In vertebrate embryos, a process called somitogenesis lays the foundations of the adult spine. This process involves elongation and segmentation of the paraxial mesoderm to form somites. Although the segmentation aspect of this has been widely studied, the elongation aspect is not well understood. Posterior growth is widely assumed to be the main driver, but there is very little evidence for this – particularly in fast-developing species like zebrafish. In this thesis, I present the first long term, multi-scale, 3D characterisation of the zebrafish paraxial mesoderm, and show that this tissue elongates through some form of convergent extension, not through growth. In fact, the tissue is compressed over time, and so decreases in volume. I suggest that these processes may be functionally linked, and thus propose a novel mechanism of “compression-extension”. Cell tracking, agent-based modelling, and perturbations show that this form of convergent extension does not involve PCP-dependent directional intercalation but, instead, involves convergent flows of cells towards the midline and non-directional intercalation. The cause of compression is not clear, but perturbation experiments suggest that extrinsic forces from the neural tube and TGFβ signalling may be involved. Comparative work in cichlids, chickens, and catsharks suggests that tissue convergence is not unique to zebrafish, and instead is a conserved feature of paraxial mesoderm elongation – even in species that undergo high levels of growth during somitogenesis. This suggests that the relative contributions of growth and tissue convergence to the process of paraxial mesoderm elongation have evolved differently across vertebrate lineages, resulting in a spectrum of elongation strategies.