You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Psychology earns its place among the sciences because of how it gathers evidence. A claim such as "violent video games increase aggression" or "eyewitnesses can be misled by leading questions" is only worth anything if it rests on a study designed to test it fairly. Component 01 of the OCR A-Level — worth 30% of the whole qualification — is where you learn to design, critique and interpret those studies. Everything you meet in the twenty core studies of Component 02 and the applied topics of Component 03 was built using the toolkit introduced here.
This opening lesson surveys the methods and techniques available to a researcher: the different kinds of experiment, the family of observational methods, the self-report methods (questionnaires and interviews), and correlational research. For each we examine what it is, when a researcher would choose it, and — crucially, because AO3 marks live here — its characteristic strengths and weaknesses. By the end you should be able to read the "Method" section of any study and name the technique it used, justify why that technique suited the aim, and identify what the researcher gave up by choosing it.
| This lesson covers | OCR H567 Component 01 sub-area | AO focus |
|---|---|---|
| Laboratory, field and quasi-experiments | 1.1 Methods & techniques — experiment | AO1 knowledge; AO3 evaluation |
| Observations (structured/unstructured, naturalistic/controlled, participant/non-participant, overt/covert) | 1.1 — observation | AO1; AO2 choosing a method for a scenario |
| Self-report: questionnaires; structured/semi-structured/unstructured interviews | 1.1 — self-report | AO1; AO3 |
| Correlation (positive, negative, none); correlation ≠ causation | 1.1 — correlation | AO1; AO3 analysis |
The specification is referenced descriptively throughout; consult the official OCR H567 specification document for the exact published wording. This lesson develops AO1 (defining and understanding each method), AO2 (selecting the appropriate method for a novel research scenario, a recurring Component 01 question style) and AO3 (evaluating the trade-offs each method involves).
An experiment is the only method that can establish a cause-and-effect relationship, because it is the only method in which the researcher deliberately manipulates one variable and measures the effect on another while attempting to hold everything else constant. The manipulated variable is the independent variable (IV); the measured outcome is the dependent variable (DV). If the IV is changed, everything else is controlled, and the DV changes, the logic of the experiment allows us to attribute that change to the IV.
The word "experiment" is not a synonym for "study". Many famous pieces of research are not experiments at all. What makes something an experiment is the manipulation of an IV, not the setting, the sample size, or the use of statistics. OCR distinguishes three types of experiment by where and how the IV is manipulated.
A laboratory experiment is conducted in a controlled, often artificial environment where the researcher manipulates the IV and rigorously controls extraneous variables. "Laboratory" does not necessarily mean a room with white coats and equipment — it means a controlled setting the participant would not normally be in, arranged so that the researcher governs the conditions.
Loftus and Palmer's (1974) eyewitness-memory study — a Component 02 core study — is a laboratory experiment: participants watched filmed car crashes and were asked about speed using different verbs ("smashed" versus "hit"). The verb was the manipulated IV; the estimated speed was the DV; the lab setting let the researchers show everyone the same film under the same conditions.
Strengths. High control over extraneous variables gives high internal validity — we can be confident the IV caused the change in the DV. Standardised procedures make the study replicable, so other researchers can repeat it and check the finding.
Weaknesses. The artificiality can lower ecological validity — watching a filmed crash is not the same as witnessing a real accident with its shock and adrenaline. The unfamiliar setting can provoke demand characteristics, where participants guess the aim and change their behaviour.
In a field experiment the researcher still manipulates the IV, but does so in the participant's natural, everyday environment. Piliavin et al.'s (1969) "Subway Samaritan" study — again a Component 02 core study — is the classic example: confederates collapsed on a New York subway train, and the researchers manipulated whether the victim appeared drunk or carried a cane (the IV), measuring whether and how quickly passengers helped (the DV). The train carriage is the participants' real environment; they had no idea they were in a study.
| Feature | Laboratory experiment | Field experiment |
|---|---|---|
| Setting | Controlled/artificial | Natural/everyday |
| Control of extraneous variables | High | Lower |
| Ecological validity | Often lower | Often higher |
| Demand characteristics | More likely | Less likely (often unaware) |
| Informed consent / ethics | Easier to obtain | Often problematic (covert) |
Strengths. Higher ecological validity because behaviour occurs in a real setting; reduced demand characteristics when participants do not realise they are being studied.
