Questions: Validity in Psychological Measurement

Question 1

A new social anxiety scale correlates strongly with clinician ratings of anxiety (r = .72) and weakly with IQ scores (r = .08). This pattern primarily supports which type of validity?

• A. Internal validity
• B. External validity
• C. Construct validity
• D. Ecological validity

Answer: (C)

Construct validity is supported by convergent evidence (strong correlation with a related measure — clinician ratings) and discriminant evidence (weak correlation with an unrelated construct — IQ). Internal validity concerns causal inference within a study, and external validity concerns generalizability — neither is assessed by correlational patterns of the measure itself.

Question 2

A psychological test demonstrated to be valid for a U.S. college student sample is automatically valid for use with adults in other cultural contexts.

• T. True
• F. False

Answer: False

Validity is not a fixed property of a test — it must be established for each population and context. Measurement invariance (whether the same items measure the same construct across groups) must be tested separately. A scale measuring 'depression' through Western symptom expressions may not capture equivalent constructs in cultures with different idioms of distress.

Question 3

Describe the core tradeoff between internal validity and external validity in psychological research.

Model answer: Maximizing internal validity (e.g., tight lab controls, random assignment) often strips away the real-world complexity that determines whether findings generalize, reducing external validity. Conversely, naturalistic designs with high ecological realism sacrifice control, making causal inference harder.

A highly controlled lab experiment can definitively establish that X causes Y under those conditions, but those conditions may not match real-world settings. Field studies preserve ecological validity but cannot rule out confounds. Researchers must balance these tradeoffs based on their questions — and ideally replicate findings across both designs.