Designing Research Methods (AQA GCSE Psychology): Flashcards

Age — comparing digit-span recall in a young and an older group; gender — comparing girls' and boys' performance on an emotional intelligence task; or circumstances — comparing trained and untrained teachers.

Because they already belong to one group or the other — for example, already young or old, already trained or untrained.

Natural experiments let researchers investigate topics that would otherwise be unethical to manipulate.

Because participants can't be randomly allocated, groups may differ in ways other than the IV — an 'age' IV might be confounded by differences in health or experience between the groups, making cause and effect hard to establish.

The naturally occurring circumstance may be unique — such as a specific institution or event, for example Covid — so the exact study is difficult or impossible to replicate.

They are high in ecological validity, as the IV occurs naturally rather than being artificially staged by the researcher.

Experimental design refers to the different ways participants can be organised in relation to the conditions of an experiment.

In an independent groups design, participants experience only one condition of the IV, and are randomly allocated to it to avoid researcher bias.

Unrelated data, because the two groups are separate from one another.

Compared to repeated measures, there is no possibility of order effects. Participants are also less likely to show demand characteristics, as one condition gives them no opportunity to guess the aim.

Individual differences may interfere with the findings — one group may have a naturally better memory than the other. More participants are also required.

In a repeated measures design, the same participants complete all conditions of the IV.

Related data, because the two scores come from the same participant.

Because their performance in condition 1 can be compared to their own performance in condition 2.

False.

Participants take part in all conditions, so order effects could affect their performance. Researchers use counterbalancing to control this.

Counterbalancing controls for order effects: practice, fatigue and boredom.

Individual differences are not an issue, and fewer participants are required than for independent groups or matched pairs.

In a matched pairs design, participants are assessed and matched on the characteristics important for the research — such as age, ethnicity or gender — then randomly allocated to one condition each.

They create the perfect matched pair — one twin can be assigned the experimental condition and the other the control condition.

Order effects are not an issue, as participants take part in only one condition. They are also less likely to show demand characteristics, having no opportunity to guess the aim of the study.

It is impossible to match participants 100% — even MZ twins differ in personality and abilities. It is also the most time-consuming design, which is why it is rarely used.

A questionnaire is a type of self-report in which participants answer a range of questions designed to collect their thoughts, feelings, attitudes, attributes and opinions.

Understanding how people feel about issues such as immigration or the cost of living; measuring psychometric properties such as IQ, depression or empathy; and conducting a large-scale survey.

The aim — its purpose and how it aids the research; the length — not so short there is a lack of data, nor so long participants become bored; and question construction — clear, concise and unambiguous.

A closed question offers limited options for the participant's response — for example 'Do you agree…? Yes/No'.

By using a scale, such as the Likert scale — 'On a scale of 1–7, how far do you agree…?'

Quantitative data: the number of 'yes' responses is totalled, the number of times a participant ticked an option is totalled, or a total score is calculated from scaled questions.

An open question offers freedom of response — for example 'Tell me about one time when you felt anxious'.

Qualitative data: the transcript of an interview, the themes which emerge in conversation, and the participant's thoughts and feelings, which cannot be boiled down into neat numerical data.

Open questions allow both researcher and participant to clarify, explain and qualify their comments.

Closed questions generate quantitative data which is straightforward to analyse and can be presented graphically or as percentages. Questionnaires are also easy to replicate, so high in reliability.

False.

Forcing participants into fixed categories may not capture their true thoughts or feelings, reducing validity.

Where participants' answers are distorted rather than accurate — for example through acquiescence bias, agreeing with a question regardless of its content, or by answering carelessly in a long questionnaire.

The tendency for people to under-report negative and over-report positive aspects of themselves.

Closed: 'Do you access social media at least once a day? Yes/No'. Open: 'Tell me how you feel about the impact that social media has on your life.'