Weaknesses. Less control means confounding variables are harder to rule out (on the subway, temperature, crowding and time of day all vary). Studying people without their knowledge raises serious ethical issues around informed consent — you cannot ask a whole subway carriage to sign a form.
A quasi-experiment looks like an experiment but the IV is not manipulated by the researcher — it is a pre-existing characteristic of the participants, such as age, sex, whether someone has a diagnosis of a disorder, or which country they live in. Because the researcher cannot assign participants to conditions (you cannot make someone 8 years old, or make them left-handed), the "manipulation" is really a selection of naturally-occurring groups.
Sperry's (1968) split-brain study is effectively quasi-experimental: the IV — whether the corpus callosum had been severed — was a medical fact about the patients, not something Sperry created. Baron-Cohen et al.'s (1997) comparison of adults with and without autism is another quasi design.
Strengths. Allows study of variables that could never be ethically or practically manipulated (you cannot ethically give someone a brain injury or a disorder). Often carries real-world relevance.
Weaknesses. Because participants are not randomly allocated to conditions, participant variables may confound the results — differences between groups might be due to some other characteristic that co-varies with the IV, not the IV itself. This weakens the cause-and-effect claim.
graph TD
A["Is an independent variable manipulated<br/>and a dependent variable measured?"] -->|No| Z["NOT an experiment<br/>(observation, self-report or correlation)"]
A -->|Yes| B{"Who or what sets the IV?"}
B -->|"Researcher, in a controlled setting"| C["Laboratory experiment"]
B -->|"Researcher, in a natural setting"| D["Field experiment"]
B -->|"A pre-existing participant characteristic<br/>(age, disorder, sex) — not manipulable"| E["Quasi-experiment"]
style C fill:#2980b9,color:#fff
style D fill:#27ae60,color:#fff
style E fill:#8e44ad,color:#fff
style Z fill:#e74c3c,color:#fff
A note that examiners reward: the difference between a laboratory and a field experiment is the setting, whereas the difference between a "true" experiment (lab or field) and a quasi-experiment is whether the researcher manipulated the IV. Keeping those two distinctions separate is a reliable AO1 discriminator.
It is worth dwelling on why the experiment holds such a special place. Science is fundamentally interested in causal claims — not merely that two things go together, but that one produces the other. The experiment is the only research design that isolates a single suspected cause (the IV), changes it deliberately while holding the rest of the world as constant as possible, and watches for a consequence (the DV). If the DV then changes and the design has genuinely controlled everything else, the only remaining explanation is the IV. This is the logic of the controlled comparison, and it is why a well-run experiment can license the word "cause" where an observation or correlation cannot.
But the control that buys causal confidence has a price, and naming that price is the essence of AO3. The more tightly a researcher controls a situation, the more artificial it usually becomes, and the further it drifts from the messy real world the findings are meant to describe. This tension — control versus realism — runs through the whole of research methods. A laboratory experiment maximises control and sacrifices realism; a field experiment recovers realism and sacrifices control; a quasi-experiment sacrifices the manipulation itself in exchange for being able to study variables that could never be manipulated. There is no design that wins on every dimension, which is exactly why a researcher must choose the design that best fits the specific aim, and why examiners reward candidates who can articulate the trade-off rather than reciting a fixed list of "good" and "bad" points.
Consider a concrete illustration. Suppose we want to know whether sleep deprivation impairs decision-making. A laboratory experiment could keep participants awake in a controlled facility, standardise the decision task, and control light, noise and diet — but "decisions made on a computer task in a sleep lab" are not the high-stakes, real-world decisions we ultimately care about. A field study of, say, junior doctors after long shifts would have far greater realism, but now diet, caffeine, individual workload and countless other variables vary freely and could confound the result. The same research question pulls toward different designs depending on whether we prize internal validity or ecological validity — and a sophisticated answer recognises that the two studies are complementary rather than one being simply "better".
An observation records behaviour as it occurs, without manipulating an IV. Observation is a non-experimental technique — it describes what people do rather than testing the effect of a manipulated variable. It can, however, be used within an experiment (Bandura measured aggressive acts by observation inside an experimental design), so "observation" is best thought of as a way of measuring the DV or recording behaviour, distinct from the experimental logic itself. OCR requires you to know four contrasting pairs.
