Experimental Methods- Laboratory
Definition- A experiemnt carried out in a laboratory with high levels of control over extraneous variables.
- Highest level of control over variables
- Use sophiticated technology and equipment
- easily repeated
- Artificial- unrealistic setting reduces ecological validity
- Demand Characteristics- participants can work out hypothesis and change behaviour accordingly.
Experimental Methods- Field
Definition: An experiement carried out in a real-world setting e.g. a classroom
- more ecological validity as the setting is more realistic
- less demand characteristics that laboratory experiment
- less control over variables
- still difficult to generalise to other situations
Experimental Methods- Natural
Definition: A quasi-experiment where the allocation of participants to conditons happens randomly (naturally occuring independent variable). The reasearcher observes differences.
- high ecological validity
- no effect on participants as very little interaction
- no demand characteristics
- observes rare behaviour and statistics
- no control over variables
- difficult to repeat
- behaviour can be so rare it doesn't occur.
Non experimental methods- interviews
- detailed, rich information gathered
- relationship with researcher developed, so more information revealed
- time consuming
- misinterpation of data collected
- social desirability bias
- interpersonal variables
Non experimental methods- questionnaires
- easy and quick to create, dsitribute and fill out.
- self explanatory
- easily quantitative
- questions can be misenterpreted
- social desirability bias
- less people respond than asked
Non experimental methods- Observation
- high ecological validity
- preliminary tool-identiifies hypothesis for furture experiments
- researcher effects- the presence of the researcher can cause behvaiour changes
- virtually unrepeatable
- coding systems are time consuming and restrictive
- poor control
Non experimental methods- case studies
- rich, personal data
- studies rare occuring behaviour
- ecological validity
- challenge existing theories
- low population validity, as only one/ a small group of people
- extremely low contol over variables
- researcher- participant relationship can leed to subjectivity and bias
Non experimental methods- correlational analysis
- investigates relationship between two variables
- cant assume causality, a there could be a missing factor that links the two variables
- doesnt work for non-linear relationships
- reduces detail of data to numbers
When a participant is not told to true aims of the study or is deliberately misled in some way.
- Prevents full informed consent, as they don't know true aims of the study. Honesty is an important ethical principle.
- Deception is sometimes necessary to avoid demand characteristics. Most deception is harmless.
Problems and solutions:
- Deception must be approved by ethical commitee.
- Particpants need to be fully debriefed- HOWEVER, this can't turn back time, and participants may feel embarrassed.
ETHICS- Right to withdraw
Participants should be fuly aware and able to leave the experiment at any point, without being questione or pressured to stay. All their data in the experiment must be destroyed and not used.
- participants sometimes feel pressured to remain in the experiment, especially if they've been paid. They should know they wont be forced to stay. This is particularly important if the participant has been decieved.
- loss of participants may lead to bias in results, and also a waste of time and money.
Dealing with issue: Participants must know they can leave at any time.
weaknesses: participants feel pressured to stay e.g. its apart of their undergraduate psychology course.
ETHICS- Informed Consent
Paricipants should know full details of experiment before agreeing to take part.
- They want to know what they are getting themselves in to.
- informed consent could lead to demand charateristics and make the experiment meaningless.
Dealing with the issue:
- Participants should be formally asked for agreement to take part in experiment.
- Presumed consent- asking people of similar backgrounds if they would be offended by the experiment procedures.
Participants have a right to control information about themselves. It is different to confidentiality, which should be respected for the participants privacy.
- Expect to have control over their own data.
- protecting privacy can be difficult, especially in observational studies. They do not want to alert participants to the fact they are being studied.
Dealing with issue
- do not observe anyone without their consent, unless in a public place. Participants may be asked to give retrospective consent.
- no universal agreement on what a 'public place' is.
ETHICS-Protection from harm
Harm includes any negative effects on participants including physical and psychological e.g. embarrassment.
- dont want to be harmed!
- It may not be possible to measure and estimate all detrimental effects of a study.
Dealing with issue:
- Participants should not be exposed to any more harm or risks than usual mundane life.
