Types of research - experiments, observations, case studies, correlations etc.

Big Research Methods post coming up. This shall cover the two types of research (experimental and non-experimental), as well as the purpose of pilot studies.

Black: Description
Red: Negative evaluation of a design or technique.
Blue: Positive evaluation of a design or technique.
Purple: My notes/hints/tips.

IV: Independent variable, the variable that changes.
DV: Dependent variable, the variable that is measured.
EV: Extraneous variables, factors that could affect the DV but that aren't being studied.

Experimental Methods

A laboratory experiment takes under controlled conditions, with participants aware that they are being studied. The independent variable is manipulated to look for an effect on the dependent variable, and extraneous variables are controlled.

The high levels of extraneous variable control increase the internal validity of the results, the degree to which the study measures that which it claims to - as we can be more  confident that factors other than the IV aren't influencing our results.

This helps to determine cause and effect and to establish a potential causal relationship between the IV and the DV.

The standardisation of procedure, instructions and apparatus gives this experimental method more replicability, allowing us to test for reliability and consistency by carrying out repeats at other times or on different samples from the target population.

Demand characteristics are more likely to occur in a laboratory experiment - being fully aware that they are in an experiment and an artificially designed setting, participants could alter their behaviour to either help or confound the research. This unnatural behaviour reduces the internal validity of the results.

The use of an artificial, controlled, and often unrealistic setting reduces the ecological validity of the results - the extent to which a study's results can be applied to a real life situation.

A field experiment takes place in a natural environment, with the participant usually unaware that they are taking part. The experimenter still manipulates the IV to look for an effect on the DV, but has less control over EVs than in a laboratory experiment.

Field experiments have more ecological validity than laboratory experiments, the use of a natural setting meaning that measured behaviour is more realistic and representative of people's behaviour in their everyday environment. 

Demand characteristics are less likely to occur in a field experiment, as due to the natural setting and potentially being unaware of the IV manipulation, participants are less likely to modify their behaviour to help or hinder the research, an action which would reduce the validity of the results.

Less control over extraneous variables makes the establishment of cause and effect more difficult and reduce the internal validity of the results - without the control of a lab experiment, it is difficult to know if the observed change in the DV is a result of IV manipulation, or a result extraneous variables. This makes establishing a causal relationship hard.

In a natural experiment, a natural setting is used, extraneous variables are not controlled, and the experimenter takes advantage of a naturally occurring change in the IV, rather than directly manipulating it.

Natural experiments allow for studying the effects of the IV when directly manipulating the IV would be too difficult or unethical.

Natural experiments have the highest external validity, being set in a natural environment without deliberate experimental manipulation of the IV. As a result of this, their results most accurately reflect real-world behaviour and attitudes, having the highest ecological validity of the three types of experiment.

The lack of direct IV manipulation limits the ability of the researchers to establish cause and effect, making it difficult to determine a direct causal relationship between the independent and the dependent variables.

Natural experiments are usually impossible to replicate due to the highly specific conditions, relying on a natural variation in the IV to carry out the research. This limits the ability to test for the reliability and consistency of the results that could otherwise be achieved through replication.

Non-Experimental Methods

Correlational research involves the examining of a relationship between two covariables (variables being investigated - correlations do NOT use an IV and a DV - both variables change and are measured, and neither are directly set by the researcher.)

Rather than being a research method in and of itself, correlational research collects and analyses data from many other research methods - for example, observations, questionnaires, interviews and experiments. 

The relationship between two covariables can be a positive correlation (as one increases, so does the other), a negative correlation (as one increases, the other decreases) or neither - no correlation can potentially exist.

Correlational research allows us to study areas that would be unethical to study by manipulation of variables - e.g, the relationship between police funding and levels of violent crime.

It allows us to study based off already known information, and attempt to draw a correlation between that and newly acquired data.

Correlations do not necessarily indicate direct causation or the direction of the relationship - a correlation between two variables does not mean one causes the other, and even if it does, it does not tell us which one causes the other - the direction of causation is unknown.

We can never rule out a third variable that explains the relationship between the covariables - if causation exists, it is not always direct, but often a result of a third, "intervening" variable.

Observations as a research method involve watching behaviour to look for patterns - often using a predefined set of behavioural categories and seeing how often behaviours from those categories occur. They can be participant or non-participant, covert or overt, and controlled or naturalistic.

Controlled observations determine duration and a list of behaviours to observe, rate and categorise first, whereas naturalistic observations are more "seeing what happens" - with no set behaviour ratings or categories.

Covert observations are carried out without the knowledge or consent of those being studied, in a situation where they would not reasonably expect to be observed. There are ethical issues with the lack of consent here, but "participants" will behave more naturally as they are unaware that they are being studied - meaning results have higher ecological validity.

In an overt observation, the participants have consented to being observed, in a situation where they would not reasonably expect to be. This is more ethically sound than a covert observation, as participants have given informed consent to be studied, but are likely to change their behaviour due to demand characteristics as they are aware of the researchers observing them, and could try to either help or hinder the research or engage in more socially desirable behaviour - reducing the validity of results.

Participant observations are where the researcher takes part in the activity being observed, becoming a member of the group. This allows them to more deeply understand behaviours and dynamics that may be a part of the group's culture and confusing to outsiders, but they may affect their own results by influencing the group that they're observing.

In non-participant observations, the researcher is observing the group without being a part of it. They may misinterpret behaviour or dynamics, being unable to fully understand them from an outsider's perspective, but will not affect their own results by influencing the behaviour of the group they're observing.

