You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Before a psychologist collects a single data point, two decisions have already shaped what the study can and cannot show: who will be studied, and how the variables will be defined. Get sampling wrong and even a perfectly executed experiment generalises to no one; get the variables wrong and the cleverest analysis measures the wrong thing. This lesson deals with both. It covers the target population and the five sampling methods Edexcel expects you to know (random, systematic, stratified, opportunity, volunteer), the twin problems of representativeness and bias, the three forms of hypothesis (directional, non-directional, null), and — pulling the threads together — the discipline of operationalising variables so that a study is replicable and its measures are defensible. These are among the most reliably examined skills in Paper 3: an item will typically describe a study and ask you to identify a sampling method and evaluate it in context, or to write a fully operationalised hypothesis for the scenario. This lesson builds both.
Key Definition: A sample is the group of participants actually studied, drawn from a larger target population. The aim is a representative sample, so that findings can be generalised to the population it was drawn from.
By the end of this lesson you will be able to:
Edexcel 9PS0 — Paper 3: Psychological Skills (Research Methods). This lesson develops the sampling, hypothesis and variable-operationalisation strands assessed in Section A of Paper 3. Our sequence deliberately leads with the logic of generalisation (population → sample → representativeness) before cataloguing methods, so the structure reflects our own teaching rationale rather than the specification's ordering.
| Our lesson covers | Edexcel 9PS0 research-methods area |
|---|---|
| Target population; random, systematic, stratified, opportunity, volunteer sampling | Sampling and selection of participants |
| Representativeness, sampling bias, generalisation | Bias and the limits of generalisation |
| Directional, non-directional and null hypotheses | Aims and hypotheses |
| Operationalising the IV and DV (and co-variables) | Variables and their measurement |
Assessment Objectives. These items are AO2-dominated (identify the sampling method used in a described study; write an operationalised hypothesis for a given aim) and AO3-weighted (evaluate the representativeness of a sample, or the consequences of a particular bias for generalisation), on an AO1 base of accurate definitions. As across Paper 3, applied and evaluative marks outnumber pure recall, so answers must engage the specific scenario.
Connects to…
Almost no study can test everyone the researcher is really interested in. A psychologist studying, say, exam stress in UK sixth-formers cannot test all several hundred thousand of them; instead they study a sample and hope to generalise the findings back to the whole target population. Two technical terms make this precise:
Key Definition: The target population is the entire group about whom the researcher wishes to draw conclusions (e.g. "UK sixth-form students"). The sampling frame is the actual list from which the sample is drawn (e.g. "the enrolment register of three named sixth-form colleges").
Whether a generalisation is legitimate depends on whether the sample is representative — whether it mirrors the target population on the characteristics that matter (age, sex, ethnicity, ability, and so on). A representative sample supports confident generalisation; an unrepresentative one does not, no matter how large it is. This is the central idea of the topic, and it drives the evaluation of every sampling method below.
Key Definition: Sampling bias is any systematic tendency for a sample to over- or under-represent certain members of the target population, so that the sample is not representative and the findings cannot safely be generalised.
Note the word systematic. Random variation between a sample and its population is unavoidable and shrinks as samples grow; bias is different — it is a built-in tilt produced by how participants were selected. Volunteer samples, for instance, are biased towards the more motivated and cooperative; opportunity samples are biased towards whoever happened to be around when the researcher was collecting data. Recognising the direction of a bias — and reasoning about who is missing and how that distorts the conclusion — is exactly the evaluative move that earns AO3 marks.
Edexcel expects fluency with five methods. It is useful to split them into two families. Random, systematic and stratified all use a defined selection procedure and aim (with varying success) at representativeness; opportunity and volunteer trade representativeness for speed and convenience.
