Observations, Self-Report, Correlations and Case Studies

Not every question in psychology can be answered by manipulating an independent variable. Some behaviours cannot be created in a laboratory, some experiences can only be reported by the person having them, and some questions are about how two things vary together rather than whether one causes the other.

For these questions psychologists use non-experimental methods. The AQA specification names five you must be able to describe, evaluate and compare:

Method	What it does	Typical data
Observational techniques	Watch and record behaviour	Quantitative or qualitative
Self-report (questionnaires)	Ask people to report on themselves in writing	Quantitative or qualitative
Self-report (interviews)	Ask people to report on themselves aloud	Quantitative or qualitative
Correlations	Measure the relationship between two co-variables	Quantitative
Content analysis	Systematically code existing material	Qualitative and quantitative
Case studies	Investigate one person, group or event in depth	Mostly qualitative

The key difference from an experiment is that none of these methods manipulates an independent variable to establish cause and effect. That single point drives most of the evaluation in this topic.

An observation involves watching and recording behaviour. The common mistake is to treat "types of observation" as one list. In fact AQA describes observations along three separate dimensions, and any real observation is a combination of one option from each.

Dimension	Option A	Option B
Setting	Naturalistic — behaviour observed in the setting where it would normally occur	Controlled — some variables are managed in a setting arranged by the researcher
Awareness	Covert — participants do not know they are being observed	Overt — participants know they are being observed and have consented
Involvement	Participant — the observer joins the group being studied	Non-participant — the observer stays separate and watches from outside

Because the dimensions are independent, a study might be naturalistic, covert and non-participant (secretly watching shoppers in a real shop) or controlled, overt and participant (openly joining a group in a room set up by the researcher).

Naturalistic vs controlled is about the setting. Covert vs overt is about awareness. Participant vs non-participant is about the observer's involvement. Keep the three questions separate and you can classify any observation.

Observations capture what people actually do, rather than what they say they do, which is their central strength.

Strengths	Limitations
Capture real, spontaneous behaviour rather than self-reported behaviour	Observer bias — what the observer records can be shaped by their expectations
Naturalistic observations tend to have high ecological validity because behaviour occurs in a real setting	Little or no control over extraneous variables, especially in naturalistic settings, so cause and effect is hard to establish
Useful where manipulation would be impractical or unethical	Covert observation raises ethical issues because participants cannot consent or withdraw

To make observations more objective, researchers define behavioural categories in advance (breaking a behaviour such as "aggression" into specific, observable actions like "pushing" or "kicking") and check agreement between two observers.

Inter-observer reliability is measured by having two or more observers record the same behaviour independently and comparing their records. High agreement suggests the behavioural categories are being applied consistently.

Covert, naturalistic observation reduces the chance that participants change their behaviour because they know they are watched, but it trades this against the ethical cost of observing people without consent.

Self-report techniques ask people to report on their own thoughts, feelings and behaviour. AQA names two: questionnaires and interviews.

A questionnaire is a written set of questions. The type of question determines the type of data:

• Closed questions restrict the answer to fixed options (yes/no, a rating scale). They produce quantitative data that is quick to analyse but limited in depth.
• Open questions let respondents answer freely in their own words. They produce qualitative data that is richer but harder to summarise.

Interviews are spoken self-report. AQA names two types — structured and unstructured — with semi-structured a common middle ground:

Interview type	Structure	Analogy
Structured	Fixed, pre-set questions asked in the same order every time	A spoken questionnaire
Unstructured	No fixed questions; the interviewer follows the conversation	A guided conversation
Semi-structured (extra context — not named by AQA 7182, but a common real-world type)	A set of core questions plus follow-ups decided as the interview unfolds	A job interview

Structured interviews are easy to replicate and compare because every participant is asked the same thing. Unstructured interviews sacrifice that consistency for depth and flexibility, since the interviewer can probe unexpected answers.

Self-report gives access to internal states such as attitudes, beliefs and feelings that cannot be observed directly. That access is its main advantage, but the same reliance on what people are willing to say is also its main weakness.

