Types of experiments
Laboratory experiments - Independent variable is manipulated by the researcher in a CONTROLLED ENVIRONMENT (lab). The prodcedure can be kept the same for each p as it's controlled so less likely to be influenced by variables such as interruptions,lighting,temp change etc (extraneous variables). When the procedure and setting are kept the same for everyone it's called STANDARDISATION. Standardisation - referred to as a control (EV under control so they don't become confounding variables i.e variables that contaminate the results). Standardisation allows researchers to replicate investigations to check for consistent results
- Can easily replicate procedures and have high control over variables
- Can generalise results.
- Yields quantitative data.
- Can use technical equipment.
- Experimenter effects
- Low ecological validity - artificial environment so may not be able to generalise to real life
- Demand characteristics
- Ethical issues such as informed consent
- Sampling bias
Types of experiments
Field experiments -
- Conducted in the NATURAL ENVIRONMENT of p's in which the IV IS STILL MANIPULATED BY THE EXPERIMENTER.
- Unlike the lab expt, it's procedures are harder (even impossible) to standardise because the environment is uncontrolled but the setting is far more reflective of everyday life. (so high ecological validity)
- High ecological validity - not an artificial environment
- Reduction of demand characteristics
- Difficult to establish high levels of control
- Cannot generalise to other situations
- May be time consuming and costly
- Harder to replicate
Types of experiments
Natural experiments (aka quasi experiments) - May take place in labs or everyday settings but the IV is NATURALLY OCCURING and NOT manipulated by the experimenter. Because of this it is sometimes argued that they are not infact 'true' expts (hence the 'quasi' meaning mock) Unlike lab expts p's CANNOT be randomly allocated to either condition.
- Reduction of demand characteristics
- Lack of direct intervention by experimenter
- Less chance of experimenter bias
- More ecologically valid as IV is naturally occuring
- IV not controlled
- No control over allocations of p's into groups.
- Harder to replicate as conditions won't ever be the same for p's
- Harder to control EV's
ALL share one important advantage - THEY ALLOW THE RESEARCHER TO ESTABLISH A CAUSE-EFFECT RELATIONSHIP BETWEEN THE IV AND THE DV.
Because experiments involve using isolating variables and looking at how they may affect each other (i.e. how the IV affects the DV). The experiment is the only method that allows this to happen.
A form of analysis applied to sets of data to establish the direction and strength of relationships between co-variables. The data used has to be QUANTITATIVE (numerical) and capable of being placed in rank order.
- POSITIVE CORRELATION = As one co-variable increases/decreases, so does the other so they change in the SAME direction
- NEGATIVE CORRELATION = As one co-variable increases the other decreases so they change in DIFFERENT directions
- ZERO CORRELATION = Is when NO relationship is found between co-variables
To assess more accurately the extent to which a relationship between the co-variable exists, a statistical analysis is performed to generate a CORRELATION COEFFICIENT. This is a numberical measure of the correlation and measurements range from:
- +1 (a perfect positive correlation)
- to -1 (a perfect negative correlation).
- The mid-point 0 denotes no relationship between co-variables.
- Avoids some of the practical and ethical problems raised by other methods as there is usually no direct involvement with or manipulation of participants
- Can reveal the direction and strength of a relationship between co-variables
- Can be used on data from ecologically valid sources e.g. crime rates
- Can not reveal cause and effect relationships - 2 co-variables may correlate but may not be casually linked. (Correlation is not causation)
- Uses only quantitative data which may be reductionist (in other words knowing what a person scores on a variable does not tell us other info that may be helpful such as why they scored as they did)
- Can only be performed on more detailed quantitative data (not data that is nominal i.e. that can be categorised)
Observation may be used as an overall method and usually refers to the kind of research that has no rigid control or manipulation but allows the researcher to access more natural, freely occuring behaviour. Observations of this kind tend to be field-based (naturalistic observations). They have a huge advantage of yielding data that's high in ecological validity.
Observations may be chosen as an appropriate method if controlled experimentation would resilt in behaviour being too artificial. It may be that the variable being observed would be impossible or unethical to manipulate.
Observation as a technique tends to refer to the way in which data is gathered within another method - e.g. observation is often used within a case study method. Within an experiment the researcher may observe p's to establish whether the IV is havin an effect. E.G. Harlow (the monkeys with wire mothers)
When being used as a technique, observations may take place in the field or a lab. Although advantageous in their ability to allow for control and good access to p's, lab based observations often lead to p's feeling dehumanised and reacting to the artificial situtation in which they find themselves. P's may behave in different ways when in a lab an it doesn't reflect real life.
Sometimes participants know they're being observed (this is called disclosed/overt observation) and sometimes they don't (undisclosed/covert observation) disclosed=ethical but affects behaviour
In observations researchers are sometimes part of the research set up, in effect they are 'observing from within' which is referred to as a PARTICIPANT OBSERVATION.
OBSERVER BIAS occurs when observers record data in a certain (subjective) way which may involve them projecting what they expect to find onto whatever they are observing. (Basically interpret data the way they expect) It may be more likely in a participant observation as researchers are more likely to become personally involved. Also it may be hard to check the reliability of the data gathered in such an observation because the researcher is often working alone and has no other researcher with whon they can check the consistency of his records.
