Using a structured literature search and meta‐regression procedures, this study sought to determine whether associations between joint attention and language are moderated by group (autism spectrum disorder [ASD] vs. typical development [TD]), joint attention type (responding to joint attention [RJA] vs. other), and other study design features and participant characteristics. Studies were located using database searches, hand searches, and electronic requests for data from experts in the field. This resulted in 71 reports or datasets and 605 effect sizes, representing 1,859 participants with ASD and 1,835 TD participants. Meta‐regression was used to answer research questions regarding potential moderators of the effect sizes of interest, which were Pearson's r values quantifying the association between joint attention and language variables. In the final models, conducted separately for each language variable, effect sizes were significantly higher for the ASD group as compared to the TD group, and for RJA as compared to non‐RJA joint attention types. Approximate mental age trended toward significance for the expressive language model. Joint attention may be more tightly tied to language in children with ASD as compared to TD children because TD children exhibit joint attention at sufficient thresholds so that language development becomes untethered to variations in joint attention. Conversely, children with ASD who exhibit deficits in joint attention develop language contingent upon their joint attention abilities. Because RJA was more strongly related to language than other types of joint attention, future research should involve careful consideration of the operationalization and measurement of joint attention constructs. Autism Res 2016, 9: 1021–1035. © 2016 International Society for Autism Research, Wiley Periodicals, Inc.