While evolutionary trends have long received much attention and have been widely disputed, new methods are now allowing the testing of directional hypotheses with increased rigor. Here, we test a general hypothesis about the way many kinds of discrete characters are thought to evolve, termed oligomerization. This is the tendency for serial structures (such as arthropod body and appendage segments) or armature (such as spines) to evolve primarily through loss and fusion. Focusing on the Crustacea, we use maximum likelihood methods to test for directional evolution in a large sample (> 500) of discrete traits, analyzed against molecular-based phylogenies. We find evidence for a significant trend toward trait loss, in accordance with the reduction principle. However, this trend is far from ubiquitous, with many characters exhibiting a reconstructed bias toward gains. These results suggest that caution must be used before drawing conclusions about which taxa are “primitive” or about the directionality of morphological shifts in the absence of phylogenetic analysis. Nevertheless, oligomerization—as a trend rather than a law—may be an important process that influences evolutionary trajectories from both morphological and functional perspectives.