A number of diamond-bearing Canyon Diablo specimens have been investigated in order to determine whether the diamonds were produced under high gravitational pressures in a meteorite parent body, or upon impact with the earth, as proposed by UREY and NININGER. respectively. Metallographic studies confirm NININGER'S observation that the diamond-bearing fragments, and only these, were reheated strongly after the formation of the Widmanstätten pattern. The metal phase appears to have been reheated to ~950°C for 1-5 sec, followed by cooling in less than 2 min. The rapid cooling rate implies that the process took place after the meteorite fragments had attained their present, small size. Thermodynamic calculations and structural studies indicate that cohenite (Fe 3C), rather than graphite is the likely precursor of diamonds in iron meteorites. This fact severely limits the possible range of conditions for diamond formation. The diamonds are always associated with troilite. which shows signs of having been heated to much higher temperatures than the metal, suggesting that the heating took place by a shock wave upon impact with the earth. The growth of diamonds must have occurred on cooling or pressure drop, but it is not yet clear whether high pressures (e.g. from the compression during impact shock, or localized stresses) were at all required, or whether the diamonds formed as a metastable product at moderate to low pressures. Possible origins of diamonds in other meteorites are discussed. It is shown that in the presence of free iron, formation under high gravitational pressures could take place only under very extraordinary circumstances. Instead, it is considered much more likely that all meteoritic diamonds were produced by catastrophic events: either upon impact with the earth, or during the breakup of the meteorite parent bodies. This mode of origin obviates the need for postulating meteorite parent bodies of lunar or planetary size, with interior pressures of 3 × 10 4 atm or greater. The laboratory formation of diamond is discussed in terms of metal carbide dissociation equilibria, and several features of the General Electric process arc explained in this manner. Some comments are offered on MOISSAN'S experiments.