Motion-sensitive neurons activated by chromatic contrast in a butterfly visual system

Abstract

A pattern of two equally bright colours contains only chromatic contrast. Unlike in flies, such a pattern elicits strong optokinetic responses in the butterfly Papilio xuthus. To investigate the neural basis of chromatic motion vision, we performed single-cell electrophysiology. We found spiking neurons exhibiting direction-selective motion sensitivity in the second optic ganglion, the medulla. We analysed the response characteristics of these neurons using two-colour stripe patterns moving vertically. We systematically manipulated the intensities of the colours so that the set of presented patterns included an isoluminant condition for the butterfly. Moving patterns containing only chromatic contrast still elicited a response in the neurons. The neurons' sensitivity profile is similar to that of the behavioural responses. Post-recording dye injection revealed that the neurons have dendrites in the ventral lateral protocerebrum and axonal processes in the medulla, suggesting a feedback role. Presumably, the neurons contribute to subtracting wide-field motion to facilitate the detection of small moving targets.

This article is part of the theme issue ‘Understanding colour vision: molecular, physiological, neuronal and behavioural studies in arthropods’.

1. Introduction

Dynamic and static features of visual scenes are often processed by separate subsystems, such as the magnocellular and parvocellular pathways in primates [1]. Similar segregation of signals also occurs in insects. Insect compound eyes typically contain two subsets of photoreceptors, the short visual fibres (SVFs) and long visual fibres (LVFs), with axons terminating, respectively, in the first optic ganglion, the lamina and the second optic ganglion, the medulla. In Drosophila melanogaster, the SVFs exhibit uniform broadband spectral sensitivity, while the LVFs are spectrally heterogeneous. Because of this organization, it was hypothesized that achromatic vision depends on the SVFs, while the LVFs provide input for colour vision [2]. Indeed, an early behavioural study concluded that D. melanogaster's motion vision is colour blind [3]. However, more recent studies have revealed that this hypothesis is too simplistic. In D. melanogaster, spectrally diverse LVFs also contribute to motion vision [4,5], and inputs from the spectrally homogeneous SVFs contribute to colour discrimination [6].

In the course of studying the colour vision of a butterfly Papilio xuthus, we found that they can use chromatic contrast to track moving objects [7]. Since the eyes of P. xuthus are spectrally more complex than those of D. melanogaster, their photoreceptors cannot be so straightforwardly divided into colour and motion vision processing roles. The compound eyes of P. xuthus bear six classes of spectral receptors, which are most sensitive in the ultraviolet (UV), violet (V), blue (B), green (G), red (R) or broadband (BB) wavelength regions (electronic supplementary material, figure S1a). We previously localized these photoreceptors in the ommatidial array and identified three spectrally distinct ommatidia types (electronic supplementary material, figure S1b) [8]. Because the seven SVFs in the P. xuthus ommatidia (R3–R9) are spectrally heterogeneous, the hypothesis that the SVFs are responsible for achromatic vision simply does not hold in this case. We initially hypothesized that motion vision is achromatic and based solely on R3 and R4, which are G-sensitive in all ommatidia types: the G receptors would offer spatially acute but achromatic vision. We tested this hypothesis by measuring head movements of tethered P. xuthus while presenting two-colour patterns [7] following Yamaguchi's isoluminance protocol [3]. Although the method was not sensitive enough to detect the LVFs' contribution in D. melanogaster, moving patterns with exclusively chromatic contrast induced clear optokinetic responses in P. xuthus (electronic supplementary material, figure S1c). In addition, we could qualitatively reproduce the results with a model postulating that the R3/R4-based motion detection system receives additional inputs from the R5–8 SVFs of all ommatidia types [7]. Conversely, the SVFs also contribute to colour vision in P. xuthus, as their vision is tetrachromatic, based on the UV, B, G and R receptors [9]. The UV and B receptors are LVFs which directly innervate the presumptive colour vision circuits in the medulla, but the G and R inputs can come only from the SVFs and must be relayed to the medulla via higher order neurons.

The interplay of SVFs and LVFs suggests that motion and colour pathways are less segregated in P. xuthus than in D. melanogaster. How do these insects see moving coloured objects? How does evolution shape parallel pathways in sensory systems in general? To address such questions, we have started investigating the neuronal mechanisms underlying the colour-dependent motion vision in P. xuthus. We analysed the motion sensitivity of the spiking neurons in the medulla using a similar two-colour isoluminance protocol to our previous behavioural study. We thus found that these neurons exhibit a similar response pattern to the chromatic contrast-driven motion detection behaviour of tethered butterflies.

