Psychologists, like anyone, observe events in their environment and ask why they happen. They seeks evidence to test and support their theories. This is 'doing research.'
Formulating research questions =
Research starts with a research question: something that a psychologists wants to find out. For example:
- How do phobias occur?
- Is it easier to remember sounds or images?
- What makes some people more obedient than others?
There are a number of possible answers to each question, which is why research needs to be carried out. Obedience levels may depend on personality. They may depend on setting. They may depend on who is giving the orders.
It would be difficult to investigate all of these factors at once, so a psychologist would narrow down their findings of research to come up with an AIM.
An Aim is a general statement which describes what a psychologist intends to investigate.
For example, an aim would focus on one factor affecting obedience. The aim of their research would therefore be 'to investigate whether a person's confidence level has an effect on the likelihood of them obeying.'
Similarly, a psychologist may aim 'to investigate whether people recall more words depending on whether thay are presented acoustically or visually.'
Having stated an aim, a psychologist then formulates hypotheses. The aim is based on a theory, and the theory should allow predictions to be made.
For example, a psychologist may predict an individual who is more confident is less obedient. This does not consitute a hypothesis yet.
A hypothesis is a testable statement making predictions about what will happen in an investigation.
A hypothesis also needs to operationalise (offer a clear set of criteria to describe how something will be set up or assessed or measured) the variables (factors that change/vary) being tested.
In the example, there are two variables being tested: level of confidence and likelihood of obedience. The psychologist may decide to determine level of confidence by using a questionnaire on participants, and measure the likelihood of obedience by observing whether the same participants follow an unreasonable request made by a stranger in uniform. This has been operationalised.
Now, the psychologist can make a very precise prediction, known as a research hypothesis.
This is a hypothesis that predicts a difference in the measured variable (change in DV due to the IV) or correlation between variables.
The experimenter's research hypothesis may be: 'Participants who score above average on a questionnaire measuring confidence and significantly less likely to follow an unreasonable request made by a stranger in uniform compared to participants who score below average.'
The psychologist would then test this hypothesis by collecting data, through carrying out an investigation. They have to make a key research decision. If the results suggest the prediction was correct, then the research hypothesis is retained. If the results suggest the prediction was wrong, the research hypothesis is rejected.
If it's possible that the psychologists may reject their research hypothesis, they will need to also form a null hypothesis, which predicts no difference, rather than a difference.
A null hypothesis: a hypothesis that predicts no difference in a variable or no correlation between variables.
For example: 'there is no significant difference in the number of times participants follow an unreasonable request from a stranger in uniform, whether they score above or below average on a questionnaire measuring confidence.'
The null hypothesis is retained if the research hypothesis is rejected.
It makes sense to use the word 'significant' when predicting a difference between conditions.
A difference needs to be 'significant' to be deemed good research.
So when writing, predict with a SIGNIFICANT hypothesis.
BUT HEDGE. Use words like 'more likely' or 'less likely'
For example: RH - 'Participants scoring above average when tested on a questionnaire measuring confidence are significantly more likely to have blue eyes compared to those who score below average.'
NH - 'There is no significant difference in participants eye colour, whether they score above or below average on a questionnarie measuring confidence.'
All of the hypotheses so far have predicted differences. However, it is also possible for a hypothesis to predict a correlation.
A correlation = a relationship between two variables (ie, the IV and DV)
Eg - 'there is a significant correlation between participants' scores on a self-esteem test and the number of friends they have.'
The null hypothesis would be 'There is no significant correlation between participants' scores on a self-esteem test and the number of friends they have.'
The hypothesis predicting no correlation or no difference is always the null hypothesis. The hypothesis predicting a difference is not always a research hypothesis (I know, frustrating.)
If the research hypothesis is formulated specifically for an experiment, it can be known as an experimental hypothesis.
This is a hypothesis used only in experiments which predicts a difference and has an IV and DV.
Experimental hypotheses only predict differences as correlations CANNOT be tested experimentally.
Although less common, a psychologist may predict no difference before carrying out their research.
Eg: 'there is no significant difference in the ability of men and women to solve anagrams under timed conditions.'
The second hypothesis is known as the ALTERNATIVE hypothesis.
This is a hypothesis that predicts a difference or correlation following a null hypothesis which predicts no difference or correlation
This is the variable that is manipulated or set up in an experiment.
Ie: Whether participants are male or female
This gets its name because, if the experimenter's prediction is right, this varibale depends on the activity of the independent varible.
The IV has an effect on the DV!!
The dependent variable is the thing that is measured after the IV may have had an effect on it.
IE: 'number of words recalled in two minutes.'
CORRELATIONS do not have an IV or a DV. They are simply predicting a relationship between two varibales but isn't stating which variable affects which.
(Limitation to correlation - can't establish cause or effect - doesn't know which is IV which is DV)
Offers one of the most common ways of collecting evidence in psychology.
Experimental methods do not all apply one particular method of data collection - there are several types. Experimenters may use observation or more direct questioning.
With experiments, it is the process which is important, and not how the data is collected. What all experimental methods have in common is the fact that an independent variable (IV) is manipulated to cause a change in an dependent variable (DV). The DV is then measured.