Alternatively, researchers may be observing without being personally involved but as outsiders - this is non-participant observation. Although, the researcher is less likely to be biased in their observations, this is because they tend not to become as close to participants and this may limit their access to data.
Structured v unstructured observations
Some observations involve the researcher gathering data in a ad hoc manner by spontaneously noting behaviours/events they regard as important as they happen (Rather than having a pre planned list of specific behaviours that are expected) Data is therefore more likely to be qualitative/descriptive (in contrast to it being quantitative/numerical)
Some observations are STRUCTURED in the way that they gather data. Often this is done by using BEHAVIOUR CATEGORIES or other means to consistently record data. E.g. of behaviour categories to observe aggressive play in children: hitting, punching, breaking toys etc.
Rating behaviours on scales or coding behaviours accoding to predefined principles are other ways that data can be gathered in a structured way in an observation.
It's essential that observers are familiar with and practised at using whatever structure has been agreed upon to gather data. This ensures good agreement between observers known as INTER-OBSERVER RELIABILITY.
If gathering data was UNSTRUCTURED the researchers may simply try record what p's actually do as it occurs so instead of waiting for certain things they record anything that appears aggressive
Strengths of observations
- Allow researchers to study what people actually DO, not what they say they do (which may or may not be valid)
- Naturalistic observations are more likely to be ecologically valid
- Lab observations allows easy access to p's as well as the use of equipment with which to record data
- Unstructured observations yield descriptive data that is not reductionist
- When structured the behaviour is recorded in a consistent manner so the data is reliable and as the process is standardised it could be replicated so findings could be checked to see if future findings are consistent (if they are it's externally reliable)
- If observers are trained and practised in the use of whatever structured system is being used there should be high inter-observer reliability (consistency between observers' records)
Weaknesses of observations
- Observations don't allow us to establish a cause-effect relationship between variables when used as a method of investigation
- Researchers may misinterpret what they see which leads to invalid and unreliavle data
- In structured observations the behaviours that researchers have set out to observe might not occur. Also, behaviours that researchers had not expected to observe might occurs but they may have no consistent or structured way of recording these so data is 'lost'
- Establishing high inter-observer reliability takes time and may never be perfect
- Structures limit out full appreciation of the richness and meaning of human behaviour in a social context, yet in unstructured observations recording data may be far more difficult e.g. on a practical level (writing it all down)
- Analysis of descriptive data from unstructured observations may be difficult and time consuming
- Unstructured observations there's lower chances of inter-observer reliability when comparing records because observers are not following a pre-determined checklist.
- Because the researcher may only be observing participants they may know what is happening (and how frequent) but not why it's happening
Designing of naturalistic observations
Needs to be planned systematically to ensure data gathered is valid & reliable as possible. Especially the case when the setting is uncontrolled (naturalistic). The findings will be more useful and valuable. The steps an investigator would use are: 1) Formulate a hypothesis based upon an aim & a 'hunch' about what's expected
- 2) Plan WHERE to observe p's (field v lab) thinking carefully about the PRACTICAL & ETHICAL REQUIREMENTS of the study - when the observation is naturalistic particular care must be given to the ethics in terms of whether the location is private/public, consent & debrief. Thought should be given to practical aspects of the observation-e.g. whether 1 venue is enough an researchers ease of access regarding the venue and participants.
- 3) Plan WHAT to observe and crucially(in structured observations) which behaviours constitute what is being observed and how these will be coded - this involves operationalising the necessary variable(s) by developing BEHAVIOURAL CATEGORIES. These involve investigators deciding which behaviours are to be recordede and are incorporated into a coding system in the form of a tally chart. This should enable observers to record data easily & in a way that is valid an reliable. Behavioural categories must be OBJECTIVE+UNAMBIGUOUS.
- Objective - Recording FACTS not opinion.
- Unambiguous-Clear. Have a single clearly defined meaning-leading to only 1 conclusion.
If such categories aren't used & records are invalid due to OBSERVER BIAS they will not be reliable. Observer bias is when an observer interprets&records behaviour in a certain way because they allow their expectations or own biases to influence what they see.
Observers can't observe & record everuthing, so they sample behaviour (refers to behaviour being observed not the sample of p's)
TIME SAMPLE of behaviour = Observing a participant for a certain predetermined period. E.g. record what a child's doing during the first 10 seconds of every 3 minutes for an hour.
POINT SAMPLE = Recording one p's behaviour before moving straight to the next. E.g. observing in a playground of children & recording what each child does before targeting the next child.
EVENT SAMPLE = Record only examples of the behavioural event each time it occurs. E.g. if the behaviour being investigated was 'fighting'-only fights would be recorded. What caused a 'fight' would need to be operationalised & what aspects of the event were to be recorded would have to be predetermined. Features such as how many kids were involved,gender,time of day etc may be worthy of note.
Prior to observatioon being conducted the researchers must decide upon how to sample behaviour.
