Sue Llewellyn's target article proposes the striking hypothesis that rapid eye movement (REM) dreaming shares some similarities with mnemonic processes underlying old mnemonic techniques (ancient art of memory [AAOM]) in elaborative encoding of episodic memory.
The idea that memory could benefit from the synergistic action of waking and sleeping is suggestive and probably true. Nevertheless, several points need to be clarified regarding the synergistic role of the different sleep stages and dreams (REM and non–rapid eye movement [NREM]) in the elaborative encoding of episodic memory. The neurobiological underpinnings of REM sleep and some specific NREM sleep stages (stage 2, slow-wave sleep [SWS]) are well documented and understood, thanks to the physiological, psychophysiological, and neurological literature of the past few decades. On the other hand, empirical data about dreaming and dreams are scant and inconsistent.
With regard to REM dreams, if we consider dream reports collected at night (with the exception of the first REM), we agree with the majority of authors who define REM dreams as characterized by vivid perception, frequent bizarre associations, narrative structure, self-embodiment, and location. In that case, the hypothesis that these characteristics of REM dreams could be similar to devices underlying the AAOM is very suggestive. In fact, it could explain a mnemonic function of REM dreaming – that is, the elaborative encoding of recent memories with emotional distinctive remote episodes though hyperassociative activity, visual imaging, and bizarre links among objects, places, people, and so on, similarly to AAOM.
The argument is based also on neurobiological evidence associated with REM and NREM sleep. That this isomorphism between neurobiological and cognitive networks (Stickgold et al. Reference Stickgold, Hobson, Fosse and Fosse2001) is related to NREM sleep, though, is an overall questionably strong assumption. NREM sleep cannot be considered a unique type of sleep, although there are important differences in biological characteristics between early stage 2, stage 2 during sleep, stage 2 near the end of sleep, and SWS. Llewellyn seems to collapse neurobiological aspects of different NREM sleep stages and NREM generic dream characteristics (thoughtlike, brief, trivial) into a unique, comprehensive concept of NREM.
This reasoning is supported by neurobiological data showing that the associative areas involved in information processing are similarly activated in REM and NREM (all of the stages), whereas differences are found in the hippocampal and parahippocampal regions, which are primarily involved in memory systems and in the emotional participation in dreams.
On the other hand, the analysis of the dreams at night and the comparison between mental activity at sleep onset (NREM stage 2), in NREM stage 2 at night, in NREM SWS, and in the REM shows many similarities. Our experimental data (Cavallero et al. Reference Cavallero, Cicogna, Natale, Occhionero and Zito1992; Cicogna et al. Reference Cicogna, Cavallero and Bosinelli1992; Reference Cicogna, Natale, Occhionero and Bosinelli1998) showed that NREM and REM dream reports cannot be distinguished if the dream reports' length is the same (the quantitative difference is predominant). Moreover, NREM dream reports collected near awakening are frequently difficult to distinguish from early-night REM dreams (Cicogna et al. Reference Cicogna, Natale, Occhionero and Bosinelli1998; Fosse & Domhoff Reference Fosse, Domhoff, Barratt and McNamara2007). That could suggest the engagement of a unique cognitive system involved in dream generation, irrespective of EEG sleep stage (as Foulkes [Reference Foulkes1985] proposed in his model). In accordance with Solms (Reference Solms2000), dreams occur all night. Therefore, it seems reasonable to state that dreaming does not depend exclusively on REM physiology. Dreaming depends on a complex functional system involving regions of forebrain (parieto-temporo-occipital carrefour, mesial frontal white matter, limbic system, and associative visual areas). Central Nervous System (CNS) activation certainly modulates cognitive processes quantitatively but not necessarily qualitatively. The dream generation could use the same cognitive mechanisms in every sleep stage. The qualitative differences could be an epiphenomenon of the quantitative ones. This is our speculative hypothesis, based on empirical data, but we hope that further data will better clarify and explain the NREM–REM “dialogue.”
According to Llewellyn, the interplay between NREM and REM is the instantiation during NREM sleep of internal cortical junctions by the hippocampus, retained as indices. The instantiation of junctions enables the elaborative encoding of recent and remote emotional episodic memories during REM dream generation. The author emphasizes that the different sleep stages are distinct but integrative in memory and dream processes. This hypothesis is interesting and suggestive because, like our hypothesis, it enables one to disregard models that reserve an exclusive role for REM sleep in dream generation (see Stickgold et al. Reference Stickgold, Hobson, Fosse and Fosse2001).
However, Llewellyn assigns much importance to the role of cortical spindle activity in the formation of cortical junctions, so the role of slow-wave sleep remains an open question. In this model, the function of SWS is unclear in the absence of electrophysiological features (spindle activity, sharp waves) considered critical to the instantiation of cortical junctions.
Concluding it is the role of SWS–NREM sleep is not persuasive. It requires much explanation inside the dream generation processes either from a physiological or a cognitive perspective. We understand the good intentions aimed at finding a synergy among all the different levels of vigilance (REM–NREM sleep–wakefulness) in encoding episodic memory and in REM dream generation. Moreover, we think that the empirical data do not allow the determination of a persuasive integration of neurophysiological and cognitive characteristics of NREM sleep stages.
Finally an interesting question that could be addressed by future research regards the dream recall of the dream episodes encoded. If we assume the “state dependence effect” on episodic mnemonic processes (see encoding specificity principle, Tulving & Thompson Reference Tulving and Thomson1973), fragments of dreams might be retrieved and used as a base of implicit knowledge for new dream generation. In this way, elaborative encoding during sleep could make accessible what is not available to waking awareness.
