Researchers almost never study everyone they care about. So three words have to be kept apart.
The target population is the broader group the researcher wants to draw conclusions about. The sample is the people who actually take part. Sampling is the process used to select them.
Sampling matters because some techniques make certain people more likely to be included than others, and that shapes both bias and what the findings can be applied to.
The confusion that costs the most marks
Random sampling is not random allocation.
Random sampling concerns who enters the sample from the target population. It mainly affects representativeness and generalisability.
Random allocation concerns which experimental condition a participant enters after they have already been recruited. It mainly protects internal validity and the causal claim.
A study can use opportunity sampling and still use random allocation. Almost every class practical does exactly that.
If you can keep those two apart in an unseen study, you will avoid the single most common error in Section B answers about experiments.
The five techniques
The syllabus names five. For each one you need to recognise it in an unseen study, state one advantage and one disadvantage, and explain what it does to generalisability.
| Technique | How it works | Main advantage | Main disadvantage |
|---|---|---|---|
| Random | Every member of the target population has an equal chance of selection, usually from a full list | Reduces systematic selection bias and can improve representativeness | Needs a complete list of the population, which rarely exists. A small random sample can still be unrepresentative by chance |
| Stratified | The population is divided into subgroups such as year group or gender, and participants are drawn from each in proportion | Ensures key subgroups appear in the right proportions. Strong for generalisability | Time-consuming. You can only stratify on characteristics you thought of in advance |
| Opportunity | Participants are whoever is available and willing | Fast, cheap, practical. This is what almost every class practical uses | Highly unrepresentative. Usually drawn from one narrow setting |
| Self-selected (volunteer) | Participants respond to an advert, poster or online call | Easy to obtain, and participants are usually motivated and cooperative | Volunteer bias. People who volunteer often differ systematically from those who do not, in confidence, interest or free time |
| Snowball | Existing participants recommend others who fit the criteria | Reaches hidden or hard-to-access populations, for example people with a stigmatised experience | The sample is drawn from linked social networks, so it can be unusually homogeneous |
Naming the technique earns you nothing
This is the part that separates a sampling point from a sampling sentence.
Weak. They used an opportunity sample, which is a limitation.
That earns almost nothing. It names a technique and asserts that it is bad. It does not say what happened to the evidence.
Strong. The researchers recruited students from a single sixth-form psychology class, which is an opportunity sample. Everyone in the study was therefore of similar age, in the same school, and already studying psychology, so they may have been unusually alert to what the study was testing. The finding can reasonably be applied to psychology students in that setting, but not confidently to adolescents in general, which limits how far the conclusion travels.
The mark comes from the consequence chain, not the label.
The template that produces it
Use this shape until it is automatic.
The researchers used [technique]. This means [what it does to who ends up in the study]. As a result, [what kind of person is over- or under-represented]. Therefore the finding can reasonably be applied to [narrow group] but not confidently to [broader group], which limits [generalisability or transferability].
Four sentences. The technique, the mechanism, the distortion, the consequence for the claim.
Notice the last word offers a choice. In quantitative work you are usually talking about generalisability. In qualitative work the appropriate term is transferability, and using the quantitative word for qualitative research is the kind of mismatch examiners notice.
Bias is not a verdict
Finding a sampling problem does not mean the finding is wrong.
A study of one population is not invalid. It is a study of that population. The bias lives in the gap between what was studied and what is being claimed.
So the move is almost never to dismiss the study. It is to restate the claim at the size the evidence actually supports. That is a more sophisticated thing to write and it is also more accurate.
The other sampling words worth knowing
Sampling bias is the systematic exclusion of certain people by the recruitment method.
Attrition is the loss of participants over the course of a study. This one matters most in longitudinal work, because the people who drop out are rarely a random subset of the people who started. If the participants who leave are systematically different, the sample at the end is not the sample you began with.
A representative sample is one whose relevant characteristics mirror the population. Relevant is doing work in that sentence. Representative on age does not mean representative on everything.
One more that is worth holding onto: size and representativeness are independent. A very large sample recruited through one platform shares the characteristics of people on that platform. A large biased sample can be more misleading than a small one, because the narrow margin of error creates false confidence. Precision is not accuracy.
What to record for your own practicals
For each of your four class practicals, write down four things.
The target population you were interested in. The actual sample, with the number. The sampling technique, named specifically. And one honest sentence about who your sample over-represents.
Almost every class practical uses opportunity sampling, and there is nothing wrong with saying so. What earns marks is being able to say what that did to your evidence, rather than apologising for it.
If you were running an experiment, record the allocation method separately from the sampling technique. They are different decisions doing different jobs, and question 1a rewards you for showing that you know it.
If you want the full research reference plus worked models for every method, that is what the PsychExcel Paper 2 Student Guide is built around.