The observation using the proposed observers and apparatus - this would involve observers gathering some initial data, checking their coding system is easy to use and then compaing their records to check for inter-observer reliability. This is done by correlating observers' results-the strength of the positive correlation reflects the degree of inter-observer reliability
Any problems or issues concerning the procedure are addressed and in particular any relating to the behavioural categories/coding and recording of behaviour. If a behavioural category proves to be ambiguous (and has been interpreted differently by different observers) this is amended accordingly until strong reliability is found (continued piloting of these behavioural categories may be necessary)
Conduct the 'observation proper' debriefing p's at the end if possible
Analyse the data and relate this to the original hypothesis. Formulate a conclusion.
Self report techniques(method/data collection)
Self report technique allows the p to provide info about themselves in the form on an INTERVIEW or QUESTIONNAIRE. Also maybe a DIARY-e.g. sleep diary. There are diff types of questions that can be asked:
OPEN ENDED Q'S = allow p's to respond freely & in depth if necessary. These lead to qualitative data. E.g. 'what can you remember from the robbery you witnessed?'
CLOSED (FIXED) Q'S = restrict the responses that p's can give & result in quantitative data being gathered. E.g. 'the burglar was carrying a weapon' - agree,unsure,disagree.
Interviews-form of self report
- May be face to face or conducted electronically (over the phone or as a 'conference interview' online) depending upon what is most convenient & appropriate.
- Format can be STRUCTURED or UNSTRUCTURED.
Structured interview involves:
- Researcher asking p's a predetermined set of questions (as in a survery);
Semi structured interview involves:
- When p's are asked some predetermined set of questions with some scope for the researcher to add spontaneous questions or allow p's to elaborate when answering.
Unstructured interview (aka non-directive interview) involves:
- Researcher supporting the p as they talk about what they need to - this type of interview is most often used therapeutically as a way to help a client rather than gather data.
Questionnaires - form of self report
- Is a set of pre-written questions and tends to be concerned with people's behaviour, attitudes and opinions.
- May be administered on paper of electronically and involves p's being asked one or several types of questions.
- Allow researcher to clarify q's if they are ambiguous (unclear) because they're asking the q directly of the p and can respond to their queries.
- Allow a relationship to build between the p and researcher especially when face to face which may in turn encourage p's to be more open and honest.
- Semi-structured interviews allow the researcher to digress or probe further if needed because of their less structured format.
- The p may be more reluctant to open up or tell the truth because they are answering to an individual-this may be the case especially when face to face with the researcher. The p may feel awkward and embarassed if the subject matter is sensitive.
- Risk of investigator effects is increases - it may not be simply what the p is questioned about that reflects the researchers beliefs about what is important but also how they ask the questions in terms of the words they use and non-verbal cues (tone of voice, eye contact etc)
- Takes longer to gather the data because p's are questioned individually by the researcher.
- Take less time to gather data because a group of p's can be questioned simultaneously.
- P's may be more honest when subject matter is sensitive because they're able to answer anonymously and can avoid having to 'look another person in the eye' as they answer.
- Researchers don't need to be present throughout so p's may complete the questionnaire when and where it is convenient for them.
- Easy to standardise and replicate.
- Because the researcher is often absent when p's respond, any ambiguous(unclear) q's cannot be clarified and unexpected/interesting responses cannot be expanded upon simultaneously.
- May appear less personal so it's more difficult to establish trust & warmth between participant and the researcher.
- The quantitative data generated by fixed choice graded response questions provides a somewhat false impression of precision - the values allocated to a particular responses when q's are being designed are subjective only.
Validity and self reports
An important issue with self reports is whether p's respond in a valid way-if they do not data is invalidated. Things that can affect internal validity of responses p's given during self reporting&solutions:
- Social desirability effect (when p's answer dishonestly to avoid creating a bad impression): Anonymous questionnaire, reassure them that it's 100% confidential and so is their identity.
- Ambiguous questions (P's may not fully understand the q,more likely with questionnaires when they have less chance to seek clarification): Pilot studies. Have someone check the questionnaires over.
- Response bias (More likely if questionnaire is lengthy+same question types used throughout):Shorter q's but more detaild. Give breaks. Different ways of gathering info.
- Forced choice format (P's may not be able to express how they really feel or what they really do because the option is unavailable due to narrow choice of options provided):Multiple choice q's.
- Demand characteristics/screw-you effect (Depending on how the q's are phrased the p's may be able to work out what is ecpected of them in terms of their response: Use distractor q's. Distractor task.
- Researcher bias (researchers get/find the results they'd expect to find. Which may be in the form of the range&nature of q's or how they're phrased-could be leading. In interviews researchers can bias p's through non verbal cues):Pilot the questionnaire.
- Characteristics of the researcher(may influence the way the p responds (e.g their gender,ethnicity etc):Double blind trial
Reliability and self reports
Another word for reliability is consistency. In terms of self reports this can be considered in several ways:
Internal reliability - Refers to whether the self report is consistent within itself. So we would need to ask whether the q's within it are consistent and measure the same variable throughout. A way to assess the internal validity of a questionnaire is to use the SPLIT HALF APPROACH which involves splitting the q's into 2 groups randomly or diving them into odd&even numbers.-Each comprising half the questionnaire. If the questionnaire is internally reliable each half should yield similar scores. The 2 scores can be correlated to assess this further (a positive correlation near the top of the possible range (0-1) would reflect reliability.
