Colouration is an important tool for systematists inferring species limits and phylogenetic relationships of teleost fishes, but the use of colouration variation in trichomycterine catfish systematics has generated some controversy. We first report and describe the occurrence of four, geographically disjunct colour morphs in Trichomycterus albinotatus, endemic to south-eastern Brazil, as well as ontogenetic colouration change in each morph. A phylogenetic analysis using a cytb fragment (1098 bp) for 23 specimens representing all colour morphs and four outgroups did not support any correlation between colour morphs and lineages, with different colour morphs sharing identical haplotypes. This study indicated that young adult specimens found in lighter habitats had white and brown to black spots on the flank, whereas similar-sized specimens inhabiting darker habitats had white spots inconspicuous or absent and dark brown or black spots expanded. Individuals above about 65 mm SL of all populations had flank white marks less conspicuous or absent and cryptic habits during daylight, contrasting with smaller individuals with white marks and actively swimming above the substrate. Literature data indicate that ontogenetic colouration and habit changes occur in different trichomycterid lineages. Our data thus show that colouration may be problematic in taxonomical studies, although often being consistently used to diagnose species and clades. We conclude that colouration should not be discarded a priori as evidence of trichomycterine relationships and species limits, but should be used with caution in systematic studies, being necessary additional evidence, such as osteological characters or molecular data.