Soluble oligomers formed by amyloidogenic intrinsically disordered proteins are some of the most cytotoxic species linked to neurodegeneration. Due to the transient and heterogeneous nature of such oligomeric intermediates, the underlying self-association events often remain elusive. NMR relaxation measurements sensitive to zero-frequency spectral densities (J(0)), such as the N-15 - R-2 rates, are ideally suited to map sites of self-association at atomic resolution without the need of exogenous labels. Such experiments exploit the dynamic exchange between NMR visible monomers and slowly tumbling oligomers. However,N-15 - R-2 rates are also sensitive to intrinsic monomer dynamics, and it is often difficult to discern these contributions from those arising from exchange with oligomers. Another challenge pertains to defining a hierarchy of self-association. Here, using the archetypical amyloidogenic protein alpha synuclein (alpha S), we show that the temperature-dependence of N-15 - R-2 effectively identifies self-association sites with reduced bias from internal dynamics. The key signature of the residues involved in self-association is a nonlinear temperature-dependence of N-15 - R-2 with a positive Delta R-2/Delta T slope. These two hallmarks are systematically probed through a thermal R-2 correlation matrix, from which the network of residues involved in self-association as well as the hierarchy of alpha S self-association sites is extracted through agglomerative clustering. We find that aggregation is initiated by residues within the NAC region that is solvent inaccessible in alpha S fibrils and eventually extends to the N-terminal segment harboring familial PD mutations. These hierarchical self-association maps help dissect the essential drivers of oligomerization and reveal how amyloid inhibitors affect oligomer formation.