An interview is a type of self-report in which a participant answers a range of questions put to them by a researcher — a one-to-one process, though it can happen by phone or online.

The thoughts, feelings, attitudes and opinions of the participant.

In a structured interview the researcher uses a series of pre-prepared closed or open questions, and does not veer from the 'script'.

Quantitative data only — such as the number of 'yes' responses to a specific question.

In an unstructured interview the researcher comes with no pre-prepared questions, keeping an open mind, and treats the interview as a conversation.

Qualitative data only.

In a semi-structured interview the researcher uses some pre-prepared questions, but also asks further questions based on the respondent's answers.

If the participant says something interesting or unexpected, is finding it difficult to discuss a topic, or is not being forthcoming and needs extra encouragement.

False.

They produce both quantitative and qualitative data, as they are a mixture of structured and unstructured styles.

Being one-to-one, participants can express themselves in their own words, without the constraints of fixed-response questions — making interviews high in validity. They are also more flexible than questionnaires.

Particularly unstructured ones, interviews can generate large amounts of qualitative data that is time-consuming to analyse and difficult to compare across participants.

Especially unstructured ones, interviews often rely on small, unrepresentative samples, which limits how far the findings can be generalised.

The interviewer's tone, body language or expectations can unconsciously influence how a participant responds — known as interviewer effects.

A case study is an in-depth, detailed and highly idiosyncratic account of a single participant, a small group of participants, or an institution — such as a school.

Interviews with the participant and people connected to them; naturalistic observations of the participant alone or interacting with others; and diary or journal entries written by the participant.

Psychometric tests such as IQ, personality and mental health measures; scores on memory tests; and clinical data such as MRI scans or blood tests.

Most case studies are longitudinal, meaning a participant's experience is tracked and measured over time.

HM developed anterograde amnesia following brain surgery, and was studied for 55 years.

A case study of dissociative identity disorder, which lasted several years.

They allow researchers to investigate rare or extreme conditions — such as severe brain damage or dissociative identity disorder — that would be unethical to create deliberately. Because the condition already exists, studying it raises no additional ethical concerns.

They amass a wealth of information which is rich, detailed and in-depth, giving them high validity.

Researchers often don't start with a fixed hypothesis or narrow aim, so they are more open-minded and can pick up unexpected findings, rather than being limited by what they expected to find.

False.

Generalising is very difficult, as they tend to feature only one participant or a small group, so the findings may not apply to anyone else.

They rely heavily on qualitative data, which the researcher must interpret. This interpretation can be influenced by the researcher's own views or expectations, which challenges the validity of the conclusions drawn.

The researcher may become closely involved with the participant over months or years, which could introduce bias and cause them to lose their objectivity.

False.

An observation involves watching and recording behaviour as it happens, rather than manipulating an IV.

They cannot infer motive, intention, feeling or thought. Only observable behaviours — what they can actually see — can be recorded.

A naturalistic observation is one in which the researcher observes and records behaviours in a natural setting, away from the lab, with no manipulation of an IV.

A controlled observation is one in which the researcher implements a level of control over the setting or procedure.

Zimbardo's (1973) prison study randomly allocated participants to the role of prisoner or guard in a simulated prison; Ainsworth's (1970) Strange Situation observed mothers and toddlers through a standardised sequence of controlled episodes.

Because they must be recruited for the study and then set a specific task, which is likely to be quite removed from their everyday activities.

In an overt observation, participants are aware they are being observed and may have been informed in advance — as in Zimbardo's prison study, where participants knew they were being filmed and observed.

In a covert observation, participants are not aware they are being observed, have not been informed in advance, and cannot see the researcher.

The only ethical way to conduct a covert observation is to observe behaviour in the public context.

In a participant observation, the researcher joins the group they are observing, becoming part of it — so it is highly likely to be covert.

Rosenhan (1973) and 7 others, known as 'pseudopatients', had themselves admitted to psychiatric hospitals and recorded what they observed during their stay.

