Butterfly wings create illusions that may confuse predators in flight

Butterfly wing patterns may do far more than advertise or conceal: according to a new study published in Nature on 30 September 2026, the high-contrast markings on butterfly wings interact with the insects’ flight dynamics to generate powerful illusory motion cues that could confuse attacking predators. The research, led by George R. A. Hancock of the University of Exeter with colleagues including Emmanuelle S. Briolat, Anna E. Hughes, Laura A. Kelley and Jolyon Troscianko, combines high-speed filming, biologically informed modelling, phylogenetic analyses, in silico evolution experiments and human behavioural tests to build a multi-stranded case for a long-debated idea.

The hypothesis has a distinguished pedigree. The dazzling stripes of zebras and snakes have long been thought to interfere with predator motion perception, but the authors note that this idea has received little empirical support so far. Butterflies, with their enormously diverse and high-contrast wing patterns, have inspired intense theoretical debate about what those patterns are for. The new study suggests an answer: that butterfly wing patterns work in tandem with flight dynamics to create visual illusions that confuse predators, potentially representing one of the most successful visual defence strategies in moving animals.

The first line of evidence came from the laboratory. The team recorded the take-offs of five Euro-African butterfly species, two of which were filmed for both sexes, giving seven morphotypes in total: the orange-tip Anthocharis cardamines, the diadem Hypolimnas misippus, the ochreous morph of the mocker swallowtail Papilio dardanus ochracea, the common swallowtail Papilio machaon and the red admiral Vanessa atalanta. Take-offs were chosen because most predator attacks begin on stationary butterflies. Footage was captured against a green-screen background at 1,057 frames per second using a CHRONOS 1.4 camera, with video durations ranging from 174 to 399 frames, or roughly 0.16 to 0.38 seconds of flight.

Those recordings were then passed through a custom-written, biologically informed elementary motion detector, or EMD, model built in ImageJ. The model works by spatiotemporal correlation of pixel intensity values across consecutive frames, quantifying motion in four cardinal directions: forwards, backwards, left and right. Crucially, the videos were filtered to mimic what a bird predator would actually see. Spatial contrast limits were based on small passerines observing at distances of 34 to 193 centimetres, depending on butterfly size, and temporal limits were based on a 100-hertz critical flicker fusion frequency for a passerine in a forest habitat. These distances were chosen because they are close enough that an avian predator would be likely to initiate its ballistic attack. Without spatial and temporal filtering, the model outputs would have been dominated by ‘wagon wheel’-type motion artefacts that would not occur in nature.

From the model outputs, the researchers calculated two key metrics. Forwards confusion, the Michelson contrast of backwards versus forwards motion energy, estimates how patterning alters the perceived speed of a moving object: greater values indicate a greater weighting of motion backwards compared with forwards, and hence a slower perceived speed. Sideways confusion, based on the mean of left and right energy compared with forwards, estimates how patterning alters the perceived turning rate of an object. Failure to accurately evaluate the speed or turning rate of an object can result in failure to intercept a target, the authors note. Each filmed take-off was analysed under four pattern treatments: the natural pattern, an averaged luminance, black and white, allowing the effect of the pattern itself to be separated from the effect of wing shape and movement.

To assess which features of butterfly wing patterns influence early-stage motion detection more broadly, the team turned to controlled three-dimensional simulations. They scanned the pages of the Collins Butterfly Guide, extracting dorsal images of 397 European butterfly species and 757 unique morphotypes, including subspecies and sexes. Each morphotype was rendered in Blender as an animated three-dimensional butterfly with five planes, a body, two forewings and two hindwings, rigged for flapping flight that matched the clap-and-fling kinematics of a filmed H. misippus. Wingbeat frequency was estimated from wing area using a near-linear relationship between butterfly mass, wing area and wingbeat frequency, because smaller wing area corresponds with higher wingbeat frequency across Lepidoptera, and higher frequencies are likely to interact with patterning to increase motion blur. Animations were rendered at 2,000 frames per second from a bird’s-eye view, with each butterfly flying forwards at a speed proportional to its body length and wingbeat frequency.

