Types of Experiments
Laboratory: Conducted in a highly controlled environment. The IV is manipulated to see the impact on the DV, whilst the effects of other variables are minimised as far as possible. For example, giving researchers lists of words to remember, giving them another task to prevent rehearsal, then testing their recall of the information.
- Advantages: extraneous variables are closely controlled, meaning the IV is likely to have affected the DV, increasing the internal validity of the study. Research can be easily repeated as there will be a controlled, standardised procedure, increasing the reliability of the results
- Disadvantages: artificial nature of the set-up means that the results may not reflect ‘real-life’ behaviour, so reducing the external validity of the study. Participants know they are being tested so may change their behaviour (demand characteristics). Tasks given in the research may not be reflective of everyday tasks (lack of mundane realism).
Field: The experimenter manipulates an IV in a more natural setting. For example, Piliavin (1969) got a confederate to collapse on a train when smelling of alcohol or carrying a walking stick, and seeing how many people helped in each condition.
- Advantages: higher mundane realism than lab experiments, therefore higher external validity. Often participants won’t know they are being studied, so demand characteristics are less of an issue.
- Disadvantages: harder to control extraneous variables, so harder to know if the IV has affected the DV. If participants are unaware they are being studied this raises ethical issues (lack of informed consent).
Natural: The experimenter studies the effects of a naturally occurring IV. Participants may still be studied in a lab-type setting to see the effects, but the IV is not manipulated by the researcher. For example, Williams (1986) looked at the effects on gender attitudes after the introduction of TV to a small town in Canada.
- Advantages: high external validity, as the IV is naturally occurring. The effects can be tested of factors that could not be manipulated by the researcher (e.g., the effects of lack of attachment in Romanian orphans).
- Disadvantages: even less control over extraneous variable than field experiments. Participants can’t be randomly allocated to conditions, introducing the possibility of bias. Naturally occurring IVs may be rare, so studies can’t be repeated.
Quasi: The IV is based on an existing difference between people. For example, gender differences in attitudes towards food.
- Advantages: can be tested under controlled conditions (as in the example above), increasing the scientific credibility of the research.
- Disadvantages: participants can’t be randomly allocated to conditions, introducing possible confounding variables.
Observations involve watching and recording people’s behaviour in a natural setting. Observations by themselves are non-experimental, but observational techniques can be used as part of an experiment. Types include:
Naturalistic and controlled: Naturalistic observations take place within a natural, non-manipulated environment, for example in a workplace or school. Controlled observations are more manipulated, for example the Strange Situation, so that variables are more controlled and effects of particular situations can be seen.
- Evaluation: Naturalistic - high in external validity (as they are very true-to life), but lower levels of control. Controlled- lower external validity, but more control (allowing for easier replication).
Covert and overt: Covert observations take place without the participants being aware that they are being watched. Overt observations are when the participant does know they are being watched, and have given prior consent to do so.
- Evaluation: Covert- no participant reactivity, so more truthful behaviour is shown, but there are ethical issues (lack of consent). Overt- participants may change their behaviour, but they are more ethically sound.
Participant and non-participant: Participant observations are when the researcher themselves takes part, for example by joining the workforce in a workplace. Non-participant observations are when the researcher does not actually participate, but just observes.
- Evaluation: Participant- the researcher gets a greater insight into the experiences of those being observed, but they may lose objectivity as they become part of the study, friendly with other participants, and so on. Non-participant- the researcher is more likely to remain objective, but may lack the extra insight gained from being a participant themselves.
Self-report techniques involve asking people about their behaviour.
These are sets of questions which participants complete independently, for example on their attitudes towards something or beliefs about something. Questionnaires can be used as part of an experiment (e.g. measuring locus of control through a questionnaire, and then testing the participants in some way). Questions can be open, in which the participant can answer in any way they wish- for example, ‘why do you think people follow orders?’. This produces qualitative data, which is rich in detail. Alternatively, questions can be closed, in which there are a set of answers participants must choose from- for example ‘do you think that people follow orders because of (a) the situation they are in, or (b) their personality?’. This produces quantitative (numerical) data which can be easily counted.
