IB Psychology SLTopic 5 — Research DesignPaper 1 & 2Core idea~10 min read
Generalisability: Who Else Do the Results Apply To?
Every study is done on some people, somewhere, at some time. Generalisability is the question of how far past those people, that place and that year the finding actually reaches. Get into the habit of asking it, because “the sample was too small” is not the answer examiners want.
📘 What you need to know
Generalisability = how far findings extend to a wider population, setting or time.
The target population is the group the researcher wants the findings to describe.
Sample generalisability needs a sample that is random, representative and large enough.
Inferential generalisability is about applying findings to other settings, and links to external validity.
Theoretical generalisability is about a concept informing further research, and is common in qualitative work.
Qualitative research usually aims for transferability rather than statistical generalisation.
A large sample from one narrow group still does not generalise.
Sample and population
You cannot test everyone, so you test some people and hope they stand in for the rest. The group you want to describe is the target population. The people you actually test are the sample. Generalising means carrying the finding back out from the small group to the large one.
The dark dots are the people who ended up in the study. Whether they were spread across the population or clustered in one corner decides everything.
What a sample needs
Representative. The sample should mirror the population on whatever matters: age, gender, background, and anything else that could affect the result.
Randomly selected where possible, so the researcher’s own preferences do not shape who takes part.
Large enough that one unusual person cannot drag the average around.
Size and representativeness are not the same thing. A study of 13,000 people has plenty of statistical power. If all 13,000 were nurses working in one country, the findings still describe nurses in that country and nobody else. Say both parts.
Three kinds of generalisability
The third card is the one that lets small qualitative studies matter. They generate ideas that later, larger work can test.
Sample to population
This is the everyday meaning: results from the people tested are inferred to apply to the population they were drawn from. It depends entirely on how the sample was obtained, which is why sampling technique matters so much. Small, self-selected or convenient samples cannot support this move, because they only represent the people who were around and willing.
To other settings
This is inferential generalisability, and it is really external validity under another name. A finding produced in a laboratory has to survive contact with a school, a workplace or a street before anyone can rely on it there. Studies conducted in real-world contexts start with an advantage here.
To new theory
Qualitative research usually is not trying to say “and this is true of 70 per cent of the population”. It is trying to produce an insight that helps you understand similar situations. That is theoretical generalisability, and its qualitative name is transferability: whether the insight can be carried into other contexts to help make sense of them.
🤔 Why “small sample” is a lazy criticism
If a study is qualitative, small samples are the point, not the flaw. Criticising a five-person interview study for not generalising is like criticising a microscope for not showing you the whole sky. The better criticism is to ask whether enough detail was given for a reader to judge where the findings might transfer. Save the sample-size point for quantitative work that is actually claiming to generalise.
What damages generalisability
Problem
What it looks like
Effect on the finding
Sampling bias
Only available or willing people took part
Describes that subgroup, not the population
Volunteer bias
People who answer adverts share traits
Results skewed towards the outgoing and interested
Narrow sample
All from one university, country or culture
Findings may not apply elsewhere
Artificial setting
Task bears no resemblance to real life
Behaviour may not occur outside the study
Age of the study
Social conditions have since changed
Findings may no longer hold today
Whenever you make a generalisability point, finish it with who the findings do apply to. “These results describe willing university students in one country” is far stronger than “the results cannot be generalised.”
Worked examples
WORKED EXAMPLE
Evaluate generalisability properly
A study of stress used 13,000 nurses working in one country. The researchers claim their findings apply to working adults in general. Evaluate this claim. [4]
Step 1: Give credit where it is due
13,000 is a very large sample, so the results are statistically stable.
Step 2: Attack the representativeness, not the sizeAll participants share one profession and one country.Step 3: Say why that matters here
Nursing involves shift work and emotional demands that most jobs do not, so the stress findings may be specific to it.
Generalises to nurses in that country, not to working adultssaying who it does apply to is what turns this into a full answer
WORKED EXAMPLE
Defend a small qualitative study
A researcher interviews six refugees about starting school in a new country. A critic says the findings are worthless because six people cannot represent anyone. Respond to this criticism. [4]
Step 1: Accept the narrow point
Six people cannot support statistical generalisation, and the study never claimed to.
Step 2: Name what it is aiming for insteadTheoretical generalisability, or transferability — insight that helps make sense of similar situations.
Step 3: Say what makes that possible
Rich description of methods, context and findings lets a reader judge where it transfers.
Judge it by transferability, not sample sizeit can also generate hypotheses that larger studies then test
💡 Exam tip
Always name the target population before commenting on the sample.
Attack representativeness rather than size. It is the sharper and more accurate point.
Link generalisability to external validity explicitly. They are close relatives.
Use transferability, not generalisability, when the study is qualitative.
If asked how to improve a study, stratified or random sampling is usually the answer worth stating.
Culture is a strong angle. A finding from one cultural context may not describe another at all.
⚠️ Common mix-up
Big sample equals generalisable. Only if it is also representative.
Generalisability and reliability. One is about who the findings cover, the other about consistency.
Criticising qualitative studies for not generalising. They were never trying to.
Confusing the sample with the population. The sample is who took part; the population is who you want to describe.
Saying “the sample was biased” with no detail. Say which kind of bias and what it did.
Assuming a random sample is automatically representative. By chance it can still come out lopsided.
Up next: Bias in the Researcher and the Participant — the two people in the room, and the ways each of them quietly bends the data.
Want this explained one-to-one?
Book a free session with an experienced IB Psychology tutor and get your trickiest topics made simple.
