The potential of tumour tracking for active spot-scanned proton therapy was assessed. Using a 4D-dose calculation and simulated target motion, a tumour tracking algorithm has been implemented and applied to a simple target volume in both homogenous and heterogeneous in silico phantoms. For tracking and retracking (a hybrid solution combining tumour tracking and rescanning), three tracking modes were analysed: 'no tracking' (uncorrected irradiation of a moving target), 'perfect tracking' (no time delays and exact knowledge of target position) and 'imperfect tracking' (simulated time delays or position prediction errors). For all plans, dose homogeneity in the target volume was assessed as the difference between D5 and D95 in the CTV. For the homogeneous phantom, perfect tracking could retrieve nominal dose homogeneity for all motion phases and amplitudes while severe deterioration of treatment outcomes was found for imperfect tracking. The use of retracking reduced the sensitivity to position errors significantly in the homogeneous phantom. In the heterogeneous phantoms (simulated rib proximal to target), the nominal dose homogeneity could not be obtained with perfect tracking. Adjustments in pencil beam positions could cause pencil beams to deform under the influence of the bone, resulting in loss of dose homogeneity. As retracking was not capable of reducing these effects, rescanning provided the best treatment outcomes for moving heterogeneous targets in this study.