Dreams of a Bayesian brain. A predictive processing account of dreaming
Friday 2nd May, 2014
A dissertation submitted in partial fulfilment of the requirements for the degree of Master of Science in Philosophy
Dreaming is an important aspect of our mental life which has been widely studied by
psychologists and cognitive scientists over the past decades, and it has recently gained popularity in philosophy of mind as well. As a topic of study, it has the potential of providing a good insight on how our mind works in a different condition from regular waking life. Furthermore, it’s a phenomenon that happens on a daily basis to everyone and, therefore, it does not constitute an exceptional contrasting case (like brain lesions or psychopathologies). However, the rich variety of theories about the origin, evolution and functional role of dreaming seems to suggest a lack of unified explanation. Dreaming also poses conceptual questions like how to relate it to other forms of mentation (e.g. imagination, hallucination, delusion) or if it’s possible to talk about consciousness during sleep, whereas sleep was considered in the past a sort of “mental shut-down”.
In this paper I will present an account of dreaming under the framework of Predictive Processing. The aim is to provide a simple and integrative account of the cognitive architecture underlying our mental activity during sleep. I will argue that, through Predictive Processing, we can better understand dreaming as continuous with other mental phenomena. I will also address conceptual issues related to dreaming, and I will show that it is a good case study for a critical analysis of Predictive Processing. In order to fulfil this task and for reasons of brevity, I will focus on what conventionally is understood as dreaming, i.e. the mentation occurring during REM sleep. However, this is not the only case studied in the literature: I will keep open the possibility that Predictive Processing can account for related phenomena like NREM dreaming, lucid dreaming and transitional states like hypnagogic imagery.
In Chapter 1 I will briefly summarise the state of the art of dream research and recapitulate the claim from Windt & Noreika (2011) that there is an integration problem of dreaming in a wider theory of consciousness. I will add that this problem extends to a theory of cognition in general. I will then suggest that Predictive Processing has several theoretical advantages which will emerge later in the paper and which make it a good candidate framework to adopt in order to address the Integration Problem.
In Chapter 2 I will lay out Predictive Processing (Clark, 2013; Hohwy, 2013) with a focus on the link between action and perception. Drawing on the literature, I will propose an explanation of dreaming as the result of the predictive mechanism working under impaired conditions. I will then provide more neurophysiological details on REM sleep mentation based on the model of dreaming proposed by Hobson & Friston (2012).
In Chapter 3 I will discuss three topics related to dreaming under the light of Predictive Processing. First, I will argue that the mechanism of formation of dreams is akin to the one of mind wandering. Second, I will provide an explanation about the origin of bizarreness in dreaming compatible with Predictive Processing. Finally, I will describe the peculiar characteristics of the dream narrative as an attempt by the brain to deploy actions while lacking feedback from the environment.
Lastly, in Chapter 4 I will tackle the issue of the evolutionary and functional role of dreaming. I will show how Predictive Processing is compatible with the idea of sleep as a form of “synaptic pruning” and memory optimisation process and briefly tackle the issue of dreaming epiphenomenality.
In Appendix, I will add some considerations about the role of dreaming during the ontogenesis, arguing that sleep and dreaming serve to tune and polish the predictive mechanism of each individual before and during the initial stage of its engagement with the world.