Evolutionary models have been calculated for Population II stars of 0.5-1.2 M from the pre-main sequence to the lower part of the giant branch. Models were calculated for Z = 0.017 × 10^-4 to 0.0068 ([Fe/H] = -4.31 to -0.71) to determine the effect of metallicity on the size of abundance anomalies to be expected from gravitational settling, thermal diffusion, and radiative accelerations. Rosseland opacities and radiative accelerations were calculated taking into account the concentration variations of 28 chemical species, including all species contributing to Rosseland opacities in the OPAL tables. It is shown that while radiative accelerations and gravitational settling may lead to abundance anomalies by factors of 2-10 in turnoff stars of metal-poor clusters such as M92, much smaller abundance anomalies are expected in relatively metal-rich globular clusters such as M5, M71, or 47 Tuc. Even in NGC 6397, which is only a factor of 2 more metal-rich than M92, atomic diffusion is expected to lead to smaller anomalies than in M92. In field stars with T_eff ≥ 6000 K and [Fe/H] < -2.3, the abundance anomalies might be even larger than in M92. Reduction of metallicity beyond [Fe/H] = -3.31 is shown not to cause further structural changes to models. Below that metallicity, all metals may be treated as trace elements. Comparisons are made to abundance observations in a number of clusters to determine if hydrodynamic processes competing with atomic diffusion are required by observations. For most metals the situation remains ambiguous: observations, taking into account the error bars, do not yet require additional processes. Monte Carlo simulations are used to show that the Spite plateau for Li in low-metallicity field stars remains the strongest argument for the presence of a process competing with atomic diffusion.