Research on embodied cognition has shown that action and language are closely intertwined. The present study seeks to exploit this relationship, by systematically investigating whether motor activation would improve eight-to-nine year old children’s learning of vocabulary in their first language. In a within-subjects paradigm, 49 children learned novel object manipulation, locomotion and abstract verbs via a verbal definition alone and in combination with gesture observation, imitation, or generation (i.e., enactment). Results showed that learning of locomotion verbs significantly improved through gesture observation compared to verbal definitions only. For learning object-manipulation verbs, children with good language skills seemed to benefit from imitation and enactment, while this appeared to hinder children with poor language skills. Learning of abstract verbs was not differentially affected by instructional condition. This study suggests that the effectiveness of observing and generating gestures for vocabulary learning may differ depending on verb type and language proficiency.