In a previous study (here termed Study 1), we explored the effects of working memory loads on young children's recall of passive voice structures. We found that participants (n = 36) were more likely to use the passive voice in recall responses when holding an unrelated memory load but were more likely to use the active voice when not under load. In the two new studies presented here (Studies 2 and 3), I extend these findings to explore how working memory loads impact not only children's recall but also their construction of original passive voice structures. In the first new study, I used a similar method to Adams and Cowan (2021) but gave 4- and 5-year-old participants (n = 38) more instruction and practice recalling the passive. Responses were categorized as either passive or active voice sentences (a small subset of responses could not be categorized as either syntax). Participants were more likely than in the previous work to use the passive voice overall, but use was not significantly impacted by memory loads. Participants were more likely to use the active voice when there was no load, as in the previous study. In a second study, I explored if 4- and 5-year-old children (n = 36) could form their own passive voice constructions about animations depicting transitive actions. Participants used the passive voice in around 21 percent of all responses, and used the active voice for most of the remaining responses. Working memory loads did not cause participants to speak in the passive or active voice more often. Performance on the QUILS, which combined measures of vocabulary, syntax, and processing abilities, predicted participant's ability to use the passive voice, especially when under working memory load. Combined results demonstrate the difference in task demand between recalling passive sentences versus constructing passive sentences and how working memory contributes to each.