Fleming and Michel’s account of the evolution of conscious vision elegantly ties together three observations: (1) there is often a long time lag, on the order of hundreds of milliseconds, between a stimulus and the conscious visual perception of it; (2) this lag is a major problem in marine environments, where distance vision is obstructed and you typically don’t consciously see something coming, or darting away, until it is too late; and (3) the lag is more manageable in terrestrial environments, where animals can see opportunities and risks at long range and have time to plan a response on the timescale of seconds. All this renders plausible a link between conscious vision and the transition to land.
Their account is concerned only with vision, aside from a very brief discussion of other sense modalities at the end, which leaves an open question about the relation of conscious vision to other conscious states. I am sympathetic to the idea that conscious processing is (or is closely related to) the modeling of risks and opportunities in the environment in the service of planning. However, given this general idea, we can distinguish two ways of thinking about the place of vision in its evolutionary history: vision first and vision later.
On the vision-first story, it is through vision that animals first become able to obtain useful information about the distal environment. This drives the evolution of modeling systems that support planning beyond the present moment. Once these systems exist, it then becomes a valuable to add further sensory modalities to the model. It is useful for planning to know how hungry you are, or how injured, so bodily sensations get added. It is useful to know if a fight-or-flight response has been triggered, so emotions like fear get added. It is useful to know what opportunities and risks might be hidden nearby, so olfaction gets added. This is a story in which the multisensory feast of consciousness as we know it today follows in the wake of conscious vision and is unique to lineages that evolved long-distance vision. I surmise that this is the kind of story Fleming and Michel would favor.
On a vision-later story, by contrast, planning for at least the next few seconds is no less essential for marine organisms than for terrestrial ones. They too must budget for future demands on bodily resources, constantly adjusting their investments based on the risks and opportunities of the environment, the current and predicted state of their body, and what this implies about their allostatic needs. No matter how murky the water gets, you still need to do that.
In marine environments, though, other sources of information are more useful for planning than vision. Vision for you is more of an immediate threat detection system, akin to a smoke alarm. It tells you when something has got too close. It is mainly olfaction and other forms of chemosensing that give you the bigger picture about your surroundings. Hence, you evolve a form of consciousness that includes bodily needs, emotions, and chemosensory/olfactory information about the external world, but initially not vision. There is no need to include the smoke alarm in the central planning model. It is more useful if the smoke alarm triggers fast, inflexible responses, bypassing your planning systems.
Later, after the transition to land, vision becomes more than a smoke alarm, and it becomes more relevant to planning on the timescale of seconds. Indeed, if you cannot exploit visual information for action planning, you will have no chance. So, evolution starts building a new visual pathway that leads to preexisting central models for planning, prioritization, and behavioral control.
Which story is more likely? Vision first or olfaction/chemosensing first? Because we humans are such strongly visual animals, we intuitively gravitate towards vision-centric narratives. But the more I think about the “olfaction/chemosensing-first” story, the more appealing I find it.
Here are the two pieces of circumstantial evidence. One is anatomical: right across the vertebrates, the olfactory bulb connects directly to the forebrain (the cortex in mammals, the telencephalon in fishes) without the need for any thalamic relay (Su, Menuz, & Carlson, Reference Su, Menuz and Carlson2009). On an evolutionary story that involves the forebrain gradually evolving greater capacities for modeling the external causes of sensory stimulation, olfaction is in the right place to provide input. By contrast, vision-first stories must contend with the problem that the main visual integration center in fishes is the optic tectum in the midbrain, homologous with the superior colliculus in mammals (Feinberg & Mallatt, Reference Feinberg and Mallatt2016).
A second concerns windows of integration. We can discriminate a difference in the order in which two odors are presented even when the offset between them is only around 60ms (Wu et al., Reference Wu, Chen, Xing, Huang, Zhao and Zhou2024). But the associated conscious experience is not that of two rapid odors experienced in quick succession. Unconscious olfactory discrimination can be very fast indeed, but conscious olfaction is at least as slow as conscious vision (Cooke & Myin, Reference Cooke and Myin2011). The quality of odor is constructed holistically over a period of hundreds of milliseconds based on many fine-grained, unconsciously processed features of the stimulus, including the temporal offsets between discriminable elements.
What we call a “postdictive” effect in vision never receives that name in olfaction, but only because we find it entirely routine for the quality of an experience to result from holistic integration across a few hundred milliseconds’ worth of stimuli. What seems remarkable in the visual case is unremarkable in the olfactory case. It is hardly a surprise that we integrate across the length of a sniff. We do not sniff visually, but postdictive effects suggest that conscious vision works as though we did. That is, conscious vision too involves constructing an experienced quality by integrating across a few hundred milliseconds of sensory input. There is no particular reason why the integration window of conscious vision should be approximately the duration of a sniff unless conscious vision is co-opting integrative machinery that first evolved for, and is still used in, conscious olfaction.
