This paper reports a striking misperception associated with involuntary saccadic eye movements: when subjects are instructed to look to the opposite side of a suddenly presented stimulus (antisaccade), they produce a certain number of involuntary prosaccades to the stimulus before they move their eyes to the other side by a corrective saccade of approximately twice the size. When asked to indicate at the end of each trial whether they believed that they made such a detour sequence of two saccades, one finds that, on average, 50+/-25% of these involuntary movements are not recognized. The average size and correction time for recognized prosaccades is larger than unrecognized prosaccades, while their mean reaction times are the same. The corrective saccades compensate for the size of both the recognized and unrecognized errors. When similar sequences of saccades are made voluntarily, the time spent at the stimulus side was 222 ms compared with 95 ms for unrecognized and 145 ms for recognized errors. The distributions of the corresponding correction times differ in their multimodal composition. Whether voluntary and involuntary saccades and their corrections are associated with different effects on the updating of the perceptual spatial frame and attention allocation is discussed.