Although it is evident that Taiwan has been formed by the collision of the west-facing Luzon arc with the Eurasian continental margin, there remain a lot of enigmas in this collision. The major ones are: (1) a transform fault presently connecting the Manila and Ryukyu Trenches in the Philippine Sea - Eurasia relative motion direction is missing, and instead, the Ryukyu Trench extends near offshore E. Taiwan, (2) the western edge of the intermediate-depth seismicity associated with the Philippine Sea plate subduction beneath NE Taiwan has a NNW trend, not NW, and (3) a large negative Bouguer gravity anomaly, with undulation of its amplitude in a wavelength of 60 - 80 km, exists along the southernmost Ryukyu forearc.

We propose a new model of the collision in Taiwan to resolve these enigmas, assuming that the southern Ryukyu forearc was migrating to the southwest with respect to Eurasia for the past several m.y. and the Luzon arc has been colliding with this actively migrating Ryukyu forearc. The northernmost Luzon arc is divided into two parts by the NNW line directing along the Philippine Sea - Ryukyu forearc motion from its initial intersection point with the Ryukyu Trench; the part west of this line has been obducted on the Ryukyu forearc-Eurasian margin, producing the collision orogen in Taiwan, and the part east of it has been subducted beneath the Ryukyu forearc.

This evolutionary scenario resolves enigmas (1) and (2) kinematically. This model also predicts that the South China Sea slab has to be torn by the westward component of the motion of the subducting Philippine Sea slab to Eurasia. This would have brought large lateral compression in the shallow portion of the Philippine Sea slab at its western border, which might lead to buckling of the slab causing the observed undulated gravity anomaly.