📚 Psychology Experimental Research Methods: Variable Control and Design Essentials | 心理学实验研究方法：变量控制与设计要点
Experimental research is the backbone of scientific psychology. It allows researchers to establish cause-and-effect relationships by systematically manipulating variables while controlling extraneous influences. This article provides a comprehensive guide to variable control and experimental design, tailored for A-level and IB psychology students.
实验研究是科学心理学的基石。它通过系统操控变量并控制额外因素，使研究者能够建立因果关系。本文为 A-level 和 IB 心理学考生提供关于变量控制与实验设计的全面指南。
1. Key Variables in Psychological Experiments | 心理学实验中的关键变量
In any experiment, three types of variables are central: the independent variable (IV), the dependent variable (DV), and extraneous variables (EVs). The IV is what the researcher manipulates; the DV is what is measured; EVs are unwanted factors that may influence the DV.
在任何实验中，有三类核心变量：自变量（IV）、因变量（DV）和额外变量（EV）。自变量是研究者操控的变量；因变量是被测量的变量；额外变量是可能影响因变量的非预期因素。
- Independent Variable (IV) | 自变量：
The condition or event that is changed systematically by the researcher. For example, in a memory study, the IV might be the type of word list (visual vs. verbal).
研究者系统改变的条件或事件。例如，在记忆研究中，自变量可能是单词列表的类型（视觉 vs. 言语）。
- Dependent Variable (DV) | 因变量：
The measurable outcome that is expected to change as a result of the IV manipulation. In the memory example, the DV could be the number of words correctly recalled.
预期随自变量操控而改变的可测量结果。在记忆例子中，因变量可以是正确回忆的单词数量。
- Extraneous Variables (EVs) | 额外变量：
Any variable other than the IV that could affect the DV. These include participant variables (age, IQ, mood), situational variables (noise, temperature, time of day), and experimenter effects.
除自变量外可能影响因变量的任何变量，包括被试变量（年龄、智商、情绪）、情境变量（噪音、温度、一天中的时间）和实验者效应。
To draw valid conclusions, researchers must ensure that changes in the DV are caused only by the IV, not by EVs. This is the essence of internal validity.
要得出有效结论，研究者必须确保因变量的变化仅由自变量引起，而非额外变量。这正是内部效度的核心。
2. Operationalisation of Variables | 变量的操作化定义
Operationalisation means defining variables in measurable, observable terms. For instance, "anxiety" cannot be measured directly; it must be operationalised as a physiological response (e.g., heart rate) or a self-report score on a standardised scale.
操作化是指用可测量、可观察的术语来定义变量。例如，“焦虑”无法直接测量，必须将其操作化为生理反应（如心率）或标准化量表上的自我报告分数。
IV operationalised: "Presentation of a stressful video (5 min) vs. a neutral video (5 min)"
DV operationalised: "Heart rate (BPM) measured immediately after viewing"
自变量操作化：“播放压力视频（5分钟）vs. 中性视频（5分钟）”
因变量操作化：“观看后立即测量心率（次/分钟）”
Good operationalisation improves reliability (repeatability) and validity (accuracy). Without clear operational definitions, replication becomes impossible and results are ambiguous.
良好的操作化提高信度（可重复性）和效度（准确性）。没有清晰的操作定义，重复研究变得不可能，结果也会模糊不清。
3. Experimental Designs: Repeated Measures, Independent Groups, Matched Pairs | 实验设计：重复测量、独立组、匹配组
Choosing an experimental design is a critical decision. Each design has strengths and weaknesses regarding participant variables, order effects, and demand characteristics.
选择实验设计是至关重要的决定。每种设计在参与者变量、顺序效应和需求特征方面各有优劣。
| Design | 设计 | Procedure | 程序 | Advantages | 优点 | Disadvantages | 缺点 |
|---|---|---|---|
| Repeated Measures 重复测量 |
Same participants complete all conditions 同一组被试完成所有条件 |
No participant variables; fewer people needed 无被试变量；所需人数较少 |
Order effects (practice/fatigue); demand characteristics 顺序效应（练习/疲劳）；需求特征 |
| Independent Groups 独立组设计 |
Different participants in each condition 不同被试分别进入不同条件 |
No order effects; less chance of demand characteristics 无顺序效应；需求特征可能性较低 |
Participant variables may confound results; more participants required 被试变量可能混淆结果；需要更多被试 |
| Matched Pairs 匹配组设计 |
Different but similar participants matched on key traits 不同但关键特质匹配的被试 |
Controls participant variables; no order effects 控制被试变量；无顺序效应 |
Time-consuming matching; impossible to match all traits 匹配耗时；无法匹配所有特质 |
Counterbalancing is often used in repeated measures designs to reduce order effects. For example, half the participants complete condition A first, the other half complete condition B first (AB/BA design).
