析因实验

A factorial design allows researchers to manipulate two or more independent variables and measure their effects on a single dependent variable within one study. Rather than testing potential influences one experiment at a time, this approach examines several variables simultaneously, requiring fewer participants while revealing whether different causes interact to affect outcomes.

A 2x2 factorial design uses two independent variables, each with two levels, creating four possible condition combinations. In the nonverbal signal study, self-awareness (high/low) and self-esteem (high/low) were manipulated through a mirror and false feedback. Participants experienced one of four combinations, allowing simultaneous examination of both variables' effects on emotion decoding accuracy.

Main effects focus on how a single independent variable influences the dependent variable, while interaction effects examine whether one independent variable changes another's influence on the outcome. In the study, the main effect hypothesis predicted high self-awareness and high self-esteem would improve emotion detection, whereas the interaction hypothesis predicted self-esteem's impact would depend on self-awareness levels.

Researchers use specific techniques to create different levels of each independent variable. Self-awareness was manipulated by having participants complete a quiz with or without a visible mirror. Self-esteem was manipulated through false feedback, telling participants they scored in the top 10% or bottom 50% on a geography quiz, creating distinct high and low conditions.

A two-way ANOVA (analysis of variance) reveals main effects and interaction effects in factorial designs. This test determines whether group differences exist and whether the effect of one independent variable depends on the level of another. In the study, the two-way ANOVA showed that self-awareness effects on emotion detection varied based on self-esteem levels.

Random assignment ensures that group assignments are based entirely on chance, eliminating selection bias and strengthening causal conclusions. Before participants arrive, researchers randomly organize packets containing each of the four condition combinations, guaranteeing that participant characteristics do not systematically influence which condition they receive.

Factorial designs apply across diverse research questions. Researchers use them to examine how shock probability and alcohol administration influence startle responses, how stress levels interact with exercise types to affect outcomes, and whether test format (on-screen versus written) and gender influence student performance. This design enables simultaneous investigation of multiple factors in complex real-world scenarios.