different question, and swapping them is one of the commonest ways to lose a mark.
长时记忆的容量与时长：Bahrick 对毕业照的追踪
长时记忆的容量没有测出上限，时长通常按几十年计。Bahrick 等人在 1975 年追踪一批高中毕业生，用毕业纪念册上的照片测试对同学姓名的记忆。十五年后，再认成绩仍在九成左右；四十八年后仍有约七成，自由回忆则掉得更快。这组数据说明长时记忆的存储非常耐久，衰退更多出现在提取环节而不是存储环节。复习时要能读出这个区别，因为它是后面解释遗忘的伏笔。常见的错误是把再认成绩当成回忆成绩引用，两个数字差得很远。
Bahrick et al. (1975) tracked former classmates through yearbook photographs and found recognition still near ninety per cent after fifteen years and around seventy per cent after forty-eight. Free recall fell much faster, which points to a retrieval problem rather than a storage problem. That distinction is worth stating explicitly, because it sets up later explanations of forgetting and separates durability of storage from ease of access.
多存储模型的证据与漏洞：HM、串行位置效应与双任务
支持多存储模型的证据主要有三条线。临床方面，HM 在 1953 年因癫痫切除大部分内侧颞叶后，短时记忆正常却无法形成新的长时记忆，这说明两个仓可以分开受损。实验方面，自由回忆的词表呈现串行位置效应，开头与结尾记得好、中间最差，Glanzer 与 Cunitz 在 1966 年用延迟任务把结尾优势打掉，进一步支持两个独立仓。批评同样明确：模型把短时记忆写成一个单一仓，双任务研究却显示视觉与言语任务可以并行，代价很小。复述也不是通往长时记忆的唯一路径，Craik 与 Lockhart 在 1972 年指出加工深度比复述次数更重要。
Three lines of evidence support the model: patient HM lost the ability to form new long-term memories while his short-term span survived, the serial position curve splits neatly into two components, and a delayed recall task removes the recency advantage. The criticism is equally clear, since dual-task studies show visual and verbal work proceeding in parallel with little cost, which a single short-term store cannot explain. Craik and Lockhart added a second problem in 1972 by arguing that depth of processing, not repetition, decides what is stored.
Tulving 的两种长时记忆：情节记忆与语义记忆
Tulving 在 1972 年提出长时记忆不是单一仓，至少可以分成情节记忆与语义记忆。情节记忆存的是带时间与地点标签的个人经历，回忆时会有重新经历的感觉。语义记忆存的是事实与概念，比如巴黎是法国首都，它不与某个具体时刻绑定。两者的差别还体现在遗忘速度上，情节记忆掉得更快，语义记忆更耐久。Edexcel 只要求这两种，复习时把程序性记忆提一句会更清楚，因为它正好解释 HM 为什么还能学会新的动作技能却否认自己练过。
Tulving (1972) argued that long-term memory is not one store, dividing it into episodic memory for time-stamped personal events and semantic memory for facts detached from any particular occasion. Episodic traces fade faster, while semantic knowledge survives damage that wipes out autobiographical detail. Procedural memory is worth a mention as well, because it explains how HM could acquire a new motor skill while denying he had ever practised it.
工作记忆模型：中央执行与两个从属系统
工作记忆模型由 Baddeley 与 Hitch 在 1974 年提出，用一组组件替换了多存储模型里那个被动的短时仓。中央执行是注意力的分配者，容量有限，不处理具体通道的信息。语音环路管声音，分成语音存储与发音复述两部分，前者像内耳，后者像内语。视空间画板管图像与空间位置，常被称作内眼。双任务实验是这套模型最有力的证据：一个视觉任务配一个言语任务，两者互不干扰；换成两个言语任务，成绩立刻塌下来。
Baddeley and Hitch (1974) replaced the passive short-term store with an active system: a limited-capacity central executive that allocates attention, a phonological loop for sound, and a visuo-spatial sketchpad for images and locations. The dual-task design carries the argument, because a visual task plus a verbal task runs almost cost-free while two verbal tasks collapse together. That pattern is very hard to square with a single short-term store, and it is the first thing to reach for when a question asks what the model explains better.
情景缓冲器与 KF：工作记忆模型后来补上的两块
情景缓冲器是 Baddeley 在 2000 年补上的第四个组件，容量约四个组块，负责把各通道的信息整合成连贯的情节，并与长时记忆对接。加它的原因很实际：语音环路与视空间画板各自独立，却解释不了人为什么能把画面与句子合成一段回忆。病例 KF 提供了另一类证据，Shallice 与 Warrington 在 1970 年报告他视空间短时记忆严重受损，言语短时记忆相对完好。分离性损伤支持组件各司其职的说法。这条证据链在评价题里很好用，因为它同时说明模型的解释力和它对临床数据的依赖。
The episodic buffer, added by Baddeley in 2000, holds about four chunks and binds separate streams into a single coherent episode with links to long-term memory. Without it the model cannot explain how a scene and a sentence become one recollection. Patient KF, reported by Shallice and Warrington in 1970, showed the reverse dissociation with spatial span badly damaged and verbal span largely intact, which supports separable components rather than one general store.
