Aggregation of the amyloid-beta-42 (Abeta42) peptide in the brain parenchyma is a pathological hallmark of Alzheimer's disease (AD), and the prevention of Abeta aggregation has been proposed as a therapeutic intervention in AD. However, recent reports indicate that Abeta can form several different prefibrillar and fibrillar aggregates and that each aggregate may confer different pathogenic effects, suggesting that manipulation of Abeta42 aggregation may not only quantitatively but also qualitatively modify brain pathology. Here, we compare the pathogenicity of human Abeta42 mutants with differing tendencies to aggregate. We examined the aggregation-prone, EOFAD-related Arctic mutation (Abeta42Arc) and an artificial mutation (Abeta42art) that is known to suppress aggregation and toxicity of Abeta42 in vitro. In the Drosophila brain, Abeta42Arc formed more oligomers and deposits than did wild type Abeta42, while Abeta42art formed fewer oligomers and deposits. The severity of locomotor dysfunction and premature death positively correlated with the aggregation tendencies of Abeta peptides. Surprisingly, however, Abeta42art caused earlier onset of memory defects than Abeta42. More remarkably, each Abeta induced qualitatively different pathologies. Abeta42Arc caused greater neuron loss than did Abeta42, while Abeta42art flies showed the strongest neurite degeneration. This pattern of degeneration coincides with the distribution of Thioflavin S-stained Abeta aggregates: Abeta42Arc formed large deposits in the cell body, Abeta42art accumulated preferentially in the neurites, while Abeta42 accumulated in both locations. Our results demonstrate that manipulation of the aggregation propensity of Abeta42 does not simply change the level of toxicity, but can also result in qualitative shifts in the pathology induced in vivo.