- Questionnaires can be sent to (potentially) thousands of people, without the researcher needing to be present whilst they are completed. Potentially there is access to a very large sample.
- The responses will usually be easy to analyse, especially if the questions are closed.
- The responses may be biased:
- Social desirability bias- not being truthful to try to present yourself in a better light (underestimating the amount of alcohol you drink)
- Response bias- answering all questions in a similar way and not reading the questions properly (ticking ‘yes’ for everything)
- Acquiescence bias- a tendency to agree with things, meaning that the questionnaire is measuring a tendency to agree rather than what it is intending to measure.
These are face-to-face interactions between the researcher and participant. They can be structured, where the interviewer asks a set of pre-determined questions and doesn’t deviate from them; unstructured, where the interviewer creates questions in response to the participant’s answers during the interview; or semi-structured, where there are some pre-set questions but also the opportunity to ask extra questions as well.
- Structured interviews can easily be repeated and the data are more easy to analyse, but they are inflexible and can’t include additional information.
- Unstructured interviews are difficult to repeat and hard to analyse for trends and patterns, but allow more flexibility to investigate answers in more depth.
- As in questionnaires, participants may not be honest with their answers, reducing the validity of the responses. The interviewer may be able to get more of a sense of how truthful the participant is being than in a questionnaire, however.
A correlation measures the relationship between two variables, and the strength of that relationship. They are plotted on a graph known as a scattergram/scattergraph. Correlations can be positive, meaning that as one variable increases, so does the other one, for example as the number of people present increases, so does the rate of obedience. Negative means that one variable increases whilst the other decreases, for example as the level of anxiety increases, the accuracy of eyewitness testimony decreases. No correlation means there is no link between the two variables.
Correlations differ from experiments. In an experiment, the researcher manipulates an IV, attempts to control all other variables, and records the effect on the DV. In a correlation, there is no manipulation of either variable- they are known as co-variables. Therefore, it cannot be concluded that one co-variable has caused the other to change, just that there is a correlation between them. Another variable might actually have caused the change. In an experiment, causal links can be established, as the impact of other variables are minimised.
- Correlations can be used as a starting point for further research. If no relationship is found, then there is no need to undertake experiments into the subject.
- They are quite easy to conduct- the researcher just needs to find two sets of data to compare.
- Demonstrating a cause-effect link is not possible, as other variables might be involved.
- The findings of correlational studies are often reported as facts, leading to misinterpretations, which could have consequences (for example, black people are more likely to be convicted of knife crimes than non-black people, but this does not mean that ethnicity causes knife crime- black people are more likely to come from poorer socio-economic backgrounds so are more likely to turn to knife crime due to this).
This is a type of observational research in which something that has been produced (such as newspaper articles and television adverts) is studied. The aim is to analyse the communication in order to detect trends and make conclusions.
- Content analysis is high in external validity, as what is being analysed is the material that people consume in ‘real life’. It allows for the investigation of potentially sensitive topics, without the need for consent, as the material is in the public domain.
- Content analyses may not take into account the motivations of the people that created the content in the first place, potentially weakening the validity of conclusions that can be drawn.
A case study is an in-depth investigation, usually of one person or a small group of people. Individuals studied may have particular conditions or unusual characteristics. Case studies often produce qualitative data, as unstructured interviews and observations may be used. The participant may be asked to complete laboratory experiments, which would be more likely to produce quantitative data. Often case studies are longitudinal, as the participant is studied over many years. Examples of case studies include HM, whose memory was damaged following an operation.
- Case studies produce rich, detailed, in-depth data, giving a close insight into particular behaviours. They can also help understanding of the behaviours of ‘normal’ individuals, for instance HM’s case showed that there are separate stores for long-term and short-term memory.
- It can be difficult to generalise the findings from case studies, as the sample sizes are so small. Data is often collected retrospectively, so relies on the recall of the participant or their friends or family members, which may be inaccurate. Comparing to a control group is often not possible, and the researcher may become personally involved with the participant over a number of years, making them less objective.