Purpose: The present study is a meta-analysis that examines the difference in novel word learning performance between children with primary language impairment (LI) and typically developing children. Participant and task characteristics were examined as variables that potentially moderated children's word learning. Method: Eight hundred and forty-six published studies were retrieved from conventional databases (PsycINFO, MEDLINE, and Web of Science). Of these studies, 28 met the criteria for inclusion in the meta-analysis, yielding 244 effect sizes across experimental conditions. Results: LI groups showed significantly lower word learning performance than typical age-matched groups and equivalent performance to typical language-matched groups. Moderator analyses showed that the magnitude of the group difference relative to age peers was significantly associated with participants' chronological age, receptive language and cognitive abilities, task and novel word type, and the extent of novel word exposure. Conclusion: The difference in novel word learning performance between children with LI and age-matched children is strongly affected by task and participant characteristics in the primary studies.