1. Experienced typists typed phrases containing words in which one isolated letter was typed with one hand, while the remaining letters were typed with the contralateral hand. 2. The translational and rotational motion of the fingers and wrist of the right hand were obtained optoelectronically from the location of reflective markers placed on the fingers. 3. Midway through the experiment, the key corresponding to the isolated letter was physically switched with another key on the keyboard, and subjects typed the letter in its new location (for 140 trials). The letter “n,” typed with the right index finger, was either switched with letters normally typed with the same finger (u), with a different finger but same hand (o), with the same finger of the left hand (v), or with a different finger of the left hand (w). 4. When the words were typed normally, the interkey intervals were relatively short, and the onset of movement of the right hand began before the preceding keypress with the left hand. Thus the movement of the two hands overlapped. Furthermore, the movement to the isolated key was highly stereotypical, with little trial-to-trial variability. 5. After the transposition of keys, there were prolongations in the interkey intervals, with the largest delay occurring directly before the typing of the transposed key. Switches between homologous fingers (involving mirror movements) delayed the onset of keypresses to a lesser extent than did other switches. With practice, these delays were reduced but never reached the control level. 6. After the keyswitch, the onset of movement to the isolated key did not occur on average until after the last keypress with the contralateral hand, except when the switch involved the use of homologous fingers. In the latter case, overlapping movement of the two hands was maintained. Thus the learning of a series of discrete movements does not necessarily require that each movement segment be performed sequentially. 7. After the transposition of keys, the movement pattern and time course to a given key were similar to the movement patterns for that key observed during control trials in all conditions. Thus the learning of a series of movements may involve the use of previously learned movements under new conditions. 8. The results suggest that typing movements may be organized at several levels, including the individual keystroke and word level.