Continental drift and true polar wander are distinguished. If any true polar wander has occurred, it would appear as a rigid rotation of the lithosphere relative to a fixed spin axis. A mathematical method is proposed to decompose a general displacement field on a sphere into two parts: a part due to a rigid rotation and a remaining part of random motions. The rigid rotation found is the one that best fits the observed displacement field in a least squares sense over the surface of the earth. The random motions, which are separated from the rigid rotation, are then ascribed to continental drift. Formulas are developed that give the Euler angles of the best-fitting rotation in terms of the observed plate displacements for the case of small displacements. These formulas for the Euler angles are also expressed in terms of the first-order spherical harmonic coefficients of the potential generating the displacement field. The proposed method is applied to three theoretical displacement fields having an analytic expression to illustrate the decomposition of specific fields. The displacement field of the plates is constructed for the time period since the early Tertiary, and the method is applied to determine the amount, if any, of true polar wander. The best-fitting rotation, or true polar wandering, was found to be only 2°, an amount less than the uncertainty (4°) of the mean paleomagnetic pole used.