We report a four-component partial least squares discriminant analysis (PLS) model for the prediction of blood-brain barrier (BBB) permeation using descriptors derived from 3D molecular fields. The 3D fields were transformed by VolSurf into suitable 1D descriptors, which were correlated to the ratio of blood-brain partitioning measured at steady state in rats (log C(brain)/C(blood)). The model so obtained sheds light on molecular properties influencing BBB permeation. It can also be used in the virtual screening of new chemicals.