The target article by Fleming and Michel offers a novel evolutionary account of conscious vision, arguing that it emerged during the transition from water to land due to the broader sensory horizons. From the role of conscious vision in model-based learning and action selection, they develop a theory of consciousness facilitating the higher order process of “reality monitoring”: “the capacity to determine whether internal signals reflect external reality or endogenous activity uncoupled from sensory input”.

While we find their arguments for the role of consciousness in reality monitoring compelling, we worry about the limitations that come from an excessive focus on vision. While F&M acknowledge that they are really just offering an account of visual consciousness, the inferences and theorising at times suggest otherwise. While they recognise that there is more to consciousness, including other types of perceptual processing and affective states, they are addressed only incidentally.

In this commentary, we’ll examine how these factors may challenge their evolutionary proposal, requiring one to think about the range of activities and perceptual experiences available to many animals. In particular, we’d like to highlight two limitations of the account presented. The first is the claim that visual consciousness is too slow for immediate action selection. The second is the possibility for reality monitoring across a far wider range of perceptual modalities, which may undermine the significance of the transition from water to land.

In arguing that conscious visual processing is too slow to guide immediate action selection, F&M have assembled a rich and convincing collection of empirical evidence about the speed of visual consciousness. As they note, Dennett & Kinsbourne (Reference Dennett and Kinsbourne1992) have argued in their influential “Time and the observer” that there is a difference between the time a mental state occurs and the time of an event represented by the mental state. From an evolutionary perspective, F&M’s insistence that conscious vision is too slow for rapid actions, such as fleeing from an ambush predator, or responding to fast-moving prey is almost certainly correct. However, we think they over-emphasise the importance of these specific rapid actions within an animal’s entire behavioural repertoire. Part of the problem here is the very nature of experiments on vision, which are almost always highly restrictive, limiting options, time, and conflicting variables for human and animal subjects (Veit Reference Veit2023). Many decisions made by animals occur at slower speeds: e.g. moving between foraging patches, finding shelter, interacting with conspecifics. These behaviours could be guided by visual input without the need for a rapid response. It is therefore unclear why visual consciousness needs to explain something beyond traditional action selection of this type. Indeed, there is already a lot of evidence that conscious vision can help animals to survey environments and build mental models of the location of key resources, and use this to guide behaviour over longer timescales.

Second, there is no reason to think that reality monitoring is necessarily or uniquely visual. F&M briefly note at the end of the paper that the same abilities could indeed arise for other sensory modalities, which could be the subject of other similar projects. While we appreciate that they deliberately focused on vision alone, we think some of the conclusions are weakened when this possibility is explored seriously. Animals, particularly aquatic animals, have a wide range of perceptual capacities beyond vision. These include hearing, lateral line vibration sensing, electroreception, and chemoreception. These modalities give the animals a far broader underwater sensory horizon than vision does, and therefore, if the sensory horizon is as important as the authors argue, they are highly likely to contribute to model-based learning and behavioural control leading to reality monitoring capacities and, presumably, consciousness.

Both of these give reasons to think the transition to land may not be as significant as the authors seem to imply. Slower-scale visual processing could play a role in much of the behaviour of aquatic animals, who could also be surveying their environments to gather information to feed into decision-making. As they note, “the cluttered terrestrial-like environments of coral reefs and tide pools may be unusually rich environments for the evolution of model-based visual cognition”. Additionally, some fish rely on visual processing for complex decisions: cleaner fish will visually detect parasites and even recognize marks on their own body with a mirror (Kohda et al., Reference Kohda, Sogawa, Jordan, Kubo, Awata, Satoh, Kobayashi, Fujita and Bshary2022). Pelagic fish who rely on stalking and ambush tactics rather than pursuit, such as many sharks (Munroe, Meyer, & Heithaus, Reference Munroe, Meyer and Heithaus2022), could reasonably use conscious visual input to guide their decision-making while hunting. Widening the scope of perceptual modalities also means that even pelagic fishes, who may not often rely on vision, could still have complex mental models that require real-time reality monitoring across other senses. Even if it were true that these types of aquatic animals specifically lack conscious vision, the likelihood of them having conscious processing across the other modalities decreases the significance of this difference. While the Cartesian theatre will always remain a tempting model, it is one we need to continue to resist (Dennett & Kinsbourne, Reference Dennett and Kinsbourne1992), and this includes its emphasis on vision (Veit, Reference Veit2022a). While vision may be central to human conscious experience, the evolutionary history of consciousness is likely to be much older (Veit, Reference Veit2022b).