2. Material and methods

(a) . Animals

We used spring-form adults of the Japanese yellow swallowtail butterfly, P. xuthus. We captured female butterflies in the field around Sokendai, Hayama, Japan. We let the females lay eggs and raised the hatched larvae on citrus leaves at around 26°C under a day : night cycle of 10 : 14 h, which induced pupal diapause. After four months of chill treatment at 4°C, we let the butterflies emerge at 26°C. We used adults more than two-days old for the experiments.

(b) . Electrophysiology

(i) . Sample preparation

For electrophysiology, we used beeswax to mount a butterfly on a custom-made stage with its dorsal side up and head pointing forward (electronic supplementary material, figure S2a). We fixed the right antenna to the thorax, cut its tip and inserted thin AgCl wire into the antenna's cut end as the reference electrode. Then, we removed the left antenna, opened the head capsule around it, removed the large trachea to expose the brain and filled the head capsule with physiological saline (147 mM NaCl, 1.3 mM KCl, 4.0 mM CaCl₂ in 10 mM NaHCO₃). We unsheathed the brain before inserting the electrode into the nervous tissue.

We used borosilicate glass micro-electrodes, pulled using a P-1000 puller (Sutter Instrument, CA, USA) and filled the tip with 2.5% neurobiotin in 1 M KCl. We used a WR-6 micromanipulator (Narishige, Tokyo, Japan) to insert the electrode into the left medulla, where we probed neurons producing spikes upon light stimulation. The electrode resistance in Ringer's solution was 70–150 MΩ.

(ii) . Stimuli and recording

We produced visual stimuli using a Lightcrafter 4500 digital light processing (DLP) projector (Texas Instruments, Dallas, Texas, USA) and a rear-projection screen set in front of the animal (electronic supplementary material, figure S2a). The projector has red, green and blue LEDs, peaking at 626, 545 and 455 nm, with the full-width-half-maximum of 16, 88 and 20 nm, respectively (electronic supplementary material, figure S2b). We measured the photon flux of the light using a USB2000 spectrometer (Ocean Insights, Orlando, Florida, USA) and calculated the integral of the photon flux in the range of 300–700 nm at a resolution of 1 nm. The integrated maximum intensity of the blue, green and red LEDs was respectively 1.19, 0.54 and 1.78 × 10¹³ photons cm⁻² s⁻¹ nm⁻¹ at the position of the butterfly's eyes. We attenuated the intensity of the three LEDs during the experiments using pulse-width modulation.

Whenever we encountered a spiking neuron, we first performed pilot experiments using black-and-white stripe (square wave) patterns moving vertically and horizontally (figure 1a). The patterns had a spatial frequency of 0.05 cycles degree⁻¹, moving at 2.5 Hz, which we determined to be most effective in eliciting a response during preliminary experiments. Before recording the motion sensitivity, we adapted the animal to the illumination level by presenting a static pattern for 2 s and then moved the pattern upwards for 3 s. We repeated the session three more times, separated with 4 s of a dark period, with the stimulus moving downwards, leftwards and rightwards, from which we could judge the neuron's preferred direction.

Figure 1.

Response of neuron I, a downward-selective motion-sensitive neuron, to the black-and-white stimulus series. (a) Responses to upwards, downwards, leftwards and rightwards motion. The top trace of each panel shows the spike train and the bottom trace shows the corresponding firing rate. The red rectangles indicate the time windows used for averaging. The circled numbers correspond to those in (b). (b) The average firing rate of the neuron for the static (light grey) and moving stimuli (dark grey) for each direction. The dashed line shows the spontaneous activity level measured before starting the first stimulus. (c) Differences of firing rate (moving minus static) for four directions. The values are normalized to the largest absolute value.

After judging the preferred direction, we presented chromatic stimuli using two-colour stripes (square wave) patterns with varying contrasts. The control was the patterns of two intensities of green (G/G pattern) because the motion pathway appeared most sensitive to green light [7]. We first tested with the G/G pattern and then with green/blue (G/B), green/red (G/R) and blue/red (B/R) patterns. The two-colour patterns had the same temporal and spatial frequency as the pilot experiment. We moved two-colour patterns only in the neurons' preferred direction because of the difficulty of maintaining quality recordings over extended periods.

