"Music helps you revise." How could you test that?

From a claim to a prediction

• An aim says what a study will investigate, e.g. whether music affects memory.
• A hypothesis is a precise, testable prediction.
• The alternative hypothesis predicts a difference or relationship: "Students recall more words in silence than with music."
• The null hypothesis predicts no difference or relationship: "There will be no difference in recall with music and in silence."
• Hypotheses can predict a difference, a correlation or no pattern (the null).

What you change and what you measure

• The independent variable (IV) is what the researcher changes (manipulates) to create the conditions.
• The dependent variable (DV) is what the researcher measures.
• Operationalise each variable: define it precisely so it can be measured.
• A measure that really tests what it claims to (here, memory) has construct validity.
• A correlation has no IV: two co-variables are both measured, e.g. hours of sleep and test score.

What you keep the same

• An extraneous variable is any other variable that could affect the DV.
• If it isn't controlled, you can't tell whether the IV caused the change in the DV.
• Situational variables come from the setting: noise, time of day, the word list.
• Participant variables come from the people: age, memory ability, tiredness.
• Control them by keeping them the same for every participant. This is called standardisation.

Lab or real classroom: which experiment's result should you trust?

The laboratory experiment

• The researcher manipulates the IV and measures the DV in a controlled, artificial setting.
• Participants know they are taking part.
• Strengths: extraneous variables are tightly controlled, so cause and effect can be shown. Standardised procedures make it easy to replicate.
• Weaknesses: artificial tasks may not reflect real life (low ecological validity). Participants may guess the aim and change their behaviour (demand characteristics).

Field and natural experiments

• In a field experiment, participants often don't know they are in a study.
• In a natural experiment, the researcher compares groups that already differ, e.g. pupils at schools that already start the day at different times.

Trading control for realism

• Reliability means consistency: repeat the study and get similar results (external reliability).
• Validity means the study measures what it intends and applies beyond the study.
• Field: higher ecological validity and fewer demand characteristics, but less control and harder to replicate. Participants often can't give consent.
• Natural: can study IVs that would be unethical or impossible to manipulate, but participants aren't randomly allocated, so participant variables may explain results.

Same people in both conditions, or different people: which is better?

Different people in each condition

• In an independent measures design, each participant takes part in only one condition.
• Strengths: no order effects, and participants are less likely to guess the aim.
• Weaknesses: participant variables: the groups may differ (e.g. one has better memories), and that could cause the result. More participants are needed.
• Control: random allocation to conditions (e.g. drawing names from a hat) spreads participant variables across the groups.

The same people in every condition

• In a repeated measures design, each participant takes part in every condition.
• Strengths: no participant variables between conditions, and fewer participants are needed.
• Weaknesses: order effects: a practice effect (better second time) or a fatigue effect (worse through tiredness or boredom). Participants may guess the aim (demand characteristics).
• Control: counterbalancing: half do condition A then B, half do B then A, so order effects balance out.

Keeping the procedure fair

• Standardised procedures: every participant gets the same tasks, materials, timings and setting, and standardised instructions (the same words).
• Randomisation: chance decides things like the order of word lists, so the researcher's choices can't bias the results.
• This is standardisation. It also reduces experimenter bias: the researcher's expectations affecting results.

You can't test every UK teenager. Are 20 from your class enough?

Population and sample

• The target population is the whole group a researcher wants to find out about, e.g. UK 13 to 16-year-olds.
• The sample is the group who actually take part.
• A representative sample has the same mix of characteristics as the target population.
• Then the findings can be generalised: applied to the whole target population.
• Sample size: small samples are less likely to be representative; one unusual person can skew results.

Random and opportunity sampling

• Pulling names from a hat is random only if every member's name is in the hat.

People who put themselves forward

• In self-selected sampling, people choose to take part, e.g. by replying to an advert, poster or online post.
• Strength: easy, and participants are willing. An advert can reach many people.
• Weakness: people who volunteer tend to share traits, such as being more helpful or more interested in the topic, so the sample may be unrepresentative (volunteer bias).

Population validity and bias

• Population validity: how far findings apply to people beyond the sample. Unrepresentative samples lower it.
• Gender bias: results from one gender (e.g. an all-male sample) are assumed to apply to everyone.
• Cultural bias: results from one culture are assumed to apply to all cultures.
• Age bias: results from one age group, e.g. university students, are assumed to apply to all ages.

How do you study a crowd, a childhood, or what people really think?

Questionnaires

• A questionnaire is a set of written questions, on paper or online.
• It can reach a large sample quickly and cheaply, and everyone gets the same questions.
• Rating scales ask for a number, e.g. 1 (strongly disagree) to 5 (strongly agree), giving quantitative data.
• Internal reliability: all the questions on one topic measure it consistently.

Interviews

• An interview asks questions face to face, or by phone or video.
• Structured: the same questions in the same order for everyone. Easy to compare and repeat, but no follow-up.
• Unstructured: like a conversation; the interviewer follows up answers. Rich detail, but hard to compare.
• Interviewees and questionnaire respondents may give socially desirable answers: ones that make them look good.
• Bias in questioning: leading questions push people towards an answer.

Observations

• In an observation, researchers watch and record behaviour as it happens.
• Categories of behaviour are decided in advance, e.g. "pushes", "shares a toy", and tallied each time they occur.
• Inter-rater reliability: two observers record separately, then compare. High agreement means the records are reliable.
• Strength: shows what people actually do, not what they say. Weakness: shows what happens, not why.
• Observer bias: observers' expectations affect what they record.

