Research Methods
Fundamentals of Hypothesis and Variable Definition
Research in psychology begins with the formulation of a theory, which serves as a suggested explanation for a specific behavior. From this theory, a researcher develops an aim, which is a general statement explaining the purpose of the study. This purpose is then tested using specific variables. The Independent Variable (IV) is the factor changed or manipulated by the experimenter to observe its effects. It can have various conditions, which refer to the number of specific changes made to the variable. The Dependent Variable (DV) is the factor that the researcher measures to determine the outcome of the experiment.
To ensure scientific rigor, researchers must operationalize their statements, which means being specific and precise about what is being studied. For example, in the statement "Students who have 8 hours sleep before an exam will perform better on test scores than students who have 4 hours of sleep before the exam," the independent variable is the number of hours of sleep (8 hours vs. 4 hours) and the dependent variable is the performance on test scores. To operationalize this, a researcher might state: "Students who sleep between 10pm and 6am for the 8 hours before an exam at 9am will have higher raw marks than students who have 4 hours of sleep between 2am and 6am before the exam."
A hypothesis is a formal statement that describes the relationship between the variables being investigated. When formulating a hypothesis, it must be clear and precise, stating the relationship between the variables. Specifically, the independent variable must change at least twice, and the dependent variable must be clearly identified. Hypotheses typically follow the format "There is a difference…" or "There is no difference…"
Types of Hypotheses
There are two primary types of hypotheses used in research: the Alternative Hypothesis and the Null Hypothesis. An Alternative Hypothesis states that there is a relationship, correlation, or difference between variables. For example, "There is a difference in the number of balls thrown in a bucket by participants performing with an audience of 30 people or performing the task alone." Conversely, a Null Hypothesis states that there is no relationship, correlation, or difference. An example would be, "There is no difference in the number of balls thrown in a bucket by participants performing with an audience of 30 people or performing alone."
Extraneous Variables and Control Measures
Extraneous variables (EV) are any variables other than the independent variable that could potentially affect the dependent variable. These must be controlled to establish a clear cause-and-effect relationship in an experiment. For instance, if a researcher is studying the effect of diet (IV) on weight/mass (DV), other factors like exercise, water intake, and fatigue act as extraneous variables. Other common examples include age, alcohol consumption, mental illness, the type of exam, the time of the exam, the baseline ability of the participant, wake/sleep times, and the total time allowed for an exam.
Extraneous variables are categorized into three types. Situational variables are aspects of the environment, such as lighting, noise, or temperature. Participant variables are individual characteristics of the subjects, including mood, concentration, skill level, and fatigue. Experimenter or Investigator effects are aspects of the experimenter’s behavior that might influence the results, such as friendliness or accidentally giving more instructions to one group than another. To improve a hypothesis like "A psychologist wants to see if participants work better if they've had a drink of coffee," one could state: "There is a difference between participants' percentage test scores if they have a cup of caffeinated coffee 1 hour before their exam starting at 10am than participants who don't have a drink of coffee 1 hour before the test."
Controls for these variables are essential for validity. These include using participants of the same age, prohibiting alcohol consumption for 2 days before a test, setting strict sleep and wake times (e.g., 10pm to 8am), and ensuring that the control group only drinks water if they are in the "no coffee" group. Researchers may also use IQ tests to screen participants, excluding those whose scores are too high or over a certain threshold to maintain a baseline level of ability.
Standardisation and Randomisation
Standardisation involve using standardized procedures, which means providing the exact same methods and information to every participant. This is designed to remove experimenter effects from the data. These procedures are written carefully before the experiment begins and are either read from a script or provided as a handout. A script is a set of instructions that participants follow. Every participant should receive the same information at the start, identical instructions throughout, be tested in the same environment, and be informed that they have the right to withdraw or quit at any time.
