In this study, we investigated how discussions about picture books among a group of young children would affect their narrations about the picture books. Thirty 5-year-old children participated in this study. The children received a verbal working memory task that required them to compare two non-words given sequentially over a headset. They also engaged in an activity involving picture books under one of two conditions. First, a female adult read one of two picture books to groups of children. Second, they answered questions about the picture book individually (the non-group activity condition) or in a group (the group activity condition). Finally, they were required to narrate a story about the picture book they were shown. We divided participants’ narrations into idea units (IUs): there are six types of IUs (basic IUs, point IUs, sophisticated IUs, picture IUs, erroneous IUs, and other IUs). Two university students independently classified the IUs. The results show that the partial correlations between the scores of the verbal working memory task and point IUs (r = .40) and between verbal working memory and other IUs (r = −.40) were significant when controlling for age in terms of months. This finding suggested that children with better verbal working memory would understand the main points of the stories. Although children were significantly more likely to report three types of IUs in the group activity condition than in the non-group activity condition, the differences were mediated by the content of the picture books: one picture book was more likely to stimulate the group activities, which facilitated the children’s narrations.