Because the diversity of butterfly wing patterns is a product of unique genetic and developmental pathways, which could either limit adaptive potential or produce motion-confusing patterns by coincidence, the team also ran in silico evolution experiments. Using a modified version of the CamoEvo Toolbox, they evolved artificial butterflies with independently evolving forewing, hindwing and body patterns under three different selection pressures: forwards confusion, sideways confusion and forwards energy. Three distinct wing shapes from three butterfly families were used, a skipper, a swallowtail and a nymphalid, with populations of 24 butterflies evolving for 20 generations, in which the top eight individuals survived and were replaced by mutant recombinant offspring. Control populations evolved under random selection to ensure patterns did not arise simply through drift. In total, 120 populations were evolved. The researchers report that these experiments independently converged on high-contrast, butterfly-like wing patterns, and that butterflies selected for motion confusion evolved to be more similar to real butterflies than randomly generated or unevolved ones.

Finally, the team validated their model predictions behaviourally. They built a high-framerate touch-screen butterfly capture game, rendered at 240 hertz, and recruited 100 anonymous volunteer players from the University of Exeter’s Penryn Campus, aged between 18 and 70 with normal or corrected-to-normal vision. Five butterfly phenotypes were selected from across the motion-confusion gradient, including Brintesia circe, Papilio alexanor, which has the highest sideways confusion in the set, Pyrgus sidae, Gonepteryx cleobule and Lycaena tityrus. Butterflies flew across the screen at 40 centimetres per second but with the equivalent wingbeat frequency and blur of a butterfly travelling at double that speed, matching the range of speeds seen in the free-flying take-offs. Each player undertook 30 trials, six replicates of each phenotype, with random flight paths, giving 3,000 trials in total. The players were told to attack the butterfly with a single finger, and the game recorded where they clicked relative to the butterfly’s position and heading.

The behavioural results supported the model’s predictions. The researchers tested whether the motion-confusion measures influenced where players clicked relative to the butterflies, using linear mixed models with the vector distance as the response and the axis and predicted confusion measures as predictors, with player identity, trial order and path identity as random effects. Their expectations were that forwards confusion should influence the click distance parallel to the butterfly’s heading and sideways confusion the perpendicular distance. The analysis also examined how the phenotypes shaped the distribution of clicks using two-dimensional kernel density estimation, and whether the EMD measures influenced the likelihood of clicking the forewing or hindwing. The authors note a limitation in these measures: because the butterfly’s orientation was not always aligned with the direction of movement, changes in orientation introduced noise into the confusion measures, and forwards confusion was on average higher for P. alexanor than for B. circe, probably owing to movements perpendicular to the start and end heading.

Phylogenetically controlled analyses, incorporating the complete European butterfly phylogeny of Wiemers and colleagues through principal coordinate analysis of cophenetic distances, demonstrated that motion confusion is an evolutionarily widespread strategy across the group. Taken together, the authors argue, the multiple independent lines of enquiry, free-flight films showing misleading motion cues, evolutionary analyses showing the strategy is widespread, simulations converging on butterfly-like patterns, and behavioural experiments validating the model’s predictions, show that butterfly wing patterns and flight dynamics create powerful illusory motion effects. The study was approved by the University of Exeter ethics committee, and the data, code and phylogenetic resources are openly available through the team’s GitHub and figshare repositories. As with any modelling-based approach, the findings rest on assumptions about predator vision, here calibrated to passerine birds, and the authors emphasise that the illusionary effects represent a potential defence whose full consequences in natural predator-prey encounters remain a subject for future work.

Subject of Research: The role of butterfly wing patterns in creating illusory motion cues that may confuse predators during flight

Article Title: Butterfly wing patterns in flight create powerful illusory motion cues

Article References: Hancock, G. R. A., Briolat, E. S., Hughes, A. E., Kelley, L. A., & Troscianko, J. (2026). Butterfly wing patterns in flight create powerful illusory motion cues. Nature, 658(8137), 958-962. https://doi.org/10.1038/s41586-026-11062-w

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11062-w

Keywords: butterflies, wing patterns, motion illusions, predator defence, visual ecology, elementary motion detector, flight dynamics, in silico evolution, high-speed filming, animal camouflage, Nature, behavioural experiment