Introduction

Does brain activity during REM resemble activity in the awake state?

Brain activity during slow wave sleep

Brain activity during REM sleep

Is REM sleep equivalent to dreaming?

References

Scientists and philosophers have discussed about the problem of consciousness for centuries. Recently, the neuroscience community has become particularly interested in this fascinating question. Specific hypothesis and experiments have been suggested in order to address rigorously what the neuronal correlates of visual consciousness are . Crick and Koch have claimed that there must be a specific group of neurons that explicitly represent the contents of our moment-to-moment aware perception.

In this regard, the study of sleep and, in particular, dreams may provide interesting and relevant data. From a subjective point of view, it seems evident that we are perfectly conscious during our dreams while we are unconscious during the rest of the sleep period. Since most dreams have strong visual content (in non-congenitally blind people) we can claim that we are visually aware during our dreams. We could therefore include one more constraint for the neurons that correlate with our visual perception. The neurons that represent our visual awareness should be activated when we see something, when we close our eyes and imagine it or when we are asleep and dream about the same thing.

PET (positron emission topography) offers a non-invasive way to study the changes in blood flow in the human brain during sleep. Another widespread imaging technique, fMRI (functional magnetic resonance imaging) has the disadvantage of the inherent difficulty in recording EEG simultaneously with the fMRI signal.

Here we summarize some of the relevant imaging, neurological and eletrophysiological literature of dream sleep that we find particularly relevant in regard to the problem of visual consciousness. For an introductory overview of PET, see .

A fascinating question of sleep research that investigators have studied for decades is whether the brain activation during REM is similar to that during the wake state. Using EEG measurements it is not easy to distinguish the awake state and REM sleep. The EOG (electro-oculogram) and the EMG are therefore required in order to distinguish the rapid eye movements and loss of muscle tension. Researchers had suggested that REM is the same as the waking state except for a closed input-output.

However, important differences can be found at the single-neuron level as well as by pharmacological interventions [Hobson, 1995 #586]. Imaging studies provide strong evidence for important differences in brain activation during these two states. Many areas that are active during REM are a subset of those during the awake state. And there are some areas that are more active during REM (reviewed in ; see also the section on Brain activity during REM sleep).

The traditional distinction between REM and SWS sleep is based on the pattern of activity obtained from scalp electrodes attached to the scalp. This electroencephalographic measure (EEG) shows increased amplitudes and higher power at lower frequencies during SWS sleep. The EEG pattern during REM sleep is basically indistinguishable from that in the awake state and is characterized by low amplitudes and higher frequencies. Eye movements (ROG) and muscle tension in the chin (EMG) are used to differentiate the REM state from the awake state.

Imaging studies support this differentiation between REM and SWS sleep. Broadly, there is a general decrease in blood flow measured by PET . A negative correlation has been found between cerebral blood flow (rCBF) and the occurrence of SWS in the brainstem reticular formation, basal forebrain/hypothalamus, thalamus, orbitofrontal cortex, anterior cingulate cortex and the medial aspect of the temporal lobe on the right side.

The role of the brainstem reticular formation and the basal forebrain in regulating the sleep/wake cycle has been extensively studied by lesions, single neuron recordings and pharmacological intervention in cats (see for an overview).

It is particularly interesting to note that the thalami, as a sensory gate to the cortex seem to be overall de-activated during SWS sleep.

With respect to the thalami, it is important to note that Steriade and colleagues have shown with single-unit recordings as well as models that the thalamus plays an important role in the low-frequency oscillations characteristic of SWS.

Very little is mentioned in these papers about the changes in activity in the temporal or occipital lobes during slow wave sleep compared to the awake state or REM sleep. Macquet et al indicate a decrease of activity in the right mediotemporal lobe correlated with SWS. It should be noted that other researchers have shown strong single-neuron activity during SWS in some areas of the temporal lobe like the hippocampus (see for example, )

From a subjective point of view we could claim that we are unconscious during the night except for those periods during which we are dreaming or awake. This reflection depends on an essential assumption that involves trusting our memory. That is, it is likely that we do not remember most of our dreams. And that does not necessarily imply that we were not conscious while we were dreaming. Therefore, while the above data may suggest a correlation between the lack of activity in certain brain areas and the lack of conscious subjective awareness, it is essentially based on the above-mentioned assumption. We therefore need to look also at the positive data of what happens in the brain while we dream.

There are two particularly relevant studies that analyzed blood changes during REM sleep by PET while performing simultaneous EEG recording . The anterior cingulate cortex (mainly Brodmann's area 24), the thalamus and the pontine brainstem are activated during REM. It is interesting to note that activity in these areas was decreased during SWS in the previously described studies. These results can be expected based on the reticular activating system theory of sleep in which brainstem mechanisms (particularly but not exclusively in the pons) are responsible for activating the cortex through the thalamus during the episodes of REM sleep .