In experiments, te ideal is to control all relevant variables whilst changing only the IV. Every attempt is made to even out confounding variables (a variable beside the IV which may have affected the DV), such as balancing the number of men and women in each condition, and to eliminate any extraneous variables (a variable beside the IV that could affect the DV), which may effect results, such as noise, poor eyesight of participant.
If all other variables are controlled, only the IV can be responsible for changes in the DV. Therefore, the experiment is the only method which can reliably state that one thing (the IV) has caused another thing to change (DV).
Experimenters need to control unwanted varibles which may affect the DV, otherwise this questions the reliability of the research.
These unwanted variables are called extraneous variables. Sex, intelligence levels could be extraneous variables if comparing girls and boys performance in a math test.
If certain variables are not controlled (eg, who their math teacher is) or cannot be controlled (eg, if some boys feel sick on the day), they may later affect the outcome of a piece of research.
When variables have had an impact on findings, then these are described as confounding variables. Confounding variables make results less reliable.
So the effects of extraneous variables should be eliminated to stop them from becoming confounding variables. However, this isn't always possible, in which case, it is common to attempt to minimise their effect. There are three wats of controlling extraneous variables:
The process of keeping variables the same
This means making things the same (or standard) across conditions. An experimenter may standardise factors such as particpants, the environment, tasks, measures and instructions.
Imagine an experimenter comparing recall of images and words. If participants were given different instructuions for recalling the images rather than words, then this may affect their performance rather than the nature of the material.
The process of ensuring variables occur occur in all possible combinations an equal number of times.
This is often used in repeated measures designs, that is where the same participants are used in all conditions. In such experiments, there is a danger that order effects (where behaviour is affected because participants take part in two or more conditions in a particular order) may affect results. Order effects include:
- The practice effect = where participants' performance improves across conditions through familiarity with a task or environment
- The fatigue effect = where participants' performance worsens across conditions because of tiredness/boredom
- Recognising demand characteristics = features or cues in an experiment which help participants work out what is expected of them (the aim of the experiment). Helpful participants may respond according to what they think is being investigated
If order effects arise, this does not get a true measure of behaviour
Counterbalancing simply means that the order in which conditions are encountered is balanced out across all participants. This means every possible combination of order of conditions occurs the same number of times.
For example, with two conditions (eg, A and B), 50% of participants would do this condition. A and then condition B, and 50% would do B then A.
This doesn't get rid of order effects, but it does stop the same condition being affected every time. If counterbalanced, the confounding variable wil be balanced across both conditions.
The process of deciding the order or use of variables by chance.
Eg, a psychologist wants to test whether where a word in a list has an effect on its chance of being recalled. They predict the first and last three in a list of 12 will be better recalled than the middle six. But they want to ensure the first three and less three aren't recalled more as they're easier words to remember = an extraneous variable. It therefore makes sense to control this variable by changing where each word occurs in the list. Rather than try to work out all the possible combinations, the psychologist would be likely to randomise the order of words. This would mean each word would have an equal chance of appearing at any position on the list. Probabiliy would suggest that each word would occupy most places in the list over a number of trials
This means controlling variables by use of chance - if researchers use chance to control variables, this means they cannot be accused of biasing the investigation.
Randomisation can also be used to decide the order of conditions as an alternative to counterbalancing
Types of experiment
There are three different types of experiment:
- Labroratory experiment
- Field experiment
As they're all experiments, they all involve the use of an IV and a DV, however, they differ in the levels of control they have over variables.
This is an experiment carried out in a controlled environment
Many memory and perception experiments are carried out under controlled conditions as psychologists are keen to control situational variables which may affect attention and thinking, such as noise, heat etc. These factors can be kept constant under lab conditions.
A lab experiment also allows random allocation (allocating participants by chance; each one has an equal chance of ending up in each condition) of participants to conditions where appropriate. This makes it a true experiment as the psychologist can't influence who participates in which condition. Findings cannot be accused of bias.
Behaviourists favour lab experiments as controlled conditions allow them to manipulate one stimulus to elict a response, whilst keeping all others the same.
Since a lab setting makes it easy to control most extraneous variables, it has the advantage that it's easier to reliably establish cause and effect. If an experimenter observes a change in the DV, it's highly likely to be due to the IV (the only variable that should've changed). However, this gives an artificial setting and may not occur in real life. This means the findings from lab experiments may lack ecological validity (the extent to which a situation reflects real life).
An experiment carried out in a natural environment
An alternative to conducting an experiment in the lab, by conducting in the field. A field experiment is one carried out in the natural environment of those being investigated - eg, in a school or street. The IV is still manipulated by the experimenter.
A number of obedience experiments have been carried out in natural environments. For eg, research assistants would dress in uniforms and approach people in shopping centres to see if the would follow a request.This gives findings more ecological validity as the people are in real- life settings and are more likely to behave as they normally would do.
These experiments also have the advantage of having some control over extraneous variables. The psychologist could keep the person in the same uniform, for example, and ensure participants are asked to do the same task the same way. BUT the psychologist will not have control over environmental factors, such as how many people are around at the time - making it more difficult to reliably establish cause and effect, compared to a lab experiment.
