4.2.3.1 - Sampling

Population, sample and generalisation

Sampling is the bridge between a researcher's question and the people who actually take part. Most psychological research cannot test every person the researcher is interested in, so the researcher studies a smaller group and then decides how far the findings can be applied.

For AQA Psychology, "population" usually means the target population.

It might be "all UK sixth-form students", "16-18-year-olds at one college", "adults with a diagnosis of depression" or "psychology teachers in England". The population must be defined clearly because the sample should come from the group the researcher wants to generalise to.

Suppose Maya wants to study sleep and revision in A-level students. If her target population is all A-level students in the UK, a sample of 40 students from her own Psychology class is easy to collect but narrow. If her target population is Psychology students at her college, the same 40 students may be more appropriate, although one class could still miss relevant differences between classes. The same sample can be strong or weak depending on the population being claimed.

Two other terms help you evaluate sampling.

This is why sampling is an AO3 issue as well as an AO1 issue. A study might use a careful procedure and valid measures, but if its sample is very narrow, the results may not generalise well. A memory study based only on high-achieving volunteers from one school may tell us something useful, but it should be applied cautiously to all teenagers.

In exam answers, keep the chain clear:

Random, systematic and stratified sampling

Random, systematic and stratified sampling are usually more planned than opportunity or volunteer sampling. They normally require a sampling frame.

For example, a school register could be a sampling frame for students in that school.

In simple random sampling, the researcher might:

1. Define the target population.
2. Obtain a complete list of members.
3. Give each member a number.
4. Use a random number generator to choose the required number of distinct students, without replacement; ignore a duplicate number and draw again.

Random sampling reduces researcher selection bias because the researcher does not personally choose who looks useful, friendly or easy to test. It can support generalisation if the sampling frame is complete and enough people respond. However, it can be impractical for a large population, and chance can still produce an unbalanced sample, especially if the sample is small. A random sample of 20 students from a large school might, by chance, include very few Year 13 students.

Systematic sampling often starts by calculating a sampling interval. If a school has 1,000 students and a researcher wants 100 participants, the interval is 10, so the researcher could select every 10th name after randomly choosing a starting position from 1 to 10. If the start is 4, select positions 4, 14, 24 and so on until 100 distinct students are selected. If the starting point is random and the list has no hidden pattern, systematic sampling can be efficient and reasonably representative.

The main warning is periodicity: a repeating pattern in the list can bias the sample. If a sports club list alternates junior and senior members and the researcher selects every second name, one age group may be over-represented or missed.

Stratified sampling is useful when the researcher knows that particular subgroups matter. Imagine a sixth form with 60% Year 12 students and 40% Year 13 students. If the researcher wants a sample of 50 students, a proportional stratified sample would include 30 Year 12 students and 20 Year 13 students, selected randomly within each year group. The calculation is $50 \times 0.60 = 30$ and $50 \times 0.40 = 20$. More generally, multiply the desired sample size by the subgroup’s fraction of the population.

The proportional allocation and the selection rule are separate decisions. After working out how many people are needed from each stratum, the researcher still has to select them. A lottery, a suitable systematic rule or approaching accessible people within each stratum are possible procedures, but approaching accessible people leaves more risk of selection bias. Merely filling subgroup quotas does not guarantee that each subgroup sample is representative.

Stratified sampling can improve representativeness because key subgroups are deliberately included. It is especially useful when a simple random sample might accidentally under-represent a small but important group. The limitation is practical: the researcher needs accurate information about the population and must choose sensible strata. If the wrong strata are chosen, the sample can still be biased.

One exam trap is confusing random sampling with random allocation. Random sampling is about who gets selected from the population. Random allocation is about how selected participants are placed into experimental conditions. They solve different problems.

Working out a proportional sample

A college has 240 students: 150 in Year 12 and 90 in Year 13. For a proportional sample of 48, the Year 12 fraction is $150/240 = 0.625$, giving $48 \times 0.625 = 30$. The remaining 18 places go to Year 13. Selecting 30 Year 12 and 18 Year 13 students already states the strata and the correct proportions; repeating the same arithmetic is unnecessary unless working is requested. A full recruitment plan also needs a selection procedure, such as randomly drawing the required distinct names from each year-group register.

For systematic sampling of 48 from the same list, the interval is $240/48 = 5$. A random start of 3 gives positions 3, 8, 13 and so on to 238. Check the list order: if every fifth position belongs to one tutor group, that interval could systematically over-select it. If a proportional allocation gives fractions of people, use a stated rounding rule that keeps the total correct and proportions as close as possible.

Opportunity and volunteer sampling

Opportunity and volunteer sampling are often easier to use in real psychology studies, but they usually create more risk for bias and weaker generalisation.

Opportunity sampling is sometimes called convenience sampling in wider research-methods writing. For AQA, use the term opportunity. A researcher might ask people in a school common room, shoppers passing a display or students already waiting outside a laboratory. It is quick, cheap and useful when time or access is limited. It can also be sensible for a pilot study where the aim is to test whether materials work before a larger investigation.