1) An experiment is a way of conducting research in a controlled way.
2) The aim is to control all relevant variables except for one key variable, which is altered to see what the effect is. This variable is the independent variable.
3) Lab experiments are conducted in an artificial setting, e.g. Milgram's study.
- Control - The effects of confounding variables (those that have an effect in addition to the variable of interest) are minimised.
- Replication - Strict controls mean you can run the study again to check the findings.
- Causal relationships - Ideally it's possible to establish whether one variable actuallyh causes change in another.
- Artificial - Experiments might not measure real-life behaviour (i.ie they may lack ecological validity)
- Demand characteristics - Participants may respond accordingly to what they think is being investigated, which can bias the results.
- Ethics - Deception is often used, making informed consent difficult.
In field experiments behaviour is measured in a natural environment. A key variable is altered so that its effect can be measured.
- Causal relationships - You can still establish causal relationships by manipulating the key variable and measuring its effect, although it's very difficult to do in a field experiment.
- Ecological validity - Field experiments are less artificial than those done in a laboratory, so they relate to real life better.
- Demand characteristics (participants trying to guess what the researcher expects from them and performing differently because of it) - these can be avoided if participants don't know they're in a study.
- Less control - Confounding variables may be more likely in a natural environment.
- Ethics - Participants who didn't agree to take part might experience distress and often can't be debriefed. Observation must respect privacy.
A natural experiment is a study that measures variables that aren't directly manipulated by the experimenter. For example, comparing behaviour in a single-sex school and a mixed school.
- Ethical - It's possible to study variables that it would be unethical to manipulate, e.g. you can compare a community that has TV with a community that doesn't to see which is more aggressive.
- Participant allocation - You can't randomly allocate participants to each condition, and so confounding variables (e.g. what area the participants live in) may affect results. The lack of control over variables makes it difficult to establish causal relationships.
- Rare events - Some groups of interest are hard to find, e.g. a community that doesn't have TV.
- Ethics - Deception is often used, making informed consent difficult. Also, confidentiality may be compromised if the community is identifiable.
Naturalistic observation involves observing subjects in their natural environment. Researchers take great care not to interfere in any way with the subjects they're studying.
- Ecological validity - Behaviour is natural and there are no demand characteristics, as the participant is unaware of being observed.
- Theory development - Can be a useful way of developing ideas about behaviour that could be tested in more controlled conditions later.
- Extraneous variables - Can't control variables that may affect behaviour.
- Observer bias - Observers' expectations may affect what they focus on and record. This means the reliability of the results may be a problem - another observer may have come up with very different results.
- Ethics - You should only conduct observations where people might expect to be observed by strangers. This limits the situations where you can do a naturalistic observation. Debriefing is difficult. Observation must respect privacy. Getting informed consent can be tricky.
Correlation means that two variables rise and fall together, or that one rises as the other falls - but not always that one variable causes a change in the other, e.g. as age increases so might intelligence, but ageing doesn't cause intelligence.
- Causal relationships - These can be ruled out if no correlation exists.
- Ethics - Can study variables that would be unethical to manipulate, e.g. is there a smoking relationship between the number of cigarettes smoked and incidences of ill health?
- Causal relationships - These cannot be assumed from a correlation, which may be caused by a third, unknown variable.
- Ethics - Misinterpretation can be an issue. Sometimes the media (and researchers) infer causality from a correlation.
A set of questions used to gather information.
- Practical - Can collect a large amount of information quickly and relatively cheaply.
- Bad questions - Leading questions (questions that suggest a desired answer) or unclear questions can be a problem.
- Biased samples - Some people are moire likely to respond to a questionnaire, which might make a sample unrepresentative.
- Self report - People sometimes want to present themselves in a good lifht (social desirability bias). What they say and what they actually think could be different, making any results unreliable.
- Ethics - Confidentiality can be a problem, especially around sensitive issues.
Structured interviews follow a fixed set of questions that are the same for all participants.
Unstructured interviews may have a set of discussion topics, but are less constrained about how the conversation goes.
- Rich data - Can get detailed information as there are fewer constraints than with a questionnaire. Unstructured interviews provide richer information than structured interviews.
- Pilot study - Interviews are a more useful way to get information before a study.