An experiment where the experimenter doesn't directly control the IV
This may take place in a lab or in the field. Like other experiments, they have an IV but the experimenter doesn't directly manipulate the IV.
Some IVs aren't open to manipulation as some conditions are predecided by fixed characterisitcs. For example, if a psychologist was comparing men and women's driving skills, she could not randomly allocate participants to be male or female! The IV is naturally occuring.
It may be naturally occuring in the sense that it's reliant on the forces of nature. Such as if the weather was the IV, one would have to wait for it to occur naturally.
Sometimes, conditions are set up, also giving a quasi-experiment, but not by the psychologist. Eg, a school may be testing a new school regieme with some students, and the psychologist may visit the school to study this - they wouldn't be directly in control of the IV.
Some participants prefer quasi's for ethical reasons as participants aren't being manipulated as much as in lab or field experiments.
Any experiment will have at least two conditions - often an experimental (the condition where a variable is actually tested) and the control condition (the condition that acts as a comparison, where nothing changes). The experimental condition describes the condition where a variable is actually being tested, whereas the control condition is where nothing is manipulated and all things are kept the same as normal.
Having established conditions, the next decision an experimenter needs to make is how they willl organise participants across conditions. They have to choose an experimental design.
There are three types of experimental design:
- Repeated (or related) measures design
- Independent groups design
- Matched pairs design
Repeated (related) measures design
In a repeated measures, the same participants are used in each condition. Ie, participants' recall is tested with cues and then without cues.
- Any differences between conditions are likely to be due to changes in the IV and are not due to participant variables (the differences between the characteristics of participants)
- Fewer participants need to be recruited as they are used twice (or more!)
- Order effect (eg, practice effect, fatigue effect, recognising demand characteristics) as participants take part in all conditions
However, order effects can be reduced (not elimitated) by counterbalancing or randomising the order of conditions.
Independent groups design
An experimental design where different participants are randomly allocated to different conditions. This would normally be decided by random allocation, through things like the toss of a coin. There is no attempt to match participants across conditions
- There are no order effects as participants only take part in one condition, so cannot get better through practice, or under-perform from fatigue, or change their behaviour
- It allows task variables to be controlled, eg, participants can be given the same word list in each condition so that this doesn't become a confounding variable
- Any differences between conditions could be due to the individual differences of participants, eg, some participants may have more motivation, which may effect results
However, the larger a sample (group of people who're selected) is, the lower the probability of a significant difference in participant characterisitcs.
Matched Pairs Design
An experimental design where different participants are used in each condition, but where they're matched on key characteristics.
This is where there are different pariticpants but they're related in the sense they're matched up on important psychological characterisics. These characteristics will depend upon the nature of the study - typical ones are gender, age, intelligence and personality.
For every participant in one condition, they have a partner they've been matched with in the second condition.
- there are no order effects as participants only take part in one condition
- Individual differences between conditions are reduced as participants have been matched up
- It's time consuming and expensive to match participants
Strengths of experimental method
- Experiments offer a high level of control over extraneous variables, especially lab experiments where the environment is also controlled
- Control over variables makes it easier to reliably establish cause and effect, that is to be surer the IV is affecting the DV
- If cause and effect is established, it's possible to predict and control behaviour. This is the goal of scientific research making experiments highly scientific
- Experiments are also objective as they're not easily influenced by the experimenter once set up. This means results aren't open to bias
Limitations of Experimental method
- Many experiments are lab based, meaning the environment is artificial
- Findings therefore lack ecological validity (however, this problem is solved by field experiments that take place in natural environments)
- Since experiments are highly controlled, they measure variables in very precise ways. However, this gives results that lack construct validity as variables are often assessed more narrowly than they would appear in real life
- Participants are often aware they are taking part in experiments. They may then respond to demand characteristics of the experiment by behaving differently from normal. Demand characteristics tend to be an issue with experiments as they are so 'set-up' the aim often becomes obvious
Non Experimental Methods
Experiments allows researchers to reliably establish cause and effect as they are highly controlled. However, do experiments involve too much control?
Some researchers believe that controlling variables distorts reality because in real life, many variables operate together to affect human thought and behaviour. They would say isolating some variables and manipulating others is too artificial. Therefore, they prefer to use more valid methods of investigation that reflect real life more accuratley.
These are sometimes named as non experiemental methods. They are also seen as more ethical. They don't necessarily manipulate and change participant's behaviour in the same way that experiments can do.
Sometimes, experiements are not an option and non-experimental methods have to be used anyway - some situations cannot be set up experimentally. Eg, a child cannot be made autistic so the effects of the disorder can be studied - this can only happen naturally.
Non experimental methods are essentially methods that don't involve the direct manipulation of situation or behaviour. Instead they investigate phenomena as they occur
Non Experimental Methods
There are a number of methods that can be catergorised as non experimental methods:
- Self-report methods, where participants report on their own thoughts or behaviour using specific methods such as questionnaries and interviews
- Observational studies, where participants' behaviour is recorded through watching
- Correlation studies, where two naturally occuring variables are measured to establish if there is a relationship between them
- Case studies, where one person, group or organisation is studied in detail
- Content analyses, where secondary material is analysed in order to give insight into human thought or behaviour
Methods of Self Report
Methods of self report simply require participants to report on themselves. This is done by getting them to answer questions.
