There is much about Sue Llewellyn's article that is appealing, not least the bridges it builds with the Humanities and its careful use of real psychological data: an individual dream report with personal associations. Cognitive neuroscience has much to gain from articles like this. It therefore pains me to say that Llewellyn's main hypothesis in unsupportable, for common-and-garden empirical reasons.

Her hypothesis is that rapid eye movement (REM) dreaming serves a similar functional purpose to the ancient art of memory (AAOM) – that is, it largely supports episodic memory consolidation “through imaginatively associating the to-be-remembered material in a mental image” (sect. 2, para. 3). She attributes this imaginative process to “hippocampal associational function during REM dreams” (abstract, para. 2). The empirical evidence adduced in support of this hypothesis accordingly relates the function of dreaming to the physiology of the REM state and the psychology (and functional anatomy) of the hippocampus.

The premise that any function of dreaming may be derived from physiological processes unique to the REM state is untenable, since dreaming and REM sleep are certainly doubly dissociable states, notwithstanding that they are on average highly correlated (Solms Reference Solms2000). Curiously, considering how fundamental this premise is to her entire argument, Llewellyn simply says at the outset that REM/dream isomorphism “is the position taken in this article” even though she simultaneously acknowledges, for example, that “early-night REM and late-night NREM dreams cannot be distinguished” (sect. 1, para. 1). This last fact alone disqualifies all the specific physiological and psychological evidence regarding the REM state that she subsequently cites to support her main hypothesis concerning the function of dreams. Thus, for example, there is no valid reason to assume that Llewellyn's own sample dream, which she subjectively “self-identified” (sect. 4, para. 8) as a REM dream, was indeed a REM dream.

Nevertheless, this still leaves open the theoretical possibility that REM-like dreaming (regardless of sleep stage) could serve the hypothesized function of consolidating episodic memories. More fatal for the hypothesis, therefore, is the clinical fact that bilateral hippocampal lesions – causing total loss of the episodic memory functions that Llewellyn attributes to dreaming – have no demonstrable effect on the occurrence of REM-like dreams, as I reported long ago (Solms Reference Solms1997). In this regard, I hope shortly to be able to describe in detail the vivid dream reports of the celebrated patient HM (S. Corkin, personal communication, July 27, 2008). By the same token, total loss of dreaming following suitably localized cerebral lesions has no demonstrable effect on episodic memory (Solms Reference Solms1997; Yu Reference Yu2006). This applies also to memory for pre-sleep episodes (Marchbank Reference Marchbank2013). Llewellyn predicts the opposite (sect. 6, para. 6).

These facts have perhaps not been sufficiently widely reported and adequately investigated. Nevertheless, my own (now extensive) experience with such cases leaves me in no doubt that the effect – if any – of loss of dreaming upon episodic memory consolidation is sufficiently slight and subtle to escape routine clinical investigation – including standardized neuropsychological assessment. It is, in short, very difficult to support Llewellyn's hypothesis that dreaming performs a function which seems to be impervious to cessation of dreaming, and a function the loss of which appears to be compatible with essentially normal dreaming. (Of course this does not mean that dreaming plays no part in episodic memory consolidation; it means only that the two processes do not causally depend upon each other.)

But the most puzzling aspect of Llewellyn's hypothesis is something which most commentators will surely remark upon, too: its curious lack of face validity.

AAOM was an active strategy used to produce episodic memory cues and deliberately composed idiographic associations that could later be used to assist in the voluntary and explicit “re-collection” of (usually semantic) material committed to memory. This psychological process differs from dreaming in several important respects, two of which seem fundamental. Firstly, the AAOM is an executive strategy (apparently relying on active prefrontal integration – what Llewellyn calls “organization” – of hippocampal hyper-associations), whereas dreaming is a passive – what Llewellyn calls “associative” – experience in which the prefrontal lobes play little or no part (Braun et al. Reference Braun, Balkin, Wesensten, Carson, Varga, Baldwin, Selbie, Belenky and Herscovitch1997; Solms Reference Solms1997). Secondly, the creative mental images and linkages forged in AAOM are explicitly brought back into working memory, to support the recollection process, whereas the vast majority of dreams are immediately and permanently forgotten – at least as far as explicit recollection is concerned.

Why, then, does Llewellyn equate them? The answer seems to be because they share some formal features: visualization, bizarre association, organization, narration, embodiment, and location. But apart from bizarre association, it is hard to see which of these features distinguishes dreaming from most other forms of everyday conscious experience. To paraphrase Llewellyn, “The stuff of cognition is the stuff of memory.”

Llewellyn addresses the face-validity problem of dream forgetting by suggesting that if its hyper-associational junctions were to become conscious, they would precipitate psychosis. It is hard to understand why this should be so for dreams but not for the AAOM. And when we do remember dreams in everyday life, why do we have so little difficulty staying sane?