Observed molecular clouds often appear to have very low star formation effi
ciencies and lifetimes an order of magnitude longer than their free-fall ti
mes. Their support is attributed to the random supersonic motions observed
in them. We study the support of molecular clouds against gravitational col
lapse by supersonic, gasdynamical turbulence using direct numerical simulat
ion. Computations with two different algorithms are compared: a particle-ba
sed, Lagrangian method (smoothed particle hydrodynamics [SPH]) and a grid-b
ased, Eulerian, second-order method (ZEUS). The effects of both algorithm a
nd resolution can be studied with this method. We find that, under typical
molecular cloud conditions, global collapse can indeed be prevented, but de
nsity enhancements caused by strong shocks nevertheless become gravitationa
lly unstable and collapse into dense cores and, presumably, stars. The occu
rrence and efficiency of local collapse decreases as the driving wavelength
decreases and the driving strength increases. It appears that local collap
se can be prevented entirely only with unrealistically short wavelength dri
ving, but observed core formation rates can be reproduced with more realist
ic driving. At high collapse rates, cores are formed on short timescales in
coherent structures with high efficiency, while at low collapse rates they
are scattered randomly throughout the region and exhibit considerable age
spread. We suggest that this naturally explains the observed distinction be
tween isolated and clustered star formation.