Randomisation is the process of using chance to control bias when creating a research study, thereby stopping experimenter bias. An example of this is tossing a coin to decide group placement. In experiments involving the Primary Recency Effect—where people recalling words from a list tend to remember only the first and last words while forgetting the middle—randomisation can be used to shuffle the word list. This ensures the experimenter does not accidentally or intentionally put easier words at the beginning or end of the list, which would act as an extraneous variable.
Sampling Techniques in Research
The target population is the large group of people about whom the researcher wants to make a statement. Because it is impossible to test an entire population (e.g., 600,000 people), a smaller group called a sample (e.g., 1,000 people) is used. A representative sample is one that accurately reflects the characteristics of the target population.
Random Sampling gives every member of the target population an equal chance of being selected, such as using a random name generator or picking names out of a hat. While it is representative and avoids bias, it is time-consuming, not guaranteed to be perfectly representative, and small minority groups (e.g., blind people) may distort results. Opportunity Sampling involves targeting participants who are accessible and willing to take part. This is easy and inexpensive but often unrepresentative and biased as participants may share similar characteristics.
Systematic Sampling involves selecting every person from a list (e.g., every 4th person). It is unbiased if the list is already randomized and is generally representative; however, if the original list was not randomized, the sample could be biased (e.g., all male or all female). Stratified Sampling divides the target group into sections or "strata" based on key characteristics that should be present in the final sample. This avoids misinterpretation caused by random sampling and ensures characteristics are present proportionally. However, it requires more time and resources to plan and may be biased if a key characteristic is missed across a strata.
Volunteer Sampling occurs when an advertisement is placed (e.g., in a newspaper) and people choose to participate. While easy for the researcher, it is usually unrepresentative because it is self-selecting. Most sampling methods eventually become volunteer samples because individuals have the right to say no to participating.
Experimental Design Frameworks
Experimental design refers to how participants are allocated to the different conditions of the independent variable. In an Independent Groups design, separate groups are used for each condition. This design avoids order effects (where the sequence of tasks affects the outcome) and practice effects (where doing a task multiple times improves performance). However, it is susceptible to participant variables, where one group might naturally be more skilled than another, lowering validity. To deal with this, researchers use systematic allocation (placing people in groups based on their arrival order) or random allocation (tossing a coin).
Repeated Measures design involves participants taking part in all conditions of the experiment. This eliminates participant variables and requires fewer participants, lowering costs. However, it suffers from order effects, practice effects, and demand characteristics (where participants try to guess the study's purpose and change their behavior to please the experimenter). To handle order effects, researchers use counterbalancing, where half the sample performs the control condition first while the other half performs the experimental condition first, then they swap.
Matched Pairs design is a "halfway house" between the other two. Participants only complete one condition, but they are paired based on a mini-test (e.g., matching the two highest scorers, then the next two). These pairs are then split between the experimental and control groups. While it minimizes participant variables and avoids order effects, it is very time-consuming and there is no perfect way to match people.
Types of Scientific Experiments
Experiments can be classified by their setting and how the independent variable is handled. A Lab Experiment is conducted in a controlled environment (like a classroom) where the experimenter changes the IV. It allows for high control over extraneous variables and standardized procedures but may lack validity because people know they are being tested and the setting is not like everyday life. A Field Experiment is conducted in a natural setting where the experimenter still changes the IV. It is more realistic and participants are often unaware they are being studied, increasing validity, though the researcher may lose control over some extraneous variables.
A Natural Experiment occurs in a natural setting where the IV is not changed by the experimenter but changes naturally (e.g., gender or weather). These have high validity and can follow standardized procedures, but they rely on rare natural events, are time-consuming to wait for, and are prone to extraneous variables due to unique participant characteristics that cannot be controlled via standard sampling.
Ethical Considerations in Research
Ethical issues are guided by the BPS (British Psychological Society) code. Informed Consent requires that participants (or parents/guardians for those under 16) are given information about the purpose of the study so they can decide whether to take part. They must be told they can leave at any time. Deception involves lying to participants about the aims; while mild deception is sometimes acceptable for adults, a full debrief must be given at the end. Protection from Harm ensures physical and psychological safety remains the same before and after the experiment. Privacy involves the right to control information about oneself, and Confidentiality ensures personal data is protected and participants remain anonymous (e.g., using numbers or letters instead of names). If distress occurs, researchers may offer counseling during the debrief.
