The present study describes how the frequency receptive fields (RF) of auditory thalamus neurons are modified when the state of vigilance of an unanesthetized animal naturally fluctuates among wakefulness (W), slow-wave sleep (SWS), and paradoxical sleep (PS). Systematic quantification of several RF parameters-including strength of the evoked responses, response latency, acoustic threshold, shape of rate-level function, frequency selectivity, and RF size-was performed while undrugged, restrained guinea pigs presented spontaneous alternances of W, SWS, and PS. Data are from 102 cells recorded during W and SWS and from 53 cells recorded during W, SWS, and PS. During SWS, thalamic cells behaved as an homogeneous population: as compared with W, most of them (97/102 cells) exhibited decreased evoked spike rates. The frequency selectivity was enhanced and the RF size was reduced. In contrast during PS, two populations of cells were identified: one (32/53 cells) showed the same pattern of changes as during SWS, whereas the other (21/53 cells) expressed values of evoked spike rates and RF properties that did not significantly differ from those in W. These two populations were equally distributed in the different anatomical divisions of the auditory thalamus. Last, during both SWS and PS, the responses latency was longer and the acoustic threshold was higher than in W but the proportion of monotonic versus nonmonotonic rate-level functions was unchanged. During both SWS and PS, no relationship was found between the changes in burst percentage and the changes of the RF properties. These results point out the dual aspect of sensory processing during sleep. On the one hand, they show that the auditory messages sent by thalamic cells to cortical neurons are reduced both in terms of firing rate at a given frequency and in terms of frequency range. On the other hand, the fact that the frequency selectivity and the rate-level function are preserved suggests that the messages sent to cortical cells are not deprived of informative content, and that the analysis of complex acoustic sounds should remain possible. This can explain why, although attenuated, reactivity to biologically relevant stimuli is possible during sleep.