External reliability - Refers to whether the self report STAYS THE SAME FROM ONE USE TO ANOTHER. We would expect a questionnaire measuring a psychological vairable that is relatively stable to yield a consistent score when applied to the same person over a short period of time. If not it lacks external reliability. A way to asses the external reliability of a self report is to question p's on one occasion then again some time later. The 2 sets of scores would then be correlated to determine the extent to which pairs of scores from the same person at 2 diff times are similar. This method of checking is called the test-retest approach.
Reliability and self reports
Reliability can be affected by:
- How questions are asked.
- How they are responded to by participant or scored/interpreted by the researcher.
Reliability and asking questions
This may relate to how the questions are phrased or how they're asked by the interviewer. Reliability may be more of a problem when asking questions in interviews than questionnaires because in interviews the questions are being asked by another person who may not be fully consistent in the way they ask the questions across participants. So reliability may be threatened if the same interviewer behaves differently with different participants, or if different interviewers ask questions in different ways.
Involve gathering data on 1 individual or one group of individuals (e.g. a family,tribe,gang) It's often the method of choice is the participant or group is undergoing some unusual experience that is of interest to the researcher. It may also be used when the subject matter being investigated changes or develops over time (longitudinally) E.g. it may be used to study the development of an ability or illness. Case studies tend to involve the collection of lots of data using a variety of techniques (triangulation) - e.g. the investigator may use self report data&observational data.
Strengths of case studies
- Provide detailed, in depth data using a variety of techniques.
- In some cases the info gathered is high in ecological validity as it related to everyday life.
- Sometimes it's the only possible method available tostudy example of behaviour because what is to be studied is reliable or unusual.
- Due to its longitudinal nature it allows us to see how behaviours or conditions develop and change over time.
- Involve a limited sample, so results cannot be generalised to others. This limits the usefullness of the findings.
- Their longitudinal nature makes it a length process, a problem because the data takes time to gather and if a participant decides to withdraw, research may be incomplete or end.
- In examples of 'action research', the aims of the research may sometimes compromise the needs of the participants involved. This is a problem because the needs of the researcher may be placed before those of participants. This raises ethical issues such as psychological harm or that p's feel forced to continue.
- No 2 case studies are the same so this method is often criticised for lacking consistency or reliability.
- Case studies often rely heavily retrospective data (demanding the p to remember past events/times) which they may not be able to do in an accurate way meaning that data could lack validity.
Aims and hypotheses
All research starts with an aim that tells us why the research is taking place. This then becomes the basis of a specific prediction or HYPOTHESIS. A hypothesis is a testable statement in which the variables are clearly stated in measurable terms. This is called OPERATIONALISING the variables. (To operationalise a variable is to define it - specifically)
E.g. a broad aim: 'To investigate the effects of TV viewing on mood' but an experimental hypothesis might state that 'people who watch over 30 hours of TV viewing per week score lower on a mood rating scale that people who watch under 30 hours per week' hypothesis is MORE PRECISE than the aim.
Hypothesis: An unambiguously (clear) phrased, testable statement/prediciton.
E.g. 'boys are more aggressive than girls' this is vague & doesn't state the IV or DV. Therefore, must be more specific and given an operationalised hypothesis (So IV&DV clearly specified)
E.g. 12 year old males will score more highly on an aggression questionnaire than 12 year old females.
Hypotheses can be directional or non-directional.
- A directional hypothesis is one tailed. You assume that by manipulating the independent variable there will be one specific change in the dependent variable. You can predict if this change will be positive or negative.
For example if you ask someone to say la la la la while trying to remember a list of words (the IV) you can assume that this will have a negative impact on their ability to recall the words (the DV). Another example, males are more aggressive. - certainty,how certain are you about the hypothesis?
- Non directional - A non-diectional hypothesis is two tailed. You assume that by manipulating the independent variable there will be a change in the dependent variable. You cannot predict if this change will be positive or negative.
For example if you ask someone to roll a ball in their hands while trying to remember a list of words (the IV) this could either have a positive or negative impact on their ability to recall the words (the DV). Another example, there's a sex diff in aggression(one thing may affect another)
A good way to express a directional hypothesis:
Participants who (condition A of the IV) will (refer to the DV and its expected direction of change) than participants who (condition B of the IV)
E.g. Students who have attended 3 or more revision sessions (condition A of the IV) will score significantly more in their psyc exam (DV result-expected direction of change) than students who have not attended 3 or more revision sessions. (condition B of the IV) 'more' = direction of change.