- Self report - Can be unreliable and affected by social desirability bias.
- Impractical - Conducting interviews can be time-consuming and requires skilled researchers.
- Ethics - Confidentiality can be a problem, especially around sensitive issues
Case studies allow researchers to analyse unusual cases in a lot of detail, e.g. Milner et al's study of HM.
- Rich data - Researchers have the opportunity to study rare phenomena in a lot of detail.
- Unique cases - Can challenge existing ideas and theories, and suggest ideas for future research.
- Causal relationships - The researcher has very little control over variables.
- Generalisation - Only using a single case makes generalising the results extremely difficult.
- Ethics - Informed consent can be difficult to obtain if the subject has a rare disorder.
Although the aim states the purpose of the study, it isn't usually precise enough to test. What is needed are clear statements of what's actually being tested- the hypotheses.
1) Research Hypothesis - Proposed at the beginning of a piece of research and is often generated from a theory, e.g. Bowlby's research hypothesis was that maternal deprivation causes delinquency.
2) Null Hypothesis - What you're going to assume is true during the study. Any data either supports or opposes the assumption. If the data doesn't support this hypothesis, an alternative hypothesis is suggested instead. Very often, the null hypothesis is a prediction that there will be no relationship between key variables in a study- and any correlation is due to chance.
3) Experimental Hypothesis (or Alternative Hypothesis) - If the data forces you to reject your null hypothesis, you accept your experimental hypothesis instead. So if the null hypothesis was that two variables aren't linked, the alternative hypothesis would be that they are linked.
4) Directional Hypothesis - A hypothesis might predict a difference between the exam results obtained by two groups of students. If it states which group will do better, it is making a directional prediction.
5) Non-Directional Hypothesis - A non-directional hypothesis would predict a difference, but wouldn't say which group would do better.
Independent Groups Design
An independent groups design means there are different participants in each group. Here, for example, one group does the task with an audience and another group does it alone. This avoids the problem that if all participants did the test in both conditions, any improvement in performance might be due to them having two goes at the task (which would be a confounding variable).
- No order effects - No one gets better through practice (learning effect) or gets worse through being bored or tired (fatigue effect).
- Participant variables - Differences between the people in each group might affect the results (e.g. the 'without audience' group may just have people who are better at the task- so the groups can't be safely compared).
- Number of participants - Twice as many participants are needed to get the same amount of data, compared to having everyone do both conditions.
Repeated Measures Design
A repeated measures design is where, e.g., all participants do the task both with an audience and then without. You can compare the performances in each condition, knowing the differences weren't due to participant variables.
- Participant variables - Now the same people do the test in both conditions, so any differences between individuals shouldn't affect the results.
- Number of participants - Fewer participants are needed to get the same amount of data.
- Order effects - If all participants did the 'with audience' condition first, any improvements in the second condition could be due to practice, not the audience's absence.
Matched Pairs Design
A matched pairs design means there are different participants in each condition, but they're matched on important variables (like age, sex and personality).
Some studies use control groups. These groups have not experienced any of the manipulations of the IV an experimental group might have. This allows the researcher to make a direct comparison between them. In the example above the group that didn't have an audience would be the control group.
- No order effects - There are different people in each condition.
- Participant variables - Important differences are minimised through matching.
- Number of participants - Need twice as many people compared to repeated measures.
- Practicalities - Time-consuming and difficult to find participants who match.
A sample is a selection of participants from a target group. It should be representative and should not be biased.
When every member of the target group has an equal chance of being selected for the sample. This sampling method is fair and likely to be representative. However, this method does not guarantee a representative sample.
When researchers sample whoever is available and willing to be studied. This is a quick and practical way of getting a sample. But the sample is unlikely to be representative of a target group or population as a whole. This means the findings of the research can't be confidently generalised.
When people actively volunteer to be in a study by responding to a request for participants advertised by the researcher. The researcher may then select only those who are suitable for the study. This allows a large number of people to respond, giving more participants to study. This may allow more in-depth analysis and more accurate statistical results. Even though a large amount of people may respond, it will only be people who actually saw the advert. Volunteers may be more cooperative than others. This means the sample is unlikely to be representative of the target population.