The yeast inheritable phenotype [URE3] is thought to result from conformational changes in the normally soluble and highly helical protein Ure2p. In vitro, the protein spontaneously forms long, straight, insoluble protein fibrils at neutral pH. Here we show that fibrils of intact Ure2p assembled in vitro do not possess the cross beta-structure of amyloid, but instead are formed by the polymerization of native-like helical subunits that retain the ability to bind substrate analogues. We further show that dissociation of the normally dimeric protein to its constituent monomers is a prerequisite for assembly into fibrils. By analysing the nature of early assembly intermediates, as well as fully assembled Ure2p fibrils using atomic force microscopy, and combining the results with experiments that probe the fidelity of the native fold in protein fibrils, we present a model for fibril formation, based on assembly of native-like monomers, driven by interactions between the N-terminal glutamine and asparagine-rich region and the C-terminal functional domain. The results provide a rationale for the effect of mutagenesis on prion formation and new insights into the mechanism by which this, and possibly other inheritable factors, can be propagated.