Autopsy is the diagnostic gold standard for progressive supranuclear palsy (PSP). The National Institute of Neurological Disorders and Stroke and Society for Progressive Supranuclear Palsy (NINDS-SPSP) criteria for the clinical diagnosis of "probable" PSP are thought to possess high specificity and low sensitivity. The NINDS-SPSP criteria for "possible" PSP are considered to increase sensitivity at the expense of specificity. The Neuroprotection and Natural History in Parkinson Plus Syndromes (NNIPPS) criteria are intended to improve sensitivity while maintaining high specificity. The aim of this study was to conduct a clinicopathological evaluation of the NINDS-SPSP and NNIPPS criteria in tertiary neurological centers. Defined clinical features and their year of onset were recorded by chart review in neuropathologically diagnosed patients with PSP, Parkinsons's disease (PD), MSA parkinsonism and corticobasal degeneration from four European brain banks. Fulfillment of the clinical diagnostic criteria was verified for each year after disease onset and for the final antemortem record. We analyzed 98 PSP patients and 46 disease controls. The NINDS-SPSP "probable" criteria yielded shorter time to diagnosis, slightly higher specificity and positive predictive value (PPV), and similar sensitivity, compared with the NNIPPS criteria. Unexpectedly, the NINDS-SPSP "possible" criteria yielded the lowest sensitivity, specificity, and PPV. A combination of NINDS-SPSP possible and probable criteria yielded the highest sensitivity. We suggest that the NINDS-SPSP probable criteria might be preferred for recruitment of patients for clinical trials, where an early and specific diagnosis is important. For routine clinical care, where high sensitivity is crucial, a combination of NINDS possible and probable criteria might be preferred. Copyright © 2013 Movement Disorder Society.