There are three key methods which allow psychologists to ask questions:
- Structured interviews
- Unstructured interviews
Questionnaries are made up of a list of pre-determined questions to which participants respond. Questions may focus on:
- Opinions. Eg, do you think first impressions are important?
- Past experiences, eg 'have you ever voted for something?
- Certain scenarios, eg, how would you respond if a teacher told you to stand out in the cold?
Questionnaries can be administered in a number of different ways, for example:
- Face-to-face in a private or public setting
- En masse to a group in a particular setting
- Through the post
- Via the internet
- Over the phone
Evaluation of Questionnaries
- Large numbers of questionnaries can be administered at once, making them more cost-efficient and less time consuming than interviews. It is also easier to reach a wide range of people if methods such as postal or internet surveys are used. However, if questionnaries are sent out, this relies on respondents returning them
- Response rates tend to be low, makin it difficult to generalise. Additionally, researchers often get a response bias (where respondents represent certain types of people, not others) as only certain people return questionnaires (like those with 'time on their hands') or they may be returned by people who are motivated to comment on a subject. This gives an unrepresentative sample
- Questionnaires are often completed privatley and can easily be made anonymous. This should give more honest (or valid) responses. However, it may lead to less honest responses as there are no researchers monitoring answers
- Another issue is that respondents may misunderstand or misinterpret questions without a researcher present. This, again, could lead to invalid responses.
Interviews involve the researcher directly asking participants questions and recording their responses. This is often done on a one-to-one basis
Structured interviews = these use pre-determined questions. The interviewer has already decided what they are going to ask about
Unstructured interviews = These may start with some common questions, but generally, the interviewer just has a topic that they want to cover. The interviews are like conversations, with a broad framework to guide discussion. They tend to be directed by the participant rather than the researcher. The questions are determined by the answers that the interviewee gives.
Comparing structured/unstructured interviews
- Structured interviews have set questions which make it easier to compare interviewee's answers. This makes it easier to identify patterns and trends in responses.
- However, becuase unstructured interviews do not have pre-set questions, the interviewer can follow new lines of enquiry. Interviewees may introduce relevant ideas that the researcher wouldn't have though to ask about
- Unstrucutured interviews allow interviewees to go into more depth, giving more valid results. It is likely to give researchers a clearer idea of their participant's view of the world. Critics of the structured interview argue that this is preferable to the researcher imposing their view of the world on participants through their questioning. However, if the researcher is in control, they're more likely to get info they want, rather than irrelevant information.
Open vs Closed questioning
Self report methods clearly rely on questioning. Questions can be open or closed.
Closed questions are questions where participants are offered a fixed set of responses to choose from. Common ways of closing questions are:
- yes/no responses
- Rating scales, eg agree 1 2 3 4 5 disagree
- Multiple choice, eg, Very often || Often || rarley
Open questions do not restrict responses. Participants are free to answer a question how they wish.
Open questions are clearly better for exploring answers in more depth, which may help researchers get closer to the truth. They may also allow researchers ot discover new lines of enquiry.
Closed questions make it easier to compare answers, and to identify patterns and trends. They will also help researchers stay focused on the aims of the investigation
This is a small scale, trial study.
Researchers may want to test their questions to ensure they are valid measures of the concept under investigation. It may simply be a case of ensuring questions are understood, or that all options are covered by a closed question.
When a researcher trials their question, this is a pilot study. It generally involves a smaller sample of people and sometimes just a sample of questions.
Pilot studies aren't just used for self-report methods. Psychologists can practise any part of any method. The general purpose of a pilot study is to identify any factors that might negativley affect the outcome of a study.
This saves the researcher from wasting time and money on a piece of research that could be unreliable.
Evaluation of self report
- Unlike observations, it's possible to access people's thoughts and feelings through asking questions
- Questions allow researchers to find out what people would do in certain situations without having to set them up (however, p's may lie)
- Methods of questioning need participants to possess a number of qualities to be reliable. They can be ineffective if participants are dishonest, inarticulate, lacking confidence, lack insight or have poor memory
- It is possible that p's responses are influenced by researchers when using interviews or questionnaries. Eg, when using face to face interviews, participants may feel pressured to give socially desirable responses, or where questions and possible answers are pre-set, this may lead to participants giving certain responses.
These may simply involve watching and recording people's behaviour. This may be done in a number of ways including:
- using a scoring system, such as rating a teacher's level of discipline
- Using a check list of criteria, like checking how many certain behaviours an autistic child displays
- Keeping a tally, like counting the number of times a doll is picked up by a girl or boy
- Making notes
- Video recording
When a psychologist decides to do an observation, there ar a number of decisions that they have to make about how they will carry out their observation:
- Should it be in a lab or natural setting?
- Should it be covert (undercover, people not aware being watched) or overt (open, people are aware) ?
- Should it be participant or non participant (participant is when the researcher observes people whilst joining in their activities or situation) ?
Laboratory vs Natural Observations
Observations can take place in a lab or natural setting
Observations that take place in a laboratory would basically be lab experiments. It is an observation in a controlled environment.
