This study is an examination of the melodic improvisations of a group of children (N = 62) for 3 years, from ages 7 through 9. Participants improvised as part of a class rondo for Orff instruments, in ABACADA form, in which the B, C, and D sections were 8-measure alto-xylophone improvisations in C pentatonic. Each participant improvised three melodies per year. A total of 558 improvisations were collected. Participants took the IMMA and a researcher-designed measure of mallet skill each year. Improvisations were examined for the presence of repeated and developed melodic motives, repeated and developed rhythmic motives, pulse adherence, phrases, and antecedent/consequent phrases. Data were analyzed using a repeated-measures MANCOVA. Significant age effects emerged for repeated melodic and rhythmic motives, sense of pulse, and antecedent/consequent phrases. Post-hoc Tukey Pairwise Comparisons revealed significant differences between ages 7 and 8 in the presence of each of these characteristics. Significant differences also occurred between ages 7 and 9 for the presence of antecedent/consequent phrases, presence of repeated rhythmic motives, and pulse adherence. No significant differences emerged between ages 8 and 9 for any of the characteristics. (Contains 2 endnotes.)