We calculated the intensity contrast of each pattern, I_c, by

$$I_c={\log }_{10}\left({\displaystyle \frac{c_{\hspace{0.17em}\mathit{v}\mathit{a}\mathit{r}}}{c_{cst}}}\right),$$ 2.1

where c_var and c_cst are the intensities of two lights of a stimulus pattern. We systematically increased one colour's intensity (c_var) while keeping the other constant (c_cst). The control being a set of stripes with the same intensity of green at I_c = 0; no motion was visible throughout the experiments: the stimulus was a static green screen. The colours that changed in intensity (green in the G/B and G/R patterns and blue in the B/R patterns) had 11 intensity steps (electronic supplementary material, table S1).

(c) . Anatomy

After recording the responses, we injected neurobiotin into the neuron by applying a positive current of 1.5 nA for up to 7 min. If the response was still stable after injection, we continued recording if necessary. The butterfly was then kept on the stage for up to 20 min to let neurobiotin diffuse in the cell.

We opened the head capsule's top and back, isolated and immersed the head in 4% paraformaldehyde, 2.5% glutaraldehyde and 0.2% picric acid in 0.1 M phosphate buffer (PB) at 4°C overnight. We then removed the remaining cuticle and embedded the brain in gelatin-containing albumin. We further fixed the brain-containing gelatin block in 4% paraformaldehyde in 0.1 M PB at 4°C overnight.

The following day, we made 40 µm vertical brain sections using a vibrating microtome (Leica VT 1000 S, Wetzlar, Germany). After being rinsed in 0.1 M PB (20 min × 3 times), we incubated the sections in horseradish peroxidase in 0.1 M PB saline containing 0.5% Triton X at room temperature for about 18 h. Next, we rinsed the sections in 0.1 M PB and applied 0.1 mg ml⁻¹ diaminobenzidine (DAB) to visualize the neurons. Once we confirmed the staining under a dissecting microscope or after 30 min of DAB incubation, we rinsed and arranged the sections on a gelatin-covered slide. Finally, the sections were air-dried overnight, rinsed with distilled water, dehydrated with a graded ethanol series and covered using Mount-Quick (Daido Sangyo, Tokyo, Japan) for observation.

(d) . Data analysis

We analysed the electrophysiological data using a custom-written python program. First, we calculated the firing rate (inverse of the inter-spike interval) across time (figure 1a). Next, we determined the spontaneous activity of the neuron by taking an averaged firing rate over 2–5 s of activity before presenting the first stimulus. Then, for every stimulus, we calculated the average firing rate when the pattern was static and moving (red rectangles in figure 1a). We excluded the first and last 0.25 s as the neurons often showed strong phasic responses. We plotted the average firing rate for the static and moving intervals of the stimulus (figure 1b). Finally, we subtracted the corresponding static response from each motion response to categorize the neurons based on the apparent profile (figure 1c).

3. Results

(a) . Response of motion-sensitive neurons to black-and-white stripes

Figure 1 shows a typical result of the pilot experiment. The firing rate increased when the stimulus was moving downwards, while upward motion inhibited the activity: we judged the neuron to be motion sensitive with selectivity to downward motion. The neuron also responded (negatively) to horizontal motion but exhibited no directional selectivity in this plane.

Some apparent features of the responses are as follows. First, the neurons transiently respond when the pattern appears. A transient response also occurs when the stimulus starts moving, and when the pattern disappears. Second, the firing rate during the static stimulus presentation was typically higher than the spontaneous activity level in the dark, presumably because of the overall brightness. Third, after the on-transient response, the firing rate gradually returns to the level of spontaneous activity while the static stimulus is on.

Electronic supplementary material, figure S3 shows the selectivity profiles across the 15 neurons (I–XV) in pilot stimuli, whose distribution is summarized in the electronic supplementary material, figure S4. The average of the selectivity index was of 0.50, and it was significantly different than 0 (Student's t-test; t = 11.48, p-value = 1.64 × 10⁻⁸). The nine downward-selective neurons (I–IX) all showed decreased activity when presented with upward motion. For the upward-selective neurons (X–XV), the inhibition by downward motion was less pronounced. The neurons arranged in the same manner with the identities (IDs) in the electronic supplementary material, figures S3, S5 and S6.

(b) . Response of motion-sensitive neurons to two-colour patterns

Figure 2 shows the responses of a downward-selective neuron I (figure 2a) and an upward-selective neuron X (figure 2b) to four two-colour patterns moving in their preferred directions. (The same graphs appear in the electronic supplementary material, figure S5.)

