With the increase of laboratory facilities, molecular phylogenies are playing a predominant role in evolutionary analyses. However, understanding the evolution of morphological traits remains essential for a comprehensive view of the evolution of a group. Here we present a new approach based on co-inertia analysis for identifying characters which variations are dependent to the phylogeny, a prerequisite for analyzing the evolution of characters. Our approach has the advantage of treating the full data set at once, including qualitative and quantitative variables. It provides a graphical output giving the contribution of each variable to the co-structure, allowing a direct discrimination among phylogenetically dependent and independent variables. We have implemented this approach in deciphering the evolution of morphological traits in a highly specialized group of Neotropical catfishes: the Loricariinae. We have first inferred a molecular phylogeny of this group based on the 12S and 16S mitochondrial genes. The resulting phylogeny indicated that the subtribe Harttiini was restricted to the single genus Harttia, and within the subtribe Loricariini, two sister subtribes were distinguished, Sturisomina (new subtribe), and Loricariina. Among Loricariina, the morphological groups Loricariichthys and Loricaria + Pseudohemiodon were confirmed. The co-inertia analysis highlighted a strong relationship between the morphological and the genetic data sets, and identified three quantitative and eight qualitative variables linked to the phylogeny. The evolution of quantitative variables was assessed using the orthogram method and showed a major punctual event in the evolution of the number of caudal-fin rays, and a more gradual pattern of evolution of the number of teeth along the phylogeny. The evolution of qualitative variables was inferred using ancestral states reconstructions and highlighted parallel patterns of evolution in characters linked to the mouth, suggesting co-evolution of the traits for adapting to divergent substrates.