This study evaluates the performance of various methods for testing indirect effects in complex models with multiple mediators and indirect paths. The research compares traditional z tests with alternative methods such as the M test, empirical-M test, and resampling methods like the percentile and bias-corrected bootstrap. The study also examines contrasts of indirect effects. The study found that the bias-corrected bootstrap had the least biased confidence intervals, the greatest power to detect nonzero effects and contrasts, and the most accurate Type I error rate. However, all tests had less power to detect three-path effects and more inaccurate Type I error compared to two-path effects. Confidence intervals were biased for mediated effects, as found in previous studies. Results for contrasts did not vary greatly by test, although resampling approaches had somewhat greater power and might be preferable because of ease of use and flexibility. The study used a simulation with 12 parameter sets and three sample sizes (50, 100, 200) to evaluate the performance of five mediation tests and three contrast tests. The results showed that the bias-corrected bootstrap had the most accurate Type I error rates, followed by the M test and the empirical-M test. The percentile bootstrap had greater power and more accurate Type I error rates than the z test but did not outperform the M test. For contrasts, the percentile bootstrap was closest to the nominal error rate across the types of contrasts and sample sizes. The z test underestimated Type I error rates when both effects had only zero or small effect sizes in their paths. The bias-corrected bootstrap had the most accurate Type I error rate across all sample sizes for this type of contrast, followed by the percentile bootstrap. The study also found that the number of paths in the effect and the type of test used significantly influenced Type I error rates. The percentile bootstrap had the most accurate overall Type I error across all other conditions, followed by the bias-corrected bootstrap and then the z test. Contrasts with two two-path effects had an overall error rate slightly below the nominal .05 level, whereas contrasts with a three-path effect were slightly above .05. These two effects made up the only significant interaction. The z method was inconsistent, with lower than .05 error rate with two paths and greater than .05 when there were three paths.This study evaluates the performance of various methods for testing indirect effects in complex models with multiple mediators and indirect paths. The research compares traditional z tests with alternative methods such as the M test, empirical-M test, and resampling methods like the percentile and bias-corrected bootstrap. The study also examines contrasts of indirect effects. The study found that the bias-corrected bootstrap had the least biased confidence intervals, the greatest power to detect nonzero effects and contrasts, and the most accurate Type I error rate. However, all tests had less power to detect three-path effects and more inaccurate Type I error compared to two-path effects. Confidence intervals were biased for mediated effects, as found in previous studies. Results for contrasts did not vary greatly by test, although resampling approaches had somewhat greater power and might be preferable because of ease of use and flexibility. The study used a simulation with 12 parameter sets and three sample sizes (50, 100, 200) to evaluate the performance of five mediation tests and three contrast tests. The results showed that the bias-corrected bootstrap had the most accurate Type I error rates, followed by the M test and the empirical-M test. The percentile bootstrap had greater power and more accurate Type I error rates than the z test but did not outperform the M test. For contrasts, the percentile bootstrap was closest to the nominal error rate across the types of contrasts and sample sizes. The z test underestimated Type I error rates when both effects had only zero or small effect sizes in their paths. The bias-corrected bootstrap had the most accurate Type I error rate across all sample sizes for this type of contrast, followed by the percentile bootstrap. The study also found that the number of paths in the effect and the type of test used significantly influenced Type I error rates. The percentile bootstrap had the most accurate overall Type I error across all other conditions, followed by the bias-corrected bootstrap and then the z test. Contrasts with two two-path effects had an overall error rate slightly below the nominal .05 level, whereas contrasts with a three-path effect were slightly above .05. These two effects made up the only significant interaction. The z method was inconsistent, with lower than .05 error rate with two paths and greater than .05 when there were three paths.