Sue Llewellyn's target article proposes the striking hypothesis that rapid eye movement (REM) dreaming shares some similarities with mnemonic processes underlying old mnemonic techniques (ancient art of memory [AAOM]) in elaborative encoding of episodic memory.
The idea that memory could benefit from the synergistic action of waking and sleeping is suggestive and probably true. Nevertheless, several points need to be clarified regarding the synergistic role of the different sleep stages and dreams (REM and non–rapid eye movement [NREM]) in the elaborative encoding of episodic memory. The neurobiological underpinnings of REM sleep and some specific NREM sleep stages (stage 2, slow-wave sleep [SWS]) are well documented and understood, thanks to the physiological, psychophysiological, and neurological literature of the past few decades. On the other hand, empirical data about dreaming and dreams are scant and inconsistent.
With regard to REM dreams, if we consider dream reports collected at night (with the exception of the first REM), we agree with the majority of authors who define REM dreams as characterized by vivid perception, frequent bizarre associations, narrative structure, self-embodiment, and location. In that case, the hypothesis that these characteristics of REM dreams could be similar to devices underlying the AAOM is very suggestive. In fact, it could explain a mnemonic function of REM dreaming – that is, the elaborative encoding of recent memories with emotional distinctive remote episodes though hyperassociative activity, visual imaging, and bizarre links among objects, places, people, and so on, similarly to AAOM.
The argument is based also on neurobiological evidence associated with REM and NREM sleep. That this isomorphism between neurobiological and cognitive networks (Stickgold et al. Reference Stickgold, Hobson, Fosse and Fosse2001) is related to NREM sleep, though, is an overall questionably strong assumption. NREM sleep cannot be considered a unique type of sleep, although there are important differences in biological characteristics between early stage 2, stage 2 during sleep, stage 2 near the end of sleep, and SWS. Llewellyn seems to collapse neurobiological aspects of different NREM sleep stages and NREM generic dream characteristics (thoughtlike, brief, trivial) into a unique, comprehensive concept of NREM.
This reasoning is supported by neurobiological data showing that the associative areas involved in information processing are similarly activated in REM and NREM (all of the stages), whereas differences are found in the hippocampal and parahippocampal regions, which are primarily involved in memory systems and in the emotional participation in dreams.
On the other hand, the analysis of the dreams at night and the comparison between mental activity at sleep onset (NREM stage 2), in NREM stage 2 at night, in NREM SWS, and in the REM shows many similarities. Our experimental data (Cavallero et al. Reference Cavallero, Cicogna, Natale, Occhionero and Zito1992; Cicogna et al. Reference Cicogna, Cavallero and Bosinelli1992; Reference Cicogna, Natale, Occhionero and Bosinelli1998) showed that NREM and REM dream reports cannot be distinguished if the dream reports' length is the same (the quantitative difference is predominant). Moreover, NREM dream reports collected near awakening are frequently difficult to distinguish from early-night REM dreams (Cicogna et al. Reference Cicogna, Natale, Occhionero and Bosinelli1998; Fosse & Domhoff Reference Fosse, Domhoff, Barratt and McNamara2007). That could suggest the engagement of a unique cognitive system involved in dream generation, irrespective of EEG sleep stage (as Foulkes [Reference Foulkes1985] proposed in his model). In accordance with Solms (Reference Solms2000), dreams occur all night. Therefore, it seems reasonable to state that dreaming does not depend exclusively on REM physiology. Dreaming depends on a complex functional system involving regions of forebrain (parieto-temporo-occipital carrefour, mesial frontal white matter, limbic system, and associative visual areas). Central Nervous System (CNS) activation certainly modulates cognitive processes quantitatively but not necessarily qualitatively. The dream generation could use the same cognitive mechanisms in every sleep stage. The qualitative differences could be an epiphenomenon of the quantitative ones. This is our speculative hypothesis, based on empirical data, but we hope that further data will better clarify and explain the NREM–REM “dialogue.”
According to Llewellyn, the interplay between NREM and REM is the instantiation during NREM sleep of internal cortical junctions by the hippocampus, retained as indices. The instantiation of junctions enables the elaborative encoding of recent and remote emotional episodic memories during REM dream generation. The author emphasizes that the different sleep stages are distinct but integrative in memory and dream processes. This hypothesis is interesting and suggestive because, like our hypothesis, it enables one to disregard models that reserve an exclusive role for REM sleep in dream generation (see Stickgold et al. Reference Stickgold, Hobson, Fosse and Fosse2001).
However, Llewellyn assigns much importance to the role of cortical spindle activity in the formation of cortical junctions, so the role of slow-wave sleep remains an open question. In this model, the function of SWS is unclear in the absence of electrophysiological features (spindle activity, sharp waves) considered critical to the instantiation of cortical junctions.
Concluding it is the role of SWS–NREM sleep is not persuasive. It requires much explanation inside the dream generation processes either from a physiological or a cognitive perspective. We understand the good intentions aimed at finding a synergy among all the different levels of vigilance (REM–NREM sleep–wakefulness) in encoding episodic memory and in REM dream generation. Moreover, we think that the empirical data do not allow the determination of a persuasive integration of neurophysiological and cognitive characteristics of NREM sleep stages.
Finally an interesting question that could be addressed by future research regards the dream recall of the dream episodes encoded. If we assume the “state dependence effect” on episodic mnemonic processes (see encoding specificity principle, Tulving & Thompson Reference Tulving and Thomson1973), fragments of dreams might be retrieved and used as a base of implicit knowledge for new dream generation. In this way, elaborative encoding during sleep could make accessible what is not available to waking awareness.