- Its hard to predict the effect of studies on individuals.
ETHICS- Dealing with issues
Ethical guidelines- These are issued by the BPS, and provide a guidelines for every experiment.
Ethical commitees- Every experiment must have their procedure approved by an ethical commitee before it goes ahead.
cost-benefit approach- Ethical commitees often look to see if the advantages of doing the experiment outweigh the disadvantages; is it worth putting a few participants through stress for groundbreaking results?
debriefing-Participants should be fully debriefed after every experiment, and this help counteract irritation from lack of informed consent or deception. However, it can't turn back time and the participants could be harmed.
presumptive consent- getting a group of people of similar demographic to particpants to decide if they would be offended/distressed by the experiemental procedures.
Sampling Techniques- Random
Every person or item in the target popukation has an equal chance of getting chosen to participate.
- high population validity
- less bias
- completely random could lead to a accidental unfair distirubution of characteristics - BIAS.
- Unlikely to get the target population for your experiment.
Selecting anyone who is available to take part in the study from a given population.
- quick and easy
- large sampling size
- biased- only certain people will be in that place at that time, and lifestyle choices that lead them there could also influence the studies results.
- therefore not representative of the whole population
Sampling Techniques- Volunteer
No random process- the experiment is advertised and volunteers offer themselves as participants.
- less time consuming- you dont have to go out and find participants
- very biased- volunteers will have key characteristics and values that could effect experiements results
- unlikely to reach target population
investigator effects- these are differences between conditions or within conditions caused by the investigator themeselves. It could hinting at a answer, changing the phrasing of a question, or simply being there, observing. These could all effect participants behaviour, causing demand characteristics, making the experiment internally invalid.
demand characteristics- these occur when the participants deduce the aim/hypothesis of the experiment, and then alter their behaviour to either aid or ruin it.
social desirability bias- This is when participants alter their behaviour/answers so as to appear more desirable/acceptable in the researchers eyes. Questionnaires are particularly susceptible to this, and participants sometimes don't realise they are doing it!
extraneous variables- these are variables that affect the dependent variable, other than the independent variable. Natural experiments are particularly susceptible to it. They stop us from assuming a cause-and-effect relationship.
The study uses a small amount of participants, or participants of a similar demographic (e.g. all students, all over 80 years old.) This wouldn't be a fair representation of the world's population, so we cannot generalise the findings.
This is when the procedure, settings or resources used in the experiment are artificial or unrealistic. Laboratory experiments are susceptible to this. If the participants are not acting as they would in a real life situation, this means the experiement is ecologically invalid.
Predicts the score of each condition e.g.
'x will score higher than y'
Only predicts there will be a difference between the two conditions e.g
'there will be a difference in scores between x and y'
This is making your hypothesis specific to the experiment e.g
'students who revised 12 hours a week will score better in eventual GCSE tests than students who did not revise at all'
rather than 'students who revised will do better in tests'
Experimental Design- Independent groups
Different participants are used in each condition.
- order effects are impossible
-demand characteristics reduced, as participants aren't exposed to both conditions
-individual differences can lead to bias in groups (e.g. all men in one group, all women in the other) and extraneous variables.
-uneconomic use of participants; double the amount of people needed to collect same amout of data using repeated measures design.
Experimental Design- Repeated Measures
Same participants are used in each group.
-very efficient use of participants.
-No individual differences.
-Bigger chance of demand characteristics as participants are exposed to all conditions.
-Order effects: These are when participants improve in tests due to practice from condition to condition. This causes bias in results.
-Restricted in use, as it is not suitable for some experiments e.g. different teaching methods for kids
Experiemental design- Matched pairs
different participants in each group, but they are matched on various characteristics e.g. gender, income, race
-Combines advantgaes from both other designs: less individual differences, no order effects.
-Time consuming to match people
-Choice of participants may not allow you to match them efficiently. e.g. only one 70 year old man.
-Might not be matched on the right characteristic, and another feature could cause individual differences.