Observations are fairly subjective and unscientific - two researchers might interpret the same behaviour in completely different ways. To alleviate this, a method of testing for inter-rater reliability is used - raters observing the behaviour agree on a rating and categorisation system before carrying out their observation, so the same behaviour will be interpreted in the same way consistently between different observers.

Interviews are a verbal method of gathering data, where a participant will be asked a series of questions by a researcher and their responses will be recorded. They can be structured, unstructured or semi-structured.

Structured interviews have a set question list, which the researcher does not deviate from.

The researcher can make sure that they'll address the most important and desirable issues, while avoiding too much irrelevant information.

It is easier to compare responses between participants when using a structured interview, as all participants are asked the same questions.

However, there is no room for participants to elaborate upon and to explore any important points that arise - they are more restricted and can give less in-depth information.

The interviewee has very little control over the progression of the interview - this may make it feel more like an interrogation, and make them more reluctant to give answers.

Unstructured interviews have a topic of discussion, but no set questions - letting the conversation take a natural route.

The interviewee has more control over the direction of conversation, potentially being more willing to give information.

There is room for elaboration and for exploration of areas of interest, meaning that a more in-depth reply can be given if a topic of particular interest surfaces.

It is difficult to compare answers between participants, since not everyone will have been asked the same questions due to the natural flow of conversation.

Lots of irrelevant information can be given over the course of an unstructured interview, wasting both the participant and the researcher's time.

The researcher cannot be sure that all topics of interest will be discussed.

A semi-structured interviews has a topic of discussion and a number of key points to address, but lacks the direct questions used in a structured interview and the free flow of conversation used in an unstructured interview. It functions as a balance between the other two styles.

Interviews tend to give us rich, qualitative data that gives us in-depth insight into people's views and beliefs. 

However, this data is difficult to quantify - being difficult to statistically analyse, compare and look for trends.

Generally, interviews suffer from social desirability bias as people feel pressured to give socially acceptable answers that don't reflect their true beliefs and behaviours - this reduces the validity of the research.

Open and closed questions can be used - open questions give rich, detailed and qualitative responses that are difficult to analyse and compare, whereas closed questions give objective, shallow and quantitative data that can be numerically analysed with relative ease. A Likert scale can be used for quantifiability in a questionnaire or interview - presenting a statement, and asking the participant to indicate on a numerical scale how much they agree or disagree.

Questionnaires can use open or closed questions, and involve the participant responding to a range of questions about a particular topic of interest.

Especially through use of the internet, questionnaires can reach a huge range of people very quickly - this makes them easy and quick to analyse results and draw conclusions from a large sample size.

However, questionnaires have a poor response rate, and suffer from self-selection bias - biased towards the specific type of person with a helpful temperament that is more likely to reply to questionnaires. This makes them unrepresentative and lacking in validity, not representing a range of temperaments and personalities. Generally, only people with a strong opinion either way reply - so researchers could lack insight into the views of the middle ground.

Questionnaires measure opinion rather than behaviour - reducing validity by not accurately reflecting the way people behave, but rather the way they consider themselves to feel. They can also suffer from social desirability bias, as participants might be likely to give the answers that present them in the best light, rather than the most truthful answers.

Case studies are in-depth studies of an individual or a small group who are particularly of interest in a certain area of research, often over an extended period of time. They use a range of other research methods to gather data into the participants, including interviews, observation, questionnaires and content analysis of diaries.

They provide in-depth, qualitative and unique insight into the lives of psychologically significant individuals, studying things that otherwise be too obscure to study in the general population, or would be unethical to induce in order to study.

They are often longitudinal, allowing the researcher to study change over a long period of time - such as the progression or degeneration of a specific condition. The longitudinal nature allows a trusting relationship to build up, making the participant more willing to give sensitive information.

However, as a trust relationship builds up between the researcher and case subject, the study loses objectivity as the researcher seeks to present them in a certain light, meaning results start to lack internal validity.

Case studies' longitudinal nature gives them high levels of participant attrition - becoming unrepresentative through self-selection, biased towards the kind of participant who doesn't drop out.


Content analysis involves analysing the results of questionnaires, interviews, experimental results, speech, literature and other methods of data collection to look for meaning and to identify trends. Discourse analysis is a sub-type that involves specific focus on speech patterns and conversations. Content analysis uses quantitative research methods to analyse qualitative data and make inferences into the media, the writers, the audience, the culture and time of the content. Content is broken down into manageable categories on a variety of levels such as words, word sense, phrase, sentence and theme - then quantitatively examined.


This method allows us to quantify data that would otherwise be qualitative, and far less scientific.


Content analysis uses pre-existing information to draw conclusions, meaning that it is very open to measures of inter-rater reliability - the same source can be analysed by multiple researchers.


However, if inter-rater reliability is not used, then this a very subjective research method, as researchers could come to completely different conclusions about the same piece of content.

Pilot Studies

Pilot studies involve carrying out the proposed research on a small group within the target population before carrying it out on the main sample, for a number of reasons.

• Identifying any issues with the instructions and research - such as confusing or ambiguous tasks, and offensive or irrelevant questions.

• Checking all raters are using rating scales or content analysis techniques correctly, agreeing on what constitutes a certain behaviour - to increase inter-rater reliability.

• To check the suitability of the stimuli - tasks are not too difficult or skillful for participants to carry out, nor too boring that they lose concentration and drop out halfway through.