Using larval data of zoeae from selected genera of majids, we determined tree topologies, levels of homoplasy, and frequencies of reduction under three different assumptions of character argumentation: ordered reduction events, unordered reduction events, and outgroup comparison. Under each assumption we provided a phylogenetic hypothesis for some majid genera and evaluated the assumption that structural reduction can be assumed a priori as a criterion to infer character transformation polarity in phylogenetic reconstruction of decapods. The results indicate that the a priori assumption of “reduction” as the derived condition is not justified because under this assumption, reduction is not always maintained throughout the resulting phylogenetic hypothesis. Furthermore, we also found that this criterion fails to provide the most parsimonious explanation of the data set. Therefore, we reject the use a “reduction=derived” criterion to infer polarity in phylogenetic reconstruction. Phylogenetic analysis using outgroup comparison provided a phylogenetic hypothesis with a better fit and a lower frequency of reduction events. However, we found that statements of homology may be problematic when the number of larval stages in the outgroup differ from those of the ingroup. To overcome this problem, we suggest that, in the absence of evidence for developmental homology, all larval stages should be considered as potential homologues. Using this approach to homology of larval stages, we provide a new phylogenetic hypothesis for 15 genera of majids based on larval morphology. Within Majidae, representative members of Majinae formed a highly nested monophyletic group with the following topology: ((Jacquinotia+Notomithrax) (Leptomithrax+Maja)). In contrast, the Oregoniinae (Hyas+Chionoecetes) formed a basal monophyletic group. Contrary to established ideas for the monophyly of Inachinae,Macrocheirais basal to the Oregoniinae. Other taxa did not form monophyletic groupings based on classical assignment to subfamilies.