It has recently been suggested that the nervous system employs forward models for the purpose of motor control. The evidence for this hypothetical computational structure comes from various sources, theoretical and experimental. The purpose of this commentary is to distinguish between three different structures that are used to support the forward model hypothesis: (1) output predictor, (2) state estimator, and (3) distal teacher. It is possible that the brain employs all these structures for control. However, the general term forward model could be misleading since evidence for one structure could be misinterpreted as supporting evidence for the other structures. q 2002 Elsevier Science Ltd. All rights reserved.