It is suggested that, if most star formation in the Galactic halo occu
rred in objects the size of globular clusters, then most of the spalla
tion reactions producing beryllium could also have occurred in such ob
jects, and that this could have led to the observed approximately line
ar relationship between beryllium and oxygen abundances. Furthermore,
it is plausible that the relative spallation rates in the halo and the
disc could lead to the beryllium-oxygen relationship observed at high
er metallicities.