Widespread use of 40Ar 39Ar incremental heating of hornblende as a geochronometer has prompted a detailed investigation of the mechanisms of Ar release from this mineral in vacuo in order to accurately interpret the isotopic data. A combination of high-temperature, in situ, in vacuo X-ray diffraction, and 40Ar 39Ar isothermal and step-heating runs was conducted on three hornblende samples with distinct compositions. A comparison of Ar retentivity in hornblende with chemical composition suggests that Mg-rich samples will lose Ar in vacuo at higher temperatures than Fe-rich samples; this may indirectly be a result of the smaller ionic radius of Mg2+ in comparison to Fe2+, as well as internal stress caused by Fe2+ oxidation. The oxidation of Fe2+ as a result of dehydrogenation appears to play an important role in controlling the release of Ar from the crystal structure by retarding structural decomposition of the mineral, and may also explain the one- vs. two-pulse nature characterizing the release of Ar from various hornblendes. Isothermal 40Ar 39Ar runs at different temperatures indicate that: (a) at temperatures at or above the structural decomposition of the hornblendes, Ar is not released via a volume diffusion- or reaction-controlled mechanism; and (b) D/a2-values from the low-temperature isothermal runs decrease exponentially with time, yielding estimated diffusivities that do not differ greatly from those obtained from hydrothermal experiments. This decrease in diffusivity with time also cannot be explained by either volume-diffusion or chemical-reaction processes, but might be explained by the presence of a diffusion mechanism much faster than volume diffusion, termed short-circuit (SC) diffusion. Very rough estimates of SC diffusion parameters for Ar yield a pre-exponential coefficient D0 ≈ 5.55·10-3 cm2 s-1 and activation energy E ≈ 45.0 kcal mol-1. More importantly, these results demonstrate that, in general, the incremental heating of hornblende in 40Ar 39Ar experiments will not necessarily reflect the true Ar distribution within the sample, because structural change and decomposition initiated by Fe oxidation and dehydroxylation will tend to homogenize any spatial gradients that originally exist in the mineral. For thermochronological studies, this means that plateau dates inferred from 40Ar 39Ar hornblende age spectra may not accurately reflect the time at which the mineral passed through its Ar closure temperature.