Canopy parkour: movement ecology of post-hatch dispersal in a gliding nymphal stick insect, Extatosoma tiaratum

Flightless nymphal and adult arthropods compose the majority of arboreal invertebrate fauna (Basset, 2001; Basset et al., 2012); these taxa locomote on vegetational structures (e.g. tree trunks, branches, lianas and leaves) for various behaviors including locating resources, dispersal and homing after dislocation (Cunha and Vieira, 2002; Basset et al., 2003; Yoshida and Hijii, 2005). Nevertheless, legged movements by small invertebrates can be challenged by the three-dimensional landscape and dynamic environment of the canopy space. Irregular habitat formed by different vegetational structures (e.g. foliage, branches and stems) precludes straight displacement, and may also obstruct directed aerial movement within canopy gaps. Complexity and dynamics of the physical environment can also impede navigational processes and locomotor performance. Diverse vegetational structures form heterogeneous visual environments under sunlight, with sunflecks and shadows projected dynamically onto various surfaces (Evans and Coombe, 1959). The visual environment thus changes with time and with the observer's perspective, potentially compromising the use of long-range visual cues. In addition, vertical movement between the forest floor and canopies can be energetically costly given the relatively long distances (e.g. forest canopy height ranges globally from 8 to 40 m; Simard et al., 2011). For ascent in particular, the efficiency of transport largely depends on the animal's endurance capacity (e.g. the duration of movement at a given speed; Full, 1987; Garland and Else, 1987). Endurance can also be influenced by environmental temperature (e.g. Full and Tullis, 1990), and the daily cycle of temperature in forests (e.g. ∼10°C day–night difference in Amazonian rainforest; Shuttleworth, 1985) may constrain the timing and extent of activity. Exposure to predators and parasitoids, many of which are specialized for either searching for or ambushing targets on vegetation (e.g. insects, spiders and birds; Southwood et al., 1982; Russell-Smith and Stork, 1994; Ford et al., 1986; Basset, 2001) may also threaten invertebrates' movement through the canopy space. In addition to biological threats, environmental perturbations (e.g. rain and wind gusts; see McCay, 2003) can interfere with directed movements. Overall, arboreal transport by small invertebrates can be sensorily and energetically challenging, and such demand may be particularly strong in vertical displacement given the concomitant variation of microhabitats and ecological communities (see Erwin, 1988; Shuttleworth, 1985).

Despite such spatiotemporal variation in the forest environment, both gravity and sunlight can serve as useful long-range cues in canopy space. Terrestrial arthropods generally perceive gravity with gravireceptors and integrate it with other signals (e.g. vision) for navigation (Büschges et al., 2001; Brockmann and Robinson, 2007; Robie et al., 2010). Also, spatial light gradients may be used to infer sun direction in the presence of foliar cover (Bjorkman and Ludlow, 1972; Endler, 1993; Wolken, 1995), with the strongest signal and lowest temporal variation occurring at midday (i.e. a peak plateau of solar altitude with <5 deg variation between 11:00 h and 13:00 h). In addition, vegetational structures under sunlight likely provide short-range visual cues for spatial discrimination. For example, insects identify contrast edges in landing response (Kern et al., 1997; Zeng et al., 2015). The use of gravitational and visual cues is potentially more general than chemical cues, which generally depend on a strong source (e.g. odor from flowers) or permanence (e.g. pheromone trails used by ants; Jackson et al., 2007).

Arboreal invertebrates exhibit various locomotor behaviors within canopy space. Jumping is a common and energetically efficient strategy for crossing air gaps (designated ‘gaps’ hereafter; see Graham and Socha, 2019), as shown in many wingless arthropods (e.g. spiders, nymphal mantids and ants; Parry and Brown, 1959; Burrows et al., 2015; Ye et al., 2020). Accidental loss of foothold (e.g. via perturbation by wind or predators) and defensive self-dropping (e.g. startle response: Haemig, 1997; Humphreys and Ruxton, 2019; Dudley and Yanoviak, 2011), along with failed landings after jumping, can transition to directed aerial descent (hereafter termed ‘gliding’; see Dudley et al., 2007; Socha et al., 2015), which reduces height loss (see Dudley and Yanoviak, 2011; Yanoviak et al., 2011). An arboreal spatial niche may not entirely underpin the evolution of gliding. For tree-dwelling invertebrate taxa, the diurnal niche permits visually mediated aerial maneuvers and landing, which may be essential for the evolution of gliding (Yanoviak et al., 2011; Zeng et al., 2015). Studying the ecology of movement in canopy space, and also initiation mechanisms for aerial descent, can thus help to address the utility of gliding within specific ecological contexts.

The newly hatched nymphs of the Australian stick insect Extatosoma tiaratum (Macleay 1826) disperse by ascending from the forest floor, where eggs are deposited, by climbing vegetation into the canopy (Fig. 1A,B). In contrast to the typically nocturnal lifestyle of phasmids, E. tiaratum nymphs hatch during the middle of the day (i.e. 11:00 h to 15:00 h) in the rainy season (November to January; Carlberg, 1981, 1983, 1984a,b; Brock and Hasenpusch, 2009; Brock, 2001), and immediately start to rapidly move. The nymphs also possess ant-mimicking coloration and exhibit comparable behaviors (e.g. body shaking during crawling; Fig. 1; Movie 1) while climbing and exploring their surroundings (Carlberg, 1981, 1983; Rentz, 1996; Brock, 2001). Ant mimicry (i.e. myrmecomorphy) presumably facilitates dispersal by deceiving visual predators (e.g. birds and small reptiles). These traits collectively form a disperser's syndrome (Ronce and Clobert, 2012) that potentially facilitates ascent and the search for suitable post-hatch microhabitats. The hatchlings can also glide to reduce height loss if falling (see Zeng et al., 2015). Such post-hatch dispersal behavior is exhibited only during the first 3–5 days after hatching, following which the nymphs become nocturnally active and lose their ant-mimicking morphology (Fig. 1B) (Brock, 2001, 1999).