Strengths	Limitations
Direct access to thoughts, feelings and opinions that observation cannot reach	Social desirability bias — participants answer in a way that presents them favourably rather than truthfully
Questionnaires can reach large samples quickly and cheaply, including by post or online	Response bias, such as always choosing the middle option or agreeing regardless of the question (acquiescence)
Closed questions give quantitative data that is easy to analyse and compare	Interviewer effects — the interviewer's presence, tone or characteristics can influence answers, especially in unstructured interviews

Questionnaires with closed questions scale well but capture little depth; unstructured interviews capture depth but are time-consuming and difficult to analyse and replicate. The choice depends on whether the aim is breadth or richness.

A leading question (one that suggests a particular answer) or an ambiguous question reduces validity. Good questionnaire construction avoids both, and a pilot study is used to check questions are clear before the main study.

A correlation does not manipulate anything. Instead it measures the relationship between two co-variables — two variables that are both simply measured, not controlled. Because there is no manipulated independent variable, a correlation is not an experiment.

There are three possible directions of relationship:

Type	What happens	Everyday example
Positive	As one co-variable increases, the other increases	More hours revised, higher exam score
Negative	As one co-variable increases, the other decreases	More hours revised, fewer marks lost
Zero / no correlation	No consistent relationship between the co-variables	Shoe size and exam score

The strength and direction of a correlation are expressed by a correlation coefficient, a number between -1 and +1:

• +1 is a perfect positive correlation, -1 is a perfect negative correlation, and 0 means no correlation.
• The sign (+ or −) gives the direction; the distance from zero gives the strength. So a coefficient of −0.8 is stronger than +0.4, even though it is negative.

The single most important fact about correlations: a correlation shows that two co-variables are associated, not that one causes the other. A third, intervening variable may be responsible for the relationship. This inability to establish cause and effect is the key difference between a correlation and an experiment.

The final two methods deal with material that already exists and with individual cases.

Content analysis is a technique for analysing qualitative material — media, texts, adverts, diaries or interview transcripts. The researcher systematically codes the material into pre-defined categories, then usually counts how often each category appears. This turns qualitative source material into quantitative frequency data, so content analysis produces both kinds of data.

Case studies are in-depth investigations of a single individual, small group or event. They are often longitudinal (following the case over time) and typically use several methods together, such as interviews, observation and psychological testing. Famous examples include patient HM, whose memory loss after brain surgery informed theories of memory, and Little Hans, Freud's detailed case study of a young boy's phobia.

Method	Strengths	Limitations
Content analysis	Uses existing material, so few ethical issues; can be replicated if coding categories are clear	Coding categories can be applied subjectively, introducing observer bias
Case studies	Rich, detailed data; allow study of rare cases that could not be created experimentally	Findings from one case cannot be generalised; risk of researcher bias and reliance on retrospective, subjective evidence

Case studies often depend on retrospective data — asking participants or relatives to recall past events — which may be incomplete or inaccurate. This, together with the small sample, is why conclusions from a single case must be drawn cautiously.

1. Saying a correlation proves cause and effect

A correlation only shows that two co-variables are associated. Because nothing is manipulated and extraneous variables are uncontrolled, an intervening variable could explain the link. Write "is associated with" or "is linked to", never "causes".

2. Confusing the three observation dimensions

Naturalistic vs controlled is about the setting; covert vs overt is about awareness; participant vs non-participant is about the observer's involvement. A naturalistic observation is not automatically covert — the dimensions are independent, and a study takes one option from each.

3. Mixing up structured and unstructured interviews

A structured interview uses fixed, pre-set questions in a set order, like a spoken questionnaire. An unstructured interview has no pre-set questions and flows like a conversation. Do not describe one using features of the other.

4. Thinking content analysis only produces qualitative data

Content analysis starts with qualitative material but codes it into categories and counts them, producing quantitative frequency data as well. It is a way of turning qualitative material into numbers, not just describing it.

5. Over-generalising from a single case study

A case study examines one individual, group or event, so its findings cannot be generalised to the wider population. Praising the rich detail is fair; claiming the result applies to everyone is not.

6. Reading a correlation coefficient by sign alone

The sign gives direction and the distance from zero gives strength. A coefficient of −0.9 is stronger than +0.3. Do not treat a negative coefficient as automatically weaker than a positive one.