Methods (Edexcel A-Level Psychology): Revision Notes
Experiments
Introduction to experiments in cognitive psychology
Memory is difficult to observe or accurately measure using self-report methods. Cognitive psychologists therefore use experiments to objectively quantify the capacity and duration of each memory store. Traditionally, laboratory experiments have been the primary research method in memory research. However, field experiments are increasingly being used to understand memory in more everyday contexts.
An experiment is an investigation where a variable is manipulated or altered and its effect can be measured, whilst maintaining control over other variables that might interfere with this situation.
Experiments 'set up' a situation where participants are required to perform a task and the performance of this task is measured. The extent to which this task reflects real life or is conducted in a realistic situation depends on the type of experiment being conducted.
Laboratory experiments
A laboratory experiment is an experiment conducted in a controlled environment. Removing participants from their natural environment eliminates the potential for extraneous variables affecting their behaviour. Exposing the objective truth by stripping away the context ensures a human characteristic can be studied in an objective and value-free way. An artificial context provides the researcher with the level of control over relevant variables necessary to achieve a more scientific approach and ensure causality.
Laboratory experiments have high levels of control and standardisation, meaning studies can be repeated to check the findings are consistent. However, they may lack validity because they do not reflect real-life behaviour. Participants are often invited to take part in laboratory research, so they are aware of their participation which can lead to demand characteristics and expectancy effects. The presence of the researcher during the experiment may influence the behaviour or performance of participants, so experimenter effects are more likely to have an influence on the results.
Key Trade-off in Laboratory Experiments:
Laboratory experiments provide excellent control and the ability to establish causality, but this comes at the cost of ecological validity. The artificial setting may not reflect how participants would behave in real-world situations.
Field experiments
A field experiment is conducted in a natural environment where the independent variable is manipulated and the dependent variable measured. Participants are tested where they would normally display the behaviour being studied; this may be a classroom, supermarket or high street, and they may not be aware that they are taking part in an experiment at all. This means field experiments have greater ecological validity as participant behaviour will be more natural and the environment in which they are tested is more realistic.
If participants are unaware of their participation in the experiment they will not show demand characteristics. However, because the research is not conducted in a controlled environment, there is greater chance of extraneous variables having an effect on the dependent variable. There may also be ethical problems if a participant is unaware that they are taking part, as they have not given consent and do not have a right to withdraw from the experiment. In such cases, the experimenter may choose to debrief them after the experiment and offer them the right to withdraw their data from the study.
Balancing Control and Realism:
Field experiments offer a middle ground between complete control (laboratory) and complete naturalism (naturalistic observation). Researchers must carefully consider which approach best suits their research question.
Aim
The aim of an experiment is a general statement about what area or topic is being researched. An aim typically begins with "To investigate...". The aim is a concise and to-the-point statement that directs the overall ambition of the study.
Example aims:
- To investigate whether the type of food given to cats affects their purring
- To investigate whether praise affects the time children spend washing dishes
The aim is an important part of a research investigation because it clearly signposts the topic being investigated. An aim should be as clear and precise as a hypothesis. Once you have read about hypotheses and operationalisation, it is important that you apply the same detail and accuracy to an aim.
Hypotheses
Experiments begin with a prediction of what is likely to happen in the investigation based on previous knowledge, research or theory. This prediction of a likely outcome is known as the experimental hypothesis. An experimental hypothesis is a type of alternative hypothesis. An experimental hypothesis is a clear and precise statement predicting the results of the experiment.
Sometimes we can be certain of the outcome of an experiment because, perhaps, there is strong evidence or research to suggest the outcome may happen or we have based our prediction on a robust theory. In such cases a directional hypothesis will be predicted. When we are not certain of the outcome of an experiment, because there are conflicting theories or a lack of relevant evidence, a non-directional hypothesis will be predicted. A non-directional hypothesis predicts that a difference or relationship will be found, but not the direction that the difference or relationship will take.
Examples of experimental hypotheses
Directional hypothesis:
- Cats will purr for longer when they are fed tinned food compared to dry food.
- Children will spend longer washing dishes the more praise they receive.
Non-directional hypothesis:
- There will be a difference in the length of time a cat purrs when given tinned and dry food.
- Praise will affect the time children spend washing dishes.