A good way to express a non-directional hypothesis:
It is predicted there will be a significant difference in (the DV) between (condition A of the IV) and (condition B of the IV)
E.g. There will be a significant difference in the scores on a psyc exam (DV) between students who have attended 3 or more revision sessions (condition A of the IV) and those who have not attended 3 or more revision sessions.(condition B of the IV)
As well as constructing the experimental (or alternate) hypothesis the researcher must also contruct the null hypothesis. This is because for research to be considered 'scientific' the hypothesis should be capable of being shown to be wrong. It should be possible to refute (prove wrong) (therefore the null exists to be refuted). The null should state that any difference found between 2 sets of data has NOT been caused by the IV, or that correlations or associations shown in data are not meaningful (significant)
Null hypotheses are always NON DIRECTIONAL.
Framework for writing a null-directional hypothesis:
There will be no significant difference (in the DV) between (condition A of the IV) and (condition B of the IV). Any difference found will be due to chance.
E.g. There will be no significant difference in the aggression questionnaire scores between males and females.
Independent Measures Design (aka Independent Groups Design) Involves using different p's for each condition in the investigation, i.e. if investigating the impact of chewing on learning a list of words, 1 group of p's will carry out the task whilst chewing, the other will carry out the task without chewing. Counterbalance- group 1 does one task first & other second and group 2 do opposite.
- Can be quicker to carry out, commitment from p's is reduced
- The same task can often be used in each condition
- P's will be less likely to respond to demand characteristics
- Twice as many p's are needed
- Participant variables may confound results
- May be difficult to keep variables constant across conditions
Possible ways of reducing the impact of the last 2 weaknesses: Make sure EV's are as controlled as possible (use matched pairs design) and use controlled environment (lab)
Repeated Measures Design - This involves the same p taking part in each condition of the IV, i.e. if investigating the impact of chewing on learning a list of words, the p's will learn one list of worlds whilst chewing, and another list whilst not chewing.
- Since the same p's are used in each condition, participant variables will not confound results.
- Fewer p's are needed.
- Results may be influenced by order effects, i.e. practice or boredom may improve or damage performance in the 2nd condition.
- P's are more likely to guess the aim of the study and respond to demand characteristics.
- P's may be 'lost' between conditions.
How to reduce the impact of the last 2 weaknesses: Counterbalance - p's do something diff at diff times. Try do it without a break.
Matched Pairs Design (aka Matched Participants Design-closest is identical twins): Participants are matched on a factor important to the experiment. E.g. if investigating the impact of chewing on memory, p's may be pre-tested to establish how good their memory is generally. Then the sample would be split in such a way that the 2 with the best memory are divided between the chewing group and non-chewing group; the next 2 would be similarly divided and so on.
- Reduces the effect of some key participant variables
- P's are less likely to guess the aim of the study and respond to demand characteristics
- There will be no order effects to confound results
- The same task can often be used in each condition
- It may be practically difficult to establish matches, and this matching takes time.
- Requires twice as many p's
The experimental method
An experiment is a research method in which the effect of one variable on another is investigated. A variable is anything that can change or vary.
INDEPENDENT VARIABLE (IV) - Is a variable thought to bring about change. In an experiment this is manipulated (changed) by the researcher. E.g. an experimenter may manipulat the way they present info to p's and test whether it affects their memory of it. So some info may be presented with pictures of objects and others may be a list of words.
DEPENDENT VARIABLE (DV) - Is what occurs as a result of the manipulation of the IV, in other words the effect. The DV is what is measured or what the results of the experiment are dependent upon (what counts as the results in an experiment). In the example above the DV would be the memory test scores of p's.
Research believes the IV influences the DV.
Extraneous variables (EV) may influence the DV. If this occurs the EV will mess up results and this is then known as a confounding variable (CV - if EV isn't controlled it becomes CV).
A pilot study is a small scale, preliminary study conducted before the main research in order to check the feasibility or to improve the design of the research. Pilot studies therefore may not be appropriate for case studies. They are frequently carried out before large-scale quantitative research in an attempt to avoid time and money being wasted on an inadequately designed project. A pilot study is usually carried out on members of the relevant population, but not on those who will form part of the final sample. This is because it may influence the later behaviour of research subjects if they have already been involved in the research.
Pilot studies are used to make sure the research design works- E.g. a questionnaire can be piloted with a small sample. This could be done to test the q's make sense and give the answers they want, also if modification is necessary it can be done after the pilot.
Pilot studies can also be useful in operationalising variables. E.g. if you were to consider aggression in the playground, it might be useful to do a prelim observation.
Pilot studies can also reveal demand characteristics and provide the researcher an opportunity to reduce them.
Pilot study may also show that the research design is not practial and send the researcher back to consider different ways of collecting data. This can be valuable: should anything be missing in the pilot, it can be added to the expt and chances are the full scale (and more expensive) expt will not have to be re done.
Variables other than the IV that affect the DV. If they're not controlled they become CV's.
1) Participant variables - These are personal variables concerning the p's in studies such as personality, age, health status, mood or intelligence etc. P variables can confound the DV if they're not controlled. We can reduce their effect by making sure p's are randomly divided between groups (conditions) in an expt. This then gives a good mix of p variables across the conditons we are comparing and ensures they are distributed unsystematically.