Data Collection: Interviews and Questionnaires
Interviews and questionnaires are used to measure variables through self-reporting. A major concern is the Leading Question, which can change a person's thinking or give too much information away, leading to bias. This is avoided by using a strict structure. Interviews can be Structured (pre-planned questions), Unstructured (a chat with few pre-planned questions), or Semi-structured (pre-planned questions with allowed follow-ups). They provide in-depth qualitative data and can reveal unexpected findings, but are difficult to analyze and subject to social desirability bias.
Questionnaires use a pre-prepared list of questions. Open questions provide qualitative data without fixed answers, while Closed questions provide quantitative data with fixed options (e.g., yes/no). Questionnaires allow for large amounts of information to be collected quickly and are easier to analyze for generalizations. However, they are prone to social desirability bias, questions may be misleading, and they can take months or years to prepare properly.
Observational Research and Case Studies
Observations record behavior as it occurs. Naturalistic observations happen in normal environments with no changes, while Controlled observations involve controlled elements. Covert observations happen without the participant's knowledge, whereas Overt observations involve informing the participants in advance. In Participant observation, the researcher joins the group, while in Non-participant observation, they remain separate. To ensure reliability, researchers use Categories of Behaviour (e.g., breaking down "flirting" into laughing, touching, mirroring, eye contact, and smiling). Interobserver Reliability is achieved when different observers compare their data and find a strong correlation; if results differ, categories must be amended.
A Case Study is a detailed investigation of an individual, group, event, or institution, often focusing on unusual occurrences. It provides qualitative data and is often longitudinal, taking place over a long period. Case history involves collecting data from the past to follow a subject's development. Case studies are highly valid and keep the researcher open-minded due to a lack of a specific prior aim.
Quantitative versus Qualitative Data
Qualitative data is non-numerical (e.g., case studies, open questionnaires) and is valued for being in-depth and detailed, though it is harder to analyze. Quantitative data is numerical (e.g., closed questionnaires, structured interviews) and is easy to analyze and draw conclusions from, although it lacks depth. Primary data is collected first-hand by the researcher to suit specific aims, making it more relevant but expensive and time-consuming. Secondary data is collected by someone else (e.g., websites); it is convenient and inexpensive but may not be as relevant or valid for the specific study.
Correlations and Descriptive Statistics
Correlations are used to investigate complex relationships and provide a starting point for future research. They are useful when experimental manipulation is unethical (e.g., you cannot give children heroin, so you must find existing addicts). However, correlations do not establish cause and effect, and intervening variables (extraneous factors like lifestyle or diet) can affect the co-variables.
Descriptive statistics summarize data. The Range is the highest value minus the lowest value, plus 1: . The Mean is the average, calculated by adding all scores and dividing by the total count; it is the most accurate but distorted by extreme scores. The Median is the middle value when scores are ordered from lowest to highest; it is not affected by extreme values but loses information about the rest of the data. The Mode is the most common number; it is easy to calculate but can be unrepresentative.
Data Interpretation and Display
Frequency tables help identify patterns. For a dataset of scores , the mean is calculated by finding the total sum of and dividing by the total frequency. Given a total score of and a frequency of , the mean is .
Data can be displayed in Bar Charts, which use gaps between bars for discrete data, or Histograms, which have no gaps to represent continuous data. On a graph, the IV is usually on the x-axis and the DV or frequency is on the y-axis. A Normal Distribution (Bell Curve) is a symmetrical histogram where the mean, mode, and median are all at the same central point; the curve never touches the x-axis. When writing a conclusion, one must determine the trend, use specific numerical examples from the data, and provide an explanation for the findings.