Magnesium diboride, MgB2, was recently observed to become superconducting at 39 K, which is the highest known transition temperature for a non-copper-oxide bulk material. Isotope-effect measurements, in which atoms are substituted by isotopes of different mass to systematically change the phonon frequencies, are one of the fundamental tests of the nature of the superconducting mechanism in a material. In a conventional Bardeen-Cooper-Schrieffer (BCS) superconductor, where the mechanism is mediated by electron-phonon coupling, the total isotope-effect coefficient (in this case, the sum of both the Mg and B coefficients) should be about 0.5. The boron isotope effect was previously shown to be large and that was sufficient to establish that MgB2 is a conventional superconductor, but the Mg effect has not hitherto been measured. Here we report the determination of the Mg isotope effect, which is small but measurable. The total reduced isotope-effect coefficient is 0.32, which is much lower than the value expected for a typical BCS superconductor. The low value could be due to complex materials properties, and would seem to require both a large electron-phonon coupling constant and a value of mu* (the repulsive electron-electron interaction) larger than found for most simple metals.