True Experiments: Comparison Groups, Control Variables, and Eliminating Confounds

Core question: Why put so much emphasis on running a true experiment?
- Only method that can establish causation with confidence.
- Brings together course‐wide themes: skeptical inquiry, valid evidence, decision-making.
Distinct from other research types
- Association (correlational) studies → identify relationships but cannot pinpoint cause.
- Descriptive work → describes phenomena but stops short of causal explanation.
- True experiments isolate variables, control alternative explanations, and test causal claims.

Causal claim syntax: “X causes Y” or “If you do X, Y will happen.”
- Requires three criteria (from earlier lectures): covariance, temporal precedence, internal validity.
- True experiments directly address all three.
Association claim syntax: “X is related to Y.”
- Can be shown with correlational data; does not prove that X is the reason Y changes.

Advertisement claim: “Wear the wrap, exercise, eat a healthy diet → lose weight.”
- Appears causal; actually mentions three causal agents simultaneously.
- Lacks clarity: Which factor(s) matter? Wrap? Diet? Exercise? All? Some subset?
Initial flawed study idea
- $n = 30$ participants do all three behaviors; outcome = weight loss.
- Finds weight reduction → still no idea which behavior mattered.
Need for a Comparison Group
- Study revision 1:
- Group 1: Wrap + Diet + Exercise.
- Group 2: No wrap + No diet change + No exercise.
- If Group 1 > Group 2 weight loss → doesn’t isolate wrap; group differed on 3 variables.
Corrected design principles
- Add control so only the wrap differs:
- Group 1: Wrap + Standardized diet + Standardized exercise.
- Group 2: No wrap + Same diet + Same exercise.
- Control additional lifestyle variables (sleep, stress, smoking, etc.) as much as feasible.

Treatment group: receives the independent variable (wrap).
Control group: identical in every way except they do not receive the critical treatment.
Interpretation hinges on comparing outcomes between these two.
- Without the control, data are uninterpretable.

Not true “variables” during the study—experimenter fixes them so they do not vary across participants.
Purpose: ensure the independent variable is the only systematic difference between conditions.
Examples across studies
- Word difficulty & test time in Sexual Arousal/Memory study.
- Using the same faces in Makeup vs. No-Makeup attractiveness study.
- Diet & exercise levels in Waist-Trainer study.

Confound: an alternative explanation; a factor that varies along with the IV and could produce the DV.
Design confound: error built into the study’s structure (experimenter’s fault).
- Waist-trainer design with diet/exercise differences → two major design confounds.
- Makeup photos where "with makeup" faces are smiling, "without" faces are neutral → confound facial expression with makeup presence.
Rallying cry: “Down with confounds!”
- Eliminating them preserves internal validity—the hallmark of a true experiment.

Sexual Arousal & Memory
- IV: Type of song (Rihanna “Skin” vs. “Yankee Doodle”).
- Control variables: number/difficulty of words, study time, test time.
Makeup & Attractiveness
- IV 1: Presence vs. absence of makeup.
- IV 2: Participant gender (between-subjects).
- Critical control: Same women’s faces used in both makeup and no-makeup images.
- Confound caution: Avoid letting smiles, pose, lighting differ across conditions.

Advertisers, political campaigns, product promoters rely on public lack of experimental literacy.
- Manipulative visuals (before/after, smiling vs. neutral) exploit design confounds.
- Consumers equipped with skeptical inquiry can spot the flaws quickly.
Science’s value: adds rigorous evidence where anecdote & marketing spin dominate.
- Well-designed experiments inform health behaviors, policy, and personal decisions.
Ethical obligation of researchers/consumers alike: demand proper controls before accepting causal claims.

True experiments are powerful because they can isolate one factor and show its causal impact.
Comparison groups + control variables = defense against confounds.
Mastery of these concepts enables critical evaluation of everyday claims—from weight-loss products to social policies—and advances scientific knowledge with confidence.