The 'amyloid hypothesis' has guided research into Alzheimer's disease (AD) for more than a decade. A detailed review of the relevant data led us to conclude that some data, particularly those from transgenic mice, are inconsistent with the predictions of the amyloid hypothesis. Instead, most data are consistent with the notion that amyloid-beta (Abeta) peptide is neuroprotective. The majority of commentators agreed with our analysis but some were unwilling to abandon the amyloid hypothesis until the outcome of anti-Abeta therapeutic trials puts the matter beyond debate. All acknowledged that we had highlighted flaws in the amyloid hypothesis which must be addressed. To stimulate a critical reappraisal of the amyloid hypothesis we have proposed the 'bioflocculant hypothesis' which posits that Abeta serves to bind neurotoxic solutes (pathogens, proteins and metal ions) so that they can be phagocytosed and prevented from causing further damage. The hypothesis makes clear predictions that are readily falsifiable, and it has already gained credibility by predicting the recent negative outcome of Abeta vaccination trials in humans.