Recent Neoproterozoic climate modelling experiments using different sea ice components have yielded opposing results. Here we have attempted to reconcile these differences through the implementation of various representations of sea ice in the University of Victoria Earth System Climate Model. In particular, using a sea ice model similar to that of the Fast Ocean Atmosphere Model version 1.4, global glaciation is not achievable, whereas it occurs with an energy-conserving version of a more physically complete thermodynamic model. We argue that using a physically incomplete representation of sea ice may be a major factor in the inability of some previous coupled models to simulate global glaciations.