Observations that take placein a natural setting are sometimes called naturalistic observations (an observation taking place in a natural environment as opposed to a laboratory. These involve observing people in their natural environment - the behaviour observed is relativley unconstrained and people have a choice in how they behave.
- Doing an observation in a lab offers a high level of control which means it is easier to reliably establish cause and effect. BUT the artificial environment means that findings lack ecological validity
- Doing an observation in a natural environment offers a high level of ecological validity as people are being observed going about their usual behaviour in a situation which is not set up. This means any findings should be generslisable to real life. Observing without intervention means there are many uncontrolled variables making it difficult to draw any conclusions about causation.
Covert vs Overt observation
Observations can either be conducted covertly or overtly.
Covert observations describe observations where the psychologist observes an individual, group or situation without people being aware of this. Alternativley, a psychologist may observe from a hidden viewpoint.
Overt observations describe observations where the psychologist is open about their observation. They make their presence obvious and people know that their behaviour is being recorded
- Doing a covert observation means participants do not know they are being watched. This means they should behave as they normally would do, giving valid results. However, there are ethical concerns surrounding this type of research (consent)
- Doing an overt observation is more ethically sound as people are now aware they are being observed, but may withdraw themselves as a consequence. The most obvious limitation here is observer effect - people may behave differently because they are being observed, giving unreliable results
Participant vs non-participant observations
Participant observations are observations where the psychologist joins in with the group or situation they are observing whilst also recording data.
Non-participant observations are observations where the psychologist is not directly involved in what is being observed and records behaviour from a distance.
- Doing a participant observation allows the researcher to experience a situation as the participants do. This gives the researcher a greater insight into what is being studied, giving more valid results. However, a researcher may become too involved, losing their ability to be objective! There is also the practical problem of recording data whilst taking part in the study - even more difficult to do if it is covert.
- Doing non-participant observation allows the researcher to remain objective as they are not directly involved. However, they may not have a true understanding of behaviour if they are too removed from the situation!
Evaluation of observational studies
- Findings from observations are more reliable as the researchers can see for themselves how participants behave rather than relying on self-reports
- Most observations take place in a natural setting and so have high ecological validity. People are more likely to behave normally if it is a covert too.
- It is difficult to make judgements about thoughts and feelings when using this method, as these features are not clearly observable
- Observer bias can be a problem as the researcher may only percieve things from a certain perspective (but, using more than one observer increases inter-rater reliablity)
- If participants are aware they are being observered, then they may act differently, giving invalid results: the observer effect.
These describe a process rather than an actual method. They use methods such as self reort or observations to collect data, but it is how data is analysed which is important.
Correlation analyses (the analysis of data to test for a relationship between two variables) look for a relationship between two variables.
They can only be done on quantitative data (numerical) as it is essentially a statistical process. They rely on such data because they actually measure the strength and direction of the relationship between two variables. They do more than simply state if two variables are related.
This relationship can be shown graphically using a scattergram (a graph for representing correlations). And can also be measured by a correlation co-efficient (a number measuring the strength and direction of a correlation) which is always between +1 and -1.
Correlations can be broadly catergorised into positive, negative and zero correlations
Positive Correlation =
If two variables show this type, it means as one variable increases then so does the other. As one variable decreases so does the other. A perfect positive correlation has a co-efficient of +1. This occurs when two variables increase (and decrease) in exact relation to one another
Negative Correlation =
It still shows an actual relationship between two varibles. It means if one varible increases, the other decreases and vice versa. A perfect negative correlation has a co-efficient of -1. This occurs when one variable increases at exactly the same time as the other decreases
Zero Correlation =
There is no clear relationship between variables. A zero correlation has a co-efficient of 0. Thic occurs if there is absolutley no indication of a pattern between variables.
Evaluation of Correlational Studies
They're often mistaken for experiments because they both employ strict measures and statistical analysis. But correlation studies have a key limitation when compared to experiments as they cannot reliably establish cause and effect.
A correlation study does not have an IV and a DV. It simply measures the relationship between existing variables and nothing is set up. However, this does demonstrate the advantage of correlation studies: they do allow researchers to statistically analyse situations that could not be manipulated experimentally for ethical or practical reasons.
Because correlation studies take place after the event, it is difficult to reliably state that one thing causes another thing to happen. To establilsh cause and effect, the reseracher needs to monitor the experiment from the beginning.
Evaluation of Correlational Studies
- Correlations can establish the strength and direction of the relationship between variables
- They allow researchers to statistically analyse naturally occuring phenomenon which could not be set up ethically or practically
- Correlations cannot reliably establish cause and effect
- Variables have to be quantified, which means measures may lack construct validity
If a correlation shows a relationship between two variables, then one probably is affecting the other. But it is not necessarily clear which.
The lack of control over other variables also causes problems. It could be that other factors account for the relationship between two variables
Sometimes, variables are not really related at all, yet they appear to be - this is a coincidence.
Case studies have two main features
- They focus on a sample of either one individual, group or organisation
- They study that sample in depth
Case studies often use unstructured interviews, observation and past records to carry out an indepth analysis of the subject (term for a participants which is still sometimes used for case studies, since the individual is under investigation and are quite passive in the research rather than activley taking part). Since samples are small, it should be possible to find the time to do this.
