Purpose: This study presents a meta-analysis of the difference in nonword repetition performance between children with and without specific language impairment (SLI). The authors investigated variability in the effect sizes (i.e., the magnitude of the difference between children with and without SLI) across studies and its relation to several factors: type of nonword repetition task, age of SLI sample, and nonword length. Method: The authors searched computerized databases and reference sections and requested unpublished data to find reports of nonword repetition tasks comparing children with and without SLI. Results: Children with SLI exhibited very large impairments in nonword repetition, performing an average (across 23 studies) of 1.27 standard deviations below children without SLI. A moderator analysis revealed that different versions of the nonword repetition task yielded significantly different effect sizes, indicating that the measures are not interchangeable. The second moderator analysis found no association between effect size and the age of children with SLI. Finally, an exploratory meta-analysis found that children with SLI displayed difficulty repeating even short nonwords, with greater difficulty for long nonwords. Conclusions: These findings have potential to affect how nonword repetition tasks are used and interpreted, and suggest several directions for future research.