The geomagnetic axial dipole (hereinafter denoted g10) is the largest component of our planet’s magnetic field. Its magnitude determines the morphology of solar-terrestrial electrical current systems and it is the most fundamental diagnostic property of the core-generated geodynamo. Elucidating past and future variations of g10(t) is consequently of central importance in geomagnetism. Previous historical geomagnetic field models, such as gufm1 of Jackson et al. [Jackson, A., Jonkers, A.R.T., Walker, M.R., 2000. Four centuries of geomagnetic secular variation from historical records. Philos. Trans. R. Soc. Lond. A 358, 957–990], used direct observations to constrain g10(t) only after 1840 A.D.; before this time a crude linear extrapolation of the post-1840 A.D. rate of change (15 nT/year) was employed. In this contribution I construct historical field models with g10(t) instead constrained from 1590 A.D. to 1840 A.D. by an archaeointensity dataset compiled by Korte et al. [Korte, M., Genevey, A., Constable, C.G., Frank, U., Schnepp, E., 2005. Continuous geomagnetic field models for the past 7 millennia. 1. A new global data compilation. Geochem. Geophys. Geosyst. 6, doi:10.1029/2004GC000800]. A range of possible linear models of the form g10(t)=g10(1840)+β(t−1840) are first explored; β=2.74±42.32 nT/year is found to explain the archaeointensity dataset with maximum likelihood, consistent with the recent findings of Gubbins et al. [Gubbins, D., Jones, A.L., Finlay, C.C., 2006. Fall in Earth’s magnetic field is erratic. Science 312, 900–902]. Relaxing the linear constraint in an effort to find more physically plausible models, I find it is necessary to artificially increase the weight given to the archaeointensity data in order to obtain acceptable models. Despite satisfactorily explaining both the historical and archaeointensity data, and possessing reasonable spatial and temporal complexity, such free evolution models perform worse than the simpler linearly constrained models when tested against the independent dataset of Gallet et al. [Gallet, Y., Genevey, A., Fluteau, F., 2005. Does Earth’s magnetic field secular variation control centennial climate change? Earth Plan. Sci. Lett. 236, 159–173]. Bayesian model comparison techniques indicate that a model (gufm1-g10c) involving no change in g10(t) between 1590 A.D. and 1840 A.D. is most probable given the presently available data and current modelling techniques. I propose that this new, empirically derived, constraint on the evolution of the geomagnetic axial dipole be incorporated into the next generation of historical field models.