The body muscles of fish are separated into segments by connective tissue sheets (myosepta) folded into a W-shape of slopes and cones. The biomechanical function of this complex shape is a long-standing question. Current hypotheses link myosepta shape to the magnitude of lateral bending during swimming, with morphological studies focusing on changes in myosepta shape across the mid- and caudal-regions of the body. We know little about how myosepta may contribute to dorsal bending during feeding, and comparative studies of myosepta morphology have been limited by the difficulty of visualising and measuring myosepta shape in 3D. How does myosepta shape vary from head to tail, and among species with differing dorsal bending during feeding? We used contrast-enhanced magnetic resonance imaging (MRI) to visualize cranial, dorsal (epaxial) myosepta in 7 fish species including different body shapes and feeding behaviours. We observed a surprising diversity of shapes across species. The more cranial myosepta of frogfish (Antennarius sp.), anglerfish (Chaenophryne draco), Cape dory (Zeus capensis), and pufferfish (Canthogaster punctissima) were relatively planar and cone-less myosepta, and bisected by a horizontal connective tissue sheet (except in pufferfish). More caudal myosepta in these species and all myosepta in trout (Oncorhynchus mykiss) and killifish (Fundulus parvipinnis) were more typically W-shaped. These results demonstrate myosepta shape can vary substantially across the spine and among species. We hypothesize that planar, cone-less myosepta may allow relatively large, local dorsal bending of the vertebral column. More data are needed to untangle the impact of functional and phylogeny on myosepta shape.