2) Situational variables - These are variables concerning aspects of the research situation such as noise, interruptions, temp and room layout. These can also confound the DV if uncontrolled. A lab expt can control situational variables (for the examples above^)
3) Demand characteristics - Are features of the study that encourage p's to think and therefore behave in a certain way and tend to relate to cues in the external research environment. E.g. p might easily work out that in a lab they're expected to concentrate more. The knowledge of what is expected by the experimenter may then influence the performance of p's and their performance may be due to this and not the IV. Demand charcteristics are closely related to social desirability effect variables.
4) Investigator effects - When the investigator influences the p's performance or behaviour by them (usually unconsciously) projecting their expectations about what will be found onto the participants. This can be very subtle -e.g. investigator may raise an eyebrow or encourage the p with a smile or eye contact. P's then behave accordingly and expected results tend to occur, but these may have reduced validity as a result of this confounding variable.
5) Order effects - Occur in experiments when p's participate in more than one condition (ie they perform more than once, each time with a different experience of the IV) and their performance is affected by the order in which they experience the conditions. E.g. a p's performance on a problem solving task may be measured when they have been given clear instructions. Their performance may change because of the clean instructions but it may also be because, second time round, they're more relaxed or practiced (practice effect). If their performance is worse it may be because they're tired (fatigue effect) or bored (boredom effect)
- Random allocation of p's to conditions
- Screening p's before they participate by asking them questions for example
- Use the SAME p's in all conditions so they are in effect being compared with themselves or use matched pairs design
- Use lab setting
- Constancy - ensure ALL p's are exposed to it (a form of standardisation)
- Deceive p's about the aim
- Keep p's naive about the aim - 'single blind technique'
- Use distraction tactics - e.g. on questionnaire at irrelevant questions
- Use field setting in which p's are unaware of their involvement in the study
- Use a third party who is unaware of the aim to test p's - 'double blind technique'
- Ensure researcher is unaware which condition p's are in - 'double blind technique'
- Use different p's in each condition
- Use a time break between conditions
- Counterbalance so some p's do condition A then condition B and others to B then A
Another popular control is standardisation which refers to keeping things the same for all. E.g. standardising instructions given to p's or standardising the amount of time p's are given to perform a task.
CONSENT - need informed consent. Under 16's have local parentis do it on their behalf
PROTECTION FROM DISTRESS - must leave in same or better state, not worse
DECEPTION - shouldn't be lied to unless there's a really good reason
DEBRIEFING - told about everything involved at the end and any q's answered
RIGHT TO WITHDRAW - p's can leave ANY TIME they want
CONFIDENTIALITY - p's have the right to expect their info will be treated confidentially
OBSERVATIONAL RESEARCH - public places only, can't invade privacy
GIVING ADVICE - only give it if psychologist is specialised in that sector, if not they refer
COLLEAGUES - if colleageus believe ethics are being broken they can take action
Explanation of ethical issues. Remember 'DIP'
- Deception is an ethical issue because it prevents that p from giving informed consent and the p may find themselves in research against their wishes.
- It is also an issue because the p's may become distrustful of psychologists in the future.
- Lack of informed consent means that the p has not agreed to be in the research & may find themselves taking part in research against their wishes.
- The p has not agreed to be in the research which breaks ethical guidelines.
- It can also apply to p's who have volunteered to take part in the research but have not been fully informed about the aims of the study.
- It may make p's distrustful of psychologists in the future.
Protection of p's - P's have the right to not be harmed as a result of taking part in research. The P should leave the research the same way they entered. If they're harmed they may suffer long term effects which could impact future lives. (E.g. Milgram)
Dealing with ethical issues
Deception: Very common in research. E.g. Mengers (1973) found that of 1000 studies reviewed, 80% didn't give the p's full info on the study. Methods to deal with ethical issue:
- Debriefing: this is where on completion of the research the true aim of the research is revealed to the p. The aim of the debrief is to restore the p to the state they came in. A p should leave in the same or better state they came in.
- Retrospective informed consent: once the true nature of the research has been revealsed the p should be given the right to withdraw their data.
Informed consent: Issue raised by research that involves children under 16 so don't understand what they're taking part in, thus impacting on their ability to give informed consent.
- Prior general consent: this solution involves obtaining the prior consent of the p's to be involved in research that involves deception. If the p agrees that they would not object to being decieved in future research studies, then in later studies where they participate it's assumed they have agreed to being deceived.
- Presumptive consent: this involves taking a random sample of the pop & introducing them to the research, including any deception involved. If they agree that they'd give consent the researchers can generalise from this an assume the rest of the pop would also agree.
- Children as p's: this is resolved by gaining the consent of the parent or those in loco parentis. E.g. the headmistress of the school.
Dealing with ethical issues
Protection of participants:
- The researcher should remind p's of their right to withdraw if at any time during the research the level of stress is higher than anticipated.
- The researcher is responsible for terminating any research that results in psychological or physical harm that is higher than expected. E.g. Zimbardo stopped his research after 6 days, although it was intended to run for 2 weeks. Some argue he should have stopped earlier.
- Debriefing is an important part of protection of p's.
