One of the currently popular theories on magma ascent is that it mainly occ
urs by propagating hydrofractures (dykes) and that magma viscosity is the p
rimary rate-controlling factor. This theory is based on mathematical models
for single hydrofractures under idealised conditions. We simulated magma a
scent with air ascending through gelatine and observed that the air ascende
d in batches, following paths made by their predecessors. Multiple batches
accumulate at obstacles along the path. Although magma viscosity may contro
l ascent rate during movement, obstacles ultimately control the size and av
erage ascent velocity of ascending batches. We propose that step-wise movem
ent of magma batches is the mechanism of primary accumulation and ascent fr
om the partially molten source rock of a magma to its first emplacement sit
e and therefore the main ascent mechanism for granitic magmas. 'Classical'
dyking is the mechanism for secondary ascent from a magma chamber.