What you'll learn

• What primary data and secondary data mean in psychological research.
• How to tell the difference between data collected first-hand and data that already exists.
• What a meta-analysis is and why psychologists use it.
• How to evaluate primary data, secondary data, and meta-analysis for AO3.

Starting point: what is data?

In psychology, data means the information researchers collect or use to answer a research question. Data might be numerical, such as reaction times or questionnaire scores, or descriptive, such as interview responses or observation notes.

Before you evaluate a study, you need to ask: where did the data come from? That is where the distinction between primary and secondary data matters.

Primary data

Primary data are “new” in the sense that the researcher gathers them directly. In psychology, this could include:

• scores from a laboratory experiment
• questionnaire responses
• interview transcripts
• observation records
• physiological measures such as heart rate or brain activity

For example, in Loftus and Palmer’s 1974 study on eyewitness testimony, participants watched film clips of car accidents and estimated the speed of the cars. The speed estimates were primary data because Loftus and Palmer collected them directly for that study.

Why researchers use primary data

Primary data are useful because they can be designed to match the researcher’s exact aim — the purpose of the study.

If a researcher wants to investigate whether sleep affects memory recall, they can design a study with the exact sleep conditions, memory task, sampling method, and procedure they need.

Strengths of primary data

Primary data are often highly relevant because they are collected for the exact research question. The researcher can carefully operationalise variables, meaning they define abstract concepts in measurable terms.

Primary data also allow researchers to standardise procedures. This helps improve reliability, meaning consistency. If all participants receive the same instructions and complete the same task, differences in results are less likely to be caused by procedural variation.

Limitations of primary data

Primary data can be expensive and time-consuming. Recruiting participants, designing materials, gaining ethical approval, collecting responses, and analysing results all take effort.

Primary data can also be affected by participant behaviour. For example, participants may show demand characteristics, where they guess the aim of the study and change their behaviour.

Primary data may also lack generalisability if the sample is small, biased, or unrepresentative.

Ethics and primary data

When researchers collect primary data from people, they must follow ethical guidelines. These include:

• gaining informed consent
• avoiding unnecessary deception
• giving participants the right to withdraw
• protecting participants from psychological or physical harm
• maintaining confidentiality
• providing a full debrief

In Loftus and Palmer’s 1974 study, participants watched accident footage. This raises mild protection-from-harm issues because the clips could have caused distress, although the risk was relatively low compared with real-life trauma research. Participants should also have been debriefed about the purpose of the study.

Secondary data

Secondary data can include:

• published journal articles
• government statistics
• school or hospital records
• historical documents
• previous experimental datasets
• case notes
• archived interviews or observations

For example, a psychologist studying mental health trends might use NHS statistics rather than collecting new data from patients. Those statistics would be secondary data.

Why researchers use secondary data

Secondary data are useful when collecting new data would be impractical, unethical, expensive, or impossible.

For instance, researchers cannot go back in time to collect childhood records from past decades. Instead, they may use archived documents, medical records, or previous research.

Strengths of secondary data

Secondary data can provide access to large-scale information. Government statistics or national surveys may include thousands of participants, improving generalisability.

Secondary data can also be valuable for studying rare behaviours, historical changes, or long-term patterns. For example, researchers interested in changes in diagnosis rates over time may use existing clinical records rather than starting from scratch.

Secondary data can reduce participant burden because researchers do not need to ask people to take part again.

Limitations of secondary data

The biggest weakness is that the data may not perfectly match the researcher’s aim. Another researcher may have used different definitions, measures, samples, or procedures.

There may also be unknown problems with reliability or validity. If the original data were collected poorly, the new researcher inherits those weaknesses.

Secondary data may also be outdated. For example, old statistics on attitudes to mental illness may not reflect current social norms.

Ethics and secondary data

Secondary data still involve ethical issues. Even if the current researcher does not meet participants, the data may contain sensitive information.

Researchers should consider whether the original participants consented to their data being reused. They must protect confidentiality, anonymise records where needed, and avoid using data in ways that could harm individuals or groups.

Comparing primary and secondary data

The difference is not about whether data are numerical or descriptive. Both primary and secondary data can be quantitative or qualitative.

For example, a researcher’s own interview transcripts are primary qualitative data. Archived interview transcripts collected by another researcher are secondary qualitative data.

A simple question helps:

Meta-analysis

Meta-analysis is a form of secondary data analysis because it uses results from studies that have already been carried out.

It is often part of a systematic review, which is a carefully planned review of research using clear search terms, databases, and inclusion criteria.

For example, a meta-analysis on attachment might include only studies that used the Strange Situation procedure, included infants under two years old, and reported attachment classifications clearly.

How meta-analysis works

A meta-analysis usually follows these steps:

1. Researchers define a clear research question.
2. They search for relevant studies.
3. They apply inclusion and exclusion criteria.
4. They extract key findings from each study.
5. They convert findings into a common measure, often an effect size.
6. They statistically combine the results.
7. They interpret the overall pattern.

A study might find a statistically significant result at $p<0.05$, but the effect could still be small. Meta-analysis helps researchers look beyond one study and consider the overall strength of the evidence.

Example from psychology

A well-known example is van IJzendoorn and Kroonenberg’s 1988 meta-analysis of cross-cultural attachment studies. They combined findings from 32 studies using the Strange Situation across several countries.

They found that secure attachment was the most common classification across cultures. However, variation within cultures was often greater than variation between cultures.

This is useful because one small study in one country may give a narrow picture, whereas a meta-analysis can identify broader patterns.

Strengths of meta-analysis

Meta-analysis can increase confidence in conclusions because it combines many studies rather than relying on one sample. This usually gives greater statistical power, meaning a better chance of detecting a real effect if one exists.

It can also help resolve contradictory findings. If some studies find an effect and others do not, a meta-analysis may show whether the overall evidence supports the effect.

Meta-analysis is also useful for real-world application. For example, clinicians and policy-makers may prefer conclusions based on many studies rather than one isolated piece of research.

Limitations of meta-analysis

A meta-analysis is only as good as the studies included. This is sometimes called the “garbage in, garbage out” problem. If the original studies have weak methods, biased samples, or poor measures, combining them does not magically fix those problems.

Meta-analyses can also suffer from publication bias.

This matters because a meta-analysis based mainly on published studies may overestimate an effect. Non-significant studies may be hidden in researchers’ files, sometimes called the “file drawer problem”.

Another issue is heterogeneity, which means variation between studies. If studies use different procedures, samples, measures, or definitions, it may be difficult to combine them meaningfully.

For example, combining studies of “stress” may be problematic if one study measures exam stress, another measures workplace stress, and another measures traumatic stress.

AO2: applying the distinction

In scenario questions, focus on the source and purpose of the data.

If a teacher gives pupils a new questionnaire for her current investigation, that is primary data. If she analyses old school attendance records, that is secondary data.

If a researcher combines published results from several studies of obedience, that is likely to be a meta-analysis.

AO3: evaluation points for essays

For primary data, strong evaluation often focuses on control, relevance, reliability, ecological validity, cost, time, and ethical demands.

For secondary data, evaluate efficiency, access to large or historical datasets, lack of control over original methods, and whether the data fit the current aim.

For meta-analysis, evaluate increased power and broader conclusions against publication bias, poor-quality original studies, and heterogeneity.