Figure 2.

Responses of downward- (neuron I, (a)) and upward- (neuron X, (b)) sensitive neurons to four sets of two-colour stripe patterns. The green, blue, red and purple curves, respectively, correspond to the patterns of G/G, G/B, G/R and B/R. SPA, spontaneous activity. (c) The difference in firing rate between the static and moving stimuli for all values of intensity contrast in neuron I. The green, red, blue and purple columns, respectively, correspond to the patterns of G/G, G/R, G/B and B/R.

In neuron I, the control G/G pattern elicited spikes at a frequency (21 Hz) slightly above the spontaneous activity (16 Hz) when the intensity contrast, I_c, was zero (green curves). This must be attributable to the increase in overall scene brightness, as no motion is visible in this instance; a similar pattern is observed in the upward-selective neuron X (electronic supplementary material, figure S5). The firing rate increased as the absolute value of I_c increased, giving a V-shaped curve. Here the neurons responded to brightness contrast only because the G/G patterns contained no chromatic contrast.

The responses to two-colour patterns also formed V-shaped curves. In this neuron, the G/R and the B/R pattern elicited minimal responses for negative I_c values (figure 2a), i.e. when the red light was more intense than the green or blue light: the neuron appeared to be less sensitive to red. The G/B pattern produced its minimum on the side of positive I_c, where the green light was stronger than the blue light, indicating that the neuron was more sensitive to blue than to green. The minimal response values of these three patterns were all higher than that of the G/G control.

Because the recording was often unstable, we could collect data for all three two-colour patterns only in neurons I, X, XII and XV (electronic supplementary material, figure S5). Plus, the experiments often failed to find the minima, particularly in the G/R and B/R patterns, which we partially attribute to the stimulus protocol. We initially used a range of I_c values covering both positive and negative contrast (‘high’ sequences, electronic supplementary material, table S1). Because this protocol repeatedly failed to detect the minima (see neurons XI, XIV and XV), we shifted the range of I_c values in the negative direction (‘low’ sequences). Nevertheless, a clear minimum could not always be established, presumably owing to the neurons' variability (e.g. figure 2b).

Strictly speaking, figure 2a,b do not demonstrate anything about motion: the neurons might have reacted to any contrasts in a motion-independent manner. We thus plotted the response differences between the static and moving parts of the stimulus, as in figure 2c for the pilot experiments, to confirm that the neurons indeed responded to motion. Figure 2c shows the plots for neuron I shown in figure 2a; and the electronic supplementary material, figure S6 shows the results of same analysis for all 15 neurons. Electronic supplementary material, figure S7 shows the distribution of the difference between the zero contrast point in G/G and the minimum (when clearly identifiable) for G/R, G/B and B/R. The differences were significantly larger in G/R and G/B. Unfortunately, we could not perform any statistical analysis for B/R because the minimum was clearly identifiable in only four cases (neurons I, VIII, X and XII). Note that the value was negative in the zero contrast G/G pattern in most neurons. This is because these neurons exhibited a strong phasic response at the stimulus onset, and the response gradually decayed over time because of adaptation to the ‘static’ green screen.

(c) . Morphology of the motion-sensitive neurons

Neurobiotin injection revealed similar morphology in all neurons (figure 3; electronic supplementary material, figure S8). We could complete only the pilot experiments for the neuron shown in figure 1a, which therefore has no ID.

Figure 3.

Morphology of neurons. Neurons (a) (no ID) and (c) (neuron II) are downward-selective, while (b) (neuron XV), (d) (neuron X) and (e) (neuron XIV) are upward-selective. The abbreviations follow those described in Ito et al. [10]; CA, mushroom body calyx; LA, lamina; ME, medulla, Oe, oesophagal foramen. The varicosities (white arrowheads) are possible presynaptic terminals in the medulla, while the smooth branches are possible post-synaptic dendrites in the ventral lateral protocerebrum (black arrowheads). White arrows indicate cell bodies. The scale bars are 200 µm except in (a′), (a″), (b′) and (b″) where they represent 20 µm. (a‴) and (b‴) are schematic drawings of the neurons in (a) and (b) respectively.

