There are three leaf types in the Cercideae tribe (Caesalpinioideae, Fabaceae): unifoliate, bilobate, and bifoliate. By the fusion hypothesis, the ancestral leaf type within the Cercideae was bifoliate, and the unifoliate leaf was derived from the fusion of the bifoliate leaflets. In the opposing splitter hypothesis, the unifoliate leaf was the ancestral leaf type within the Cercideae, and the bifoliolate leaf was derived from the splitting of the unifoliate leaf into two separate leaflets. Palmate venation in leaflets is problematic for the fusion hypothesis. Watari proposed derivation from a palmately compound leaf with many leaflets and interpreted the primary vein as homologous to a single midrib in a leaflet. The structure and function of the upper pulvinus in the unifoliate leaf of Cercis canadensis L. were examined to test the two hypotheses. Microscopic observations and measurement of leaf movement over a 12‐h period indicate that the upper pulvinus in the unifoliate leaf is a compound structure consisting of an apical common joint composed of three secondary pulvini and two lateral laminar joints separated by a laminar joint associated with the midrib. A lateral laminar joint consists of three tertiary pulvini, one at the base of each primary vein, and the midrib is formed from a rachilla and two or more pinnules. Structure comparisons were made between the upper pulvinus of a unifoliate leaf of C. canadensis and a bilobate leaf of Bauhinia purpurea. The secondary and tertiary nature of the upper pulvinus indicates that the unifoliate leaf in Cercis was derived from an extinct ancestor with bipinnately compound leaves. Results add support for Watari’s interpretation of the primary vein and for the splitter hypothesis.