Microtextural changes brought about by heating alkali feldspar crystals from the Shap granite, northern England, at atmospheric pressure, have been studied using transmission and scanning electron microscopy. A typical unheated phenocryst from Shap is composed of about 70 vol% of tweed orthoclase with strain-controlled coherent or semicoherent micro- and crypto-perthitic albite lamellae, with maximum lamellar thicknesses 700°C, and after >48 h at 700°C, all such regions were essentially compositionally homogeneous, as indicated by X-ray analyses at fine scale in the transmission electron microscope. Changes in lamellar thickness with time at different T point to an activation energy of ~350 kJmol−1. A lamella which homogenised after 6,800 h at 600°C, therefore, would have required only 0.6 s to do so in the melting interval at 1,100°C. Subgrains in patch perthite homogenised more slowly than coherent lamellae and chemical gradients in patches persisted for >5,000 h at 700°C. Homogenisation T is in agreement with experimentally determined solvi for coherent ordered intergrowths, when a 50–100°C increase in T for An1 is applied. Homogenisation of lamellae appears to proceed in an unexpected manner: two smooth interfaces, microstructurally sharp, advance from the original interfaces toward the mid-line of each twinned, semicoherent lamella. In places, the homogenisation interfaces have shapes reflecting the local arrangements of nanotunnels or pull-aparts. Analyses confirm that the change in alkali composition is also relatively sharp at these interfaces. Si–Al disordering is far slower than alkali homogenisation so that tweed texture in orthoclase, tartan twinning in irregular microcline, and Albite twins in albite lamellae and patches persisted in all our experiments, including 5,478 h at 700°C, 148 h at 1,000°C and 5 h at 1,100°C, even though the ensemble in each case was chemically homogeneous. Nanotunnels and pull-aparts were modified after only 50 min at 500°C following the simulated 40Ar/39Ar step-heating schedule. New features called ‘slots’ developed away from albite lamellae, often with planar traces linking slots to the closest lamella. Slot arrays were often aligned along ghost-like regions of diffraction contrast which may mark the original edges of lamellae. We suggest that the slot arrays result from healing of pull-aparts containing fluid. At 700°C and above, the dominant defects were subspherical ‘bubbles’, which evolved from slots or from regions of deuteric coarsening. The small degree of partial melting observed after 5 h at 1,100°C was often in the vicinity of bubbles. Larger micropores, which formed at subgrain boundaries in patch perthite during deuteric coarsening, retain their shape up to the melting point, as do the subgrain boundaries themselves. It is clear that modification of defects providing potential fast pathways for diffusion in granitic alkali feldspars begins below 500°C and that defect character progressively changes up to, and beyond, the onset of melting.