在重复测量设计中常使用平衡设计来减少顺序效应。例如，一半被试先完成条件A，另一半先完成条件B（AB/BA设计）。
4. Controlling Participant Variables | 控制被试变量
Participant variables such as age, gender, intelligence, personality, and motivation can become confounds if not controlled. Strategies include random allocation, matching, and using a within-subjects design.
年龄、性别、智力、人格和动机等被试变量如果不加控制就可能成为混淆因素。控制策略包括随机分配、匹配和使用被试内设计。
- Random Allocation | 随机分配：
Participants are assigned to conditions by chance (e.g., using a random number generator). This distributes participant variables evenly across conditions.
被试通过随机方式（如随机数生成器）被分配到各条件。这使被试变量在各条件中均匀分布。
- Matching | 匹配：
Participants are pre-tested on a relevant trait and then paired with others of similar scores. One member of each pair goes to condition A, the other to condition B.
先对被试的相关特质进行前测，然后将得分相近的参与者配对。每对中的一名成员进入条件A，另一名进入条件B。
- Standardised Procedures | 标准化程序：
All participants should experience identical instructions, timings, and environmental conditions. This prevents situational variables from systematically affecting one group more than another.
所有被试应接受相同的指导语、时间和环境条件。这防止情境变量系统性地对某组产生更大影响。
In practice, researchers often combine random allocation with strict standardisation to maximise control.
实践中，研究者常常将随机分配与严格标准化结合，以最大化控制。
5. Controlling Situational Variables | 控制情境变量
Situational variables are environmental factors such as lighting, noise, temperature, and time of day. They can be controlled through laboratory settings, standardised protocols, and automation.
情境变量是光照、噪音、温度和一天中的时间等环境因素。它们可以通过实验室环境、标准化方案和自动化加以控制。
Example | 示例:
All participants tested in the same quiet room at 10:00 AM, with identical lighting and seating.
所有被试在同一安静房间、上午10:00、相同照明和座位条件下进行测试。
Double-blind procedures control experimenter effects: neither the participant nor the experimenter knows which condition is being administered. This prevents subtle cues (e.g., facial expressions, tone of voice) from influencing behaviour.
双盲程序用来控制实验者效应：被试和实验者都不知道当前实施的是哪种条件。这可防止微妙的线索（如面部表情、语气）影响行为。
6. Confounding Variables vs. Extraneous Variables | 混淆变量与额外变量的区别
An extraneous variable is any unwanted variable that might affect the DV. If it does affect the DV systematically and varies systematically with the IV, it becomes a confounding variable. The key distinction is whether it actually confuses the results.
额外变量是可能影响因变量的任何非预期变量。如果它确实系统地影响因变量，并与自变量同步变化，则变成混淆变量。关键区别在于它是否真正混淆了结果。
| Aspect | 方面 | Extraneous Variable | 额外变量 | Confounding Variable | 混淆变量 |
|---|---|---|
| Definition 定义 |
Unwanted variable that may affect DV 可能影响因变量的非预期变量 |
EV that varies systematically with IV and affects DV 与自变量同步变化并影响因变量的额外变量 |
| Impact 影响 |
Adds noise or error to the data 给数据增加噪音或误差 |
Provides an alternative explanation for results 为结果提供替代解释 |
| Control 控制 |
Controlled via standardisation 通过标准化控制 |
Must be eliminated or controlled to ensure validity 必须消除或控制以确保效度 |
For example, if a researcher tests memory in the morning for condition A and in the afternoon for condition B, time of day is a confounding variable because it changes alongside the IV.
例如，如果研究者在上午测试条件A的记忆，在下午测试条件B的记忆，那么一天中的时间就变成了混淆变量，因为它与自变量同步变化。
7. Demand Characteristics and Investigator Effects | 需求特征与研究者效应
Demand characteristics are cues that lead participants to guess the hypothesis and alter their behaviour accordingly. Participants may act as "good subjects" (trying to confirm the hypothesis) or "screw-you" subjects (trying to disprove it).
需求特征是引导被试猜测假设并相应改变行为的线索。被试可能扮演“好被试”（试图证实假设）或“去你的”被试（试图推翻假设）。
Investigator effects occur when the researcher's expectations, body language, or tone unintentionally influence the participants' responses. These can be minimised by double-blind designs, automated instructions, and keeping the hypothesis hidden from participants.
