The evolution of bright “warning” colours in nontoxic animals often is attributed to mimicry of toxic species, but empirical tests of that hypothesis are elusive. Populations of a harmless sea snake species (Emydocephalus annulatus) in New Caledonia exhibit colour polymorphism, with around 20% of individuals banded rather than melanic. Stability in that proportion over 20 years has been attributed to Batesian mimicry of deadly snake species by banded morphs of the harmless taxon. This hypothesis requires that banded colours reduce a snake’s vulnerability to predation. We tested that idea by pulling flexible snake-shaped models through the water and recording responses by predatory fish. Black and banded lures attracted similar numbers of following fish, but attacks were directed almost exclusively to black lures. Our methods overcome several ambiguities associated with experimental studies on mimicry in terrestrial snakes, and support the hypothesis that banded colour patterns reduce a non-venomous marine snake’s vulnerability to predation.

We pulled fibreglass models of snakes through the water to elicit feeding responses by fishes. These models were commercially-made fishing lures (Savage Gear 3D), 300 mm long, with each lure consisting of 12 linked segments to create a sinuous movement that mimics the swimming action of a snake (see https://www.youtube.com/watch?v=PV8R3wwd9xo for video of the lure’s action). We removed the hooks and added weights (Storm Suspenstrips) to the ventral surface of the lure to ensure negative buoyancy. We used acrylic craft paint (Born Acrylic Paint Set, Officeworks) to render some lures black, and give others either white or grey bands. We used three replicate lures of each colour morph during the trials to avoid stimulus pseudoreplication.

To assess responses of free-living predatory fish, one of us (CG) pulled a lure behind her on a 5-m length of 10 kg monofilament fishing line as she snorkelled a 50-m straightline transect through water 1–3 m deep. A second person swam 3 to 5 m behind the lure (depending on water clarity) to score (a) attacks (fish seizes the lure), (b) follows (fish orients to lure and swims behind it, approaching but not seizing the lure), and (c) body lengths of all fishes based on comparison with the 300-mm lure. Follows as well as attacks appear to indicate predatory intent, because (a) all follows were by large (>300 mm long) predatory fishes; and (b) attacks were always preceded by follows.