Whether sleep-wake state and level of sleepiness differentially affect the various memory processes and systems described above is not clear. It is clear that neural processing, and consequently probably also memory function, is reorganized during NREM and REM sleep. Memory neurobiologists locate memory formation in the hippocampus and at the cellular level model it using long-term potentiation. How sleep state gates hippocampal memory functioning is not known. Furthermore, how reorganization occurs during transitions in state (i.e., wake to sleep or sleep to wake) and whether sleepiness itself is a transitional state is intriguing. How memory functions differentially change with state or state transitions may in turn reveal the process by which the transition occurs. We will review and discuss memory functions while sleepy, at transitions of sleep-wake state and during both NREM and REM sleep.

Sleep deprivation studies have demonstrated the association of sleepiness with memory impairment (3). Sleepiness affects stimulus registration. The sleepy state is characterized by microsleeps. Microsleeps are brief episodes of EEG-defined sleep that intrude on the waking EEG during prolonged sleep deprivation (4) Subjects required to press a switch when a tone or a visual stimulus repeatedly occurs do not respond during microsleeps (4). Apparently, the presence of the stimulus is not registered during the microsleep. Memory consolidation is also disrupted by sleep loss and the consequent increased sleepiness, as is the retrieval of memories. However, determination of whether or not increased sleepiness due to sleep deprivation or sleep restriction is differentially disruptive of a particular memory system has not been systemically evaluated.

The most extensively studied class of sedating drugs is the benzodiazepine (Bz) receptor agonist. These drugs increase sleepiness as measured by the Multiple Sleep Latency Test (MSLT) (5). Their amnestic effect is also well known (6). Whether the amnestic effect is due to the sedative activity, rather than direct hippocampal effects, is disputed. The debate goes to the specificity of the amnestic effects with regard to memory system and Bz receptor subtype (6,7). Given that Bz1 receptors are not densely distributed in the hippocampus, a Bz1 specific drug by the Squire hypothesis should not alter declarative memory. But, a recent study found amnestic effects for a receptor specific Bz agonist that were comparable to that of a non-specific Bz receptor agonist (8). Declarative memory, specifically a semantic memory task, was affected; that is the amnesia was system non-specific.

The sleepiness-amnesia association is also seen in sleep disorders patients. Standardized testing of neuropsychological functioning, including memory, in patients with sleep apnea syndrome has shown impairment (9). The extent of the sleep disturbance and resultant daytime sleepiness related directly to degree of impairment. The range of cognitive functions showing impairment included declarative and procedural tasks (i.e., verbal and visual memory) and both short-term and long-term memory (9).

This finding is the same for transitions to sleep after an awakening from sleep. Sedative drug studies have found that the rapidity of sleep onset after an awakening from sleep is associated with amnesia the next morning for the material presented during the awakening (12). The amnesia is also found with barbiturates as well as Bzs and is avoided if wakefulness is enforced for 15-min. The same is found in non-drug studies of awakenings from sleep. The morning recall of a stimulus word presented during an awakening improves if a brief motor task also is performed, which delays the return to sleep (13).

In the opposite state transition, from sleep to wake, there also is memory impairment. This modest and short-lived impairment has been called "sleep inertia". Any number of motor and cognitive functions, including memory, have shown the impairment (14). Determinants of the sleep inertia are the extent of prior sleep loss and the sleep stage from which the awakening occurred. Differential sensitivity of memory systems and processes has not been explored.

There is some suggestion that procedural memories in the form of classical conditioning can be formed during NREM sleep (15). The conditioned response in these studies has been EEG alpha attenuation, EEG K complexes and heart rate. It is reported that sleep stage shifts or an alpha EEG do not occur in response to the stimuli and thus EEG arousal does not account for the conditioning. On the other hand, habituation which is the simplest form of conditioning has not been consistently found (15). Thus, the extent to which procedural memory during NREM sleep is possible, without the confound of EEG arousal, needs further exploration.

Quite controversial is the question of whether memory consolidation occurs during REM. Animal studies have shown that learning during wake is followed by increased paradoxical sleep (PS) and deprivation of PS is followed by memory impairment (16). Furthermore, new associations can be learned during PS (17). With regard to specific memory systems, some studies in humans show memory for declarative tasks is not impaired by REM sleep deprivation, but that procedural memory is disrupted by REM deprivation (16). However, these results directly contradict those discussed earlier regarding memory system function in NREM sleep. The confound may be the extent of EEG arousal, which is a possible result of the REM deprivation. In contrast to these findings is the theory and evidence assembled by Crick that REM sleep is a period of forgetting, or as described by Crick "reverse learning" (18).

Retrieval of memories formed while awake does occur during sleep, both in NREM and REM sleep (e.g., stimulus information is monitored and processed during sleep). Stimuli presented during sleep evoke responses that subjects have learned during prior wakefulness and subjects also discriminate between stimuli during sleep by responding differentially to them (19,20).

To summarize, memory is not consolidated during uninterupted NREM sleep, but may be facilitated during REM sleep. Whether the various memory systems are differentially preserved or facilitated during these sleep states is not clear. At the transitions to and from sleep there is memory impairment that by much evidence is non-specific as to memory system. Finally, sleepiness during wake is also associated with impairment of memory function and again it is non-specific.

In attempting to synthesize this literature, two important factors emerge. The basal state of arousal or activation at the time of memory function, whether it be registration, consolidation, or retrival, seems critical (21). The sleepy state, state transitions, and NREM sleep may define a continuum of arousal with respect to memory function. But also, the arousing nature of the memory function itself seems important and whether different memory systems involve differing arousal is a question. Finally, there is a clear shift in neurobiology from the NREM to REM state and whole mode of memory function may also change (21).