Plant clades may exhibit little or wide morphological variation as a result of (1) the retention of ancestral characteristics or phylogenetic signal, (2) character displacement, or (3) random phenotypic drift or convergence. Understanding the taxonomy and systematics of many plant lineages has been challenging due to continuous intra- and interspecific morphological variation. To assess which evolutionary hypothesis could explain the morphological diversity in the genus Geonoma (Arecaceae), we performed a Mantel test between phylogenetic and morphological distances of 54 taxa, and tested for phylogenetic signal using Blomberg's K-statistic on continuous variables, and a randomization of character states. To obtain a phylogenetic (patristic) distance matrix for Geonoma, we constructed a molecular phylogeny of tribe Geonomateae using three nuclear DNA regions. A positive relationship between the patristic and a 26-discrete-character distance matrix (R2 = 0.55, P < 0.001) supported the phylogenetic signal hypothesis. The randomization test showed that signal was present in 16 characters. No relationship was evident using a 17-quantitative-variable distance matrix (R2 = 0.07, P = 0.13), supporting the random drift hypothesis or convergence, and all 17 K-values were close to 0, suggesting less phylogenetic signal than under the Brownian model. If most morphological variables traditionally used to classify Geonoma evolved randomly or convergently, it might explain Geonoma's challenging taxonomy. © 2012 The Linnean Society of London, Biological Journal of the Linnean Society, 2012, 106, 528–539.