Sitonce
Country: US
Show exams for United States Hong Kong
Sign in

Psychology Research Methods: Praxis Psychology 5391

Updated 10 min read
Key takeaway

Psychology research methods turn questions about behavior and mental processes into testable evidence.

  • Define variables, choose a suitable design, consider sampling and ethics, and evaluate measurement.
  • Random sampling differs from random assignment; correlation shows association, while an experiment may support causal inference when its controls justify that conclusion.
On this page14 sections
  1. How to use this topic
  2. Worked distinction
  3. Apply the concept
  4. Evidence and limits
  5. Practice the reasoning
  6. Identify variables before judging conclusions
  7. Choose a design that fits the question
  8. Separate sampling, assignment and generalization
  9. Statistics summarize rather than certify
  10. Ethics are built into the design
  11. Operational definitions and construct validity
  12. Replication and cumulative evidence
  13. Ethics and interpretation together
  14. A practical validity checklist

How to use this topic

Psychology research methods translate questions about behavior and mental processes into evidence that can be observed, measured, and evaluated. Praxis 5391 expects candidates to understand psychology as an empirical science, explain scientific methods, choose appropriate designs, identify variables, use basic descriptive statistics, consider ethics, and evaluate assessment. A method is useful when it fits the question and supports the inference being made. A polished graph cannot repair a design that did not measure the stated concept.

Praxis domain
Methods, Approaches, Ethics and Assessment
Exam format
Selected response
Use
Apply concepts to a new example

Start with a research question and hypothesis. A hypothesis is a testable prediction, not a guess without evidence. An operational definition specifies how a concept will be measured or manipulated. For a study of test anxiety, one operational definition might be a score on a defined questionnaire before an exam; another might be a physiological measure. Each captures a different aspect, so the researcher should explain why the measure fits the question and acknowledge what it leaves out.

Variables describe features that can differ. In an experiment, the independent variable is the condition manipulated or compared, while the dependent variable is the measured outcome. Control variables are held steady or accounted for to reduce alternative explanations. Suppose researchers compare retrieval practice with rereading and measure later recall. Study method is the manipulated condition, and number of correctly recalled terms is the outcome. If the groups have different study time, prior knowledge or instructor support, those differences may confound the comparison.

Worked distinction

Random assignment and random sampling are not interchangeable. Random sampling selects people from a population to participate and can improve representativeness. Random assignment allocates participants to conditions and can help distribute preexisting differences across groups. A study can use one without the other. A convenience sample randomly assigned to two conditions may support a causal comparison within that sample while limiting generalization. A random sample observed without manipulation may be representative but not establish causation.

Experiments manipulate conditions and measure outcomes; correlational designs measure variables as they naturally occur. A correlation can be positive, negative, or near zero and describes a pattern of association. It cannot by itself show that one variable caused the other because a third factor or reverse direction may explain the relationship. A survey asks participants for information; an observation records behavior; a case study examines one person or small unit in depth. Each method answers different questions and has distinct limitations.

Longitudinal designs follow the same participants or cohort over time, while cross-sectional designs compare groups at one point or within a shorter period. Longitudinal research can show change within individuals but may face attrition and time costs. Cross-sectional work is faster but can confound age with cohort differences. A question about development should identify which design suits the question and what alternative explanation remains.

Apply the concept

Basic descriptive statistics summarize data. The mean is the arithmetic average; the median is the middle value after ordering; the mode is the most frequent value. A high outlier can pull the mean more than the median. For values 2, 3, 3, 4, and 18, the mean is 6, median 3, and mode 3. The three measures describe different features. Choose one based on the distribution and question rather than assuming the mean always represents a typical score.

Standard deviation describes spread around the mean; correlation describes how two variables co-vary. A high standard deviation means observations vary more within a dataset, not that the measurement is invalid. Correlation is often represented by a coefficient from -1 to +1. A coefficient near zero indicates little linear association, but nonlinear relationships can exist. Do not infer causation or practical importance from a coefficient alone; sample, measurement, context and design matter.

