Systems and methods for sleep state classification involve detecting conditions related to sleep, including at least one condition associated with rapid eye movement (REM) sleep. Additionally, a condition modulated by the sleep-wake status of the patient may be detected. A medical system that is partially or fully implantable incorporates sensors and circuitry for detecting and processing the sleep-related signals. A sleep state processor classifies the patient's sleep state based on the sleep-related signals. Sleep state classification may be used in connection with the delivery of sleep state appropriate therapy, diagnostic testing, or patient monitoring.