The sample should be representative of the population as a whole (to allow generalisability)
Opportunity sampling: Sampling technique most used by psyc students. It consists of taking the sample from people who are available at the time the study is carried out and fit into the criteria you are looking for. E.g. first 20 students in college canteen.
- Strengths: Easy and quick to find p's.
- Weaknesses: Biased due to small selection and unrepresentative (so can't generalise)
Random sampling: A sample in which every member of the population has an equal chance of being chosen. This involves identifying everyone in the target pop and then selecting the number of p's you need in a way that gives everyone in the pop an equal chance of being chosen. E.g. pull names of college students out of a hat.
- Strengths: Everyone has equal chance. Unbiased
- Weaknesses: Very difficult to conduct if the size of pop is large & needs lots of time&money.
Self selected sampling: Consists of p's becoming part of a study because they volunteer when asked or in response to an advert (E.g. Milgram)
- Strengths: quick & easy to get p's
- Weaknesses: sample is biased and unrepresentative of pop.
Data analysis and presentation
Presentation and interpretation of quantative data
Data can more easily be analysed if presented visually. Tables, graphs (e.g. bar charts) and scattergrams enable us to visually present the data collected.
Points to remember when visually displaying data:
- All tables, graphs and scattergrams must be fully labelled and titled. The axes must be correctly labelled.
- Only one graph/chart should be used to illustrate a set of data.
- An appropriate scale should always be used as inappropriate scales can (and do) mislead.
- Raw data should never be presented using a graph or a chart. (Scattergrams are an exception). Instead they should be used to summarise the data.
Graphical Descriptive Statistics
How do psychologists summarise their data pictorially?
Bar charts - Shows data only for those categories that the researcher is interested in comparing. E.g. 2 or more conditions in an experiment.
Histograms - Shows data for all categories, even those with 0 values. The column width for each category interval is equal so the area of the column is proportional to the no. of cases it contains of the sample.
Frequency polygon (Aka line graph) - Is similar to the histogram, except it allows 2 or more sets of data to be shown on the same graph.
Pie charts - Show the proportions of all scores gained by various categories. Proportions of the pie chart are calculated by the following formula: degrees of pie chart = score of category x degrees of circle / by total score. E.g. 90 = 25 x 360 x 100
Scattergrams - Plot pairs of scores against each other to show their correlational relationship. Emergent patterns or trends in the data can be calculated to show a line of best fit.
Quantitative data (in the form of numbers)
Methods that produce quantitative data: IQ tests, correlations,closed questions (questionnaires)
Strengths of quantitative data:
- Allows us to analyse in standardised mathematical ways so that meaningful comparisons between sets of data can be made.
- Allows us to use interential statistical tests on sets of data. These tell us how likely it is that our results occured due to chance (rather than an IV)
- Enables us to identify patterns and trends quickly. E.g. scattergrams.
- Tends to be objecteive and doesn't require the same degree of interpretation that qualitative data demands. So it's also higher in reliability than qualitative data.
Weaknesses of quantitative data:
- Lacks detail and reduces some aspects of human behaviour down to simple numbers & some richness and complexity of behaviour is lost (reductionist)
- Numerical data often tells us what occured but not why it occured.
Qualitative data(in the form of description)
Methods that produce qualitative data: case studies, interviews (open questions)
Advantages of qualitative data analysis:
- Is usually generated in natural social contexts so the meanings of actions or interactions are illuminated.
- Allows p's to express themselves on their terms. This subjectivity should protect the validity of the data because an investigator is not imposing their interpretation on it.
- Qualitative accounts might be used to compare different views of the same thing. E.g. we could use qualitative data gathered in an interview about reasons for aggression. These could be compared with the same p's behaviour in a structured observation in the playground. This is called triangulation. It's a way of assessing the reliability of data.
Disadvantages of qualitative data:
- Not as easy to quantify and therefore analyse. So it's not as 'scientific' as quantitative data.
- It's more difficult to draw comparisons between p's.
- Involves personal accounts which may be biased and possibly invalid due to many factors.
- Time consuming to gather this type of data.
Qual + Quant data
Qualititative data strengths + weaknesses:
- Allows p's to express themselves on their terms.
- Might be used to compare diff views of the same thing.
- Not as easy to quantify and therefore analyse.
- More difficult to draw comparisons between p's.
Quantitative data strengths + weaknesses:
- Allows us to identify patterns and trends quickly.
- Higher reliability than qualitative data.
- Lacks detail (reductionist)
- Tells us what occurs, not why it occured.
Analysis and interpretation of quantitative data
Both descriptive and inferential statistics are used to help analyse data further. By descriptive statistics we mean statistical procedures that describe, organise and summarise sets of data. (Allows us to display data and identify patterns & trends. E.g. tables, bar charts etc)
The following are all forms of DESCRIPTIVE STATISTICS:
- Measures of central tendancy (mean, median mode)
- Measures of dispersion (Range, standard deviation)
How to calculate...
- Mean: Add up all the numbers & divide by how many there are.
- Median: Put the numbers in order & then find the middle number.
- Mode: Most common number.
