Hypotheses and variables
A hypothesis is a clear prediction we can test.
- Alternative hypothesis: there will be a difference or effect. 'People recall fewer words with music than in silence.'
- Null hypothesis: there will be no difference. 'Music makes no difference to the number of words recalled.'
Variables are things that can change.
- Independent variable (IV): what the researcher changes (music or silence).
- Dependent variable (DV): what is measured (number of words recalled).
- Extraneous variables: other things that might affect the DV (noise, time of day, tiredness). They must be controlled. If they are not, they can become confounding variables and spoil the result.
To operationalise a variable means to say exactly how it will be measured: 'number of words out of 20 recalled in 2 minutes'.
Sampling
The target population is the whole group we want to know about. A sample is the smaller group we actually test. A good sample is representative (like the population), so we can generalise.
- Random: every person has an equal chance (names from a hat or a computer). Fair, but may still be unbalanced.
- Opportunity: whoever is available and willing. Quick, but often biased.
- Systematic: every nth person from a list (every 4th). Simple, little researcher bias.
- Stratified: the sample has the same proportions as the population. If 60% are girls, 60% of the sample are girls. Very representative, but slow.
Designing research
Experimental designs
- Independent groups: different people in each condition. No order effects, but groups may differ.
- Repeated measures: the same people do every condition. Fewer people needed, but order effects (practice, boredom).
- Matched pairs: people are paired on a key quality; one of each pair goes in each condition. Fewer differences, but matching takes time.
Types of experiment
- Laboratory: controlled setting; easy to repeat; may feel unreal.
- Field: IV changed in a real setting; more natural; less control.
- Natural: the IV changes on its own (a town gets internet); researcher just measures.
Other methods
- Interviews (structured = fixed questions; unstructured = free conversation) and questionnaires (written questions; open or closed).
- Case study: deep study of one person or small group.
- Observation: watching behaviour; can be covert or overt, participant or non-participant.
- Correlation: measures two variables to see if they are linked. Positive, negative or none. It does not prove cause.
Procedures, reliability, validity and ethics
- Standardisation: everyone gets the same instructions and conditions.
- Randomisation: using chance, e.g. to put people in groups or order word lists.
- Counterbalancing: in repeated measures, half do A then B, half do B then A (ABBA). Order effects cancel out.
Reliability = consistency: repeat it and get the same result. Validity = accuracy: it measures what it claims to measure. A test can be reliable but not valid, like a scale that always reads 2 kg too heavy.
Ethics (guidelines from professional bodies such as the British Psychological Society):
- Informed consent: people agree knowing what will happen. Parents consent for under-16s.
- Deception: avoid it; if used, debrief afterwards.
- Right to withdraw: people can leave at any time, with their data.
- Confidentiality: names and data kept private.
- Protection from harm: no more stress than in everyday life.
Try it: plan a mini study
Question: 'Do people remember more words in silence or with music?'. Write: your alternative and null hypotheses, the IV and DV, how you will sample 10 people, your design, one extraneous variable you will control, and one ethical step. Check each choice with the 3D steps.
Key formulas and definitions
- Key term: IV = what is changed; DV = what is measured
- Key term: Null hypothesis = no difference; Alternative hypothesis = a difference
- Sampling: Random · Opportunity · Systematic (every nth) · Stratified (same proportions)
- Designs: Independent groups · Repeated measures · Matched pairs
- Reliability = consistent; Validity = accurate (measures what it claims)
- Stratified share = (group size ÷ population) × sample size
Worked examples
1. A study tests whether eating breakfast affects test scores. Name the IV and DV and write a null hypothesis.
IV: eating breakfast or not. DV: test score (marks out of 50). Null: 'Eating breakfast makes no difference to test scores.'
2. A school has 300 boys and 200 girls. You want a stratified sample of 20. How many girls?
Girls are 200 ÷ 500 = 40% of the population. 40% of 20 = 8 girls (and 12 boys).
3. Every participant learns a word list in silence, then another list with music. What design is this, and what problem may happen? How do you fix it?
Repeated measures. Order effects (practice or boredom) may change the second score. Fix by counterbalancing: half do music first, half silence first.
Common mistakes
- Swapping IV and DV. The IV is changed; the DV is measured.
- Saying opportunity sampling is random. It just picks whoever is nearby.
- Thinking correlation shows cause. It only shows two things are linked.
- Mixing reliability and validity. Reliable = same each time; valid = measures the right thing.