Main Effects & Interactions
Makeup & Attractiveness Experiment (Psych 321, Spring 2016)
- Question: Does makeup increase perceived attractiveness?
- Design: between-subjects 2 \times 1
- IV1: Makeup (present vs absent)
- IV2: Rater Gender (male, female)
- Stimuli: 5 identical female faces; attractiveness scale 1–10
- Overall means (collapsed across gender)
- With makeup: 7.4
- Without makeup: 7.3 ⇒ no main effect of makeup
- Split by gender
- Male raters: 7.9 (makeup) vs 7.3 (no makeup)
- Female raters: 6.8 (makeup) vs 7.3 (no makeup)
- Conclusion: Significant Makeup × Gender interaction; effect of makeup depends on rater gender (↑ for males, ↓ for females)
Core Concepts: Main Effect vs Interaction
- Main effect: consistent influence of an IV on the DV after averaging over other IVs
- Interaction: effect of one IV changes across levels of another IV
\text{"Effect of IV}1\text{ depends on IV}2" ⇒ answers become “it depends”
Cell-Phone Use & Driving Study
- Question: Does hands-free phone use impair braking, and does this vary with age?
- Design: 2 \times 2 between-subjects
- IV1: Phone Use (hands-free vs none)
- IV2: Age Group (Younger 18–25 vs Older 65–74)
- DV: Brake-onset time (ms) after obstacle appears (lower = better)
- Findings
- Main effect of Phone: Phone \approx1000 ms vs No phone \approx850 ms
- Main effect of Age: Older > Younger in all conditions
- No Phone × Age interaction: phone slows both age groups by similar margin
- Notable: Younger drivers on phone ≈ braking speed of older drivers off phone (functional “aging” of \sim50 years)
Quick Checklist for 2-IV Experiments
- Examine graph/means:
- Non-parallel lines → possible interaction
- Vertical gap between averaged lines → main effect
- Reporting order: each main effect (significant/not) → interaction (significant/not)
- Interpretation guide:
\text{Interaction significant} \Rightarrow describe pattern within each subgroup; main effects may mislead
\text{No interaction} \Rightarrow main effects generalize across other IV
Exam Takeaways
- Interactions reveal conditional relationships; always test them
- Absence of a main effect can conceal opposing subgroup trends
- When results answer “it depends,” a significant interaction is likely present