Case studies are often used to investigate atypical behaviour, or unusual situations. Studying what happens 'when things go wrong' can give insight into normal patterns of behaviour. This includes investigations into the effect of brain damage on memory, effect of the nazi party etc. Each case depicts a rare event, and do not affect many individuals. When they do happen, they may should psychologists how they stop the holocaust happening again etc.
Case studies do not always have to focus on unusual subjects. Sometimes they're used because a psychologist wants more insight into a particular individual.
Evaluation of Case Studies
- Offer high levels of validity as they go in depth and give insight
- They allow researchers to study events that they could not practically or ethically manipulate
- Case studies are efficient as it only takes one case study to disprove a theory
- Based on small samples, so difficult to generalise
- Researcher can become too involved and lose their objectivity. They misinterpret or influence outcomes
- Since case studies are often 'picked up' after an event (eg, after someone has suffered brain damage) it can be difficult to establish cause and effect!
Content analyses are different from other research methods in the sense that they are a process where people are studied indirectly rather than directly. It is the process of studying secondary material, such as material other people produce, than investigating actual people themselves.
For example, exploring media output or examining graffiti in public toilets
Content analysis can be quantatative which means data is collected numerically. This may involve counting the occurence of a particular feature, or coding material (operationalising variables for analysis) in some way
For example, keeping a tally, ratings or percentages on certain material.
Content analysis can be qualitative which means data collected is more descriptive. For example, deconstructing themes, uncovering thinking behind letters, identifying unconscious thoughts in diaries etc etc
Evaluation of Content Analyses
- Allows researchers to study people they have little or no access to
- They also have few ethical issues as there is little or no contact wiht participants
- Since researchers have little direct contact with participants, their thoughts and behaviour could be misinterpreted
- Qualitative content analyses are particularly open to interpretation as they're mainly based on opinion (useful to have a number of researchers to come to agreement)
- Quantatative content analyses are more objective because coding systems are used. These may lack construct validity.
(construct validity = ability of a measurement tool to actually measure the subject)
Qualitative vs Quantitative
Content analyses show how research can be both qualitative and quantitative. It is a method that can produce descriptive (qualitative) data and numerical (quantatative) data.
Other methods can also produce both types of data, whilst others may be more one or the other.
Quantatative data strengths - Can be easily summarised into graphs or statistics, so the researchers can identify patterns and trends. More objective as scoring systems are not open to interpretation.
Qualitative data strengths - has more construct validity. Describing behaviour (rather than scoring it) is what hapens in real life. It is also richer and more detailed, giving more depth and insight into a subject.
Data produced by different methods
Experiment = Mainly quantatative as DV has to be measured to establish cause and effect. To measure a DV, the researcher often needs to quantify it.
Self Report = If closed questions used, this can produce quantitative data. Open questions tend to produce qualitative data as responses tend to be individualised in detail.
Observation = If structured observations are carried out, then quantitative data can be produced. When researcheres simply describe what they see, this produces more qualitative.
Correlation = Has to produce quantitative data so that a correlational analysis can be performed. This is a statistical process, so can only be applied to scores
Case Studies = Generally produce qualitative since they are in-depth and detailed investigations
NOT a research method - psychologists need to decide who they are collecting the data from
Population - before deciding who to collect data from, psychologists need to identify who they are generally intersted in investigating. Who they want to generalise their findings to. This is known as the target population
In many investigations, the target population is the human population as psychologists want to comment on people generally. In other investigations, the target population may be more specific.
Unless a target population is very small, it is impossible for a psychologist to study everybody they want to apply their findings to. Even when the target population is small, not everyone in that population will necessarily be identifiable or accessible. It is often enough for a psychologist to investigate a sample of a population. The ideal is to obtain a sample of people that are representative of a population. If a sample is representative, it reduces the need to investigate everybody, because all necessary views or behaviours will be represented by that group.
Sometimes samples may not represent the population that well. If a sample is not representative, and there are more certain types of people than others then it can be described as a biased sample (a sample that is not representative of the target population).
It would seem that a larger sample is a more representative sample. Logically, the more of the population that is in the sample, more representative? Not always true. The size of the sample is not as important as the composition (may have 50 asian people!) The composition of the sample partly depends on the technique used to choose the participants from the target population. First a sampling frame (a section of the target population from which the sample is drawn), then a number of technqiues can be used to select from the sampling frame.
Made up of participants chosen mathematically, using chance. Every person in the sampling frame has an equal chance of being selected from the sample. This is equivilent to choosing 'names out of a hat.'
- It avoids bias as researcher has no control over who is selected
- The law of probability says that the researcher will normally get a representative sample
- There is a chance, of a 'freak' sample, that would not be representative
- Compared to other methods, it is time consuming because all potential participants have to be identified in the first place before the 'draw' can be made
Made up by participants chosen mathematically. This is done by taking every 'nth' person in the sampling frame for the sample. Very similar to random sampling but, theoretically is not random. This is becuase each person in the sampling frame does not have the same chance of being selected as the researcher will have decided they will take every 3rd, 5th etc person.
