A main effect is the overall impact of one independent variable on the outcome, averaging across all levels of the other independent variables in the study. In a factorial design, where researchers test two or more variables at the same time, each variable gets its own main effect. A 2×2 factorial design with two independent variables produces two main effects and one interaction effect, all from a single experiment.
How a Factorial Design Works
A factorial design examines the effects of two or more independent variables on a single dependent variable. Instead of running separate experiments for each variable, factorial designs test them simultaneously. This is efficient, but the real payoff is that it lets you see whether the variables influence each other, something separate experiments can never reveal.
The standard statistical test for analyzing factorial data is a two-way ANOVA (analysis of variance). For a design with two independent variables, this test produces three results: a main effect for variable one, a main effect for variable two, and an interaction effect between them. Each result gets its own F-statistic and p-value, so each is evaluated independently for statistical significance.
What a Main Effect Actually Tells You
A main effect answers a simple question: does this variable matter, regardless of what’s happening with the other variable? The key phrase is “regardless of.” When a main effect is significant, it means the overall averages across the levels of that variable are meaningfully different.
Consider a clinical trial testing two treatments for depression: a drug (medication vs. placebo) and a therapy (cognitive behavioral therapy vs. waitlist). A significant main effect for the drug means that, regardless of whether patients received therapy, those on the medication had lower depression scores than those on placebo. A significant main effect for therapy means that, regardless of which pill patients took, those in therapy improved more than those on the waitlist. Each main effect isolates one variable while collapsing across the other.
Marginal Means: The Math Behind It
To identify a main effect, you look at marginal means. These are the averages for each level of one variable, calculated across all levels of the other variable. In the depression example, you would average the depression scores for everyone who took the medication (whether or not they also got therapy) and compare that to the average for everyone who took the placebo.
If the medication group averaged a depression score of 12.6 and the placebo group averaged 19.3, that gap is what the main effect captures. The ANOVA then tests whether a gap that large is statistically significant or could have occurred by chance. When the resulting F-statistic is large enough to produce a p-value below .05, the main effect is considered significant.
One important nuance: main effects in a factorial ANOVA are not identical to running a separate t-test or one-way ANOVA for each variable. The factorial ANOVA partitions out the interaction effect first, so the main effect reflects the influence of that variable after accounting for any interplay between the two variables.
A Concrete Example: Reaction Time
In a study on attention, researchers measured how quickly people could name the color of a word. The two independent variables were congruency (whether the word matched its ink color, like “RED” printed in red vs. “RED” printed in blue) and posture (standing vs. sitting). The dependent variable was reaction time in milliseconds.
The main effect of congruency was significant, with F(1, 46) = 739.94, p < .05. People responded faster when the word and color matched (average 815 ms) than when they didn’t match (average 922 ms). That’s a difference of about 107 milliseconds, collapsed across both postures.
The main effect of posture was also significant, with F(1, 46) = 41.44, p < .05. People standing up responded faster (average 856 ms) than people sitting down (average 881 ms). That 25-millisecond gap held up regardless of whether the words were congruent or incongruent. Both main effects existed in the same experiment, each telling its own part of the story.
How to Spot Main Effects on a Graph
Factorial results are often displayed as line graphs or bar charts with one variable on the horizontal axis and separate lines (or bar groups) for the other variable. You can read main effects directly from these visuals.
For the variable on the horizontal axis, a main effect exists when both points on one side are consistently higher or lower than both points on the other side. For the variable represented by the separate lines, a main effect exists when one line sits consistently above or below the other. If the green line is entirely above the red line across the graph, the variable that distinguishes those lines has a main effect.
Parallel lines indicate no interaction, meaning the effect of one variable is the same at every level of the other. When lines cross, or would cross if extended, that signals a possible interaction effect, which changes how you should interpret the main effects.
When a Main Effect Can Be Misleading
A significant main effect can paint an incomplete picture when a significant interaction is also present. An interaction means the effect of one variable changes depending on the level of the other variable. In that situation, the overall average captured by the main effect may not accurately describe what’s happening within each specific group.
For example, imagine a study on caffeine and personality type. The main effect might show no overall difference between caffeine and no caffeine, and no overall difference between introverts and extroverts. But an interaction could reveal that caffeine helps extroverts while hurting introverts. The two effects cancel each other out when averaged, making both main effects non-significant even though caffeine clearly matters.
The reverse can also happen. A main effect might be significant, but only because the effect is very strong in one subgroup and absent in another. A study on moral judgments found that people in a messy room made harsher judgments than those in a clean room, producing a significant main effect for room condition. But the interaction showed this was only true for half the participants. Saying “messy rooms increase harsh judgments” would be technically accurate as a main effect statement, but it overstates the case for the groups where the pattern didn’t hold.
This is why researchers are advised to interpret main effects cautiously whenever a significant interaction is present. The interaction tells you the story is more complicated than the main effect alone suggests, and the group-level patterns deserve closer examination.
Main Effects vs. Interaction Effects
The distinction is straightforward. A main effect is about one variable in isolation, averaged over everything else. An interaction effect is about how two variables work together, whether the impact of one depends on the level of the other.
In a 2×2 design, you get exactly two main effects and one interaction effect. In larger designs (say, a 2×3 or 2×2×2), the number of main effects equals the number of independent variables, but the number of possible interactions grows. A three-variable design has three main effects, three two-way interactions, and one three-way interaction.
Each of these effects is tested separately, and any combination can be significant or non-significant. You can have two significant main effects with no interaction, or a significant interaction with no main effects, or all three significant at once. The pattern depends entirely on the data, and each result adds a different layer to your understanding of how the variables influence the outcome.