Key Research Methods in Psychology

---

Why This Matters

Research methods are the backbone of psychology as a science—they're how we move from "I think this is true about human behavior" to "Here's the evidence." On the AP Psychology exam, you're being tested on your ability to evaluate research designs, identify their strengths and limitations, and recognize which method is appropriate for different research questions. Understanding these methods connects directly to every unit in the course, from how we study biological bases of behavior to how researchers measure attitudes, learning, and cognitive processes.

The key concepts you need to master include experimental control, causation versus correlation, internal and external validity, and ethical considerations. When you encounter a research scenario on the exam—whether in multiple choice or an FRQ—you need to quickly identify the method being used and evaluate what conclusions can (and cannot) be drawn. Don't just memorize definitions; know why researchers choose each method and what trade-offs come with that choice.

---

Establishing Cause and Effect: Experimental Methods

Experiments are the gold standard when researchers want to determine whether one variable actually causes changes in another. The key mechanism is manipulation and control—by changing only one thing and holding everything else constant, researchers can isolate cause-and-effect relationships.

Experimental Design

• Random assignment distributes participants into groups by chance, ensuring pre-existing differences don't contaminate results—this is what allows causal conclusions
• Control groups provide a baseline for comparison against the experimental group, isolating the effect of the independent variable
• Operational definitions specify exactly how variables are measured, making the study replicable and allowing other researchers to verify findings

Variables in Research

• Independent variables (IV) are what the researcher manipulates—think of it as the input or cause you're testing
• Dependent variables (DV) are the measured outcomes—the output that may change in response to the IV
• Confounding variables are uncontrolled factors that vary systematically with the IV, threatening internal validity by offering alternative explanations for results

---

Measuring Relationships Without Manipulation: Correlational and Descriptive Methods

When experiments aren't possible—due to ethical concerns, practical limitations, or the nature of the research question—psychologists turn to methods that observe and measure without manipulating variables. These methods can identify patterns and relationships but cannot establish causation.

Correlational Studies

• Correlation coefficients range from $-1.0$ to $+1.0$, indicating the strength and direction of a relationship between two variables
• Correlation does not equal causation—a relationship between variables could be explained by a third variable or reverse causality
• Predictive value is the main strength; correlations help psychologists anticipate outcomes even when they can't explain why the relationship exists

Naturalistic Observation

• Ecological validity is high because behavior is observed in real-world settings without artificial laboratory constraints
• Observer bias occurs when researchers' expectations influence what they notice or record—addressed through blind observation protocols
• Lack of control over extraneous variables means researchers cannot rule out alternative explanations for observed behaviors

Case Studies

• In-depth analysis of rare or unique cases (like Phineas Gage or H.M.) provides rich qualitative data impossible to obtain through other methods
• Hypothesis generation is a key function—detailed case observations often inspire larger-scale studies
• Limited generalizability means findings from one individual may not apply to the broader population

---

Gathering Data at Scale: Surveys and Sampling

Surveys allow researchers to collect information from large groups efficiently, but the quality of survey research depends entirely on how questions are designed and who gets asked. Sampling methods determine whether findings can be generalized beyond the study participants.

Surveys and Questionnaires

• Self-report measures collect data on attitudes, beliefs, and behaviors directly from participants—used extensively in research on implicit attitudes and stereotype threat
• Social desirability bias occurs when respondents answer in ways they think are more acceptable rather than truthfully
• Question wording effects can dramatically alter responses; leading questions or ambiguous terms compromise data quality

Sampling Methods

• Random sampling gives every member of the population an equal chance of selection, maximizing generalizability to the broader population
• Stratified sampling divides the population into subgroups and samples from each, ensuring representation of key demographics
• Convenience sampling selects whoever is easily available (like college students), which is efficient but limits external validity

---

Studying Change Over Time: Developmental Research Designs

Understanding how behavior and mental processes develop requires tracking change—either by following the same people over time or by comparing different groups at a single moment. Each approach offers different insights and comes with different methodological challenges.

Longitudinal Studies

• Same participants tracked over time allows researchers to observe actual developmental changes and establish temporal sequences
• Attrition (participant dropout) threatens validity, especially if those who leave differ systematically from those who stay
• Cohort effects are controlled because all participants share the same generational experiences—but findings may not generalize to other cohorts

Cross-Sectional Studies

• Different age groups measured simultaneously provides a snapshot comparison that's faster and cheaper than longitudinal designs
• Cohort effects confound age effects—differences between groups might reflect generational experiences rather than developmental change
• No individual change data means researchers can identify age-related patterns but cannot track how specific individuals develop

Twin Studies

• Identical vs. fraternal twin comparisons help separate genetic from environmental influences on behavior and mental processes
• Heritability estimates are calculated by comparing trait similarity between identical twins (who share 100% of genes) and fraternal twins (who share about 50%)
• Shared environment limitations mean that even identical twins raised apart may have more similar environments than assumed, complicating nature-nurture conclusions

---

Synthesizing and Evaluating Research: Meta-Analysis and Quality Standards

Individual studies provide pieces of evidence, but psychology advances by combining findings across many studies and holding all research to rigorous quality standards. Reliability and validity are the criteria by which we judge whether research findings are trustworthy and meaningful.

Meta-Analysis

• Statistical combination of multiple studies increases statistical power and reveals patterns that single studies might miss
• Effect size calculations quantify how large or meaningful an effect is across the research literature
• Garbage in, garbage out—the validity of a meta-analysis depends entirely on the quality of the included studies

Reliability and Validity

• Reliability refers to consistency—whether a measure produces the same results across time (test-retest), items (internal consistency), or raters (inter-rater reliability)
• Validity refers to accuracy—whether a study measures what it claims to measure (construct validity) and whether findings apply beyond the study (external validity)
• Internal validity is the degree to which a study establishes a causal relationship, while external validity is the degree to which findings generalize to other populations and settings

Statistical Analysis

• Descriptive statistics summarize data (mean, median, standard deviation), while inferential statistics test whether findings are likely due to chance
• Statistical significance (typically $p < .05$) indicates that results are unlikely to have occurred by chance alone
• Effect size complements significance by indicating the practical magnitude of findings—a result can be statistically significant but trivially small

---

Protecting Participants: Ethics in Research

Ethical guidelines exist because psychology's history includes studies that harmed participants—from Milgram's obedience experiments to the Stanford Prison Study. Modern research must balance the pursuit of knowledge with the protection of human welfare.

Ethics in Psychological Research

• Informed consent requires that participants understand the study's purpose, procedures, risks, and their right to withdraw at any time without penalty
• Institutional Review Boards (IRBs) evaluate research proposals before studies begin, ensuring ethical standards are met and risks are minimized
• Debriefing occurs after participation, especially when deception was used, to explain the true purpose and address any concerns or harm

---

Quick Reference Table

---

Self-Check Questions

1. A researcher finds a strong positive correlation between hours of sleep and exam performance. Why can't she conclude that more sleep causes better performance? What third variable might explain this relationship?

2. Which two research methods would be most appropriate for studying a rare psychological condition, and what are the trade-offs of each approach?

3. Compare random assignment and random sampling: Which protects internal validity, which protects external validity, and why does this distinction matter for interpreting research findings?

4. A developmental psychologist wants to understand how memory changes from childhood to old age. Compare the advantages and disadvantages of using a longitudinal versus cross-sectional design for this research question.

5. A study finds statistically significant results ($p < .05$) but a very small effect size. What does this tell you about the findings, and why do psychologists report both significance and effect size?