Llewellyn claims very plausibly that dream imagery may be related to the processes involved in memory consolidation during sleep. She cites many of the reviews and original works in this field, to which may now be added claims that also during wake there is offline memory reprocessing (e.g., Peigneux et al. Reference Peigneux, Orban, Balteau, Degueldre, Luxen, Laureys and Maquet2006) involving spontaneous thought processes (daydreams) and the brain's default network (Wang et al. Reference Wang, Yu, Xu, Qin, Li, Xu and Jiang2009).

Llewellyn refers to the dream-lag effect as indicative of sleep-dependent memory consolidation (Nielsen et al. Reference Nielsen, Kuiken, Alain, Stenstrom and Powell2004). This effect refers to a 5- to 7-day delayed incorporation of waking sources into dreams, which follows the more well known immediate incorporations (day residues) from the previous day. Llewellyn is correct to point out that a mnemonic function for rapid eye movement sleep would be supported by evidence that the dream-lag effect is specific to REM rather than non–rapid eye movement (NREM) sleep. A recent study has indeed shown the specificity of the dream-lag effect to REM sleep rather than N2 sleep (Blagrove et al. Reference Blagrove, Fouquet, Henley-Einion, Pace-Schott, Davies, Neuschaffer and Turnbull2011a).

However, in supporting a systems approach to the memory consolidation function for REM sleep, the author does not counter the claim for slow-wave sleep having this function, with REM sleep having instead a synaptic function. In addition, that N2 and spindles may be associated with memory enhancement and integration for declarative memory (e.g., Tamminen et al. Reference Tamminen, Payne, Stickgold, Wamsley and Gaskell2010) does raise questions for claims of the involvement of REM sleep or REM dreams in declarative memory consolidation.

There are additional problems with Llewellyn's proposal in the analogy between REM dream characteristics and the bizarre, narrative, hyperassociative, embodied, spatial characteristics of the ancient art of memory (AAOM) method. The analogy fails when it comes to the voluntary and deliberate intention behind the construction of images in the AAOM method compared with the lack of effort/intentionality/choice in producing dream images. Llewellyn cites evidence that “self-generated encodings are better retained than provided ones” (sect. 3, para. 2), and yet dreams are precisely “provided” to us. Indeed, even in lucid dreams, much of the dream content arises independently of the dreamer's volition. The memory sources of dreams are generally unknown to us during sleep or even after waking. The emotional response to the violation of expectancies is cited by Llewellyn as a component of AAOM, and yet in dreams bizarreness is often not recognised as such. It is unclear why the author wants to analogise the processes of dream bizarreness to AAOM, which has connotations of rote learning, and the method of loci even more so, rather than towards the more creative meeting between two memories, as proposed by many, with Palombo (Reference Palombo1978) being an early example, to Stickgold and Walker (Reference Stickgold and Walker2013), or the more therapeutic views of narrative discontinuities causing changes in self-perception after waking (Kuiken Reference Kuiken1995).

To test her model, the author proposes several approaches. First, she states that “Loss of dreaming would be anticipated to result in loss of identity, loss of emotional concerns, and loss of the ability to plan for the future.” This is similar to Jouvet's (Reference Jouvet1991) hypothesis on the role of paradoxical sleep (REM sleep) in psychological individuation. The negative correlation between alexithymia and ream recall frequency (Nielsen et al. Reference Nielsen, Ouellet, Warnes, Cartier, Malo and Montplaisir1997) can be explained by this model (but also by others).

Secondly, she proposes to test for memory retention in the absence of dreaming. Several studies have already tested memory retention in the absence of dream recall (Göder et al. Reference Göder, Seeck-Hirschner, Stingele, Huchzermeier, Kropp, Palaschewski, Aldenhoff and Koch2011; Solms 1997) and by comparing subjects with high-frequency and low-frequency dream recall (for reviews, see Blagrove & Pace-Schott Reference Blagrove and Pace-Schott2010; Ruby Reference Ruby2011), but little relationship with memory abilities has been found. However, again one cannot be sure that low-frequency dream recallers dream less than high-frequency dream recallers. The literature on REM sleep suppression may thus be relevant, although unclear. For example, Göder et al. (Reference Göder, Seeck-Hirschner, Stingele, Huchzermeier, Kropp, Palaschewski, Aldenhoff and Koch2011) found no negative effects of a decrease in REM sleep on memory performance in patients taking antidepressants, whereas Watts et al. (Reference Watts, Gritton, Sweigart and Poe2012) found that antidepressant suppression of REM sleep impairs hippocampus-dependent learning in rats.

Thirdly, the author rightly proposes to investigate phenomenologically the relationship between immediate and delayed incorporation of waking life events into REM dream content. In a recent study Blagrove et al. (Reference Blagrove, Henley-Einion, Barnett, Edwards and Seage2011b) did investigate the incorporation of recent events into dreams, although sleep stage on awakening was not known. However, regarding Llewellyn's proposal that judges could attempt to match memory sources and dreams, that study indeed found this to be possible for day-residue incorporations, but not for delayed incorporations. The authors suggested that delayed incorporations may involve metaphorical or personal knowledge such that independent judges cannot identify source material in the way the person who had the dream can do.

Llewellyn's placing of Wamsley et al.'s (Reference Wamsley, Tucker, Payne, Benavides and Stickgold2010) finding that dream content is related to improved performance on a spatial learning task overlooks the possibility that the dream imagery is not related to any memory processing during sleep, but is instead related to concerns about pre-sleep task performance – that is, the dream reflects an emotional concern rather than a cognitive function. Such an argument could lead to an epiphenomenal view of dreaming (Blagrove Reference Blagrove2011). Nevertheless, a memory-consolidation function for REM sleep can still be supported by relating dream content to a higher-level learning of social interaction and interpersonal experiences as supported by Nielsen and Lara-Carrasco's (Reference Nielsen, Lara-Carrasco, Barrett and McNamara2007) character self-interaction model of dream content.

However, counting against REM dreams having some lasting elaborative encoding effect is the fragility of memory for dreams. Eichenlaub et al. (Reference Eichenlaub, Bertrand, Morlet and Ruby2013) have found that high-frequency dream recallers show double the amount of intrasleep wakefulness compared with low-frequency dream recallers, indicating that dreams are only recalled if encoded during wakefulness. Furthermore, randomly presented sounds were found to elicit different brain responses in high-frequency and low-frequency recallers during both wakefulness and sleep, with the higher brain reactivity in high recallers hypothesised to contribute to the higher frequency of dream reports by increasing intrasleep wakefulness. Further justification is thus needed for the claim that unremembered dreams have any consequent effects. Dreams may thus be made of memories, but without having a memory-consolidation function.