Previous studies have shown a decrease in rapid eye movement (REM) frequency during desynchronized sleep in recovery nights following total or partial sleep deprivation. This effect has been ascribed to an increase in sleep need or sleep depth consequent to sleep length manipulations. The aims of this study were to assess REM frequency variations in the recovery night after two consecutive nights of selective slow-wave sleep (SWS) deprivation, and to evaluate the relationships between REM frequency and SWS amount and auditory arousal thresholds (AAT), as an independent index of sleep depth. Ten normal males slept for six consecutive nights in the laboratory: one adaptation, two baseline, two selective SWS deprivation and one recovery night. SWS deprivation allowed us to set the SWS amount during both deprivation nights close to zero, without any shortening of total sleep time. In the ensuing recovery night a significant SWS rebound was found, accompanied by an increase in AAT. In addition, REM frequency decreased significantly compared with baseline. This effect cannot be attributed to a variation in prior sleep duration, since there was no sleep loss during the selective SWS deprivation nights. Stepwise regression also showed that the decrease in REM frequency is not correlated with the increase in AAT, the traditional index of sleep depth, but is correlated with SWS rebound.