Dealing with problems in experimental design
Individual Differences- solved by RANDOM ALLOCATION
Participants are put into conditions completely randomly, so individual differences do not cause bias. This avoids any concious or subconcious bias in the allocation of participants e.g. an researcher putting all academic people in the 'did revise' condition so as to helped the results show his hypothesis. With completely random allocation you are not likely to find a bias or link between participants within the condition.
Order Effects- solved by COUNTER BALANCING
Equal numbers of participants experiencing each condition in different orders e.g. 50% of participants experience condition A first, and the other 50% experience condition B first. This diminishes order effects as each condition is subject to equal amounts of fatigue or practice.
A pilot study is a small scale trial study that is carried out before the main experiment.
Reasons to do pilot studies:
- To identify extraneous variables
- To test if participants guess the aims of the study (demand characteristics)
- To check that participants correctly understand or interpret questions
- To check that procedures are accurate.
- To check if participants are harmed physically or psychologically.
- Check for a significant result; experiments are costly and time consuming, its best to know if you will find anything intresting before starting.
Used to show scores on a continuous scale e.g. IQ
Used to shop frequency or scores in seperate catergories e.g. eye colour
The bars are not joined together because they represent separate catergories.
Used to show the strength and direction of correllations.
Data Analysis: Central Tendency- Mean
Add up all scores and divide by how many there are.
- uses all scores
- May not be a number represented in the data
- Affected by outliers
Data Analysis: Central Tendency- Mode
The most commonly occuring score in the data set.
- Always a number in the data set.
- Not affected by outliers.
- Doesn't use all the data
- May be more than one
Data Analysis: Central Tendency- Median
The middle number of the set when its in numerical order. If it is a even-numbered set, the two middle numbers are added and divided by two.
Not affected by outliers
Could be more than one.
Doesn't represent the whole data set.
Data Analysis: Dispersion
The largest value in the data set, minus the smallest. This find the difference between the two. It shows us how representative the median is. The larger the range, the less representative the median is, because values have a wide spread.
This tells us the average distance of EACH score from the MEAN. A high standard deviation tells us the mean is not representative, because, on average, each score is very far away from it. A low standard deviation shows the mean is representative.
- Quick to make and distribute
- Doesn't require skill to fill in
- Easy to keep anonymous, providing more truthful data
- Susceptible to demand characteristics and social desirability bias, as participants can see links between questions and might want to impress the researcher. They might not even realise they are doing it!
- Easy to misunderstand or misinterpretate. This would lead to invalid results.
- If distributed freely, rarely everyone will return/fill out the questionnaire.
Interviews: Structured v unstructured
- quicker to complete and train interviewers, more quantitative date (less time to analyse).
- less freedom so participants can't express their full ideas.
- more rich detailed data
- A relationship can develop between interviwer and interviewee leading to more honest data.
- difficult/expensive to analyse, more time consuming,
- interviewers have to be trained more.
- more personal relationship, so more truthful relationship
- less room for misinterpretation
- Time consuming to complete and then analyse
- Less privacy- stops full detailed data
Good observations should have:
- a clear aim
- Behavioural catergories
- Trained observers
- standardised timing
- video recording
- Objective; the participants themselves are not recording the data
- Flexibility and diversity; observation can vary in structure and application.
- extremely time consuming and expensive
- researcher effects- if people know they are being observed they may act differently
- ethics- if you dont want participants to know they are being observed thsi could infringe on their privacy and ethics.
Correlations investigate the strength and direction of an relationship between two variables. It is expressed using a co-efficient.
- Very scientific method.
- Shows relationship
- Useful for indicating whether further research would be useful
- Doesn't imply causality/cause and effect.
- Doesn't work for non linear relationships.
This is used to convert qualitative data into quantitative data.
- Draw up coding/checking system. These must be clear.
- Select material to code.
- Code the data into catergories/units. You can get more than one person to do this to help consistency and validity.
- Record frequency of each catergory.
- Put into a table/chart.
- Draw conclusions.
Allows us to use rich, detailed qualitative data in a scientific way.
Very time consuming