The parallax e †ect in ground-based microlensing (ML) observations consists of a distortion to the standard ML light curve arising from the EarthÏs orbital motion. This can be used to partially remove the degeneracy among the system parameters in the event timescale,In most cases, the resolution in t 0 . current ML surveys is not accurate enough to observe this e †ect, but parallax could conceivably be detected with frequent follow-up observations of ML events in progress, providing the photometric errors are small enough. We calculate the expected fraction of ML events where the shape distortions will be observable by such follow-up observations, adopting Galactic models for the lens and source distributions that are consistent with observed microlensing timescale distributions. We study the dependence of the rates for parallax-shifted events on the frequency of follow-up observations and on the precision of the photometry. For example, we Ðnd that for hourly observations with typical photometric errors of 0.01 mag, 6% of events where the lens is in the bulge, and 31% of events where the lens is in the disk (or B10% of events overall), will give rise to a measurable parallax shift at the 95% conÐdence level. These fractions may be increased by improved photometric accuracy and increased sampling frequency. While long-duration events are favored, the surveys would be e †ective in picking out such distortions in events with timescales as low as days. We study the dependence of these fractions on t 0 B 20 the assumed disk mass function and Ðnd that a higher parallax incidence is favored by mass functions with higher mean masses. Parallax measurements yield the reduced transverse speed, which gives both v8 , the relative transverse speed and lens mass as a function of distance. We give examples of the accuracies with which may be measured in typical parallax events. Fitting ML light curves, which may be shapev8 distorted (e.g., by parallax, blending, etc.), with only the three standard ML parameters can result in inferred values for these quantities that are signiÐcantly in error. Using our model, we study the e †ects of such systematic errors and Ðnd that, due primarily to blending, the inferred timescales from such Ðts, for events with disk lenses, tend to shift the event duration distribution by B10% toward shorter Events t 0 . where the lens resides in the bulge are essentially una †ected. In both cases, the impact parameter distribution is depressed slightly at both the low and high ends.