Many insects’ motion vision is achromatic and thus dependent on brightness rather than on colour contrast. We investigate whether this is true of the butterfly Papilio xuthus, an animal noted for its complex retinal organization, by measuring head movements of restrained animals in response to moving two-colour patterns. Responses were never eliminated across a range of relative colour intensities, indicating that motion can be detected through chromatic contrast in the absence of luminance contrast. Furthermore, we identify an interaction between colour and contrast polarity in sensitivity to achromatic patterns, suggesting that ON and OFF contrasts are processed by two channels with different spectral sensitivities. We propose a model of the motion detection process in the retina/lamina based on these observations.
Many animals possessing colour vision nevertheless appear to perceive motion in an achromatic manner. Drosophila, for instance, exhibits no optomotor response to patterns of alternating blue and green stripes of a particular relative luminance, indicating that their motion detection mechanism is colour blind . This behavioural observation can be explained in anatomical terms by the finding that motion vision is mediated in insects by the short visual fibres (SVFs) , the set of photoreceptors that terminate in the lamina. In Drosophila, the SVFs are the outer photoreceptors R1–6, which all have the same broad-band wavelength sensitivity. Similarly, bees have spectrally homogeneous SVFs  and exhibit achromatic motion vision .
In this study, we consider a swallowtail butterfly, Papilio xuthus. Like many butterflies, this animal exhibits rather complex retinal organization (table 1). Papilio possess eight distinct photoreceptor classes with different spectral sensitivities, a subset of which appears in one of three fixed configurations in each ommatidium of the compound eye . Within each ommatidium, the photoreceptors are divided into distal (R1–4) and proximal (R5–8) tiers. R1/2 are short-wavelength-sensitive long visual fibres (LVFs) peaking in the ultraviolet (UV) to blue range. R5–8 are tuned to longer wavelengths: they are red-, broad-band- and green-sensitive in ommatidial types I, II and III, respectively. R3/4 are green-sensitive in all ommatidial types. A ninth photoreceptor (R9) also exists, but, owing to its small size, little is known about its physiology. Based on wavelength discrimination ability, Papilio's colour vision appears to be tetrachromatic and is thought to be mediated by the UV and blue receptors of R1/2 and the red and green receptors of R5–8 .
Unlike most insects, Papilio's SVFs are spectrally heterogeneous. If it is the case that all the SVFs retain their presumed ancestral motion-detecting role, then one would expect Papilio to possess a rather unusual polychromatic motion vision system. Alternatively, it has been proposed on anatomical grounds that motion is exclusively mediated by R3/4 ; in that case one would expect it to be achromatic, as in flies and bees. In this study, we resolve this quandary by measuring Papilio's optomotor responses to moving contrasts of various colour combinations.
2. Material and methods
Laboratory-reared spring-form P. xuthus were used for all experiments.
A 37″ (82 × 46 cm) plasma screen was used to present the stimuli. Plasma was chosen because it suffers markedly less from viewing angle artefacts than other display technologies (electronic supplementary material, figure S1). The spectra of the colours produced by the screen are shown in the electronic supplementary material, figure S2. The animal was suspended by its wings, with its head approximately 27 cm from the centre of the display, corresponding to a retinal size of approximately 112 × 80°. A camera was positioned to the side of the butterfly to record its responses.
Stimuli were generated using a custom-written program. In the case of edge-type stimuli, a horizontal edge would wipe vertically over the screen at constant velocity over the course of 2 s. For stripe-type stimuli, a 5 cm wide horizontal band would do likewise; note that the motion is slightly faster in the latter case, because the object covers 46 + 5 cm in the 2 s interval.
Images of the butterfly were captured immediately prior to the stimulus onset and at its offset 2 s later. From these, we estimate the head pitch by measuring the angle of the antennae (see the electronic supplementary material). The difference in the angles before and after the stimulus, averaged over the two antennae, was taken as the behavioural metric Δθ (figure 1a,b). Positive Δθ values correspond to upward movement.
To quantify the animal's sensitivity to a particular colour, we identify the lowest three consecutive intensity values all yielding a Δθ > 0.2°. In the case where this never occurred over the range we tested, all intensities above the highest we presented were assumed to elicit superthreshold responses.
Experiments were performed in batches of up to 14 stimuli separated by 12 s gaps. Within a batch, the stimulus type (edge/stripe), direction (up/down) and colour combination were kept constant. In stripe batches, the background colour intensity was held constant and that of the stripe varied. Within each edge batch, the polarity (ON/OFF) was kept constant, and the intensity of the colour varied (the dark side of the edge was always black). The sequence of trials in each batch was randomized, except for the final stimulus, which was a maximum contrast control; if this failed to elicit a Δθ > 0.5° (except in the case of downward control experiments), the whole batch would be repeated, or after several such failures, abandoned. Batches were in turn presented in a randomized sequence. Each individual completed up to eight batches, all of which involved different stimulus conditions.
