Debate regarding migration of granitic magmas arises from the fact that the physical link between migmatites and plutons and batholiths is rarely documented. Within the Karakoram shear zone, Ladakh, NW India, the synkinematic transfer of magma can be traced from the anatectic source region to the Karakoram Batholith through a complex dike swarm. Melting is due to water fl uxing at upper-amphibolite facies, producing Miocene hornblende and garnet-two mica leucogranites. The anatectic zone is characterized by synmagmatic folding and shearing of migma tites, and by pervasive, irregular magma migration paths forming injection complexes. These irregular networks are linked to dike networks that are interpreted to represent a magma transfer zone characterized by a range of styles, such as (1) dike swarms parallel to the shear plane, (2) dikes following the two major foliation planes in the shear fabric, (3) magma sheets in conjugate pairs of ductile fractures, or (4) chaotic dike complexes. Network geometries are controlled by the regional stress fi eld, strain distribution, preexisting anisotropies, and rheological contrasts. Dikes give rise to anastomosing systems with rare crosscutting relationships, and with intersections oriented parallel to the dominant mineral stretching lineation. Dikes feed a number of sills, stocks, plutons, and, ultimately, the leucogranitic Karakoram Batholith. We propose that magma developed an interlinked, continuous dike swarm from source to sink by interacting with deformation and preexisting anisotropies. Unlike theoretical and laboratory predictions, dikes are typically oriented at high angles to the maximum shortening axis, and unlike expectations, water-fl uxed melting of the crust produced mobile magma that migrated to form large granitic bodies.