Meta-analytic structural equation modeling (MASEM) is an increasingly popular technique for summarizing findings from multivariate correlational research. The goal of MASEM is to fit and interpret structural equation models in order to explain the (pooled) correlations between variables. With SEM, one can test for different types of models, including factor models, path models, and full SEM models. A common issue in meta-analytic research is the dependence of effect sizes. For instance, dependency can occur due to multiple informants (e.g., mother- and father-report on parenting practices) or measurement occasions (e.g., pre- and post-test measures). Over time, several (ad hoc) solutions have arisen to overcome the issue of dependency when applying MASEM. These methods have not always been justified or well-examined for their statistical properties. The objective of the present study was to compare several (commonly used) approaches, such as ignoring dependency, aggregation, and elimination, to a recently developed approach by Wilson et al. (2016), further referred to as the WPL-approach. The different approaches are illustrated with data from 114 independent samples, reporting on 772 bivariate correlations between parental delinquency, parenting behaviour, and child delinquency. Results showed that different approaches of dealing with dependency lead to substantially different results when conducting MASEM analysis. Ignoring dependency of effect sizes led to consistently smaller standard errors, which increases the likelihood of Type I errors. With aggregation and elimination of effect sizes, the standard errors seemed to be consistently larger, which may have increased Type II errors. The standard errors of the WPL-approach depended on both the amount of information available and the level of within- and between-study variability. Based on our results we provide practical recommendations for researchers who want to apply MASEM with dependent effect sizes.