It manipulated the victim's appearance — drunk vs ill — as an IV, and measured whether and how quickly people helped as the DV. It used participant observation to measure the DV, rather than being a pure observational study.

The researcher stays separate and apart from the group. Ainsworth (1970) observed toddler–mother interactions through a one-way mirror — a controlled, non-participant, overt observation.

Those which are naturalistic and covert are high in validity, as they measure real, unforced behaviour rather than behaviour produced for a researcher. Using multiple observers also allows inter-observer reliability to be checked.

They raise ethical concerns — there is no informed consent — which makes this type of research problematic to justify.

Where the researcher brings their own bias to the process, observing what fits their expectations rather than what is actually happening.

They do not manipulate an IV, so a researcher can only establish that certain behaviours occurred, not why.

Categories of behaviour are the observable behaviours which researchers will be looking out for during an observation.

Which specific behaviours should be recorded, so that all observers are in agreement beforehand.

So they are described precisely enough that any observer would code the same behaviour in the same way.

Behaviour categories are mutually exclusive when they do not overlap, so a behaviour fits into only one category — if two categories could both apply, observers won't know which to tally.

PA = physical aggression; ABL = aggressive body language; NAB = non-aggressive behaviour.

To make sure they are specific and cannot be confused with anything else — 'aggressive body language' subdivided into 'pointing', 'shaking fist', 'baring teeth'.

By recording whether it is boys or girls behaving aggressively or non-aggressively, and to whom that behaviour is directed.

Inter-observer reliability is the level of consistency between two or more trained observers when they conduct the same observation.

To avoid conformity — one observer may otherwise be influenced by one or more of the others.

They compare the two independent data sets, often designed as a tally chart, and then test the correlation between them.

False.

A strong positive correlation (+0.8 or more) shows good inter-observer reliability, and that the behaviour categories are reliable.

Establishing good inter-observer reliability means there is less chance that observer bias has interfered with the observation.

'Drops litter' and 'Puts litter in the bin'.

False.

A correlation is not a research method — it is an analysis of the possible relationship between co-variables.

Two co-variables are measured and compared to look for a relationship. The variables are not manipulated, so there is no IV.

School attendance, measured as days present in Year 11, and the number of GCSEs achieved.

Two — for example an average of 4 hours of sleep per night correlated with 58 cups of caffeine consumed in a week.

A positive correlation is where one co-variable increases as the other increases, though not necessarily at the same rate — such as calories consumed and weight gained.

A negative correlation is where one co-variable increases while the other decreases — such as hours spent watching television and level of fitness.

A zero correlation is where there is no relationship between the variables — such as hair colour and IQ.

A curvilinear relationship is one where the relationship between co-variables changes direction rather than staying constant — such as alertness rising through the day, peaking, then falling towards bedtime.

As points on a scatter diagram. The pattern and direction of the points shows what type of correlation exists between the two variables.

The correlation coefficient is a numerical value representing both the direction and the strength of the relationship between two variables.

A perfect positive correlation is expressed as +1, and a perfect negative correlation as −1.

0.2 is a weak positive correlation; −0.8 is a strong negative correlation.

They show a relationship between variables, which can be a good starting point for further experimental research — e.g. a positive correlation between temperature and aggression could lead a researcher to test this further with an experiment.

Correlational analysis can investigate more complex relationships than a simple straight line, such as curvilinear relationships — giving it wider use as a technique for understanding how variables are associated.

They cannot show cause and effect — so a positive correlation between stress and days off work doesn't tell us whether stress is causing illness, or whether being ill is causing the stress.

A third, 'in-between' factor that may really be causing an apparent link between two co-variables — e.g. a stressful job might lead someone to sleep less, smoke or drink more, and it is this behaviour, not the stress itself, that increases their chance of becoming ill.

Because a correlation involves no control of extraneous variables, the conclusions drawn may be misleading.