Spectral circuits for figure-ground segmentation in motion vision

We will elucidate the role of "colour" information in supporting motion vision. These two fundamental abilities of eyes are usually considered in isolation. However, both from basic physics of how light travels in the water, and from looking at the evolution of vision, the two must be fundamentally entwined.

Background. Vision evolved first in the water. First came light sensitivity, enabled by the evolution of opsins some 800 million years ago. Soon after probably came a rudimentary sense of "colour vision", enabled by the diversification of opsins into variants that were sensitive to different wavelengths ("colours") of light. Primitive animals likely would have been able to use this newfound sense to tell the "colour" of their surroundings, albeit without knowing its spatial structure - after all, image forming vision, requiring ordered arrays of photoreceptors, screening pigment and eye optics had not yet evolved. Nevertheless, even without knowledge of space, colour alone can be useful. For example, it can inform about water depth: Light from the sun penetrates water in a "colour-dependent" manner: Blue and UV light is rapidly lost, while green and red light penetrates much deeper. Accordingly, if the environment is blue/UV-rich, chances are you are near the surface.

It would take another ~250 or so million years before early "colour-vision" systems would evolve into full-flown eyes. This critical step probably happened some 540 million years ago during the Cambrian explosion, when a newly found sense of "image forming vision" is thought to have centrally enabled the emergence of neurally complex animal life as we know it today. Suddenly, animals could use their eyes to navigate their surroundings much more efficiently, stabilise their bodies, and visually spot potential prey and predators. These newfound abilities were made possible by new neural circuits within the eyes and brains of our early ancestors that computed complex types of information in the visual scene.

Perhaps most critical of all was the ability to sense motion. Motion of the background would tell animals how they themselves were moving through the environment, while motion of the foreground would highlight potential nearby objects to interact with. Animals must be able to be able to tell the two apart. This is generally thought to be achieved by relatively complex and far from understood circuits of the retina and brain that constantly compare brightness changes over time across different parts of visual space. However, looking back at how our very earliest ancestors might have told water depth simply based on the "colour" of their surroundings, the very same principle of basic physics should serve equally well to tell the distance of objects in the water. In other words, the "colour" of an object alone should tell animals if it is near, or far. What is more, since "colour" vision almost certainly predates motion vision, circuits enabling the latter would have necessarily had to evolve on top of pre-existing colour circuits. It would then be very surprising indeed if colour information were not fundamentally inbuilt into circuits that extract visual motion, including in animals that are alive today.

Objectives. We will work on the experimentally amenable larval zebrafish which allow unrestricted optical access to any part of the eyes and brains in the live animal, and which inhabit shallow freshwaters not too dissimilar from the world where vision first evolved. We will combine videography data from the field, behavioural observations, genetic manipulations of retinal circuits, and state-of-the-art neurophysiological recordings of 1,000s of individual nerve cells at a time to ask if and how zebrafish use colour information for motion vision.

Impact. Understanding the true evolutionary origins and possible interplay of colour and motion vision systems will inform how "vision" works in a very general sense, including in our own eyes.

The evolution of image-forming vision during the Cambrian ~540 million years ago is thought to have centrally driven speciation on our planet. However, at this time a basic capacity for "colour vision" was likely already in place, based on spectral comparisons that could be made between different opsin and photoreceptor lineages that evolved ~800 mya. Four ancestral photoreceptor lineages (SWS1: UV, SWS2: blue, RH1/2: rods+green, LWS: red) survive in the eyes of extant vertebrates, with three of them (UV, rods, red) driving vision in mammals including humans.

Yet, one early purpose of spectral computations was probably not for what we might call "colour vision" today, but rather for "figure-ground" segmentation - a much more fundamental ability of vision which allows e.g. spotting predators, prey, and obstacles against the background. This is because under water where vision first evolved, light scatters in a strongly wavelength selective manner: red penetrates deep, but UV is rapidly lost with distance. UV-light therefore primarily serves foreground vision, but red-light can serve both fore- and background vision. Contrasting UV against red at a retinal circuit level then accentuates the foreground against the background. I therefore hypothesise that this may present the true origin of vertebrate colour vision as we know it today.

Working on zebrafish as a model, we will combine natural imaging, behaviour, genetic cone-ablations, and in vivo physiology to establish the basic principles by which UV-red contrast can be leveraged to accentuate figure-ground segmentation in aquatic visual scenes.

Because zebrafish share ancestral red and UV cone-photoreceptor systems with nearly all extant vertebrates, it is likely that any findings from fish will readily translate to other species, including humans. Our results will fuel future thinking and experiments into the origins of vision, and feed into the design of diverse camera technologies.