The muscle segments of fish have a folded shape, termed a chevron, which is thought to be optimal for the undulating body movements of swimming. However, the mechanism shaping the chevron during embryogenesis is not understood. Here, we used time-lapse microscopy of developing zebrafish embryos spanning the entire somitogenesis period to quantify the dynamics of chevron shape development. By comparing such time courses with the start of movements in wildtype zebrafish and analysing immobile mutants, we show that the previously implicated body movements do not play a role in chevron formation. Further, the monotonic increase of chevron angle along the anteroposterior axis revealed by our data constrains or rules out possible contributions by previously proposed mechanisms. In particular, we found that muscle pioneers are not required for chevron formation. We put forward a tension-andresistance mechanism involving interactions between intra-segmental tension and segment boundaries. To evaluate this mechanism, we derived and analysed a mechanical model of a chain of contractile and resisting elements. The predictions of this model were verified by comparison with experimental data. Altogether, our results support the notion that a simple physical mechanism suffices to self-organize the observed spatiotemporal pattern in chevron formation. ©2014. Published by The Company of Biologists Ltd.