Choosing Between Directional and Non-Directional:
Use a directional hypothesis when previous research or strong theory supports a specific outcome. Use a non-directional hypothesis when research is inconclusive or theories conflict about the expected direction of results.
Null hypothesis
A null hypothesis is a default prediction that is supported if there is a greater likelihood of the results occurring by chance. When we conduct research we often find some difference or relationship; it is rare that we would find nothing, but sometimes the difference or relationship found is too small or insignificant to be due to anything other than chance variation. For example, if we are investigating whether praise affects a child's inclination to tidy their bedroom, it is unlikely that we will find no/zero effect of praise on bedroom tidying. However, the change observed in bedroom tidying may be too small or insignificant to be due to praise alone and could be due to chance.
Example null hypotheses:
- There will be no difference in the length of time cats spend purring when fed tinned or dry food. Any difference found will be due to chance factors.
- There will be no effect of praise on the time children spend washing dishes. Any effect found will be due to chance factors.
Independent and dependent variables
An experiment always has an independent variable and dependent variable. The independent variable (IV) is the variable that is manipulated or changed by the researcher in order to demonstrate a difference between the experimental conditions. The dependent variable (DV) is the variable that is measured or the result of the experiment. The dependent variable measures any changes that occur because of the independent variable. This allows causality to be established (cause and effect).
Example: IVs and DVs
'A researcher wished to investigate whether participants will recall more words from an organised list compared to a random list'
In this example, the researcher will have to change which list participants have to learn and recall from. This is manipulated by giving one set of randomised words and one set of organised words. The type of word list is the IV.
The researcher will then ask participants to recall the list of words and record how many words they remember. This is the measured variable or outcome of the investigation, so is referred to as the DV.
Operationalisation
Once the IV and DV have been decided, it is very important to make these variables precise and specific by operationalising them. This means deciding exactly how you are going to manipulate the IV and exactly how the DV will be recorded. Operationalisation of the IV and DV means that the study can be precisely replicated to check the conclusions are reliable. Operational definitions of the DV can increase objectivity in research; this is because the outcome is measured in the same way by all researchers, and the outcome is not open to interpretation. It also means that other psychologists can assess whether or not the researcher has conducted valid research.
Why Operationalisation Matters:
Without clear operational definitions, different researchers may measure variables differently, making it impossible to replicate studies or compare findings. Operationalisation is essential for scientific reliability and validity.
Good and poor operationalisation
Poor operationalisation:
A researcher thought that children who came to school without a healthy breakfast had problems during literacy hour. The researcher decided to ask the children what they had for breakfast and split them into healthy and unhealthy breakfast groups. She then watched them read a book and decided how well they could read.
This is an example of poor operationalisation because the way in which the healthy and unhealthy breakfast groups are defined is unclear. It is also not clear how the researcher measured reading skill. A study such as this example would be difficult to replicate exactly to check for reliable findings. If more than one researcher was involved in the research, it would not be clear what is meant by reading skill, so they may reach different conclusions for the same child. A different researcher would not be able to assess how healthy breakfasts were defined or how reading skill was defined, so could not be certain that the study was valid.
Good operationalisation:
A researcher thought that children who came to school without a healthy breakfast had problems during literacy hour. The researcher asked the children and parents to make a record of what they ate for breakfast over the course of a week. A nutritionist was asked to categorise the breakfasts as health and unhealthy. Breakfasts with over the recommended meal allowance for salt, fat and sugar were defined as unhealthy. The researcher then timed how long a child took to read a story during literacy hour. All children read the same story out loud to the researcher, who timed the children and recorded any errors they made.
This is an example of good operationalisation because the IV (healthy and unhealthy breakfasts) are clearly defined and the DV (reading speed and errors) can be measured exactly without any ambiguity. This study is replicable and it would be easy for a different researcher to assess whether the definition of healthy and unhealthy breakfasts and reading ability were measuring what was intended.
Experimental variables
Extraneous and confounding variables
An experimenter should try to establish control over factors that may have an unwanted effect on the dependent variable. These other variables are known as extraneous variables. Sometimes an extraneous variable can influence the dependent variable and make it look as though the effect was from the independent variable; this is called a confounding variable. This variable confounds the results of the study in such a way that you are no longer measuring the effect of the IV on the DV.
Extraneous and confounding variables can be divided into two types: situational variables and participant variables.
