Sampling, Ethics, Reliability & Validity (AQA GCSE Psychology): Flashcards

Using participants who are, for example, all from the same town, the same age, or who share similar personality traits — which makes the results difficult to generalise.

Stratified sampling selects participants from different subgroups (strata) in the target population — such as age, gender, education or religion — in proportion to each subgroup's frequency in that population.

If the whole population has 18% of people who are male aged 30–40, then the representative sample will also have 18% of people who are male aged 30–40.

The sample is representative of the target population, making it easy to generalise, and selection is unbiased as it is based on the subgroups in society.

Knowing the subgroups, dividing the population into categories and selecting participants to match is very time-consuming — and the researcher needs knowledge of the subgroups, which may not be available.

Random sampling is selecting participants so that everyone has a fair chance of being selected — for example putting all names in a hat, or using a computer-generated method.

It is time-consuming and impractical, as a list of all members of the target population must be obtained. It may also be non-representative — for example, one gender could end up being selected entirely by chance.

Systematic sampling is selecting every nth person from a list to make a sample.

They calculate how many people they need in order to know which nth to use — every 10th, 100th or 1000th on a register, database or roll.

It is a quicker sampling method compared to stratified sampling.

False.

The selection process is not always truly unbiased — you could end up with all men or all women in the sample, making it non-representative of the target population.

Opportunity sampling is selecting those who are most convenient, willing and available to take part — such as shoppers in a mall at 11am, or students at one of the researcher’s lectures.

Opportunity sampling is also known as convenience sampling.

It is an easy way to find participants, and a time-efficient way of obtaining a sample.

You cannot generalise, as the sample is likely to be unrepresentative — for example, some university students could be in another lecture or at the library, and people at work are unlikely to be shopping at 11am.

The BPS — British Psychological Society. Ethics boards at universities and research institutions review proposed research against the BPS code of conduct.

Informed consent refers to the information that must be given to participants before the procedure, so they can make an informed decision about taking part.

How long the task will take; what they will be required to do; any potential hazards; and the purpose of the study.

False.

They are told the purpose, but not necessarily the aim — as knowing it could affect the validity of the study.

If participants are below the age of 16, parental consent must be obtained.

They may withdraw at any point in the research process, and their data is destroyed.

Deception refers to any attempt to mislead or lie to participants.

They were told they were taking part in a study about punishment and learning, when it was their obedience that was being measured. This would not be allowed today.

To protect the validity of the study — too much information about the aim could give rise to demand characteristics. Had Bartlett revealed his interest in unfamiliar cultural detail, participants would have been primed to look out for it.

It takes place after the procedure: giving the aim if it was withheld; telling participants about the other group if independent groups was used; and offering extra information or helplines and counselling.

Restore participants to the psychological and emotional state they were in before the study.

Harm is any physical, psychological or emotional damage inflicted on participants as a result of their participation in the research.

Many of the men suffered huge stress, anxiety and strain, with more than one suffering a seizure during the task.

Privacy is any invasion of an individual’s private space beyond what is acceptable for the topic. Dickerson et al. (1992) had female confederates hide in a swimming pool shower block and time participants’ showers.

Participants’ data must not be disclosed outside the research process, and participants must remain anonymous to all but the researcher — numbers instead of names, and published research giving no indication who they were.

Reliability is achieved if a study has been set up so that the IV can be seen to affect the DV, and replicating the procedure produces similar results.

That the measure is consistent and not liable to fluctuate to any significant degree.

Lab experiments; closed questions on a questionnaire; and observations which use a tally chart to record the frequency of specific behaviours.

The procedure can be standardised and controlled, so it can be replicated in exactly the same way for every participant, and it generates quantitative data that is straightforward to compare and analyse for consistency.

They involve less control over extraneous variables in the real-life setting, which makes exact replication harder.

They use a naturally occurring IV the researcher cannot control, which makes them harder to replicate exactly.

They use all or mainly closed questions generating quantitative data, use a standardised set of questions, and show consistency — people are unlikely to radically change their answers between sessions.

With closed questions the choice of answers is fixed, whereas with open questions a person might phrase their answer slightly differently each time, even if their underlying view hasn't changed.

The test-retest method measures external reliability: the same participants are given the same questionnaire at separate time intervals.

In the test-retest method, participants are given the same questionnaire again after a gap of around 6 months.

The split-half method measures internal reliability: the researcher splits the test in half and compares the responses given to the first half with those given to the second half.

The level of consistency between two or more trained observers when they conduct the same observation.

They compare two independent data sets, often a tally chart, and test the correlation between them. A strong positive correlation shows good inter-observer reliability.

False.

The observation is conducted separately by each observer, to avoid conformity — one observer may otherwise be influenced by the others.

They are difficult to repeat in exactly the same way — in an unstructured interview, for example, the interviewee talks freely about their thoughts and feelings, so the interview is unlikely to unfold the same way twice.

Although the questions are the same every time, they may still be delivered differently by different interviewers, whose own style can prompt different replies.

Qualitative data — such as interview transcripts or case study material — is much more open to interpretation than numerical data, so two different researchers might draw different conclusions from the same material.

Validity is the extent to which a piece of research can be said to be ‘true’ — a realistic measurement or explanation of behaviour within the context of the research.

Internal validity is whether we can be confident the IV genuinely caused the change in the DV, rather than some other, uncontrolled factor.

Ecological validity is how well the findings generalise beyond the study itself, to real-life behaviour outside the research setting.

False.

Lab experiments use artificial tasks in a controlled environment, which can lower ecological validity — the setting may not reflect real-life behaviour, and participants know they're being measured, which can make them behave less naturally.

By randomly allocating participants to conditions, which reduces the risk of participant variables acting as an extraneous variable and threatening internal validity.

Order effects can threaten internal validity — counterbalancing addresses this: half the participants complete condition A then B, and the other half B then A.

By matching participants on key variables across conditions, attempting to address both individual differences and order effects — though matching is never perfect.

Taking place in a more natural setting can improve ecological validity — but the increased extraneous variables make it harder to be certain the IV caused the change in the DV, weakening internal validity.

By reducing complex human behaviour to a single numerical score — such as a digit-span score in a memory experiment — which may give little insight into the wider behaviour being studied.

They tend not to be representative of the target population — researchers often use the easiest sample available, such as students at their own university, which limits generalisability.

It only uses participants available at the time of the study, so it will always exclude those who were not available to participate.

One of the most representative sampling methods is stratified sampling, which matches the target population proportionally.

They deal in descriptions rather than numerical data. Words add expression, explanation and meaning, whereas numbers simply show what happened, how many, or what percentage.

They allow data to be rich, in-depth and detailed — case studies sometimes use just one participant. The more a method matches real experience, the more valid it is — particularly in ecological validity.

They generate huge amounts of data which can be difficult to analyse and interpret, because it is so subjective — a particular tone of voice could mean bored, angry, upset, or something else.

Their expectations can influence how they interpret the data, which can reduce the validity of the conclusions drawn.