3. Results and discussion
We presented moving stimuli to restrained Papilio and quantified the head movements they performed to track the visual motion. In the first experiment, a variable-intensity horizontal stripe of one colour moved vertically across the screen against a fixed-intensity background of another. Results are shown in figure 1c. Clearly, upward motion elicits positive Δθ values and downward negative, indicating that the response is directional, and thus motion-dependent. The appearance of a high-contrast static stripe elicits a small, barely significant downward response (electronic supplementary material, figure S3). There was no significant effect of presentation order within a batch (repeated measures ANOVA, p = 0.21), thus habituation is negligible. A strong effect of individual on Δθ existed (ANOVA, p < 0.0001). Taking the median Δθ for each individual across all three-colour pairings revealed no difference between the sexes (Wilcoxon signed-rank test, p = 0.39, n = 31).
For all three-colour pairings, we see a U-shaped response, with the minimum presumably corresponding to where the luminance contrast between the stripe and background is lowest, but Δθ always remains significantly above zero. We cannot conclude from this result that Papilio can detect motion using purely chromatic contrast, because the plasma display suffers from motion artefacts that generate unwanted luminance contrast (electronic supplementary material, figure S4). Quantifying this effect with a high-speed camera, we find that the artefact is considerably larger for blue/green than it is for green/red. However, the opposite pattern is observed in our behavioural data (figure 1c): the individual-wise minimum Δθ is significantly lower for blue/green than green/red (Wilcoxon signed-rank test, p = 0.0001, n = 31). Therefore, while this artefactual luminance contrast may contribute to the responses we observe, we consider it implausible that they can be fully attributed to it.
Further evidence of a polychromatic motion detection system comes from experiments using monochromatic edge-type stimuli, i.e. a transition from black to a colour (ON) or vice versa (OFF). To ensure that responses can be attributed to visual motion, rather than simply the overall change in luminance, we performed each trial with both upward and downward motion and subtracted the latter Δθ from the former (electronic supplementary material, figure S5). Figure 1d shows responses to blue, green and red contrasts of both polarities. For blue and green stimuli, the animals are significantly more sensitive to OFF than ON contrasts, but no significant difference is found for red (figure 1e). This strongly suggests that (at least) two distinct motion detection pathways with different spectral tunings exist. In flies, visual motion signals are split into ON and OFF channels, mediated by different lamina monopolar cells (LMCs) . While little is known about the anatomy and physiology of Papilio LMCs, we hypothesize that a similar separation exists, but that the distal and proximal tiers of SVFs (R3/4 and R5–8, respectively) contribute unequally to each channel.
Based on these findings, we implemented a two-channel model of Papilio visual motion detection (figure 2a and see the electronic supplementary material, Methods). One channel is purely green-sensitive, modelling R3/4, and is biased towards OFF edges; the other responds preferentially to ON contrasts, and is based on R5–8, giving it a complex spectral sensitivity profile that is heterogeneous across ommatida of different types. For the chromatic contrast experiment, the model quantitatively predicts the approximate intensity ratios that elicit the minimal response (figure 2b). Furthermore, it explains why responses are so small in the blue/green condition: these colours are rather close together in the model's trichromatic colour space (figure 2c) and thus generate less chromatic contrast than the other pairings. The model also exhibits the colour-dependent polarity effects on monochromatic sensitivity seen in the animal (figure 2d).
Recent studies on flies indicate that the visual systems for colour and motion are not as anatomically distinct as previously thought. SVFs are involved in colour discrimination , and conversely, LVFs contribute to the motion pathway . We consider it likely that the latter point is true of Papilio too, as synapses from R1/2 to R5–8 exist in the lamina ; this could explain why the animal is relatively more sensitive to blue than our model predicts (cf. figures 1d and 2d). However, we have omitted the LVFs from our model in the interests of parsimony.
In any case, our model proposes that R5–8 in Papilio play a dual role in both colour  and motion vision. We have demonstrated that this allows the animal to detect moving patterns that an achromatic system could not. However, considering that this effect is only revealed using carefully contrived artificial stimuli, chromatic motion vision may be of little ecological use. Indeed, this feature could even be seen as a weakness, as it has been argued  that an optimal motion vision system should be colour blind.
For a natural scene, contrast signals obtained from our model with and without contributions from R5 to R8 are highly correlated (electronic supplementary material, figure S6). Therefore, we take the view that sophisticated colour vision is adaptive in butterflies for some other purpose (e.g. feeding or mating), and this is what drove the spectral diversification of the SVFs. As this would have had rather little effect on motion vision, and because six photoreceptors are presumably better than two in terms of signal-to-noise ratio, there may have been no evolutionary reason to cease using inputs from R5 to R8 to detect motion.
Animal experiments were performed in accordance with MEXT guidelines.
Raw data can be found in the electronic supplementary material.
F.J.S. designed and performed experiments, conducted analysis and modelling, and drafted the manuscript. M.K. and K.A. designed experiments and revised the manuscript. All authors approve this version of the manuscript and accept accountability for its integrity.
We declare we have no competing interests.
This work was supported in part by JSPS grants-in-aid for Scientific Research to F.J.S. (26650116), M.K. (24570084) and K.A. (26251036).
We thank the anonymous reviewers for their constructive comments.
- Received August 10, 2015.
- Accepted September 25, 2015.
- © 2015 The Author(s)