Situational variables
An extraneous variable that might affect the results of a study could be found in the environment in which the study is conducted. Situational factors such as lighting, noise, temperature, other people, disturbances, time of day, etc., may all affect the results of a study so should be controlled or eliminated. Controlling extraneous variables means that they are held constant for all participants, so that the variable affects everyone equally. Eliminating extraneous variables involves removing the possibility of them occurring in the first place.
Participant variables
Participants themselves may affect the results of the study. Participants may bring different characteristics to an experiment that could have an effect on the dependent variable, such as level of motivation, personality, intelligence, experience, age and skills. It is fairly easy to control participant variables such as age and gender, but controlling motivation or experience may take more thought.
It is not necessary, and would certainly be far too time-consuming, to control for all situational and participant variables. It is only really necessary to control those variables that might have an unwanted impact on the dependent variable. For example, controlling the temperature of a room is not vital unless you are testing something where the temperature might affect performance.
Careful control
When considering control, it is common to suggest that noise, temperature and lighting generally might affect the dependent variable without carefully considering whether they are relevant extraneous variables. In extreme, these variables could affect research, but they would have minimal influence on the dependent variable in most studies. It is important to consider the nature and aims of the investigation before you decide which extraneous variables are likely to have an effect.
Experimenter effects
Experimenter or researcher effects refer to the way an experimenter may influence the outcome of an experiment by their actions or mere presence. These may be subtle cues that may influence the way a participant responds in an experimental situation. Sometimes these can be obvious effects, such as a female researcher asking a male participant about his attitudes towards gender equality, or a young researcher asking an older participant what they think about youth culture. However, some experimenter effects are more subtle.
Hawthorne effect
The Hawthorne Effect is one such example where the mere presence of a researcher can have an effect on performance.
Demand Characteristics and Expectancy Effects:
Closely related to experimenter effects is the concept of demand characteristics. This is when the effect of the experimenter causes participants to alter their behaviour to meet the expectations (whether real or imagined) of the experimenter.
Rosenthal researched this expectancy effect across many decades. Rosenthal found that psychology graduates who were told one set of rats were brighter than another set of rats resulted in the bright rats being able to learn their way out of a maze faster than the dull rats. With no actual difference between the two sets of rats, Rosenthal concluded that the students may have treated the rats differently, pressed their stopwatch earlier or reported false findings as a result of expectancy effects.
Experimenter effects may explain why a researcher finds a result that other researchers fail to replicate.
Experimental control
In experiments using human participants, a great many variables can influence outcomes. It is important to be able to identify these variables and then put into place controls to help prevent them having any effect on the experiment. Various control techniques have been established to help deal with these control issues.
Standardisation
Standardisation refers to making an experiment the same experience for all participants. Standardised instructions are a set of instructions given to all participants that can be used to eliminate experimenter effects because it removes the potential for the experimenter to give verbal or non-verbal cues to participants. Standardised procedures (stages of the experiment, timings, apparatus, etc.) ensure that all participants are treated in the same way (other than the change in condition due to the independent variable) so there is no variation in the way they experience the research that may affect the way they behave. Standardisation also improves the replicability of the experiment.
Double- and single-blind experiments
To control for demand characteristics, participants may be unaware that they are part of an experiment, or may have been deceived as to the true nature of the study. This is known as a single-blind procedure, where the participants are unaware of the study aim so it does not influence how they perform. To eliminate experimenter effects, independent researchers who are not told the aim of the study may be employed by an experimenter to conduct the study on their behalf. If neither the participant nor researcher knows the aim of the study, it is referred to as a double-blind procedure.
Single-blind vs Double-blind:
- Single-blind: Only participants are unaware of the study aims
- Double-blind: Both participants and researchers conducting the study are unaware of the aims
Experimental design
Once the independent variable has been operationally defined, the levels of the independent variable can be identified and the conditions of the experiment established.
Identifying levels of the IV and conditions of the experiment
For example: The effect of music (IV) on transcription speed and accuracy (DV)
Levels of the IV: Rock music or silence
Conditions of the experiment: Whether participants hear rock music while trying to transcribe verbally dictated information or transcribe in silence
The conditions of the experiment reflect directly the levels of the independent variable. More levels of the IV can be added, for example, classical music or popular music and therefore there will be more conditions involved in the experiment.
Often one level of the independent variable is a control group, which receives no treatment. In the above example, the control group is the group that transcribes in silence. It is used to have a control group as a baseline comparison to determine the effect of the IV on the DV.