Research ethics include informed consent, confidentiality, protection from harm, and review by an ethics committee or institutional review board as appropriate. Consent should communicate the study's nature and relevant risks in language participants can understand. Participants should know whether participation is voluntary and what happens if they decline or withdraw, subject to applicable exceptions and rules. Confidentiality means protecting identity and sensitive information; it is not the same as promising absolute secrecy in every circumstance.

Evidence and limits

Historical research contributed to the development of research safeguards. When studying a famous experiment, distinguish what the original researchers did from modern ethical standards and later commentary. A study may be historically important and ethically troubling at the same time. Avoid using a dramatic example to suggest that all research has the same risks. The relevant question asks which principles apply and how current review protects participants.

Psychological assessment asks whether a measure supports a particular interpretation and use. Reliability concerns consistency; validity concerns evidence for the intended meaning. A measure can be reliable without measuring the intended construct. Standardization concerns consistent administration and scoring, while norms provide a comparison group. A score should be interpreted for the population and purpose for which evidence supports it. Cultural or language differences can affect results, so fairness requires context and appropriate use.

A fictional assessment example: a researcher uses a vocabulary test created for English-speaking adults to infer general reasoning in multilingual teenagers. The result may reflect language familiarity as much as the construct the researcher intended. The test could score consistently and still support an invalid inference for this use. Examine the target population, instructions, norms, reliability evidence, and the consequence of the decision before accepting the interpretation.

Practice the reasoning

A second scenario distinguishes correlation from experiment. Students who report more hours of study also have higher course scores. That association does not prove study time caused the difference; prior knowledge, course attendance, motivation, or access to tutoring could matter. A randomized study assigning different study schedules may improve causal inference, but compliance and ethical implementation still matter. Use the evidence to state a limited conclusion rather than claiming more than the design supports.

On a selected-response item, identify the design first, then ask what inference it permits. If participants self-select a therapy, group differences can reflect initial severity or access. If a researcher manipulates one variable and controls the procedure, causal language may be more defensible. If a survey records two traits at one time, correlation is the safe description. The design determines the ceiling on the claim.

Research methods support other domains in 5391. Conditioning studies measure behavior after a stimulus or consequence; development studies compare age groups or follow people over time; assessment studies examine reliability and validity; treatment research tests outcomes. A candidate who understands variables and inference can evaluate these examples without memorizing every study detail. The course outline's examples indicate expected breadth, while the question stem defines the particular task.

Use the ETS Study Companion for the full 5391 methods outcomes. OpenStax Psychology 2e provides introductory discussion of empirical research and inference; the APA Ethics Code offers professional standards for psychologists. These sources support educational summaries but do not replace current institutional rules, local law, or specialized statistical references. All numerical examples here are original and illustrative.

Example question

A study finds a correlation between sleep hours and quiz scores. What conclusion is justified?

  1. The variables are associated; causation is not established by correlation alone
  2. Sleep caused the score increase
  3. Random assignment occurred
  4. The assessment is valid
Answer: A. Correlation supports an association but leaves confounds and direction unresolved.

The topic is listed in the ETS 5391 Study Companion. Introductory concepts are supported by the cited educational and professional sources. Clinical content is general and does not diagnose or prescribe.

Identify variables before judging conclusions

A research vignette becomes clearer when you label the predictor, outcome and possible confounds. In a study of evening screen use and sleep quality, screen use may be the measured predictor and sleep quality the outcome, but a correlation cannot show that the screen caused poor sleep. Work schedules, stress or health may affect both. In an experiment, researchers manipulate a condition and measure an outcome, while using comparison and control procedures to support an inference. The design, not the topic’s importance, determines what can be claimed.

