Major puzzling aspects of the New Madrid seismic zone (NMSZ) are addressed in a comprehensive seismotectonic model based on past investigations and recent results. The reason why the NMSZ exists has been a subject of both scientific curiosity and controversy. The midcontinent setting, lack of topography and small measured strain rates are at odds with historic and paleoseismologic evidence for several large earthquakes over the last ~4000 years. In the seismotectonic model, NMSZ faults in the Mississippi Valley graben (MVG) are part of a broader, northeast trending system of strike-slip basement faults. The northwest trending Reelfoot fault occurs within an axial ring dike intrusion, the presence of which necessitates the occurrence of reverse faulting. Ongoing strain accumulation in the NMSZ is driven by perturbations to a low viscosity upper mantle region below the MVG produced by heat and metasomatizing fluids emanating from a thick segment of the Farallon slab located in the mantle transition zone. Two important sources of evidence for the proposed seismotectonic model are discussed: near surface features and the presence of the Farallon slab. The NMSZ is covered by hundreds of meters of unconsolidated sediments in the Mississippi Embayment, but seismic reflection studies indicate the presence of numerous northeast trending, near-vertical faults in the crust. A new set of focal mechanisms reveals that strike-slip faulting occurs on all major NMSZ arms of seismicity including the Reelfoot fault, commonly interpreted as a left step-over reverse fault in between the northeast trending right-lateral Axial and North strike-slip faults. Strike-slip faulting is as prevalent as reverse faulting on the Reelfoot fault and significant strike-slip motion may occur during large earthquakes. The portion of the flat Farallon slab below the MVG contains the thick Hess plateau. Metasomatizing fluids emanating from the plateau rise below the thinned Embayment lithosphere and markedly reduce mantle velocities below the NMSZ. The presence of a low velocity, weak region lends credence to previously proposed models for the generation of NMSZ earthquakes that satisfy the generation of repeated large earthquakes and small surface strain rates.