Participants recruited to take part will need to be allocated to one or both conditions of the experiment. There are several designs that can be used to achieve this: an independent groups design, a repeated measures design and a matched pairs design.
Independent groups design
This is when the participants are divided into groups and are only involved in one of the experimental conditions of the experiment. A strength of this experimental design is that participants are less likely to guess the aim of the investigation as they only take part in one level of the independent variable. They do not get to know about the other conditions. This means that the chance of demand characteristics or expectancy effects is somewhat reduced. However, it does mean recruiting twice as many participants because you need separate groups and there may be individual differences or participant variables between the participants in each group that make a comparison of the groups unreliable.
One way of controlling for individual differences is to randomly allocate participants to one or other of the conditions. Random allocation means that it is probable, but not certain, that there will be an even distribution of participant variation because they all have an equal chance of being selected for each condition of the experiment.
Repeated measures design
This is when all participants take part in all conditions of the experiment. This resolves the problem of individual differences because the same participants are in all levels of the independent variable, so the same participant results in one condition can be compared with the same participant in a different condition. Fewer participants are needed for a repeated measures design, because they are used twice, so it is more economical than an independent groups design. However, the chance of participants displaying demand characteristics is greatly increased because they have knowledge of all conditions of the study, and are therefore more likely to be able to guess the aim of the study.
There is also a problem of order effects; this is when the performance of participants in one condition is influenced by the previous condition of the experiment. Order effects include practice and fatigue; a participant may learn the task in the first condition so perform better in a second condition, or become tired and performance declines in a second condition.
Controlling for Order Effects:
Order effects are a significant challenge in repeated measures designs. Without proper controls like randomisation or counterbalancing, practice or fatigue effects can confound your results, making it impossible to determine if the IV truly caused the observed changes.
Controlling for order effects
One way of controlling for the effect of demand characteristics is to use a single-blind technique. To control for order effects, randomisation or counterbalancing can be used to ensure participants experience the conditions in a different order. Randomisation involves selecting at random which of the conditions of the experiment a participant does first. This can be done by picking a card out of a hat.
Counterbalancing involves the participants being placed into either a group that does Condition A then Condition B, or a group that does Condition B then Condition A. However, if the order effect (practice or fatigue) in one sequence order (AB) is not equivalent to the order effect in a different order (BA), a more complex counterbalancing technique may be required. The ABBA design is used to balance unsymmetrical order effects by getting participants to complete the conditions twice – A, B, B then A. The mean score for both conditions A and conditions B are then taken.
When there are more than two conditions in an experiment, a Latin square can be used to designate participants to one of the combinations of ordering. This means that, although order effects still occur, they are balanced out between each group.
A simpler way to overcome order effects is to leave a time gap between participants completing condition A and condition B. The effects of fatigue are likely to be reduced, although the same may not be true of practice effects depending on what the task is.
Matched pairs design
To overcome the problems associated with repeated measures and independent groups designs, a matched pairs design can be used. This is when different participants are assigned to each condition of the experiment (similar to independent groups) but they are matched on characteristics important to the study. These characteristics are often established by pre-testing and researching the lives and backgrounds of all the participants. The pre-testing ensures that the participants in each condition can be compared fairly. This can be achieved by matching all participants on important characteristics and then randomly assigning them to each condition. It is important to match participants on characteristics central to the aim of the study; it would not be useful to match participants on hair colour, for example, in a study of driving ability, the matching would have to concern driving experience, eyesight, reaction time, or other characteristics where any variation could affect the results.
A matched pairs design ensures that the conditions can be compared more reliably and that any difference found between the results of each condition is more likely to be due to the manipulated variable, so causation can be established. However, a matched pairs design is time-consuming and many participants have to be excluded from the study because they do not meet the matching criteria. It is also very difficult to match participants on all possible characteristics that could have an effect on the dependent variable. For example, if a study was conducted into the effect of an unhealthy breakfast on reading ability, it would be useful to match participants' educational level and eyesight. However, there may be variables that are much more difficult to match, such as how many books a child has at home, the educational level of parents, how much time parents spend reading with their children, etc. Therefore a matched pairs design cannot be truly matched on all possible variables.
