2.1 Experimental and Correlational Methods

Experimental and Correlational Methods

Experimental and correlational methods are the two main approaches social psychologists use to study human behavior. Experiments manipulate variables to establish cause-and-effect relationships, while correlational studies examine naturally occurring associations between variables. Understanding when and why researchers choose each method is central to evaluating any study you'll encounter in this course.

Experimental Design

An experiment is the only research method that can establish causation. The researcher deliberately changes something, holds everything else constant, and measures what happens. That level of control is what gives experiments their power.

Key Components of Experimental Design

• Independent variable (IV): The variable the researcher manipulates. In a study on aggression, the IV might be whether participants play a violent or nonviolent video game.
• Dependent variable (DV): The variable the researcher measures to see if the manipulation had an effect. In that same study, the DV might be the number of aggressive words a participant uses afterward.
• Control group: Receives no treatment or a neutral version of the treatment. This group serves as the baseline for comparison.
• Experimental group: Receives the actual treatment or manipulation being studied.
• Random assignment: Each participant has an equal chance of being placed in any condition. This is what allows researchers to rule out pre-existing differences between groups. Without it, you can't confidently say the IV caused changes in the DV.

Validity in Experimental Research

Internal validity refers to how confident you can be that changes in the DV were actually caused by the IV, not by something else.

• Controlled environments minimize the influence of extraneous factors
• Standardized procedures keep the experience consistent across all participants
• Confound elimination is the goal: if something other than the IV could explain the results, internal validity drops

External validity refers to how well the findings generalize beyond the specific study.

• A representative sample increases the applicability of results to broader populations
• Ecological validity is a specific type of external validity that asks whether the experimental setting resembles real-life conditions. A study conducted in a sterile lab with artificial tasks may have high internal validity but low ecological validity.

There's often a tradeoff between the two: tightly controlled lab experiments boost internal validity but can feel artificial, while more naturalistic designs improve external validity but introduce more uncontrolled variables.

Experimental Design Considerations

• Counterbalancing reduces order effects by varying the sequence of conditions. If participants complete Task A before Task B, fatigue or practice could skew results. Counterbalancing has some participants do A then B, and others do B then A.
• Double-blind studies prevent both the researcher and the participant from knowing who is in which condition. This eliminates demand characteristics (participants changing behavior because they guess the hypothesis) and experimenter bias.
• Factorial designs examine the effects of two or more independent variables simultaneously, including how they interact. For example, a 2×2 design might test the effects of both sleep deprivation and caffeine on social judgment.
• Repeated measures designs use the same participants across all conditions, which eliminates individual differences as a confound. The downside is that order effects become a concern, which is where counterbalancing comes in.

Correlational Studies

When researchers can't or shouldn't manipulate a variable, they turn to correlational methods. You can't randomly assign people to experience childhood trauma or have a certain personality type, but you can measure those variables and examine how they relate to other outcomes.

Understanding Correlation

The correlation coefficient ($r$) measures the strength and direction of a linear relationship between two variables.

• $r$ ranges from $-1$ to $+1$. A value of $0$ indicates no linear relationship.
• Positive correlation: Both variables move in the same direction. As one increases, the other tends to increase too. Height and weight are positively correlated.
• Negative correlation: The variables move in opposite directions. More hours spent studying tends to correlate with lower test anxiety.
• Strength: The closer $r$ is to $-1$ or $+1$, the stronger the relationship. An $r$ of $0.85$ is a strong positive correlation; an $r$ of $-0.30$ is a weak negative correlation.

Correlation does not imply causation. This is probably the single most repeated phrase in research methods, and for good reason. Ice cream sales and crime rates are positively correlated, but ice cream doesn't cause crime. A third variable, hot weather, drives both. That's the core problem with correlational data: you can't tell why two things are related just from the fact that they are.

Challenges in Correlational Research

• Confounding variables influence both measured variables and can create the appearance of a relationship that doesn't actually exist (a spurious correlation) or hide a real one. Researchers use statistical controls to account for known confounds, but unknown confounds remain a threat.
• Third-variable problem: An unmeasured variable may be the real reason two other variables appear related. The ice cream and crime example above is a classic case. Identifying and ruling out third variables requires careful thinking about alternative explanations.
• Directionality problem: Even when two variables are genuinely related, correlational data can't tell you which one influences the other. Does low self-esteem lead to social media overuse, or does social media overuse lower self-esteem? Longitudinal studies and cross-lagged panel designs can help clarify the direction by measuring both variables at multiple time points.

Applications and Limitations of Correlational Studies

Correlational studies are valuable when variables can't be ethically or practically manipulated, such as personality traits, socioeconomic status, or traumatic experiences. They allow researchers to study naturally occurring relationships in real-world settings, which often gives them stronger ecological validity than lab experiments.

The main limitation is the inability to establish causation due to the lack of experimental control. Still, correlational findings play an important role in generating hypotheses that can later be tested experimentally, and they can reveal patterns across large populations that would be impossible to study in a lab.