Figure 3a is a downward-selective neuron (no ID) and figure 3b is an upward-selective neuron (neuron XV); the micrographs are montage reconstructions from 13 and 17 sections, respectively. Both neurons have a cell body in the midbrain region around the oesophageal foramen through which the oesophagus connects the mouth and the gut. Fine processes spread in the medulla on the side of electrode penetration. The medulla processes are varicosity-rich (figure 3a′,b′), suggesting synaptic output sites. The neurons also have smooth processes in the ventrolateral protocerebrum (VLP) that resemble post-synaptic dendrites (figure 3a″,b″). This morphology suggests that they are feedback neurons, sending information from the midbrain back to the optic lobe.

(i) . Downward-selective neurons

We found that the morphology of downward-selective neurons could be classified into two types. The first type (figure 3a) has its cell body on the posterior surface of the midbrain, slightly to the contralateral side. A thick axon connects the medulla and the contralateral VLP. The medulla processes occupy the dorsal third of the medulla, extending frontally in this neuron. Generally, the medulla processes are restricted to about a half or less of the medulla (electronic supplementary material, figure S8). The second type (figure 3c, neuron II) shares its basic morphology with the first type, but the presumptive output processes cover almost the entire medulla extending to the front as well (electronic supplementary material, figure S8).

(ii) . Upward-selective neurons

The upward-selective neurons could be divided into three morphological types (figure 3b,d,e). The first type (figure 3b, neuron XV) has its cell body on the posterior surface of the midbrain directly above the oesophageal foramen. The smooth dendritic processes spread in the ipsilateral VLP. Fine processes cover the ventral third of the medulla, extending frontally. Figure 3d shows the second type, with its cell body and the VLP dendrites on the contralateral side (neuron X). The medulla processes cover the frontal part of the medulla, extending towards the dorsal and the ventral side. The third type (figure 3e, neuron XIV) has its cell body and the VLP processes situated as in the first type. However, the medulla processes cover only a small region in the centre.

4. Discussion

Ibbotson et al. [11] identified two types of neurons sensitive to vertical motion in the medulla of Papilio aegeus, MV1 and MV2 [11]. The MV1 neurons have their cell body in the midbrain, while the MV2 neurons' cell bodies are in the optic lobe. The neurons we identified, which respond to vertical motion in a direction-selective manner, all have their cell bodies in the midbrain, so we identified them as the P. xuthus version of MV1 neurons.

We used moving stripe patterns to stimulate the MV1 neurons. In the control G/G pattern, we altered the intensity of the green light from one channel, while keeping the other channel constant. When both green channels are set at the same intensity, the G/G pattern is isoluminant as well as isochromatic and therefore no motion can be detected. Thus, the increased responses elicited by the G/G patterns when the intensity difference is non-zero can be unambiguously attributed to motion detection based on intensity contrast.

Two-colour patterns generally contain both chromatic and luminance contrast. Let us consider the neuronal responses to G/R patterns shown in figure 2a as an example. We systematically varied the intensity of the green light (c_var) while keeping the red light constant (c_cst). As equation (2.1) indicates, the I_c values become negative when the green light is less intense than the red light (electronic supplementary material, table S1). The neuron produced its minimal response at I_c = −1.24. We have interpreted from this response pattern that the green and red lights appeared (almost) equally ‘bright’ for the neuron at I_c = −1.24. In other words, the G/R pattern at this point is isoluminant for the neuron, despite the colours having different physical intensities. The subjectively isoluminant pattern still elicited spikes at 40 Hz, compared to the 20 Hz minimum for the G/G control. The difference is attributable to the chromatic contrast between the green and red stripes (and possibly some residual intensity contrast).

Because bright red light generates minimal contrast with dim green light, we can conclude that this neuron is more sensitive to green than to red. A similar pattern is observed for the B/R pattern: sensitivity is higher to blue than to red. The G/B pattern elicits a minimal response when I_c is positive, thus blue sensitivity is higher than green sensitivity. The relative positions of these isoluminant points are rather constant among neurons, irrespective of their preferred directions (electronic supplementary material, figure S5).

Electronic supplementary material, figure S3 shows the response characteristics of nine downward-selective (electronic supplementary material, figure S3I–IX) and six upward-selective neurons (electronic supplementary material, figure S3X–XV). After analysing their responses offline, we noticed some anomalies. For example, neurons II, III, IX, XIII and XIV exhibited slight directional selectivity to horizontal motion. The neurons X and XV, respectively, showed preferences for rightward and leftward motion. These effects may reflect the neurons' true directional selectivity, but could also be owing to experimental artefacts, such as misalignment of the head or accidental recording from multiple neurons. The neuron XIII strongly responded to upward motion, but downward motion failed to inhibit its activity. This might be owing to this neuron's unusually high level of spontaneous activity, 58 Hz.