- It avoids bias as they have no control over who is selected
- The law of probability says that the researcher will normally get a representative sample
- There is a chance of a 'freak' sample
- It is not as objective as random sampling, as the researcher may decide on how people are listed before selection and what number is used for the 'selection'
Made up of participants who have been selected after the sampling frame has been stratified or layered.
This means the sampling frame is divided into roups that the researcher wants to make sure are represented in the final sample. A certain number are selected from these groups so they are proportionatley reprsented
- Avoids problem of 'freak' samples, more or less guaranteeing a representative sample by making sure all characterisitcs are present
- It is relativley objective as once the sampling frame is stratified, it is normally left to chance who is selected from each strata
- The researcher may not identify all the key characteristics for stratification, meaning the sample is still not representative
A sample made up of participants who have been chosen as they are convenient. They may have been selected as they have activley voulenteered to be chosen, as they are in the locality or known to the researcher. The researcher uses anyone they have the oppurtunity to use.
- Less time consuming than other technqiues as time isn't spent planning and using sophisticated systems for selection
- Likely that the sample will be biased as only certain types of people will volunteer to be chosen. Similarly, if people come from the same locality, they may be similar in terms of their characteristics, or if they are acquaintances of the researcher, they may share similar characteristics to them (ie, middle class)
- The researcher may show bias when selecting participants, whether intentional or not.
It is not enough to simply select participants for a study. Psychologists have to consider the psychological well-being, health, values and dignitiy of their paticipants. If psychologists do not deal with their participants correctly, their work is said to be 'unethical.' This may lead to psychologists being stopped from practising psychology.
Researchers should seek to make their investigations as ethical as possible. Various organisations produce codes of ethics - in the UK, most psychologists follow the Code of Conduct and Ethical Guidelines, produced by th British Psychological Society (BPS).
- Protection from harm
Researchers should avoid unfair or prejudiced practice. They should respect people as individuals, taking account of factors such as gender, race, culture and religion.
The risk of harm should be no greater than in everyday life. If a researcher is unsure whether a risk (eg, deception, distress) is justifiable, it is recomended that they consult with individuals that are similar to the participants (socially/culturally) and ask them about how they would feel if they were a participant to the planned investigation.
The right to withdraw should be made clear to participants at the start of the research. If children avoid a situation it should be taken as their need to withdraw.
Researchers should deal with any negative effects of an investigation at the end of the research, and should be avaliable to help deal with any long term consequences.
If a researcher finds evidence of any psychological or physical problems which the participant is unaware of. In such case, the researcher has a responsibility to inform the participant if they believe that by not doing so the participant's future and well-being may be at risk. In these cases, the researcher can offer advice, or if not qualified, should recommend an appropriate professional.
Data collected on a participant should be kept confidential, so that others cannot identify it as theirs (Eg, names should not be used). If confidentiality cannot be ensured (eg, if a participant discloses something illegal, if there is a threat to someone's safety) then particpants must be warned in advance of investigation.
Audio, video and photographic recordings should only be made with consent
Participants should give informed consent (consent based on awareness of the aims of an investigation). Participants under the age of 16 should give their own consent where possible, but consent of parents or other guardians (eg, teachers) should be sought as well.
Adults with impairments should give their own consent to participate in research where possible, but the researcher should also consult with a person who is well-placed to appreciate the participant's reaction (Eg, family member)
Detained persons must have the freedom to give their own consent. Researchers shouldn't pressurise people into consent to an investigation, nor use payment to encourage them to agree to do something they wouldn't normally choose to do without payment.
It is possible to carry out an investigation without consent in situations where people would expect to be observed by strangers (eg, a night club). However, researchers need to be wary about observing people who may believe they are unobserved, even in a public area (eg, toilet)
Privacy and Protection from Harm
Be careful in situations where people do not believe they are being observed (Eg toilet) - privacy
Protection from harm = the risk of harm should be no greater than everyday life
Psychologists should avoid deception (misleading participants through lies or withholding information)
Participants should not be misled about the nature of the investigation. However, th BPS does understand that in some circumstances, research may not be valid if participants know that they are being studied or know why they are being studied.
Researchers should therefore ensure participants are informed as early as possible, that they are being studied or why they are being studied.
- Researchers should make sure participants have fully understood the true nature of an investigation
- Even if a participant has completed the investigation, the have the right to withdraw their own data or have it destroyed - this may be particularly important for participants who realise they have been decieved once debriefed
- Researchers should ask the participants about their experience of the research in order to check for any negative effects or misunderstandings that may not have been expected
- Researchers may have to use counselling as part of their briefing to ensure that participants leave in a healthy state of mind (especially if they have been decieved) or that they at least leave in the same state as they entered.
Debriefing does not necessarily exuse an unethical piece of research
Right to Withdraw
Participants have the right to withdraw (ability not to continue with an investiagtion) from the research at any time during an investigation, for whatever reason.
If non-human animals are used as part of their research, there is a number of ethical guidelines that must be stuck to:
- researchers should choose a species which is scientifically suitable for an area of study, and which will suffer the least from an investigation. Researchers should use the smallest number of animals that is possible
- Researchers should avoid procedures which cause pain, suffering, distress or lasting harm. Where such procedures are used, they need to be fully justified and require a Project Licence.
