Parallel robots with Shoenflies motions (also called 3T1R parallel robots) are increasingly being used in applications where precision is of great importance. Clearly, methods for evaluating the accuracy of these robots are therefore needed. The accuracy of well designed, manufactured, and calibrated parallel robots depends mostly on the input errors (sensor and control errors). Dexterity and other similar performance indices have often been used to evaluate indirectly the influence of input errors. However, industry needs a precise knowledge of the maximum orientation and position output errors at a given nominal configuration. An interval analysis method that can be adapted for this purpose has been proposed in the literature, but gives no kinematic insight into the problem of optimal design. In this paper, a simpler method is proposed based on a detailed error analysis of 3T1R fully-parallel robots that brings valuable understanding of the problem of error amplification.