4.2.3.2 - Primary and secondary data, including meta-analysis

Why the source of data matters

Psychologists do not only ask what type of data they have. They also ask where the data came from. For this specification point, the key distinction is between data collected first-hand for a current investigation and data that already existed before the current researcher used it.

The first category is data the researcher creates for the investigation they are doing now.

The second category is data that the researcher did not create for this current investigation.

The difference is about source and purpose. If a psychologist recruits participants, gives them a memory task and records their scores, that is primary data. If the psychologist analyses published memory studies, archived school records or an existing national survey, that is secondary data.

This distinction is separate from whether data is numerical or verbal. Primary data can be numerical, such as reaction times from the researcher's own experiment, or verbal, such as transcripts from the researcher's own interviews. Secondary data can also be numerical, such as published diagnosis rates, or verbal, such as existing clinical case notes. The lesson you are in now focuses on source, not on the separate quantitative/qualitative distinction.

AQA exam questions often make this distinction through short scenarios. The safest method is to ask: did the researcher collect the data directly for this study? If yes, call it primary. If the researcher is reusing existing information, call it secondary. The same dataset can therefore be primary for the team that collected it and secondary for a later researcher. This is about how evidence is obtained, not whether it is trustworthy.

Primary data in psychological research

Primary data can arise in experiments, observations, questionnaires, interviews, correlations and case studies when the researcher gathers original measurements for the current aim. Do not classify a study by its method name alone: a case study using an old diary is using secondary material, while a diary newly requested for the current study is primary. A new coding exercise can likewise analyse existing secondary communication material. In psychology, this is useful because research questions often need precise operational definitions.

For example, a researcher studying whether background music affects concentration can design a laboratory experiment with two music conditions, the same word-search task, standardised instructions and a clear concentration score. The resulting scores are primary data because they have been collected specifically to test that hypothesis.

The main strength of primary data is fit for purpose. The researcher can choose the sample, measures, setting, instructions, timing and ethical procedure to match the aim. This can improve validity because the data directly addresses the question being asked.

Primary data also gives the researcher more control over quality. They can pilot the procedure, standardise instructions, train observers, check consent and record exactly how data was gathered. If the research involves a sensitive topic, such as anxiety in exams or experiences of therapy, the researcher can build safeguards into the procedure from the beginning.

However, primary data is expensive in time, planning and resources. A study may need ethical approval, participant recruitment, materials, pilot testing, data collection and data cleaning before analysis can begin. Small student projects may therefore end up with small or unrepresentative samples.

Primary data can also create reactivity. Participants may change their behaviour because they know they are being studied. In an interview, they may give socially desirable answers. In an experiment, they may respond to demand characteristics. So primary data is not automatically more valid than secondary data; it simply gives the researcher more direct control over collection.

Applying primary data

Imagine Dr Okafor wants to investigate whether a new breathing exercise reduces immediate exam stress. She recruits Year 13 volunteers, measures their stress rating before and after the exercise, and uses the same script for everyone. Her data is primary because she collected it directly for this research question.

This has a strength: she can design the procedure around exam stress and use a measure that fits her aim. It also has a limitation: the volunteers may not represent all students, and they may report feeling calmer because they think that is what Dr Okafor expects.

Secondary data in psychological research

Secondary data already exists before the current researcher begins their investigation. In psychology, this might include:

• published journal articles and their reported findings;
• open datasets shared by researchers;
• government statistics, census data or school records;
• NHS or public-health records, where access is lawful and ethical;
• historical documents, diaries, court transcripts or case notes;
• organisational records such as absence logs or wellbeing surveys.

Secondary data is useful when it would be impractical, unethical or unnecessary to collect everything again. A researcher studying national changes in mental-health service use would probably need existing health records rather than trying to recruit a new national sample. A developmental psychologist interested in childhood over many years might use an existing longitudinal dataset because collecting equivalent primary data would take decades.

The main strength is efficiency. Secondary data can be quicker and cheaper than collecting new data, and it may involve much larger samples than one researcher could gather alone. It can also allow analysis of real-world behaviour recorded outside the artificial setting of a research study.

Secondary data can also reduce participant burden. If suitable data already exists, researchers may avoid asking participants to repeat sensitive questionnaires or interviews. This can be ethically valuable, provided confidentiality and permission are handled properly.

The main limitation is lack of control over how the data was collected. The original researcher or organisation may have used definitions, measures or categories that do not fit the new research question. For example, one school record system might define "absence due to anxiety" differently from another. This makes comparisons difficult.

Secondary data may also be outdated, incomplete, biased or recorded for administrative convenience rather than psychological accuracy. The current researcher must check provenance: who collected the data, why it was collected, how variables were defined, whether the sample is suitable, and whether there are ethical or legal restrictions on reuse.

Applying secondary data

Suppose Aisha wants to study whether rates of diagnosed depression changed after a local mental-health campaign. Instead of recruiting thousands of residents, she requests anonymised NHS referral records from before and after the campaign. This is secondary data because the NHS records were collected for healthcare administration before Aisha's study. Removing names protects confidentiality but does not turn secondary data into primary data.

The data may be powerful because it covers a large real-world population. But Aisha must be cautious: referral rates are not the same as actual depression rates. They can be affected by GP behaviour, stigma, awareness, service availability and record-keeping practices.