重建记忆与图式：Bartlett 的鬼魂之战
重建记忆的思路与存储模型相反，它认为回忆是把碎片重新拼起来的过程。Bartlett 在 1932 年让英国被试反复复述一个北美原住民故事《鬼魂之战》，得到的版本越来越短，陌生的细节被删掉或被换成熟悉的事物，独木舟变成了船。他把这些变化归为三类：削平、锐化与同化，动力来自图式。图式是过去经验形成的知识框架，它让理解变快，也让回忆带上预期。Allport 与 Postman 在 1947 年的实验重复了同类现象，被试把图画里的物件换成自己熟悉的版本。这个传统后来直接连到目击证词的可靠性问题，也是犯罪心理学里反复出现的接口。
Bartlett (1932) asked English participants to retell a Native American story repeatedly and watched it shorten, lose unfamiliar detail, and drift towards familiar objects such as boats. He described the changes as levelling, sharpening and assimilation, all driven by schemas that make comprehension fast and recall expectant. Allport and Postman (1947) reproduced the same pull towards the familiar in pictures, which is why eyewitness accounts deserve caution and why the tradition still matters in criminological psychology.
经典研究 Baddeley 1966b：四组词表与二十分钟延迟
经典研究 Baddeley 1966b 的任务设计不复杂，逻辑却很清楚。被试学习四类词表：声音相似、声音不同、意义相似、意义不同。学习后分两种条件测试，一种立刻回忆，测短时记忆；另一种先做二十分钟干扰任务再回忆，测长时记忆。结果分成两半：即时回忆下声音相似组最差，延迟回忆下意义相似组最差，声音效应基本消失。结论是短时记忆以声码为主，长时记忆以义码为主。评价要落到方法与生态效度上，词表是人工材料，结论能否推广到日常阅读仍有争议。
Baddeley (1966b) crossed two kinds of similarity with two retention conditions, using acoustically similar, acoustically different, semantically similar and semantically different lists. Immediate recall punished acoustic similarity, while delayed recall after a twenty-minute interference task punished semantic similarity, and the acoustic effect largely vanished. The conclusion, that short-term coding is acoustic and long-term coding semantic, rests on artificial word lists, so ecological validity is the obvious line of criticism alongside the artificiality of the task itself.
当代研究 Schmolck 2002：HM 的语义知识与颞叶损伤
当代研究 Schmolck 等人在 2002 年把语义记忆放到脑损伤的框架里检验。研究对象是 HM 以及其他有双侧内侧与外侧颞叶损伤的患者，任务是成套的语义知识测验，涵盖词汇、事实与概念。结果显示，海马与鼻周皮层受损越重，语义记忆的缺损越明显，两者呈正相关。这个发现把 Tulving 关于长时记忆分型的说法落到了具体脑区上。评价时要注意样本量小、损伤范围无法实验控制，结论只能说明相关，不能说明因果。
Schmolck et al. (2002) tested HM and other patients with bilateral medial and lateral temporal lobe damage on a battery of semantic knowledge tasks. The severity of hippocampal and rhinal cortex damage tracked the size of the semantic deficit, which puts Tulving's division of long-term memory onto specific tissue. Sample size is small and the lesions were never under experimental control, so the finding is correlational rather than causal, and generalisation beyond these patients is limited.
个体差异、临床联系与三个固定丢分点
个体差异这部分常被跳过，考试却会直接问。记忆广度在童年期随年龄上升，Sebastian 与 Hernandez-Gil 在 2012 年用西班牙样本做了数字广度的测量，跨度随年龄增长。临床对照更明显，阿尔茨海默病先侵蚀情节记忆，阅读障碍常伴随语音环路的弱点。答题时的三个丢分点很固定：只描述不评价，结论写了数据没写，编码与容量混在一句话里。AO1 描述、AO2 应用到情景、AO3 评价三条要分开写，分值高的题通常同时要评价与判断。这一步最容易失分的地方，是把 HM 当成因果证明而不是分离性证据。
Individual differences are examinable, and the developmental evidence is straightforward: digit span rises through childhood, as Sebastian and Hernandez-Gil mapped in a Spanish sample in 2012. Alzheimer's disease strips episodic memory first, while dyslexia is often linked to weaknesses in the phonological loop, so the models do real clinical work. Marks are lost in three predictable ways, by describing without evaluating, by stating a conclusion without the supporting figure, and by merging coding with capacity in one sentence; sectioning the answer into description, application and evaluation prevents all three.
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
A-Level / GCSE / IB / KS2-KS3 全科辅导
更多咨询请联系16621398022（同微信）
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图