During a brief period of 10 to 15 million years in the Middle-Ordovician, the Michigan Basin departed from its bull's-eyes subsidence pattern and tilted toward the east, opening to the Appalachian basin. This tilting is observed in maps of tectonic subsidence estimated for the Black River and Trenton Formations and extends over 300 km across the Michigan Basin and into eastern Wisconsin. Contours of constant tectonic subsidence rate are approximately parallel to the inferred position of the Laurentian-Iapetus convergent margin. The distance between the inferred position of the subduction zone to the limit of tilting is approximately 1000 km. Three alternative models for the tilting are tested, two relying on the rigidity of the continental lithosphere and a third on the viscous flow generated by a subducted slab. In the first elastic model we assume the edge of the elastic plate is simply loaded from above (by a fold and thrust sheet, volcanic pile, etc.). This model, however, cannot simultaneously satisfy the space of tectonic subsidence and subsidence rate, even for lithospheres which have a rigidity of 1028 Nm. In the second elastic model, the Laurentian continental margin descends into a trench of an eastward dipping slab, that is, a slab descending under the Iapetus ocean. This process cannot generate any significant far field displacements, even for extremely rigid plates, and must be rejected. For the third model we use finite element solutions of a negatively buoyant slab in a viscous medium with a faulted lithosphere. Such slabs can easily generate not only realistic trenches on the under thrusting plate but also significantly tilt the lithosphere as much as 1000 km from the plate margin.

The magnitude and distribution of far-field displacements depend on the age, length, and dip angle of the slab. In contrast to the elastic models, penetration of a west-dipping slab beneath the continent can reproduce both the extent, magnitude, and rate of tectonic subsidence observed in the Trenton and Black River Formations. The observed data are best fit by an old slab (140 Ma), which initially descend at a moderate dip (20¿--30¿) for 10--15 m.y., which then steepened as the slab penetrated deeper into the mantle. At the position of the previous Ordovician plate margin, there are narrow, block faulted basins which underwent rapid subsidence ~10 m.y. before the Michigan Basin tilted toward the east. We propose that these earlier subsidence events were caused by the initial descent of the slab under the preexisting Cambro-Ordovician passive margin. The time lag of ~10 m.y. may be due to the time it takes the slab to penetrate the upper mantle. This result is important for understanding the time evolution of mantle convection and mechanisms for the initiation of subduction. ¿ American Geophysical Union 1995