The choice of initial conditions in multifield inflation has a direct effect on observable quantities. Even after the attractor solution has been reached, the existence of multiple scalar degrees of freedom requires us to exclude certain inflationary trajectories when comparing their predictions to actual values. In the presence of observational uncertainties, this translates to fine-tuning for the initial values of the inflaton fields. In this talk, I will present a method that may be used to quantitatively distinguish between stable and unstable trajectories in general scalar-curvature models of inflation, and I will show how to quantify the degree of fine-tuning required for inflationary trajectories living in a curved field space.