An outstanding geophysical issue concerns the nature, and dynamical ro
le in the mantle general circulation, of the seismically fast body wav
e anomalies that have been tomographically imaged beneath continents.
In this paper, we investigate the possibilities that these seismologic
ally imaged ''roots'' represent either neutrally buoyant, chemically d
istinct material or cold, negatively buoyant, upper mantle and transit
ion zone downwelling flow. In assessing these alternatives, we first c
onstruct disaggregated models of the seismic heterogeneity in which a
component associated with subcontinental fast anomalies is isolated fr
om the global tomographic models either by employing the ''continent f
unction'' or a new ''craton function'' We find that the use of the new
craton function leads to geophysically more realistic chemical models
of subcontinental heterogeneity. The thermal and chemical density fie
lds reconstructed from the disaggregated tomographic models are employ
ed to compute the long-wavelength nonhydrostatic geoid, the free-air g
ravity field and the upper mantle radial flow pattern within the frame
work of an anelastically compressible internal loading theory We find
that the radial component of how velocity provides useful insight into
the dynamical implications of the alternative density models. However
, since this field is not directly observable, we consider the geoid a
nd free-air gravity anomaly as possible diagnostic discriminants and s
how that the free-air gravity anomaly provides a sensitive discriminan
t of the gravitational differences that characterize the chemical and
thermal models, whereas the geoid does not. By focusing on the free-ai
r gravity low over the Hudson Bay region of Canada, we are able to rul
e out the hypothesis that positively or neutrally buoyant subcontinent
al material that is chemically distinct from the surrounding mantle ex
ists below the Laurentian craton. However, when the fast body wave ano
maly is mapped into a high-density downwelling flow beneath this regio
n, we are able to fully explain the fraction of the anomalous free-air
gravity low which is inexplicable as a contribution associated with t
he existing degree of glacial isostatic disequilibrium due to the disi
ntegration of the Laurentide ice sheet. This conclusion concerning the
North American craton may be equally valid for other continental nucl
ei. We explore the general tectonophysical implications of this dynami
cal model.