Various experimental data indicate that rapid eye movement (REM) sleep is involved in learning processes. In animals, any complex task in a learning environment leads to an increase of the consecutive total REM sleep time, especially just before learning completion. In humans, the oculomotor activity during REM sleep seems to constitute an interesting marker of learning performance. In this work, we focus on the qualitative analysis of REM sleep characteristics after a Morse code learning session. Eight male subjects were polygraphically recorded during three consecutive nights. A computer aided teaching session was performed just before bedrest onset of the experimental night. The learning performance (percentage of saving) was checked on awakening. The Morse code learning led to some modifications in REM sleep components, particularly increases of REM sleep time and number of REM episodes. We did not observe any significant modification in the total number of REMs in the experimental night. However, the correlative analysis between learning performance and sleep parameters indicates a superior r for the oculomotor activity than for the tonic components. This is consistent with the information processing hypothesis in which the temporal distribution of REMs reflects the subject's ability to increase the signal-noise ratio from environmental information intake.