A structured observation uses a predetermined system of behavioural categories (a coding frame), so observers tally specific, operationalised behaviours (e.g. "hits Bobo doll with mallet"). An unstructured observation records everything of potential interest in rich, qualitative detail, useful when a researcher does not yet know which behaviours will matter.
A naturalistic observation takes place in the setting where the behaviour would ordinarily occur, with no interference from the researcher (e.g. watching children in a real playground). A controlled observation takes place in a setting the researcher has arranged, where some variables are controlled — Bandura's (1961) study used a controlled observation in a laboratory playroom.
In a participant observation the researcher joins the group being studied and takes part in its activities; in a non-participant observation the researcher watches from the outside without getting involved. Participant observation can yield deep insight but risks the observer losing objectivity ("going native").
In an overt observation participants know they are being watched; in a covert observation they do not. Covert observation reduces demand characteristics and social desirability but raises ethical problems around consent and privacy.
| Dimension | Option A | Option B | Key trade-off |
|---|---|---|---|
| Structure | Structured (coding frame) | Unstructured (rich record) | Comparability & reliability vs depth |
| Setting | Naturalistic | Controlled | Ecological validity vs control |
| Involvement | Participant | Non-participant | Insight vs objectivity |
| Awareness | Overt | Covert | Ethical consent vs natural behaviour |
Strengths of observation generally. Captures behaviour as it actually happens rather than what people say they do, so it can have high ecological validity (especially naturalistic). Useful where manipulation would be unethical or impossible, and valuable at the early stage of investigating a phenomenon, before enough is known to design a controlled experiment.
Weaknesses. Cannot establish cause and effect (no manipulated IV). Vulnerable to observer bias — observers may unconsciously see what they expect. Inter-rater reliability must be checked by having two observers code independently and comparing their records. Covert designs raise consent and privacy issues.
The four dimensions are genuinely independent choices, and a common exam error is to treat them as a single scale. A study can be, for example, structured and naturalistic and non-participant and covert all at once — a researcher tallying predefined categories of pedestrian behaviour at a crossing, from a parked car, without anyone knowing. Another study might be unstructured, controlled, participant and overt — a researcher openly joining a laboratory-based group task and writing rich field notes. Because the dimensions combine freely, describing an observation fully means specifying it on all four axes, not picking one label. When a question asks you to "describe the type of observation used", the marks accumulate across the dimensions you correctly identify.
The participant/non-participant distinction repays particular attention because it captures a genuine methodological dilemma. A participant observer, embedded in the group, can access meanings, motives and subtleties invisible from the outside — but risks losing objectivity by identifying with the group ("going native"), and cannot easily record data in the moment without breaking cover. A non-participant observer preserves detachment and can record systematically, but sees only surface behaviour and may misread it for lack of insider context. Neither is simply superior; the choice depends on whether the research question is about observable behaviour (favouring non-participant) or lived meaning (favouring participant).
Self-report methods gather data by asking participants to report on their own thoughts, feelings, attitudes or behaviours. The two families are questionnaires and interviews.
A questionnaire is a set of written questions, usually completed by the participant without a researcher present. Questionnaires can use closed questions (fixed response options, yielding quantitative data that is quick to analyse) or open questions (free text, yielding qualitative detail). They can be distributed to large samples cheaply.
Strengths. Efficient — large amounts of data from many people quickly and cheaply. Closed questions produce quantitative data that is easy to analyse and compare. Anonymity may reduce social desirability.
Weaknesses. Prone to social desirability bias (people answer to look good). Response rates can be low and those who respond may be unrepresentative. Cannot clarify ambiguous questions — the participant is on their own. Fixed options may not capture a respondent's true view.
An interview is a self-report method conducted face-to-face (or by phone/video), where a researcher asks questions directly. OCR names three types on a continuum of structure:
| Interview type | Flexibility | Replicability | Data | Best when… |
|---|---|---|---|---|
| Structured | Low | High | Mostly quantitative | You need comparable answers from many people |
| Semi-structured | Medium | Medium | Mixed | You want comparability and the ability to probe |
| Unstructured | High | Low | Qualitative | Exploring a topic in depth, few participants |
Strengths of interviews. Can gather detailed, in-depth data; interviewer can clarify questions and probe; less structured formats access meanings a questionnaire would miss.
