Inclusion of an informant yields strong associations between cognitive complaint and longitudinal cognitive outcomes in non-demented elders.