The diagnosis and treatment of diseases involving tau-based pathology such as Alzheimer disease and certain frontotemporal dementias is hampered by the inability to detect pathological forms of tau with sufficient sensitivity， specificity and practicality. In these neurodegenerative diseases， tau accumulates in self-seeding filaments. For example， Pick disease (PiD) is associated with frontotemporal degeneration and accumulation of 3-repeat (3R) tau isoforms in filaments constituting Pick bodies. Exploiting the self-seeding activity of tau deposits， and using a 3R tau fragment as a substrate， we have developed an assay (tau RT-QuIC) that can detect tau seeds in 2 µl aliquots of PiD brain dilutions down to 10-10. PiD seeding activities were 100-fold higher in frontal and temporal lobes compared to cerebellar cortex. Strikingly， this test was 10- to 10-fold less responsive when seeded with brain containing predominant 4-repeat (4R) tau aggregates from cases of corticobasal degeneration， argyrophilic grain disease， and progressive supranuclear palsy. Alzheimer disease brain， with 3R + 4R tau deposits， also gave much weaker responses than PiD brain. When applied to cerebrospinal fluid samples (5 µl)， tau RT-QuIC analyses discriminated PiD from non-PiD cases. These findings demonstrate that abnormal tau aggregates can be detected with high sensitivity and disease-specificity in crude tissue and fluid samples. Accordingly， this tau RT-QuIC assay exemplifies a new approach to diagnosing tauopathies and monitoring therapeutic trials using aggregated tau itself as a biomarker.