In the last years, paleomagnetism has been increasingly used to provide emplacement ages of loosely dated volcanics. Dating is achieved by comparison of paleomagnetic directions with a given reference curve of the paleosecular variation (PSV) of the geomagnetic field. Recently, a debate has developed on the achievable precision (the α95 value) of the paleomagnetic directions and hence on the accuracy that "paleomagnetic dating" can yield. At 39 different sites from Etna we paleomagnetically investigated 13 flows (four "test flows" with known age, and nine loosely dated flows), emplaced between 122 B.C. and 1865 A.D. We systematically drilled 12 cores per flow spaced in three (far from each other) sites and demagnetized one specimen per core by alternating field cleaning. Results from the four test flows yield age windows effectively encompassing the respective true flow ages, when dating based on Bayesian statistics at a 95% confidence level is adopted. We find α95 values for the flow mean directions ranging between 3.3¿ and 5.7¿ (4.5¿ on average), which translate into accuracies of age determinations of 136--661 years (307 years on average). Such dating uncertainty is likely underestimated, as we disregarded several kinds of errors that might affect both the fidelity of paleomagnetic recording and the PSV reference curve. The strong magnetization of both the underlying terrain and the cooling flow itself and mineral magnetic variations across the flows are the most likely sources for the scatter characterizing the recording process of the magnetic field in volcanic rocks.