Comparing Experimental Designs:
Each design has trade-offs:
- Independent groups: Simple but requires more participants and vulnerable to individual differences
- Repeated measures: Economical and controls for individual differences but prone to order effects
- Matched pairs: Combines benefits of both but is time-consuming and difficult to implement perfectly
Reliability
Reliability refers to the consistency of findings from research, and it is an important criterion for being scientific. For experiments, test-retest reliability is important.
Test-retest reliability
If findings are consistent, and can be considered reliable, we can trust that the finding will happen again and again. In order to achieve reliability, research must be replicable. This requires very tight control of extraneous variables that, if not controlled, could result in different findings when a study is repeated.
Validity
Validity refers to whether the study is measuring the behaviour or construct it intends to measure. Understanding validity is an important skill for both designing and evaluating research studies. There are two broad categories of validity: internal validity and external validity. Internal validity refers to how well the procedure of a study establishes a causal relationship between the manipulated independent variable and the measured dependent variable, or whether it has been confounded by uncontrolled extraneous variables.
Internal validity can be ensured by using standardised procedures, controlling for order effects and individual differences, and avoiding demand characteristics.
Assessing internal validity
A way of assessing internal validity is by examining construct validity. Construct validity is how well the measure of a behaviour being used is a useful indicator of what is supposed to be studied. For example, recall of a previously learned list of words may not be a useful measure of episodic memory because a participant may draw on semantic memory and make a good guess. If you are measuring what you intend to be measuring, then another way to assess internal validity is through predictive validity; the extent to which the performance on the measure can predict future performance on a similar criterion. For example, if a test of intelligence can accurately predict future academic success, then it has predictive validity.
External validity
External validity refers to how well research findings study can be generalised beyond the study itself, to other situations or other people. There are two main types of external validity; ecological and population validity.
Ecological validity refers to the extent to which the research can be generalised to other situations, for example real-life or everyday situations. Memory experiments conducted in artificial environments with artificial tasks may not be generalised to everyday use of memory.
Population validity refers to the extent to which research findings apply to other populations than those used as the sample. External validity can be improved by ensuring that the sample is representative of the population it intends to represent, and by making the context of the study as realistic as possible.
Internal vs External Validity:
- Internal validity: Does the study measure what it claims to measure within the study itself?
- External validity: Can the findings be generalised beyond the specific study context and participants?
Objectivity
Being objective refers to the need to be impartial and judgement free. It is important that the dependent variable is measured objectively, so that the opinions or judgements of the researcher do not affect how the dependent variable is recorded. For example, imagine that you are asked to guess the length of a table. Your judgement will be based on your own opinion or belief about length, and will probably differ from the guesses of others. Your guesses and those of others are subjective and therefore unlikely to be either reliable or valid. However, if you use a ruler to measure the length of the table, your recorded answer is objective, and will be exactly the same as others who measure the same table using the same ruler. This is an objective measure of the table length, and therefore will be reliable and valid.
Objectivity in Cognitive Psychology:
Cognitive psychology studies concepts, such as memory, that cannot be directly observed and measured. Cognitive psychologists would agree that we cannot objectively measure mental processes, but we can objectively observe the data produced by experiments and neuroimaging techniques.
If we conduct a short-term memory test that records a participant recall of five words, this is an objective measurement of short-term memory. If we use a PET scan to observe brain functioning during an experiment, we can objectively observe regions of the brain that are active during the task.
WIDER ISSUES AND DEBATES
Psychology as a science
The laboratory experiment is considered to be the most scientific of the research methods that psychologists can use. It is characterised by a high level of control and standardisation. This means that the study can be repeated to check the findings are consistent. As such, laboratory experiments are highly reliable. However, they may lack validity because they do not reflect real-life behaviour.
Key Points to Remember:
-
Experiments manipulate an independent variable (IV) and measure a dependent variable (DV) whilst controlling extraneous variables to establish causality.
-
Operationalisation means defining variables precisely so they can be directly tested and replicated, increasing objectivity and reliability.
-
Hypotheses can be directional (predict direction of change), non-directional (predict change but not direction), or null (predict no difference/change due to chance).
-
Experimental designs include independent groups (different participants per condition), repeated measures (same participants in all conditions - controlling for order effects is essential), and matched pairs (participants matched on key characteristics).
-
Laboratory experiments have high control and can establish causality but may lack ecological validity, whilst field experiments have greater ecological validity but less control over extraneous variables.