We compile a dataset of reliable palaeointensity estimates based both on pu
blished work and on new data from basaltic glass. The basaltic glass data m
ore than double the number of reliable (Thellier method with pTRM checks) p
alaeointensity estimates available. Although the new data dramatically impr
ove both spatial and temporal coverage, there is still a strong bias toward
the most recent past. The last 0.3 Ma claim over half of the data in our c
ombined database. We therefore divide the data into two groups, the densely
sampled last 0.3 Myr and the more sparsely sampled period of time comprisi
ng roughly half of the data from 0.3 to 300 Ma. Separating them in this way
, it is clear that the dipole moment of the Earth over the past 0.3 Myr (ca
. 8 x 10(22) A m(2)) is dramatically higher than the average dipole moment
over the preceding 300 Myr (ca. 5 x 10(22) A m(2)). Inclusion of poor-quali
ty results leads to an overestimate of the average dipole moment. Interesti
ngly, no other significant changes in the distribution of dipole moments ar
e evident over the 300 million year span of the data.