| Method | How it works | Strengths | Limitations |
|---|---|---|---|
| Random | Every member of the target population has an equal chance of selection (names drawn from a hat; random-number generator applied to a numbered list) | Free from researcher bias; the most likely of any method to be representative | Requires a complete list of the population (rarely available); can still, by chance, be unrepresentative; time-consuming |
| Systematic | Every nth member of a list is selected (e.g. every 5th name), often after a random start | Objective; simple; no researcher bias in who is chosen | Not truly random (once the interval and start are set, selection is fixed); a periodic pattern in the list could coincide with the interval and bias it |
| Stratified | The population is divided into subgroups (strata, e.g. year groups) and participants are drawn in proportion to each stratum's size, usually randomly within strata | The most representative method, because subgroup proportions are deliberately reproduced | Time-consuming; the strata must be identifiable in advance, and any characteristic not stratified can still be unrepresentative |
| Opportunity | Whoever is available and willing at the time is used | Quick, easy and economical; the most practical method | Unrepresentative (a narrow slice of the population); open to researcher bias in who is approached |
| Volunteer (self-selected) | Participants respond to an advertisement or notice | Easy access to willing, committed participants; convenient for the researcher | Volunteer bias — those who come forward may be atypical (more motivated, more curious, more time), limiting generalisation |
graph TD
A[Sampling methods] --> B[Aim for representativeness]
A --> C[Prioritise convenience]
B --> B1[Random: equal chance for all]
B --> B2[Systematic: every nth from a list]
B --> B3[Stratified: proportional subgroups]
C --> C1[Opportunity: whoever is available]
C --> C2[Volunteer: those who respond to an advert]
style B fill:#27ae60,color:#fff
style C fill:#e67e22,color:#fff
Random sampling is often held up as the "ideal", and the reason is subtle: it is the only method in which selection is independent of any characteristic of the person. Because every member of the population has an equal chance, there is no systematic reason for any subgroup to be over- or under-represented, so on average the sample mirrors the population. The catch is practical. True random sampling needs a complete sampling frame — a full list of the target population — which for most real populations simply does not exist (there is no list of "all people with insomnia"). And randomness guarantees only the absence of systematic bias, not representativeness in any single draw: a small random sample can, by chance, contain far too many of one group. This is why examiners like to distinguish random sampling from the random allocation met in the experimental-design lesson — the first decides who takes part, the second decides which condition each participant is placed in.
Systematic sampling feels random but is not, and the difference is worth stating carefully. Once the sampling interval (say, every 5th) and the starting point are fixed, the entire sample is determined — person 3, 8, 13, 18… will be chosen and no one else can be. Selection is objective and free of researcher bias, but it is not true randomness, and it carries a specific risk: if the list has a periodic structure that lines up with the interval (imagine a class register ordered boy, girl, boy, girl… and an interval of 2), the sample could systematically exclude an entire group. In practice a random start mitigates this, but the periodicity risk is the standard exam point.
Stratified sampling is the most representative method because it builds in the population's structure rather than hoping to reproduce it by chance. The researcher identifies the relevant strata (subgroups), calculates each stratum's proportion in the population, and samples that same proportion — usually randomly within each stratum. If Year 12 makes up 55% of a college and Year 13 45%, a stratified sample of 40 draws 22 from Year 12 and 18 from Year 13. The pay-off is faithful subgroup proportions; the price is effort (the strata must be known and the population divisible by them) and the fact that only the stratified characteristics are guaranteed representative — the sample could still be skewed on some variable no one thought to stratify by.
Exam Tip: A common AO2 task is to identify the sampling method in a scenario and evaluate one strength or weakness in context. Watch the trigger phrases: "put up a notice / advertised for participants" → volunteer; "asked students in the college canteen" → opportunity; "every 10th patient on the register" → systematic; "in proportion to the number of male and female nurses" → stratified; "names drawn using a random-number generator" → random. Then say who is likely missing and how that limits generalisation — that is the AO3.
The choice of sampling method is a genuine design decision with real consequences, and comparing methods on the same study makes the trade-offs concrete. Suppose a researcher wants to investigate attitudes to online learning among the 800 students at a sixth-form college.
Notice that the "best" method is not fixed: it depends on the resources available, whether a full sampling frame exists, and how much the conclusion depends on subgroup representation. This is precisely the reasoning an exam scenario invites — not "which method is best in the abstract?" but "which method suits this study, and what does the researcher's actual choice cost in terms of generalisation?"
Sampling ultimately serves generalisation — the move from "what we found in our sample" to "what is probably true of the population". The technical name for the external validity concern that a sample supports is population validity: the extent to which findings generalise to people beyond those studied. A sample can fail on population validity in ways that are easy to miss. It may be biased by age (studies on undergraduates say little about older adults), by culture (Western samples may not generalise to collectivist cultures — a recurring criticism in the science's over-reliance on WEIRD participants), by sex (Milgram's all-male sample, or the historical dominance of male participants in early research — sometimes called a beta-bias problem when sex differences are then ignored), or by era (an attitudes study from decades ago may lack temporal validity today). A skilled evaluation names the specific dimension on which a sample is unrepresentative and reasons about the direction of the resulting distortion, rather than asserting a vague "the sample was biased".
Abstract talk of bias becomes far sharper when tied to studies you already know. In Asch's (1951) conformity experiments, the participants were male American college students recruited in the early 1950s. That is a triple restriction — on sex, on culture and on era — and each restriction has been argued to inflate the conformity rate: 1950s America was an unusually conformist, McCarthy-era culture, and later replications in different decades and cultures (for example, Perrin and Spencer's 1980 UK study with science and engineering students) found markedly less conformity. The bias here is not that "the sample was small" but that the specific slice of humanity Asch tested was systematically more prone to the very behaviour being measured, so the headline figure over-states conformity in the general population.
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.