The seeding mechanism by which neoclassical tearing modes (NTMs) are triggered is crucial to understanding the onset conditions for these modes. However, it is difficult to decouple measurements of the MHD activity involved in the mode onset from the seed perturbation itself, making direct verification of the onset criteria difficult. It is important to examine this, in order to validate the underlying physics being used to predict the requirements for larger devices such as ITER-FEAT. Following earlier work on COMPASS-D identifying NTM behaviour, and development of mode avoidance and removal techniques, new studies have focused on validating models of onset criteria. Using `resonant magnetic perturbation' windings, clear, identifiable and measurable rotating islands are produced in the plasma. By varying the timing and size of the heating power ramp-up, applied during the (natural) decay of such an island, the island evolution can be observed as a function of the key parameters - the values of normalized plasma pressure and island width. The resulting variation confirms the form predicted by theory, provided threshold mechanisms are incorporated, and favours the ion polarization model in particular.