Epibranchials are among the main dorsal elements of the gill basket in jawed vertebrates (Gnathostomata). Among extant fishes, chondrichthyans most resemble the putative ancestral condition as all branchial arches possess every serially homologous piece. In osteichthyans, a primitive rod-like epibranchial 5, articulated to ceratobranchial 5, is absent. Instead, epibranchial 5 of many actinopterygians is here identified as an accessory element attached to ceratobranchial 4. Differences in shape and attachment of epibranchial 5 in chondrichthyans and actinopterygians raised suspicions about their homology, prompting us to conduct a detailed study of the morphology and development of the branchial basket of three ostariophysans (Prochilodus argenteus, Characiformes; Lophiosilurus alexandri and Pseudoplatystoma corruscans, Siluriformes). Results were interpreted within a phylogenetic context of major gnathostome lineages. Developmental series strongly suggest that the so-called epibranchial 5 of actinopterygians does not belong to the epal series because it shares the same chondroblastic layer with ceratobranchial 4 and its ontogenetic emergence is considerably late. This neomorphic structure is called accessory element of ceratobranchial 4. Its distribution among gnathostomes indicates it is a teleost synapomorphy, occurring homoplastically in Polypteriformes, whereas the loss of the true epibranchial 5 is an osteichthyan synapomorphy. The origin of the accessory element of ceratobranchial 4 appears to have occurred twice in osteichthyans, but it may have a single origin; in this case, the accessory element of ceratobranchial 4 would represent a remnant of a series of elements distally attached to ceratobranchials 1–4, a condition totally or partially retained in basal actinopterygians. Situations wherein a structure is lost while a similar neomorphic element is present may lead to erroneous homology assessments; these can be avoided by detailed morphological and ontogenetic investigations interpreted in the light of well-supported phylogenetic hypotheses.