- Where animals are kept in captivity for research purposes, researchers should ensure they are well housed and cared for. Normal feeding and breeding habits should not be disrupted
- Researchers are encouraged to investigate free-living animals as much as possible, rather than studying them under controlled conditions. When this happens, researchers should try to interfere with the natural habitat as little as possible.
Despite the guidelines, research that breaches ethical guidelines may still be carreid out. Deception is a relativley common in conformity experiments where psychologists want to test genuine behaviour, or in memory experiments etc.
In turn, this may cause distress, discomfort or embarrassment for the participant.
Unethical research CAN be carried out if the psychologist can show the means justify the ends. If the research benefits wider society, it may be carried out - a large number of people at the expense of a small minority.
This may become even less of an issue if the participants are treated ethically after the event - eg, if they are counselled, ensured privacy and given the right to withdraw data.
When research is qualitative, data is sometimes presented in its original format (eg, a transcript of an interview) or is summarised (eg, a synopsis of what happened).
Quantatative data is easier to summarise as there are various tools for doing this. Graphs offer one way of summarising or representing data.
Data can also be represented in a summary table. This is different from a table fo raw scores as it presents descriptive data, rather tan individual scores. This is normally enough to see a pattern
Bar graphs are used to represent data which is divided into categories. This is sometimes known as discrete data (divided into catergories). Examples of discrete data:
- Yes or no questions
- Identifying a person as extroverted or introverted
On a bar graph:
- Each bar represents a different category. These are listed along the horizontal (x) axis
- The frequency each category is chosen or occurs is measured up the vertical (y) axis, and is shown by the height of the bar
- Bars should be drawn SEPERATLEY to indicate that each category is seperate and discrete
- If a category is not chosen or observed, it should be still represented on a bar graph to show readers that its frequency is ZERO
Line graphs are used to represent data which is in numerical form. This is sometimes known as continuous data. Examples include:
- Number of words recalled in a memory experiement
- The age of a person
On a line graph:
- The scale of measurement would be placed along the horizontal (x) axis
- The frequency of each score is measured up the vertical (y) axis and is represented by a point on the graph
- When all points are plotted, a line is drawn between each of them, showing data is continous.
A line graph can be used to make estimation about the frequency of scores between those of the graph.
An alternative to line graphs as they also represent continous data. Bars represent each score rather than a point. Bars are drawn TOUCHING to show data is continous.
Histograms are useful when there is a large range of data. Rather than representing every score on a scale on the graph, they can be grouped and represented by one bar.
Each group of scores is known as a CLASS INTERVAL and this gives the width of the bar. A histograph should have around 6-8 bars
Line graphs do have their advantages over histograms, however:
- A line can be used to make estimaitions
- It is easier to visually represent and compare two or more sets of data on a line graph, whereas compound histograms ( a histogram representing two or more sets of data) are generally more difficult to read
Used to represent relationships between variables. This is different from the other graphs, which are used to display differences.
Both variables need to be measured using a numerical scale. One is represented on the Y and one on the X axis. It doesn't matter which way round this is because it's measuring a relationship, not a cause and effect.
Descriptive data is quantitative data that has been analysed to show patterns. This includes data presented in graphs.
Quantitative data can also be presented in a table: it is conventional to present a table that summarises data rather than one that contains all of the raw scores. Generally, people don’t want to see evidence of all of a psychologist’s research. They want to see main trends. Measures of central tendency are useful.
Measures of central tendency describe a data set by identifying on score that represents the general trend of that data – it is a measure of average. This score will tend to be somewhere ‘central’ to other scores.
Data can be summarised using one of three measures of central tendency, they are: The mode, median, mean.
The mode identifies the most common score in a data set
The median is when data is organised numerically, and locating the middle score. If there is two middle scores, it is the middle between these.
The mean is it added all up, and divided by the number of scores
All scores, including anonymous scores, affect the mean. The median and mode are not
The fact that the mean uses all the scores has strengths and weaknesses. Some argue even extreme scores should be accounted as they are equally important – it gives a valid measure.
However, other researchers argue extreme scores make measures unreliable as they misrepresent the true tendency of a data set, by skewing it so it is too high or low.
Some dislike the mean as it gives a decimal score (eg, 5.1 people) which is meaningless. The median is normally one of the scores in data set and mode is always.
The main issue with the mode is that it relies on there being a score that occurs more than other.
However, sometimes each score in data occurs only one, so there is no mode as such.
There may be two or more scores that occur equally frequently, giving a number of averages. These can be at either end of a range of data – do not really help in measuring central tendency.
Measures of Dispersion
If people are interested in how spread out a set of scores are, they can take a measure of dispersion, as they want to assess whether it is reasonable to compare two groups.
The Range - Offers a measure of dispersion and is normally used in conjunction with the median. The range analyses what is happening at either end. In a repeated measures, a researcher may expect the range of scores to be very similar across conditions as same participants are being used.
The Standard Deviation - This is a measure of dispersion that calculates how much each score deviates from the mean. It calculates the spread of scores. This is why it is often used in conjunction with the mean. When scores deviate a lot, it gives a high S.D. It gives useful info that takes us beyond the mean.