Increasing evidence of the incompatibility between the geologic thermal histories deduced from the 40Ar/39Ar incremental-heating technique and those derived from other methods has necessitated a careful evaluation of the assumptions inherent in the former approach. 40Ar/39Ar age spectra are frequently used to infer the spatial distribution of Ar in hydrous minerals and, as such, require that the mineral remains stable while heated in vacuo. Aliquots of Mmhb-1, an inter-laboratory hornblende standard used for K-Ar and 40Ar/39Ar dating, were heated in vacuo at temperatures of 500, 700, 930, and 1050 °C for 30 min. Up to 700 °C, no physical changes in the hornblende were observed with the scanning electron microscope, although X-ray diffraction data indicate that it had dehydroxylated to oxyhornblende at this temperature. The onset of the structural decomposition of the oxyhornblende, as revealed by backscattered- and secondary-electron images, occurs at ∼930 °C and is directly correlated with the initial release of 6% of its radiogenic Ar (40Ar*). The breakdown of the oxyhornblende by the internal nucleation of incongruent melt is almost (93%) complete by 1050 °C—the reaction products being titano-magnetite, melt, a "pyribole" phase, and internal voids. Because Ar is released simultaneously from these products and is subsequently homogenized in vacuo, this breakdown can yield plateau age spectra, regardless of any original Ar spatial gradients in the mineral. Consequently, independent evidence may be required to establish whether a plateau date from hornblende is truly a geologically significant age.