A new simpli®ed genetic classi®cation scheme for arcuate fold±thrust belts is proposed. Based on total strain patterns and displacement vector ®elds, we distinguish three extreme end-member models: (1)`Oroclines', pure bending of an initially straight belt, (2)`Piedmont glacier' with divergent transport directions and (3)`Primary arcs'. A simple geometric model set-up for the simulation of strain patterns in primary arcs with uniform transport direction demonstrates that divergent strain trajectories and rotations of passive marker lines do not require any divergence in displacement directions. These often quoted arguments are insu cient for the identi®cation of`Oroclinal bending' or`Piedmont glacier' type of arc formation. Only three-dimensional restorations of an arc provide the critical information about displacement directions. In their absence, arc parallel stretches and rotations in comparison with total strains provide the most useful criteria for the distinction of arc formation modes. As an example, the Jura fold±thrust belt of the external Alps is discussed. A large set of strain data includes total shortening estimates based on balanced cross-sections, local strain axes orientations from the inversion of fault populations [Homberg, C., 1996. Unpublished PhD thesis, Universite de Paris VI (France)], tectonic stylolites and micro-strains from twinning in sparry calcite. Strain trajectories (maximum shortening direction) computed from these data de®ne a strongly divergent fan with a 908 opening. A complete displacement vector ®eld for the entire Jura has been determined from balanced cross-sections augmented with three-dimensional`block mosaic' restorations [Philippe, Y., 1995. Unpublished PhD thesis, Universite de Chambe ry (France)]. Displacement vectors diverge by about 408, markedly less than strain trajectories. The non-parallelism between strain trajectories and transport directions indicates that considerable wrenching deformation did occur in both limbs of the Jura arc. Paleomagnetically determined clockwise rotations of 0±138 from ten sites (Kempf, O., et al., Terra Nova 10, 6±10) behind the right-hand half of the Jura arc and two sites with a combined 238 anticlockwise rotation behind the left-hand half of the arc are and additional argument in favor of such a wrenching deformation. We conclude that the Jura arc formed as a`Primary arc' with a minor component of`Piedmont glacier' type divergence in transport directions.