Importance: To improve the ability to develop treatments that prevent incipient Alzheimer disease (AD) from progressing to overt AD, it is important to understand the molecular basis of the earliest pathophysiological abnormalities and to determine how amyloid-β (Aβ) is involved very early in its pathogenesis. Objective: To investigate 2 specific Aβ oligomers, Aβ trimers and Aβ*56, in human cerebrospinal fluid (CSF); evaluate the effects of aging and AD; and obtain support for the hypothesis that they may be pathogenic by determining their relationships to CSF tau. Design: A CSF sampling study. Settings: The University of Minnesota Medical School in Minneapolis, Minnesota, and the Salhgrenska University Hospital, Sweden. Participants: Forty-eight older adults with mild cognitive impairment or AD (impaired group); 49 agematched cognitively intact control subjects (unimpaired group); and 10 younger, normal control subjects. Main Outcome Measures: Measurements of CSF Aβ trimers, Aβ*56, the 42-amino acid Aβ isoform (Aβ1-42), total tau (T-tau), and phospho-tau 181 (p-tau181). The hypothesis being tested was formulated after data collection. Results: We observed that Aβ trimers and Aβ*56 levels increased with age; within the unimpaired group, they were elevated in subjects with T-tau/Aβ1-42 ratios greater than a cutoff that distinguished the unimpaired group from subjects with AD. In the unimpaired group, T-tau and p-tau181 were found to correlate strongly with Aβ trimers and Aβ*56 (r > 0.63), but not with Aβ1-42 (-0.10 < r < -0.01). The strong correlations were found to be attenuated in the impaired group. Conclusions and Relevance: In cognitively intact older adults, CSF Aβ trimers and Aβ*56 were elevated in individuals at risk for AD, and they showed stronger relationships with tau than did Aβ1-42, a surrogate for Aβ fibril deposition. These findings suggest that prior to overt symptoms, 1 or both of the Aβ oligomers, but not fibrillar Aβ, is coupled to tau; however, this coupling is weakened or broken when AD advances to symptomatic stages. The uncoupling is interesting in light of the failure of experimental Aβ therapies to improve mild cognitive impairment/AD, which has prompted a shift in the timing of Aβ therapies to asymptomatic subjects. Knowing which Aβ species to target in asymptomatic subjects may enhance the success of future treatments for AD.