The Colour-Blind Animal That Matches Everything

ANo animal makes a better case for looking without seeing than the common cuttlefish, Sepia officinalis. Its skin carries some ten million chromatophores, and within a second of landing on a bed of gravel the animal can raise a coarse disruptive pattern that breaks up its outline. Its retina, however, holds a single visual pigment, which by every ordinary measure makes it colour-blind. The performance is routinely described as a decision, a word borrowed, along with the checkerboards and the scoring sheets, from a century of work on vertebrate vision. The borrowing sits badly: those procedures assume an eye that separates wavelengths, an observer that weighs what it sees against what it has stored, and a body still enough for the weighing to happen. Whether the matched skin records anything that earns the old vocabulary is the question the field keeps circling.

BInes Vaz, a visual ecologist at the Farne Institute of Marine Optics, argues that the patterns are the settled output of a rule plain enough to need no memory. Each patch of skin, on her account, answers to the local edge statistics of the image falling on the retina, meaning the density and the coarseness of light-dark boundaries, with brightness alone doing the work that colour is assumed to do. Between 2015 and 2022 her group filmed 214 cuttlefish at 200 frames per second and scored 46,000 skin images; a pattern settles, on average, about 240 milliseconds after a substrate appears. Fed nothing but those edge statistics, her model reproduced 79 per cent of the patterns her animals actually produced. Nothing in the model looks backwards: what an animal saw a minute ago enters the arithmetic only in so far as it is still in front of the eye.

CAn account of that shape makes a prediction, and Vaz has tested it. If the display is only the present image settling into muscle, the animal should be beaten wherever the useful response is one the substrate underfoot does not call for, and beaten too wherever what to show next turns on something seen minutes earlier. In delayed-substrate tanks, where the gravel visible at the start of a trial announced a second gravel that would be uncovered four minutes later, her cuttlefish succeeded on 22 per cent of trials, against 91 per cent on immediate matching of comparable difficulty. Success in the delayed tanks was confined to trials in which both substrates could be taken in from a single position. The contrast is instructive, since a wrasse handles the waiting version easily and the immediate one poorly. Vaz insists that the figures be read as a fence around one explanation rather than a verdict on the animal.

DTomas Hedlund, a neuroethologist at the Cape Aldric Marine Station, replies that a cuttlefish held on bare white sand under unchanging overhead light is performing in conditions its species never meets, and that a failure recorded there describes the tank at least as much as its occupant. His own animals are kept for twelve days on mixed gravel under a regular cycle of light and darkness, an arrangement now unremarkable but thought eccentric when he adopted it in 2018. Under that regime, 64 per cent of his cuttlefish began shifting towards the coarser pattern shortly before a substrate change that had been made every four minutes was next due, and went on doing so for two further cycles after the changes had stopped. Hedlund grants two weaknesses. His fifteen animals came from a single hatchery, and the skin-level light sensing he calls distributed seeing was inferred from opsin expression and the timing of the shifts alone, because nothing yet allows it to be measured in a living animal.

EThe argument has travelled well beyond the tanks. Engineers have taken up the matching rule as a design principle: a 2019 study turned it into an adaptive fabric that observers took 31 per cent longer to pick out, although the advantage vanished whenever the light source was moved. Textile firms in three countries have commissioned similar work, and one was criticised for advertising its output as camouflage designed by an animal. Vaz points out, uncomfortably for her own position, that the rule was fitted entirely on substrates made of printed paper and predicts nothing about a broken rock face. Hedlund, for his part, insists that a mechanism and an experience are not alternatives between which anyone has to choose, and that funding bodies read reports of skill as evidence that an animal is understood. Both regret the habit of publishing the single most striking photograph rather than the spread of the untidy ones.

FA settlement of sorts is forming in which the two accounts are given different work rather than ranked against each other. Where the substrate is in view, the skin appears to match it with nothing carried from one moment to the next, which would explain both the speed of the response and the animal's indifference to the pattern it was wearing a moment earlier. Where a display must survive a gap in time, attention has turned outwards, to the light-sensitive skin itself, which registers a substrate the eye is turned away from, a sensor kept outside the head. What remains open is whether that distributed sensing does any work on a reef at dusk, where the light is scattered and the very contrast the rule depends on largely disappears, and no field method now available can decide the matter. What both researchers reject is the assumption behind the checkerboard photographs: that an animal without colour vision must be either a poor copy of an eye that has it or nothing at all.