Accounting for intraspecific trait variation (ITV) is central to plant ecology and crucial for vegetation modeling efforts. ITV can be substantial; however, it remains unclear how ITV influences community-weighted mean (CWM) trait estimates. We use leaf and root trait data from 423 trees of 72 species from 15 Angiosperm families in combination with community data from 164 small plots comprising 582 species to evaluate the contribution of ITV to CWMs, comparing unlogged, primary forest to selectively-logged and clear-cut secondary forest. We examine the effect of gap-filling missing trait values via phylogenetic generalized linear modeling (PhyloPars) on CWMs. For six of seven traits, ITV negatively covaried with species turnover to generate larger CWM differences than observed if ITV was not integrated. For example, plot average CWM specific leaf area was 10.7 and 10.4 m2 kg− 1 for primary and secondary forest, not accounting for ITV, but shifted to 9.8 and 11.1 m2 kg− 1 after doing so. Specific root length showed a similar trend. Our results from 72-species assemblages were supported by the results from the gap-filled analysis using the entire community, where the contribution of ITV to CWMs ranged from 25 to 75%, with nearly all trait variation due to forest type attributable to ITV. Therefore, CWM trait estimates became more-conservative with forest age, whereas ITV for many traits showed an acquisitive shift, and because of negative covariation between ITV and species turnover, forest age-related CWM differences increased. Differences were unaffected, if not strengthened, by gap-filling incomplete functional trait matrices.