Amyloid formation and aberrant protein aggregation have been implicated in more than 15 different human diseases and an even wider range of proteins form amyloid in vitro. From a structural perspective the proteins which form amyloid can be divided into two classes: those which adopt a compact globular fold and must presumably at least partially unfold to form amyloid and those which are unstructured in their monomeric state. Important examples of the latter include the Aβ peptide of Alzheimer's disease, atrial natriuretic factor, calcitonin, pro-calcitonin, islet amyloid polypeptide (IAPP, amylin), α-synuclein and the medin polypeptide. The kinetics of amyloid assembly are complex and typically involve a lag phase during which little or no fibril material is formed, followed by a rapid growth stage leading to the β-sheet-rich amyloid structure. Increasing evidence suggests that some natively unfolded polypeptides populate a helical intermediate during the lag phase. We propose a model in which early oligomerization is linked to helix formation and is promoted by helix–helix association. Recent work has highlighted the potential importance of polypeptide membrane interactions in amyloid formation and helical intermediates appear to play an important role here as well. Characterization of helical intermediates is experimentally challenging but new spectroscopic techniques are emerging which hold considerable promise and even have the potential to provide residue specific information.