Electronic supplementary material, figure S1c shows the previous behavioural data replotted against the intensity contrast (I_c). We used a plasma screen for the behavioural experiments and a DLP projector for physiology to present stimuli. Because the emission spectra of the red, green and blue channels of these devices are not identical, one would not expect the results to match up perfectly. The slight differences, such as the relative positions of the minima, could be attributed to the differences in the devices. Still, the results look quite similar, indicating that the MV1 neurons serve as an element in the motion processing pathway presumably close to the final output.

One point to note is about the effect of blue light. In the behavioural experiments, we assumed that B receptors' contribution was negligible. However, the model simulation deviates considerably from the behavioural data for the G/B pattern: the response to the G/B pattern never reaches as low as that to the G/G control, whereas the model predicts the G/B and G/G patterns elicit similar responses [7]. We attribute this contradiction to possible inputs from the B-sensitive LVFs to the motion detection system. Unlike in flies, P. xuthus's LVFs make numerous synaptic contacts with SVFs in the lamina, which could modulate the responses of SVFs [12,13]. The MV1 neurons generally exhibit high sensitivity to blue when viewing the G/B pattern, suggesting such a role for the B receptors. On the other hand, the B receptors are not the only channel sensitive to blue light. The blue light (electronic supplementary material, figure S1b) would also to some extent stimulate the V, G, R and BB receptors (electronic supplementary material, figure S1a), whose contributions have to be carefully addressed.

Morphologically, the identified MV1 neurons all have similar branching patterns. They appear to receive signals in the VLP and feed information to large parts of the medulla. The eye of P. xuthus is a random mesh of three spectrally distinct types of ommatidia, each housing nine photoreceptors (electronic supplementary material, figure S1a,b). The photoreceptors construct a cartridge module in the lamina, typically with four second-order lamina monopolar cells (LMCs). In a cartridge, all the photoreceptors feed LMCs, yielding their broad spectral sensitivity. The spectral sensitivity of LMCs from different ommatidia types differ only in the UV and R parts of the spectrum [12]. The spectral difference is probably not sufficient to explain the sensitivity of motion vision to the G/B and other chromatic contrasts, and suggests that the motion signal is sensed via pathways parallel to LMCs. The motion vision's colour sensitivity might stem even at the earliest photoreceptor level. For example, the B receptors receiving negative G inputs (B + G−) could strongly respond to the B/G contrast, largely independent of the light intensity.

We assume that the MV1 feedback neurons' basic response pattern is generated somewhere in the visual pathway from the optic lobe to the VLP. The MV1 neurons would then relay the response pattern back to the medulla. There, the neurons might contribute in subtracting wide-field motion (i.e. optic flow) to facilitate detection of small moving targets, maybe together with other unidentified neurons including the motion-sensitive MV2 neurons [11]. The resulting motion information most likely eventually feeds the presumptive direction-selective motion-sensitive system in the lobula plate [14], the centre for the visual course control [15].

Are the chromatic properties of P. xuthus's motion vision adaptive, or rather a side effect of the evolutionary elaboration of the SVFs' spectral sensitivities that serves to enhance their colour vision [9]? One could argue that motion detection benefits from being achromatic, as having a colour-blind detector allows for a more straightforward comparison of input channels [16]. However, we assume that the use of chromatic contrast enhances the possibility of detecting motion in isoluminant conditions. The same probably holds for using polarization cues in an object detection context, where polarization creates brightness contrast [17,18]. Further comparative studies on whether and how the chromatic inputs improve motion detection may offer clues for the understanding of how evolution shapes the segregation of parallel sensory pathways in animals.

Ethics

We followed the SOKENDAI regulation for conducting research using animals.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material [19].

Authors' contributions

C.C.: conceptualization, data curation, formal analysis, methodology, software, writing—original draft, writing—review and editing; K.A.: conceptualization, funding acquisition, project administration, writing—review and editing; M.K.: conceptualization, funding acquisition, methodology, project administration, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the JSPS KAKENHI (grant nos. 18H05273 to K.A. and 17K07485 to M.K.) and the SOKENDAI Advanced Sciences Synergy Program (SASSP) to M.K.