- Range: Smallest value - biggest value.
- SD: Calculate the mean & subtract it from each individual score; square each of these scores; add them together; divide by the sum of the squares by the number of scores minus 1. This is the variance. Using a calculator work out the square root of the variance to give the SD.
Types of data
a) Do you smoke. (yes or no) - NOMINAL DATA. (most basic)
b) On a scale of 1 to 10 (1=not at all. 10=very) how much of a smoker are you? - ORDINAL DATA.
c) On average, how many cigarettes a day do you smoke? - INTERVAL DATA.
Measures of central tendency & dispersion
Mean (all scores divided by number of scores)
- Appropriate to use with sets of data with no outliers (extreme scores) in one direction.
- Not appropriate to use when there's outliers.
- Advantages- It uses all the data points so provides a good estimate for the central score of a data set.
- Disadvantages-It may not represent best the general trend in a set of scores.
Median (central score when all scores are ordered)
- Appropriate to use when there's outliers.
- Not appropriate to use when there's no outliers or a small number of scores.
- Advantages- Unaffected by outliers in one direction (skewed data)
- Disadvantages- Doesn't take every score into account so less sensitive than the mean. & Less representative when there's a small data set.
Mode (most frequently occuring value) - Appropriate to use with data where there's a high number of repeated scores.
- Not appropriate to use with data where there's several scores that occur as frequently as each other (especially when the data set is relatively small)
- Advantages- Shows what most p's scored. & Easy to calculate. & can use nominal data.
- Disadvantages-It can be unrepresentative if the most frequently occuring data point is very high or very low.
Measures of dispersion
There are 2 types of dispersion:
- The range - The difference between the highest score & the lowest. (lowest - highest)
- The SD - How widely spread the values in a data set are around the mean. Unlike the range it takes all the scores into account making it a powerful measure of dispersion.
What SD tells us....
- Large SD - The scores are SPREAD OUT
- Small SD - The scores are CLOSE TOGETHER
- Zero - They're all the SAME - NO VARIATION
Measures of dispersion
Range (The difference between the highest and lowest score in a set of data)
- Appropriate to use when you wish to make a basic measure of the variation within the data and the data is consistent. If there are extreme scores (outliers) the range is inappropriate as it will be a distorted measure of variation.
- Advantages - Easy to calculate.
- Disadvantages - The range can be easily distorted by extreme scores (outliers)
Standard deviation (A measure of dispersion that indicates the 'spread' of the data around the mean).
- Appropriate when you wish to make a very sensitive measure of dispersion.
- Advantages - Uses all the values. It's a sensitive measure of dispersion.
- Disadvantages - More difficult to calculate compared to the range.
Analysis and interpretation of correlational data
Correlation is a form of analysis applied to sets of data to establish the direction and strength of the relationships between co-variables. The data used has to be QUANTITATIVE and atleast capable of being placed in rank order.
Data to be correlated is presented in the form of a scattergram. If there's a strong positive or strong negative correlation between sets of data this will be apparent from the scattergram. There may be weak positive or weak negative correlations that are less easily detected just from the eyeball impression gained from the scattergram. There may be no correlation at all. There may be clusters of scores around a certain point.
An inferential statistical test will be used to arrive at a CORRELATION COEFFICIENT. This is an inferential statistic which indicates the strength and direction of a correlation ranging from -1 (a perfect negative correlation) to +1 (a perfect positive correlation). A score of 0 would indicate no correlation. Inferential tests allow us to discover the degree to which a relationship may exist between 2 variables because of chance factors (rather than because there is a meaningful relationship bewtween the variables in question)
Advantages and disadvantages of correlational anal
- May avoid some of the practical and ethical problems raised by other methods as there is often no direct involvement with or manipulation of the participants.
- Can reveal the direction and strength of a relationship between variables.
- Can be used on data from ecologically valid sources. E.g. actual health statistics.
- Can not reveal cause and effect relationships
- Uses only quantitative data which may be reductionist and not holistic
- Can only be performed on quantitative data which is ordinal or interval
Turning QUALITATIVE data to QUANTITATIVE DATA. (For content analysis p's need to be asked open ended q's)
- The researcher establishes a set of categories and simple 'counts' the number of instances the behaviour/image/word/action ect occurs.
- This is very similar to a structured observaton (Where behavioural categories are used) but instead the categories are applies to qualitative data. (e.g. answers to a questionnaire or a recording of a convo) etc.
How can reliability and investigator bias be a problem from content analysis?
Reliability - Subjective; People aren't trained adequatley; Apply categories inappropriately; No pilot study
Investigator bias - Vague (flexible) categories- more investigator bias so invalidates data.
- Constructing the categories (choice of them)
Content analysis steps
1) Read/listen/watch the qualitative data you wish to quantify and write a list of common behaviour categories/coding units. These must be clear, relevant and easy to spot.
2) Train investigators to apply the behaviour categories/coding units and conduct a pilot study to check everything works as it should.
3) Apply the categories to the data.
4) Check reliability between the investigators (for inter-observer reliability).
5) Analyse the data for patterns; formulate a conclusion.