It is interesting to note the existence of a morphological unique type of neurons in humans and great apes in the anterior cingulate . As far as we know, no one has done single-neuron recordings from this area during sleep but it would be very interesting to verify the activation during REM sleep and the turning off during NREM sleep observed in the PET experiments.

Crick and Koch have argued based on neurological, psychophysical and electrophysiological evidence that the activity in striate cortex (area V1) does not correlate with our visual awareness. It is therefore interesting to note that striate cortex does not seem to be activated during dreams . These conclusions were drawn from a comparison of REM sleep with the awake state but also during a comparison of REM and NREM sleep. Against the former data, it could be argued that subjects were imagining visual scenes in the rather ambiguous rest state. However, the direct comparison between REM and NREM sleep shows unambiguously the lack of changes in blood flow in striate cortex during REM sleep.

There is controversy in the functional imaging literature about whether there is activation of striate cortex during voluntary visual imagery . The differences may be due to the different tasks and baseline conditions. However, all the researchers have found activation in higher extrastriate visual areas including the association areas in the temporal lobe during visual imagery.

Extrastriate visual areas are activated during REM sleep . This correlates with the visuospatial vividness of dreams. The activation seems to be higher in the ventral stream compared to the dorsal extrastriate visual areas. It is notable that Goodale and Milner have suggested a clear separation between the two pathways, where the ventral pathway is more correlated with our perceptions while the dorsal pathway would be more involved in motor processing without necessary concomitant conscious processing.

Limbic and paralimbic regions of the brain are suggested to be involved in processing emotions as evidenced by neurophysiological, lesion and molecular studies in both animals and humans. These areas are more activated during REM than during the awake state or SWS. This fits with our subjective recall of most dreams as being rich in emotional content. In particular, the amygdala is strongly activated during REM sleep. There is also strong activation during REM sleep of other higher association areas like the hippocampus, parahippocampal gyrus and entorhinal cortex. . It is interesting to note that these areas receive strong direct inputs from higher visual areas in the inferior temporal cortex . Electrical activation in several parts of the human temporal lobe elicits vivid visual hallucinations in human epileptic patients . Furthermore, there are visual responsive and selective units in the amygdala and entorhinal cortex (G. Kreiman et al, unpublished)

There is a de-activation of pre-frontal cortex during REM sleep. It was suggested by Hobson that dream-amnesia could be due to prefrontal deactivation. Crick has proposed that it is the inactivation of the locus ceruleus (observed electrophysiologically and also in PET) that is responsible for the forgetfulness of dreams.

Both EEG (in cats and humans) as well as single-unit recordings (in cats) show synchronization in the gamma frequency band (broadly 30-70 Hz) during both REM sleep and waking. Hobson interprets these oscillations as a reflection of our cognitive processing during dreams and during the awake state .

Ever since the paradigm-shifting discovery of the correlation between REM sleep and dreaming by Kleitman's group it was also clear that some form of dreams can also occur during NREM sleep. These NREM dreams are typically quite different from the more common REM dreams. Upon awakening from NREM sleep, subjects usually report no dream recollection whatsoever. If they do recollect mental activity, it usually is related to plans or elaborate thoughts and lacks the visual vividness, hallucinatory and delusory components of typical dreams .

Neurological data shows that a double-dissociation can be established in some cases between REM sleep and dreams. That is, there are patients with lesions in their brains that show a lack of REM sleep (as characterized by EEG) but they still experience dreams and viceversa. Patients that do not experience visual dreams typically have lesions in the temporal lobe .

The question of whether REM is necessarily equivalent or not to dreaming is particularly important for the study of the neuronal correlate of visual consciousness. It is not clear that people dream during the whole period of REM sleep. The only exact verification we have so far of whether someone is dreaming or not is to ask them for their subjective experience upon awakening. This relies again on memory and of course it has the disadvantage of disrupting sleep.

It is possible that the phenomenon that we experience as dreams depends on the activity in a specific group of neurons, the same neurons that correlate with our visual perceptions during normal vision or during visual imagery. There may be a correlation, but not necessarily a causal link between rapid eye movements and the activity in this group of neurons so that the two can be dissociated, as is the case in neurological patients. The imaging activity reported above is certainly very promising in terms of the analogies with the data that comes from electrophysiology, neurology and imaging during vision and visual imagery. There is a considerable lack of single or multiple neuron recordings from visual areas during sleep together with EEG recordings.

Wilson, M. A., and McNaughton, B. L. (1994). "Reactivation of Hippocampal Ensemble Memories During Sleep." Science, 265, 676-679

Gabriel Kreiman

gabrielk@caltech.edu