Shrout and Bolger discuss new methods for assessing mediation in experimental and nonexperimental studies. They recommend using bootstrap methods to evaluate mediation effects, which are more powerful than traditional tests because they detect skewness in the distribution of the mediated effect. They argue that testing the X→Y association for statistical significance is not always necessary when there is a priori belief that the effect size is small or suppression is possible. Mediation models help decompose associations into causal mechanisms, aiding theory development and intervention identification. These models are relevant to both experimental and nonexperimental research, with examples in social-cognitive, developmental, and organizational psychology. Baron and Kenny's (1986) approach to mediation analysis involves four steps: testing the X→Y association, testing X→M, testing M→Y while holding X constant, and testing the mediated effect. However, critics argue that this approach has limitations, such as assuming partial mediation and relying on linear regression. Bootstrap methods, which are more flexible and powerful, are increasingly used to assess mediation effects, especially in small to moderate samples. Bootstrap methods provide more accurate confidence intervals and are less affected by the skewness of the mediated effect distribution. Bootstrap methods allow for the estimation of indirect effects through resampling, providing more reliable results than traditional methods. They are particularly useful in cases of suppression, where the direct and indirect effects have opposite signs. The article also discusses the importance of considering the strength of mediation, using the ratio of the indirect effect to the total effect (PM) to quantify mediation. This approach provides a more nuanced understanding of mediation than binary classifications of complete or partial mediation. The authors emphasize the importance of confidence intervals over significance tests in interpreting mediation effects.Shrout and Bolger discuss new methods for assessing mediation in experimental and nonexperimental studies. They recommend using bootstrap methods to evaluate mediation effects, which are more powerful than traditional tests because they detect skewness in the distribution of the mediated effect. They argue that testing the X→Y association for statistical significance is not always necessary when there is a priori belief that the effect size is small or suppression is possible. Mediation models help decompose associations into causal mechanisms, aiding theory development and intervention identification. These models are relevant to both experimental and nonexperimental research, with examples in social-cognitive, developmental, and organizational psychology. Baron and Kenny's (1986) approach to mediation analysis involves four steps: testing the X→Y association, testing X→M, testing M→Y while holding X constant, and testing the mediated effect. However, critics argue that this approach has limitations, such as assuming partial mediation and relying on linear regression. Bootstrap methods, which are more flexible and powerful, are increasingly used to assess mediation effects, especially in small to moderate samples. Bootstrap methods provide more accurate confidence intervals and are less affected by the skewness of the mediated effect distribution. Bootstrap methods allow for the estimation of indirect effects through resampling, providing more reliable results than traditional methods. They are particularly useful in cases of suppression, where the direct and indirect effects have opposite signs. The article also discusses the importance of considering the strength of mediation, using the ratio of the indirect effect to the total effect (PM) to quantify mediation. This approach provides a more nuanced understanding of mediation than binary classifications of complete or partial mediation. The authors emphasize the importance of confidence intervals over significance tests in interpreting mediation effects.