Weaknesses. Time-consuming and expensive; interviewer effects and social desirability can distort answers; less structured data is hard to analyse and less reliable.
The choice of self-report format is really a choice about the kind of data you want and can afford to analyse. Closed questionnaire items and structured interviews yield largely quantitative data — easy to summarise numerically, compare across large samples, and subject to statistical test — at the cost of depth and the risk that fixed options misrepresent a respondent's real position. Open items and unstructured interviews yield qualitative data — rich, contextualised and capable of surprising the researcher — at the cost of being slow to analyse, difficult to compare, and more exposed to the researcher's interpretation. A semi-structured interview is popular in applied psychology precisely because it straddles this divide: the core questions supply comparability while the freedom to probe supplies depth.
Two threats deserve emphasis because they recur across every self-report method. Social desirability is the tendency to answer in ways that present the respondent favourably — under-reporting socially disapproved behaviour, over-reporting virtues — which is most acute for sensitive topics and can be reduced (though never eliminated) by guaranteeing anonymity and phrasing items neutrally. Interviewer effects arise when the interviewer's presence, appearance, tone or reactions shape the responses; a nod at the "right" answer, or an interviewer whose characteristics make certain admissions uncomfortable, can systematically bias the data. These are not merely academic worries: they are the reason a self-reported figure for, say, alcohol consumption or prejudice should always be read with caution, and the reason researchers sometimes triangulate self-report with observation or behavioural measures.
A correlation is not a method of manipulation at all — it measures the strength and direction of a relationship between two co-variables that are both measured (not manipulated). Because neither variable is manipulated, a correlation cannot establish cause and effect. This single point is the most examined idea in this section.
Correlations are displayed on a scatter diagram, with each participant one point. The strength of the relationship is summarised by a correlation coefficient ranging from −1 (perfect negative) through 0 (none) to +1 (perfect positive):
−1≤r≤+1The crucial limitation: a correlation between A and B could mean A causes B, B causes A, or a third variable (C) causes both. Ice-cream sales correlate with drowning deaths, but neither causes the other — hot weather (the third variable) drives both. Confusing correlation with causation is the single most penalised error in this topic.
Strengths. Allows study of variables that could not ethically or practically be manipulated (you cannot manipulate someone's stress level to see if it causes illness). A useful first step: a strong correlation can justify a later experiment. Can use existing data.
Weaknesses. Cannot infer causation — the direction of the relationship and third variables are unknown. Can be misused by the media to imply causal claims that the data cannot support.
The three rival explanations behind any correlation are worth naming explicitly, because doing so in an exam demonstrates genuine understanding rather than a memorised slogan. Given a correlation between variables A and B: (1) A might cause B; (2) B might cause A — the direction of causality is not fixed by the data (does anxiety cause poor sleep, or poor sleep cause anxiety?); or (3) a third (confounding) variable C might cause both, producing the apparent link without any direct relationship between A and B at all. Only an experiment, by manipulating one variable, can begin to separate these possibilities. This is why the phrase "correlation does not imply causation" is not mere caution but a precise statement about what a co-relational design can and cannot license.
It is equally important to see the positive value of correlational research, so that evaluation is balanced rather than dismissive. Correlations are indispensable where manipulation is impossible or unethical — no ethics committee would approve deliberately raising participants' stress to clinical levels to test whether it causes heart disease, so the relationship must be studied correlationally. They also make efficient use of existing datasets, allow the study of naturally-occurring variables, and frequently provide the first hint of a relationship that later, more controlled work can pursue. A strong correlation is often the starting gun for experimental research, not a poor substitute for it.
To consolidate, imagine four teams investigating the same broad question — does using social media affect adolescent wellbeing? — each with a different method:
| Method | How this team would proceed | What they could conclude | Key limitation |
|---|---|---|---|
| Laboratory experiment | Randomly allocate teenagers to use a curated feed vs a neutral task for 30 minutes, then measure mood | A tentative causal claim about short-term mood | Artificial exposure; short timescale; low ecological validity |
| Field experiment | Manipulate real notification settings on participants' own phones for a week, measuring wellbeing | A causal claim with better realism | Less control; consent and privacy complications |
| Observation | Systematically record teenagers' visible engagement and interactions in a common room | Rich description of actual behaviour | No causation; observer bias; ethics of covert watching |
| Correlation/self-report | Survey daily usage hours and a wellbeing scale across many teenagers | Strength/direction of the association | No causation; direction and third variables unknown; social desirability |
No single row is the "right answer": together they triangulate the question, and a sophisticated researcher recognises that the convergence of evidence across methods is more persuasive than any one study — a point that connects directly to how science accumulates knowledge (Lesson 10).
