The current study provided longitudinal robust norms for individuals age 70 years and older for several neuropsychological tests. We compared baseline neuropsychological test performance in three groups free of dementia at baseline: a robust normative sample free of dementia for at least two post-follow-up assessments, an incident dementia sample which developed new onset dementia during the follow-up and a lost to follow-up (LTF) sample. ANCOVAs showed that the robust sample performed better on all neuropsychological tests compared to the incident dementia and LTF samples. These findings support the argument that individuals in transition to developing dementia may reduce the mean, increase the variability and therefore underestimate cognitive performance in normal aging. We suggest that longitudinal robust norms may help mitigate the limitations inherent in cross-sectional normative samples.