∙ Lognormal distributions and self-similarity are characteristics associated with a wide range of complex systems. The sequential breakage model established a link between lognormal distributions and self-similarity. We tested the hypotheses that (1) the branching structure of Japanese elm trees is statistically self-similar and that (2) the distribution of the lengths of terminal stems, the end products of the self-similar branching process, approaches a lognormal distribution.

∙ We measured the length of stems of Japanese elm trees (Ulmus davidiana) sampled from a riparian forest.

∙ We obtained three results: (1) each occurrence of branching caused variations in the lengths of the daughter stems; (2) the branches were statistically self-similar: the observed distributions of the lengths of the daughter stems relative to their parent stems were similar at all scales within each branch; and (3) the multiplicative effect of these independent variations generated a lognormal distribution of the lengths of terminal twigs.

∙ We provide the first empirical evidence that self-similarity of a plant form generates a lognormal distribution of the size of its terminal organs.