研究者效应发生在研究者的期望、肢体语言或语气无意中影响被试反应时。双盲设计、自动化指导语和对被试隐藏假设可以将这些效应最小化。
Single-blind: participant does not know condition | 单盲：被试不知道条件
Double-blind: neither participant nor researcher knows condition | 双盲：被试和研究者均不知道条件
To further reduce demand characteristics, researchers can use deception (with ethical approval), cover stories, or indirect behavioural measures rather than self-report.
为进一步减少需求特征，研究者可以在获得伦理批准后使用欺骗、掩护故事或间接行为测量而非自我报告。
8. Types of Experimental Research Designs | 实验研究设计的类型
Beyond the three participant designs, psychology recognises several broader experimental designs: laboratory experiments, field experiments, natural experiments, and quasi-experiments. Each has a different level of control and ecological validity.
除了三种被试设计，心理学还承认更广泛的实验设计类型：实验室实验、现场实验、自然实验和准实验。每种设计具有不同的控制水平和生态效度。
- Laboratory Experiment | 实验室实验：
High control, artificial setting. High internal validity, low ecological validity. Example: Milgram's obedience study (though ethically controversial).
高控制，人为环境。高内部效度，低生态效度。例如：米尔格拉姆的服从研究（尽管伦理上有争议）。
- Field Experiment | 现场实验：
IV manipulated in a natural setting. Participants often unaware. Higher ecological validity, but lower control. Example: Hofling's study of nurses' obedience to a fake doctor.
在自然环境中操控自变量。被试通常不知情。生态效度较高，但控制较低。例如：霍夫林关于护士服从假医生的研究。
- Natural Experiment | 自然实验：
The IV is a naturally occurring event (e.g., a natural disaster, a change in law). No direct manipulation. Example: comparing mental health before and after a flood.
自变量是自然发生的事件（如自然灾害、法律变化）。没有直接操控。例如：比较洪水前后的心理健康状况。
- Quasi-Experiment | 准实验：
The IV is based on a pre-existing characteristic of participants (e.g., age, gender, clinical diagnosis). Participants cannot be randomly assigned. Example: comparing memory between older and younger adults.
自变量基于被试的既有特征（如年龄、性别、临床诊断）。被试不能被随机分配。例如：比较老年人和年轻人的记忆力。
Trade-off: Internal validity vs. ecological validity | 权衡：内部效度 vs. 生态效度
Researchers must decide which design best answers the research question while balancing control and generalisability.
研究者必须决定哪种设计最适合回答研究问题，同时平衡控制与可推广性。
9. Reliability and Validity in Experimental Research | 实验研究中的信度与效度
Reliability refers to consistency: would the same results be obtained if the study were repeated? Validity refers to whether the study measures what it claims to measure. Both are essential for credible research.
信度指一致性：如果重复研究，能否得到相同结果？效度指研究是否测量了它声称要测量的东西。两者对于可信的研究都至关重要。
| Type | 类型 | Definition | 定义 | How to Improve | 如何提高 |
|---|---|---|
| Internal Reliability 内部信度 |
Do items within a test measure the same thing? 测试内项目是否测量同一概念？ |
Split-half method; Cronbach's alpha 分半法；克隆巴赫α系数 |
| External Reliability 外部信度 |
Is the test consistent over time and across raters? 测试在时间和评分者间是否一致？ |
Test-retest; inter-rater reliability checks 重测法；评分者间一致性检验 |
| Internal Validity 内部效度 |
Is the IV truly causing the change in DV? 自变量是否真正引起因变量变化？ |
Control EVs; use standardised procedures 控制额外变量；使用标准化程序 |
| External Validity 外部效度 |
Can findings be generalised beyond the study? 研究结果能否推广到研究之外？ |
Random sampling; naturalistic settings 随机抽样；自然化情境 |
A study cannot be valid unless it is reliable, but a reliable study can still be invalid if it measures the wrong construct.
一项研究除非可靠，否则不可能有效；但可靠的研究如果测量了错误的概念，仍然可能是无效的。
10. Common Pitfalls and How to Avoid Them | 常见陷阱及避免方法
Students often lose marks by confusing IV and DV, failing to operationalise variables, or neglecting to discuss order effects. Here are common pitfalls and solutions.
学生常常因混淆自变量与因变量、未能操作化变量、或忽视讨论顺序效应而失分。以下是常见陷阱及解决方法。
- Pitfall 1: Vague operationalisation | 陷阱1：操作化模糊
Solution: Define exact procedures — e.g., "memory" becomes "number of correct words recalled from a 20-item list within 60 seconds."
解决方法：定义精确程序——例如，“记忆”变为“60秒内从20个单词列表中正确回忆的单词数量”。
- Pitfall 2: Ignoring order effects | 陷阱2：忽视顺序效应
Solution: Use counterbalancing or switch to independent groups design.
