Pain in fish and molluscs - why the multiple realisability argument fails  Brian Key , Deborah Brown (School of Biomedical Sciences, The University of Queensland, Brisbane, Queensland Australia)   C11

Subjective experience of pain is often bestowed on organisms such as fish and molluscs (e.g. octopus) because they display aversive behavioural responses to noxious stimuli. Such assertions appear to gain strength from reports that these creatures also exhibit associative learning in response to noxious stimuli. Attributing pain to these animals is however begging the question since it is a priori assumed that pain is necessary for these behaviours. There is strong empirical evidence that conscious awareness of noxious stimuli is not necessarily required for either aversive responses or associative learning. To overcome difficulties in using these behavioural measures to assess whether an organism feels pain we turn our attention to the bioengineering principle that "structure determines function". That is, an animal can only feel pain if it possesses the appropriate neural architecture to monitor and become aware of its own sensory processing. An animal that is unaware of its sensory processing cannot then subjectively experience or feel. This particular approach has been vigorously challenged on the basis that function (i.e. pain) can be multiply realised by disparate structures in the animal kingdom. It is argued that neural structures responsible for pain in human brains may be vastly different to those in the fish or octopus brain. Consequently, neural architecture has been considered unreliable for determining the likelihood that an animal can or cannot experience pain. We argue that multiple realisability is level-dependent; that it may be true at some levels of abstraction but not at the level that matters. Stochastic differences in the molecular composition of neural membranes, axonal and dendritic branching patterns, synapse number and neuronal number occur in natural populations of the same species without affecting function. However, there are certain organisational structures that are essential for brains to function correctly. For instance, at its simplest, sensory processing must be hierarchically organised (either serially and/or in parallel) in a pipeline from receptor to feature processing (extraction, interpretation and recognition) and ultimately to motor output. Likewise, the neural structures responsible for pain need to be organised within a neural circuit that can execute the appropriate neurocomputations. The challenge then becomes one of identifying the generic circuitry necessary for the feeling of pain. We advance the framework of the brain as an inference machine that generates models of its own internal processes (Key and Brown, 2018). When hierarchically arranged, the outputs of these models represent progressive levels of awareness that are antecedent to feelings (i.e. the brain's experience of its own neural activity). We discuss our parallel forwards model algorithm and show that fish and molluscs lack the required neural architecture and therefore do not feel pain. Key, B. and Brown, D. (2018) Designing brains for pain: Human to mollusc. Frontiers in physiology 9:1027.