Types of observation

• Observer effect: people change their behaviour because they know they are being watched, a risk in overt observations.
• Covert observation avoids that but raises an ethical issue: no consent.

Case studies

• A case study is an in-depth study of one person or a small group, often over a long time.
• It combines methods (interviews, observations, records) and mainly gives qualitative data from a small sample.
• Strengths: rich detail; can study rare cases, e.g. an unusual brain injury.
• Weaknesses: findings may not generalise; hard to replicate; relies on memory and the researcher's interpretation.

More ice creams sold, more swimmers rescued. Is ice cream to blame?

Two things that change together

• A correlation is an association between two variables.
• Both variables are measured using quantitative data. Neither is manipulated, so there is no IV or DV.
• The two measured variables are called co-variables.
• A scatter diagram shows a correlation: one variable on each axis and one dot for each person or case.

Positive, negative or zero

• Positive correlation: as one variable increases, so does the other, e.g. if more hours of revision went with higher test scores.
• Negative correlation: as one variable increases, the other decreases, e.g. if more screen time went with less sleep.
• Zero correlation: no relationship, e.g. shoe size and test score.
• The closer the dots lie to a straight line, the stronger the correlation.

Linked, but not proved to cause

• A correlation can't show that one variable causes the other.
• A third variable may cause both, or the cause may run the other way.
• Strengths: can study links that would be unethical or impossible to manipulate; quick, often using existing data; suggests ideas to test in experiments.
• Weaknesses: no cause and effect; an unmeasured third variable may explain the link.

Psychologists have lied to participants. When is that allowed?

The rules psychologists follow

• An ethical issue arises when a study could harm participants or ignore their rights.
• UK psychologists follow the British Psychological Society (BPS) Code of Ethics and Conduct.
• Its four principles are respect, competence, responsibility and integrity.
• Researchers must weigh what a study could teach us against any cost to participants, and protect participants first.

Consent and deception

• Informed consent: participants agree to take part after being told what the study involves. Under-16s also need consent from a parent or guardian.
• Deception: misleading participants or hiding the true aim.
• Deception is acceptable only if telling the aim would spoil the results, it won't cause distress, and participants are debriefed afterwards.
• A lack of informed consent and deception are two key ethical issues.

Harm, withdrawal and confidentiality

• Protection of participants: participants should leave in the same state as they arrived, facing no more physical or psychological risk than in everyday life.
• Right to withdraw: participants can leave at any time and ask for their data to be removed.
• Confidentiality: personal information stays private; participants get numbers, not names, so they can't be identified.
• The right to withdraw and confidentiality are ways of dealing with ethical issues.

Dealing with ethical issues

• Debriefing: at the end, explain the true aim, reveal any deception, check participants are OK and remind them they can withdraw their data.

Ethics in classic studies

• Real studies show why the guidelines matter.
• In Bickman (1974), passers-by were given orders by people in uniform without knowing it was a study, so they gave no informed consent.

Two classes have the same average mark. What can the average hide?

Kinds of data

• Quantitative data can be summarised with averages, measures of spread and charts.

Averages and spread

• Mean: add the scores and divide by how many there are. Uses every score, but extreme scores distort it.
• Median: the middle score when they are in order (with two middle scores, take halfway between). Not affected by extreme scores.
• Mode: the most common score.
• With grouped data, the group with the highest frequency is the modal class.
• Range: highest minus lowest, a measure of dispersion (spread).

Charts and the normal distribution

• A pie chart shows shares of a whole; a line graph shows how a variable changes, e.g. over time.
• A normal distribution is a symmetrical, bell-shaped curve. The mean, median and mode are equal, at the centre. Most scores are near the middle; few are at the extremes.

Big ideas

• A hypothesis is a testable prediction: the alternative predicts a difference or relationship, the null predicts none.
• The IV is changed and the DV is measured; both must be operationalised. Extraneous variables must be controlled, e.g. by standardisation; in a correlation, two co-variables are measured instead.
• Lab and field experiments both manipulate an IV; a natural experiment uses an IV that varies naturally.
• Labs give control and replicability (external reliability); field and natural experiments give higher ecological validity but less control.
• Choose the type that suits the question: a lab for cause and effect, the field for real behaviour, a natural experiment when the IV can't ethically be manipulated.
• Independent measures: different people in each condition, so watch for participant variables (use random allocation).
• Repeated measures: the same people in every condition, so watch for order effects and demand characteristics (use counterbalancing).
• Standardisation and randomisation control extraneous variables and experimenter bias.
• A sample should represent the target population so the findings can be generalised (population validity).
• Random, opportunity and self-selected sampling each trade convenience against representativeness; small samples and gender, cultural or age bias lower population validity.
• Questionnaires and interviews ask; observations watch; case studies go deep into one person or small group.
• Closed questions give quantitative data and open questions give qualitative data; answers may be socially desirable.
• Watch for the observer effect, observer bias and bias in questioning; check inter-rater and internal reliability.
• A correlation measures two co-variables and shows whether they are positively, negatively or not related, usually on a scatter diagram.
• Correlations can study links that can't be manipulated, but they can't show cause and effect: a third variable may explain the link.
• Psychologists follow the BPS Code of Ethics and Conduct.
• The key issues are a lack of informed consent, psychological harm and deception; they are dealt with by debriefing, the right to withdraw and confidentiality.
• Data can be quantitative or qualitative, primary or secondary.
• Mean, median and mode describe the centre; the range describes the spread; reliability is about consistency and validity about accuracy.
• A normal distribution is symmetrical and bell-shaped, with mean, median and mode equal at the centre.