The semiconductor gallium nitride is the material at the centre of energy-efficient solid-state lighting and is becoming increasingly important in high-power and high-frequency electronics. Reducing the dislocation density of gallium nitride planar layers is important for improving the performance and reliability of devices, such as light-emitting diodes and high-electron-mobility transistors. The patterning of selective growth masks is one technique for forcing a three-dimensional growth mode in order to control the propagation of threading defects to the active device layers. The morphology of the three-dimensional growth front is determined by the relative growth rates of the different facets that are formed, and for GaN is typically limited by the slow-growing {1 - 101} facets. We demonstrate how the introduction of nanodash growth windows can be oriented in an array to preserve fast-growing {11 - 22} facets at the early stage of growth to accelerate coalescence of three-dimensional structures into a continuous GaN layer. Cathodoluminescence and Electron Channelling Contrast Imaging methods, both used to measure the threading dislocation density, reveal that the dislocations are organised and form a distinctive pattern according to the underlying mask. By optimising the arrangement of nanodashes and the nanodash density, the threading dislocation density of GaN on sapphire epilayers can be reduced significantly from 10(9) cm(-2) to 3.0 x 10(7) cm(-2). Raman spectroscopy, used to monitor the strain in the overgrown GaN epilayers, shows that the position of the GaN E-2(H) phonon mode peak was reduced as the dash density increases for a sample grown via pendeo-epitaxy whilst no obvious change was recorded for a sample grown via more conventional epitaxial lateral overgrowth. These results show how growth mask design can be used to circumvent limitations imposed by the growth dynamics. Moreover, they have revealed a greater understanding of the influence of the growth process on the dislocation density which will lead to higher performing electronic and optoelectronic devices as a result of the lower dislocation densities achieved. (C) 2017 The Authors. Published by Elsevier B.V.