We propose a simple toy model to explain the 2:3:6 QPO structure in GRS 1915+105 and, more generally, the 2:3 QPO structure in XTE J1550-564, GRO J1655-40, and H 1743-322. The model exploits the onset of subharmonics in the context of diskoseismology. We suggest that the observed frequencies may be the consequence of a resonance between a fundamental g-mode and an unobservable p-wave. The results include the prediction that, as better data become available, a QPO with a frequency of twice the higher twin frequency and a large quality factor will be observed in twin peak sources, as it might already have been observed in the especially active GRS 1915+105.