Choose a design that fits the question

A descriptive design can document behavior or estimate frequency. A correlational design estimates association. An experiment can test a causal hypothesis when the manipulation and controls support it. A longitudinal design observes change across time, while a cross-sectional design compares groups at one point. Each has tradeoffs: experiments may have artificial settings, surveys depend on self-report, and longitudinal work can lose participants. On an exam, identify the question being asked and the method that can answer it. In practice, researchers often combine approaches to strengthen evidence.

Separate sampling, assignment and generalization

Random sampling selects people from a target population; random assignment allocates enrolled participants to conditions. Sampling affects how well findings may generalize, while assignment helps make condition comparisons fair. A convenience sample can still support a carefully bounded experiment, but its participants may not represent all people. Random assignment does not make a sample representative. Explain these distinctions with a concrete study: recruiting volunteers from one campus is sampling; placing each volunteer into a condition by chance is assignment.

Statistics summarize rather than certify

Mean, median and mode describe different aspects of a distribution. A mean can be pulled by extreme values; the median is resistant to outliers; the mode identifies the most frequent value. Variability measures such as standard deviation describe spread. Statistical significance, when covered, does not alone show a large or important effect. Consider who was measured, how the variables were operationalized, and whether the design supports the claim. A statistic is evidence within a method, not a substitute for reasoning about measurement and context.

Ethics are built into the design

Research ethics include informed consent, privacy, minimizing harm and appropriate independent review. A sound method does not excuse unnecessary risk or deception without justification and safeguards. Consent needs clear information about participation, relevant risks and withdrawal conditions. Confidentiality involves protecting identifiable information; it does not necessarily mean absolute secrecy in every circumstance. The APA Ethics Code is a professional source, while legal duties vary by jurisdiction. On exam questions, identify the principle implicated by the scenario and avoid inventing a universal legal rule.

Operational definitions and construct validity

An operational definition states how a researcher measures or manipulates a concept. “Stress” might be measured by a validated questionnaire, a physiological indicator or a specific task. These measures capture different aspects and can produce different conclusions. A measure needs evidence that it represents the intended construct for the population and use. A precise operational definition improves replicability, but precision alone does not establish validity. Ask what the measure actually records and whether that is close enough to the psychological idea being discussed.

Replication and cumulative evidence

A single study provides one piece of evidence. Replication asks whether a finding appears again under comparable or deliberately varied conditions. Failure to replicate can reveal measurement problems, contextual moderators, sampling differences or an initial false positive; it does not always mean the topic is meaningless. Meta-analysis can synthesize studies, but depends on the quality and comparability of included work. For exam preparation, understand why psychology relies on multiple studies rather than treating one dramatic result as settled fact.

Ethics and interpretation together

Ethical design and scientific inference are related. A participant who does not understand consent may not be making an informed choice. A study that withholds important support may create risk. Researchers need to weigh knowledge value against burden and protect privacy. Ethics review does not guarantee that a study is methodologically sound, and strong design does not remove ethical duties. If a scenario describes a participant risk, identify the relevant safeguard and avoid assuming the study is acceptable merely because it has a useful hypothesis.

A practical validity checklist

When evaluating an assessment, ask what construct it is meant to represent, whether the items sample that construct, whether the population matches the supporting evidence, and what decision will follow. A classroom quiz can reliably measure vocabulary recall without validly measuring broad psychological reasoning. A questionnaire may be useful for research but inappropriate for high-stakes individual diagnosis. Validity is tied to an interpretation and use, not a permanent property that follows a test in every setting. This helps distinguish the quality of a measure from the quality of a decision made using it.

Common questions

Does correlation prove causation?

No. Correlation shows association; confounds and direction remain possible without a design that supports causal inference.

What is the difference between random sampling and assignment?

Sampling selects participants for a study; assignment allocates participants to conditions. They address different design questions.

What makes a measure reliable?

Reliability concerns consistency of measurement, while validity concerns whether evidence supports the intended interpretation and use.