We are still far from understanding "somitogenesis" as a whole, but there is an emerging picture of the tissue interactions and molecular mechanisms that underlie and govern various aspects of this essential multistep patterning process in vertebrates. The ability to form segmental units appears to be a property specific to the paraxial mesoderm (as opposed to lateral or limb mesoderm), and this ability is probably acquired during early development, when paraxial mesoderm is specified and emerges from the primitive streak. Signaling molecules expressed in the primitive streak and tail bud are prime candidates involved in specifying paraxial (as well as other mesodermal) fates. Increasing levels of signaling molecules may be required in posterior regions of the embryo, and combinatorial signals may be essential to specify the paraxial mesoderm along the entire anterior-posterior axis. However, most of the pivotal signals, and the ways in which they are integrated and interact, remain enigmatic. Once the paraxial mesoderm is formed, segmentation proceeds largely without the requirement for continuous interactions with surrounding tissues. Somitomeres represent a morphologic pattern in the mesenchymal presomitic mesoderm, but their significance for somite formation is unclear. Molecular patterns are established in the presomitic mesoderm and probably are of functional significance. Cell interactions within the paraxial mesoderm appear to be involved in forming segment borders and ensuring their maintenance during subsequent differentiation of somites. These interactions are, at least in part, mediated by components of the conserved Notch signaling pathway, which may have multiple functions during somitogenesis. Epithelial somites are clearly a result of segmentation, but epithelialization is not the mechanism to form segments, supporting the idea that the basic mechanisms that govern segmentation in the mesoderm of vertebrates are very similar in different species despite divergent types of resulting segments (i.e., epithelial somites versus rotated myotomes). Concomitantly with segmentation, segment polarity and positional specification are established. How these processes are linked to, and depend on, each other is unknown, as is how they are regulated and how segmentation is coordinated on both sides of the neural tube. In contrast to early patterning in the presomitic mesoderm, patterning of the mature somites during their subsequent differentiation is the result of extensive tissue interactions. Virtually all tissues in close proximity to somites provide signals that are involved in induction or inhibition of particular differentiation pathways, but how these pathways are initiated is less clear. Some of the molecules mediating inductive signals and tissue interactions are known, and a growing number of candidate genes are potentially involved in regulating various steps of somitogenesis. The roles of these genes have yet to be analyzed. In addition, the molecular genetic analysis of mutations affecting somitogenesis, which were collected in the mouse and more recently in the zebrafish (Driever et al., 1996; Haffter et al., 1996; van Eeden et al., 1996), promises to add important new insights into this process. Much remains to be done, but the tools are at hand to provide further understanding of the molecular mechanisms underlying somitogenesis.