Recent determinations of geomagnetic reversal ages using two independent dating techniques have led investigators to the same unexpected conclusion---many of the geomagnetic reversal ages estimated over the past three decades with potassium-argon (K-Ar) radiometric analysis are too young by up to 10%. The identification of this systematic error is important because an accurate and precise geomagnetic reversal time scale (GRTS) is essential for kinematic studies concerning variations in plate velocities and the surficial and mantle processes that control plate dynamics.
The first evidence that geomagnetic reversals younger than 5.3 Myr might be significantly
older than suggested by K-Ar radiometric dating came from reversal ages estimated from
marine-derived sediments exposed in southern Italy and deep-sea drilling cores
[ Shackleton et al., 1990; Hilgen, 1991]. These marine sequences contain high resolution
records of 
O oxygen isotope concentrations and CaC0
content, both
of which depend strongly on variations in insolation (solar radiation per unit area) that are
driven by variations in the earth's orbit. Such variations can be accurately predicted
by orbital models that incorporate changes in precession, obliquity, and eccentricity, each of which
varies with well-known periods. The inter-modulation of these three orbital
parameters yields a distinctive time series of insolation (and derivative quantities
such as ice volume) that can be compared to the depth records of 
O and
CaC0
in order to convert stratigraphic depth to stratigraphic age. The resultant
chronostratigraphies permit dating of magnetic reversals within the stratigraphic column.
Reversal ages determined with the relatively new 
Ar/
Ar radiometric dating technique
[e.g. Baksi et al., 1992; Spell and McDougall, 1992] strongly corroborate
reversal ages determined via the astrochronologic techniques used by
Shackleton et al. [1990] and Hilgen [1991ab].
The 
Ar/
Ar technique can be used to determine whether a mineral has
suffered loss or addition of argon gas after it has crystallized, thereby overcoming an important
source of systematic error in K-Ar dating, namely, that undetected diffusional loss or gain of argon
isotopes will lead to a biased radiometric age. The close agreement between
the astrochronologic and 
Ar/
Ar reversal ages, and their mutual disagreement with
reversal ages estimated from K-Ar age dating indicates that an important source of systematic
error in the GRTS has been discovered.
Efforts to minimize or eliminate errors in the GRTS caused by the use of reversal ages estimated from K-Ar age dating are already underway. Cande and Kent [1992] have derived ages for all Cenozoic and Late Cretaceous field reversals by fixing the ages for several key reversals to values given by reliable radiometric or biostratigraphic studies, and by further requiring that other reversal ages yield smoothly varying spreading rates along the southern Mid-Atlantic Ridge. Their work has sparked recent efforts to determine how the GRTS changes if the assumption of smoothly varying spreading between Africa and South America is relaxed or modified [ Baksi, 1993; Huestis and Acton, 1993].
To test whether or not the astronomically-based GRTS minimizes improbable large fluctuations in spreading rates over geologically brief intervals, Wilson [1993a] determined the post-5.3 Myr spreading histories of five spreading centers in the eastern Pacific and southern Atlantic. He found that seafloor spreading rates derived for four of the five spreading centers varied significantly less over the past 5.3 Myr if astronomically-derived reversal ages were used rather than ages derived from previous versions of the GRTS. Wilson's results suggest that errors in the astronomically-calibrated reversal ages are no greater than 20,000 years, allowing previous versions of the GRTS to be excluded with high confidence.
As a consequence of the well-documented need for revisions to the GRTS, the widely used NUVEL-1 model
of 3.0 Myr-average global plate velocities [ DeMets et al., 1990] has been recalibrated and renamed
NUVEL-1A [ DeMets et al., 1994]. NUVEL-1A has angular velocities 4.38% slower than NUVEL-1, as
well as decreased model covariances. The required decrease in global plate velocities
nearly eliminates [ Gordon, 1993; Baksi, 1994] a persistent 
4% difference between
plate velocities predicted by NUVEL-1 and 148 geodesic rates determined from satellite laser
ranging (SLR) or very long baseline interferometric (VLBI) measurements at 20 sites on five plates
[ Smith et al., 1992]. Geodesic rates predicted by NUVEL-1A agree remarkably well with observed
geodesic rates (Fig. 1), even though the latter are based on a decade or less of measurements.