We studied the development of narrative ability yearly in a cross-sectional design from 4 to 10 years. All children were given the same props and underwent the same procedures. Children's narratives were transcribed verbatim, and then coded for number of words, propositions, episodes and indexes of narrative cohesion and structure. Verbal IQ was controlled as appropriate. Words showed no significant effect for age. A number of propositions and episodes followed nonlinear trajectories of development, and spline regression analyses identified maxima at 8 and 7 years, respectively. However, indexes of cohesion and structure followed linear trajectories of development.