解决方法：使用平衡设计或改用独立组设计。
- Pitfall 3: Confusing correlation with causation | 陷阱3：混淆相关与因果
Solution: Only experiments with proper control can infer causation; correlation does not imply causation.
解决方法：只有具有适当控制的实验才能推断因果关系；相关并不意味着因果。
- Pitfall 4: Demand characteristics | 陷阱4：需求特征
Solution: Use single-blind or double-blind designs; hide the hypothesis with a cover story.
解决方法：使用单盲或双盲设计；用掩护故事隐藏假设。
- Pitfall 5: Inadequate sample size | 陷阱5：样本量不足
Solution: Calculate required sample size using power analysis or use at least 20–30 participants per condition as a rule of thumb.
解决方法：使用功效分析计算所需样本量，或按经验法则每条件至少20–30名被试。
Being aware of these pitfalls and explicitly addressing them in your answers demonstrates deep understanding and earns higher marks.
意识到这些陷阱并在答题中明确应对它们，能够展示深层理解并获得更高分数。
11. A Step-by-Step Guide for Writing an Experiment Proposal | 撰写实验提案的分步指南
In exams, you may be asked to design an experiment. Following a structured approach ensures you include all necessary components.
在考试中，你可能会被要求设计一个实验。遵循结构化方法可以确保你包含所有必要组成部分。
Step 1: State the aim and hypothesis
第一步：陈述目的与假设
Step 2: Operationalise IV and DV
第二步：操作化自变量与因变量
Step 3: Choose experimental design
第三步：选择实验设计
Step 4: Describe participants and sampling method
第四步：描述被试与抽样方法
Step 5: Write a standardised procedure
第五步：编写标准化程序
Step 6: Identify and control EVs
第六步：识别并控制额外变量
Step 7: State controls for ethical issues
第七步：说明伦理问题的控制
Step 8: Describe data analysis and predicted results
第八步：描述数据分析与预期结果
Always write a directional (one-tailed) or non-directional (two-tailed) hypothesis clearly. For example: "Participants who study in a quiet room will recall significantly more words than those who study in a noisy room."
始终清晰地写出方向性（单尾）或非方向性（双尾）假设。例如：“在安静房间学习的被试将比在嘈杂房间学习的被试显著回忆出更多单词。”
12. Ethical Considerations in Experimental Designs | 实验设计中的伦理考量
Ethical guidelines are central to psychological research. Participants must give informed consent, be debriefed, and have the right to withdraw. Deception must be justified and cause no harm.
伦理准则是心理学研究的核心。被试必须给予知情同意，接受事后解释，并有权退出。欺骗必须有正当理由且不得造成伤害。
- Informed Consent | 知情同意：
Participants should know what the study involves before agreeing. If deception is required, consent must be obtained after debriefing as much as possible.
被试应在同意前了解研究内容。如果必须使用欺骗，事后应尽可能获取延迟的知情同意。
- Debriefing | 事后解释：
After the study, participants must be told the true purpose, shown their data, and offered support if needed.
研究结束后，必须告知被试真实目的，展示其数据，并在需要时提供支持。
- Protection from Harm | 免受伤害：
Procedures must not cause physical or psychological distress. Stressful tasks should be monitored and terminated if necessary.
程序不得造成身体或心理痛苦。压力任务应受到监控，必要时终止。
- Confidentiality | 保密性：
Personal data must be anonymised and stored securely.
个人数据必须匿名化并安全存储。
- Right to Withdraw | 退出权：
Participants can leave the study at any time without penalty, and their data can be removed.
被试可以随时退出研究而不会受罚，其数据也可以被删除。
Ethical considerations are not extra marks — they are mandatory criteria in any credible experiment. Always integrate ethical safeguards into your experimental design.
伦理考量不是附加分——它们是任何可信实验的强制性标准。务必在你的实验设计中纳入伦理保障措施。
Conclusion | 结论
Mastering experimental methods requires a deep understanding of variables, control techniques, and research design trade-offs. By operationalising clearly, choosing an appropriate design, controlling extraneous variables, and addressing ethics, you can design rigorous psychological studies and excel in exam questions that ask you to evaluate experiments.
掌握实验方法需要深入理解变量、控制技术和研究设计的权衡。通过清晰操作化、选择恰当设计、控制额外变量并处理伦理问题，你能够设计严谨的心理学研究，并在要求你评估实验的考试题目中表现出色。
Remember the golden rule: the purpose of an experiment is to isolate the effect of the IV on the DV. Every design decision you make should serve that goal.
记住黄金法则：实验的目的是将自变量对因变量的影响分离出来。你做出的每一个设计决策都应服务于这一目标。
Published by TutorHao | Psychology Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图