Earlier studies using imaged wordless stories have shown that by 4-5 years children can produce descriptive narratives but have difficulties explaining and evaluating events, as well as making reference to the characters' internal states and to their different perspectives. This paper relates results of a conversational intervention procedure aiming to promote linguistic and evaluative aspects of children's narratives. Participants were 104 French-speaking children, aged 5:6 to 8;8 years, attending kindergarten, 1st and 2nd grades (38 in each class level) in a public school of Paris. All children were presented with five pictures telling the story of a misunderstanding between two characters, the "Stone Story". After the pictures were removed, all children were requested to tell the experimenter what they understood of the story (first narrative). Then, children in the intervention group (84) participated in a conversation with the experimenter who focused the child's attention on the reasons of the key events, while children in the control group (30) played a memory game with the story pictures and similar cards. Then, all children were asked to narrate once again the story (second narrative). One week later, children told a third narrative of the same story (to test the stability of the eventual gains) as well as a new story analogous to the stone story (to test whether the gains were generalizables). Results show that, relative to their first narrative, children in the conversation group significantly improved measures of linguistic expression, overall coherence and mind-oriented causal plot of their second narrative, while children in the control group didn't. Improvements appeared stable and generalizable. These findings point to the importance of the conversational intervention in promoting children's narrative skills and its usefulness as an assessment tool.