You have
Access
Fleming and Michel’s account of the evolution of conscious vision elegantly ties together three observations: (1) there is often a long time lag, on the order of hundreds of milliseconds, between a stimulus and the conscious visual perception of it; (2) this lag is a major problem in marine environments, where distance vision is obstructed and you typically don’t consciously see something coming, or darting away, until it is too late; and (3) the lag is more manageable in terrestrial environments, where animals can see opportunities and risks at long range and have time to plan a response on the timescale of seconds. All this renders plausible a link between conscious vision and the transition to land.
Their account is concerned only with vision, aside from a very brief discussion of other sense modalities at the end, which leaves an open question about the relation of conscious vision to other conscious states. I am sympathetic to the idea that conscious processing is (or is closely related to) the modeling of risks and opportunities in the environment in the service of planning. However, given this general idea, we can distinguish two ways of thinking about the place of vision in its evolutionary history: vision first and vision later.
On the vision-first story, it is through vision that animals first become able to obtain useful information about the distal environment. This drives the evolution of modeling systems that support planning beyond the present moment. Once these systems exist, it then becomes a valuable to add further sensory modalities to the model. It is useful for planning to know how hungry you are, or how injured, so bodily sensations get added. It is useful to know if a fight-or-flight response has been triggered, so emotions like fear get added. It is useful to know what opportunities and risks might be hidden nearby, so olfaction gets added. This is a story in which the multisensory feast of consciousness as we know it today follows in the wake of conscious vision and is unique to lineages that evolved long-distance vision. I surmise that this is the kind of story Fleming and Michel would favor.
On a vision-later story, by contrast, planning for at least the next few seconds is no less essential for marine organisms than for terrestrial ones. They too must budget for future demands on bodily resources, constantly adjusting their investments based on the risks and opportunities of the environment, the current and predicted state of their body, and what this implies about their allostatic needs. No matter how murky the water gets, you still need to do that.
In marine environments, though, other sources of information are more useful for planning than vision. Vision for you is more of an immediate threat detection system, akin to a smoke alarm. It tells you when something has got too close. It is mainly olfaction and other forms of chemosensing that give you the bigger picture about your surroundings. Hence, you evolve a form of consciousness that includes bodily needs, emotions, and chemosensory/olfactory information about the external world, but initially not vision. There is no need to include the smoke alarm in the central planning model. It is more useful if the smoke alarm triggers fast, inflexible responses, bypassing your planning systems.
Later, after the transition to land, vision becomes more than a smoke alarm, and it becomes more relevant to planning on the timescale of seconds. Indeed, if you cannot exploit visual information for action planning, you will have no chance. So, evolution starts building a new visual pathway that leads to preexisting central models for planning, prioritization, and behavioral control.
Which story is more likely? Vision first or olfaction/chemosensing first? Because we humans are such strongly visual animals, we intuitively gravitate towards vision-centric narratives. But the more I think about the “olfaction/chemosensing-first” story, the more appealing I find it.
Here are the two pieces of circumstantial evidence. One is anatomical: right across the vertebrates, the olfactory bulb connects directly to the forebrain (the cortex in mammals, the telencephalon in fishes) without the need for any thalamic relay (Su, Menuz, & Carlson, Reference Su, Menuz and Carlson2009). On an evolutionary story that involves the forebrain gradually evolving greater capacities for modeling the external causes of sensory stimulation, olfaction is in the right place to provide input. By contrast, vision-first stories must contend with the problem that the main visual integration center in fishes is the optic tectum in the midbrain, homologous with the superior colliculus in mammals (Feinberg & Mallatt, Reference Feinberg and Mallatt2016).
A second concerns windows of integration. We can discriminate a difference in the order in which two odors are presented even when the offset between them is only around 60ms (Wu et al., Reference Wu, Chen, Xing, Huang, Zhao and Zhou2024). But the associated conscious experience is not that of two rapid odors experienced in quick succession. Unconscious olfactory discrimination can be very fast indeed, but conscious olfaction is at least as slow as conscious vision (Cooke & Myin, Reference Cooke and Myin2011). The quality of odor is constructed holistically over a period of hundreds of milliseconds based on many fine-grained, unconsciously processed features of the stimulus, including the temporal offsets between discriminable elements.
What we call a “postdictive” effect in vision never receives that name in olfaction, but only because we find it entirely routine for the quality of an experience to result from holistic integration across a few hundred milliseconds’ worth of stimuli. What seems remarkable in the visual case is unremarkable in the olfactory case. It is hardly a surprise that we integrate across the length of a sniff. We do not sniff visually, but postdictive effects suggest that conscious vision works as though we did. That is, conscious vision too involves constructing an experienced quality by integrating across a few hundred milliseconds of sensory input. There is no particular reason why the integration window of conscious vision should be approximately the duration of a sniff unless conscious vision is co-opting integrative machinery that first evolved for, and is still used in, conscious olfaction.
Financial support
JB is supported by The Jeremy Coller Foundation and the Good Ventures Foundation.
Competing interests
None.