Going further. Undergraduate methods courses formalise the "control versus realism" tension as the trade-off between internal and external validity, and introduce quasi-experimental and longitudinal designs that sit between the categories here. If you are drawn to this, look up the idea of a natural experiment (a naturally-occurring event manipulates the IV — e.g. a policy change) — it blurs the line between field and quasi-experiments and is a favourite of health and economic psychology.
Common errors to avoid. (1) Calling any study conducted indoors a "laboratory experiment" — the label depends on manipulation and control, not the room. (2) Describing a study as "an experiment" when no IV was manipulated (it is then an observation, self-report or correlation). (3) Treating the four observation dimensions as one scale rather than four independent choices. (4) Writing that a correlation "shows" or "proves" one variable affects another. (5) Evaluating a method with generic points that are never tied to the study in the question — the surest way to cap marks at AO1.
graph LR
A["Two co-variables measured<br/>(neither manipulated)"] --> B{"Direction of relationship?"}
B -->|"Both rise together"| C["Positive correlation<br/>r toward +1"]
B -->|"One rises as other falls"| D["Negative correlation<br/>r toward −1"]
B -->|"No pattern"| E["Zero correlation<br/>r near 0"]
C --> F["Correlation ≠ causation:<br/>third variable may explain both"]
D --> F
style C fill:#27ae60,color:#fff
style D fill:#e67e22,color:#fff
style E fill:#7f8c8d,color:#fff
style F fill:#c0392b,color:#fff
Specimen question modelled on the OCR H567 paper format
A researcher wants to investigate whether background music affects how well students concentrate while revising. She arranges for one group of students to revise in a room with music playing and another group to revise in a silent room, then gives both groups the same concentration test.
(a) Identify the type of experiment used and justify your answer. [3 marks] (b) Explain one strength and one weakness of using this type of experiment for this investigation. [6 marks]
Mark-scheme decomposition (in our own words). Part (a) is worth 3 marks: 1 mark (AO2) for correctly identifying the experiment type given the scenario, and up to 2 marks (AO2) for a justification that refers to specific features of this scenario. Part (b) carries 6 marks split across AO1 and AO3: credit for accurate knowledge of a genuine strength and weakness (AO1) and for applying/evaluating each in the context of this particular study (AO3). Top-band answers contextualise — they do not list generic textbook points.
Mid-band response (2/3): (a) This is a laboratory experiment because the students are tested in controlled conditions with the same concentration test.
Examiner-style commentary: One mark for correctly identifying "laboratory experiment", and this response earns partial justification credit for mentioning controlled conditions and a standardised test. To reach top-band (3/3) it needs to tie the identification explicitly to the scenario — that the researcher manipulated the IV (music vs silence) in a setting she arranged, which is what makes it a laboratory rather than a field experiment. Naming the manipulated variable is the missing AO2 discriminator.
Top-band response (3/3): (a) This is a laboratory experiment. The researcher has deliberately manipulated the independent variable — whether music is playing or the room is silent — and measured its effect on the dependent variable, concentration test score. Because the students are placed in a room the researcher has arranged, rather than their own normal revision environment, and extraneous variables (same test, same room) are controlled, it is a laboratory rather than a field experiment.
Examiner-style commentary: Full marks. The response names the type, identifies the manipulated IV and measured DV from the scenario, and justifies "laboratory" by contrasting it with a field experiment on the basis of setting and control. This scenario-anchored justification is exactly the AO2 move examiners reward; a generic definition of "laboratory experiment" with no reference to the music study would cap at 1 mark.
For part (b), a strong answer might offer as a strength high control of extraneous variables (both groups sit the same test in comparable rooms, so any difference in concentration is more plausibly due to the music, giving high internal validity), and as a weakness low ecological validity (revising to a set test in an unfamiliar room is not how students normally revise, so the findings may not generalise to real revision) — each point applied to this study, not stated generically.
This content is aligned with the OCR A-Level Psychology (H567) specification.