Here, we evaluated movement ecology (Nathan et al., 2008) and canopy dispersal in newly hatched E. tiaratum (subspecies E. tiaratum tiaratum from north Queensland; see Brock, 2001) under controlled laboratory conditions, focusing on their navigational and locomotor mechanisms for traveling through the canopy. We hypothesized that: (1) dispersing nymphs use gravity, light gradients and visual contrast to navigate toward vegetational structures, which are then used for ascent; (2) newly hatched nymphs are capable of considerable vertical ascent; (3) when dispersing, newly hatched nymphs can also initiate aerial descent both volitionally and in response to tactile perturbation, leading to aerial righting and gliding (Zeng et al., 2015, 2017); and (4) jumping is adopted by nymphal stick insects for negotiating gaps, a strategy found in both wingless and winged phasmid species (Burrows and Morris, 2002; Burrows, 2008). Displacement by individual dispersing insects within vegetational structures across gaps derives from both locomotor capacity and sensory responses (Fig. 1C), and can explain both movement ecology of small wingless invertebrates in canopy space and the ecological utility of gliding within the canopy.

Eggs of E. tiaratum were incubated on vermiculite substratum at ∼70% humidity and 25–27°C air temperature. Newly hatched nymphs were collected every 12 h and were maintained in clear plastic cups (354 ml) with caps. Nymphs were provided with fresh leaves of Himalayan blackberry (Rubus armeniacus) from the day of hatching, and were lightly sprayed with water every 2–3 days. Cups were kept in an environmentally controlled room with a 12 h:12 h light:dark cycle. All experiments were conducted at air temperatures of 25–27°C. The experimenter first collected the insect from the housing cup by letting it climb onto a small wooden rod (diameter ∼0.5 cm, length ∼20 cm), and then oriented the rod toward the entrance of the experimental setup to allow the insect to enter.

A T-maze within a uniformly lit chamber was used to evaluate negative gravitaxis. A T-shaped system of wooden rods (diameter, 9 mm), consisting of a horizontal entrance beam and a vertical beam to evaluate binary choice, was placed in the center of the experimental chamber (40×40×40 cm), which was covered with white felt and surrounded by two natural-spectrum bulbs (R30FF100VLX, Verilux Inc., Waitsfield, VT, USA) with voltage control (Fig. 2A; Fig. S2). Luminance at the center of the chamber was measured with a light meter oriented horizontally (Model #401025, Extech, Waltham, MA, USA; spectral range: 400–740 nm). In each trial, the experimenter first released the insect onto the horizontal beam through the entrance window (2×2 cm), and then observed through an observation window (1.5×1.5 cm) for the insect's movement toward the intersection of T-maze and its subsequent directional choice. The insect's directional choice was indicated by its final position when reaching either end of the vertical beam. A minimum of 3 min resting time was given between trials both within and among individuals. Experiments were conducted for three age groups (0–1, 6–7 and 12–14 days old) under four environmental luminances (0, 10, 600 and 3000 lx, adjusted using the voltage controller). Each combination of age and luminance was tested with 10–26 individuals, using 4–6 trials per individual. A repeated G-test (McDonald, 2015) was used to test frequencies of occurrence of negative gravitaxis (f_g) relative to a null frequency of 0.5. A Poisson GLMM was used to test whether counts of ascent responses were significantly correlated with environmental luminance and age, using individuals as random factors.

A T-maze was used to test phototactic response of insects. The T-maze consisted of two tunnel sections (section 1: diameter 2 cm, length 20 cm; section 2: diameter 4.5 cm, length 10 cm); the interior of both sections was covered with black fabric to reduce light reflection (Fig. 3A). A natural-spectrum bulb (R30FF100VLX, Verilux Inc.) with a voltage controller was placed ∼5 cm away from each exit of the T-maze, with a paper screen (5×5 cm) positioned vertically 3 cm distant from each exit. Luminance was measured with a light meter (401025, Extech; spectral range: 400–740 nm) at two exits. Three luminance contrasts (0 lx versus 3 lx, 100 lx versus 500 lx, and 1×10⁴ lx versus 2×10⁴ lx) were tested, a range which covers the luminance variation between forest understories and canopies (e.g. ∼600 lx to 2.4×10⁴ lx; Bjorkman and Ludlow, 1972; Pearcy, 1983; Lee, 1987). Each luminance contrast was tested for insects of three ages (0–1, 5 and 10 days old), each represented by 5–20 individuals (with 5–6 trials per individual). During each trial, the experimenter first released the insect into the entrance, immediately covered the entrance with a lid and then recorded the insect's final directional choice within the light gradient, as indicated by exit from either of the two tunnels. Control experiments were conducted with one 0–1 day old group in complete darkness, for which an infrared video camera (SONY HDR-HC7) was used to observe the insects' directional choice. A repeated G-test was used to analyze the significance of directional bias relative to the null frequencies (i.e. 1:1). Poisson generalized linear mixed models (GLMM) were used to test the correlation between counts of phototactic responses and environmental luminance, using individuals as random factors.

Directional choices were examined using a cylindrical arena (height 35 cm, diameter 26 cm) decorated internally with vertically oriented contrast patterns. The patterns were made with felt sheets using shades of black, gray and white (average reflectance over 300–550 nm: black, 3%; gray, 38%; white, 47%), as measured with a spectrophotometer (USB2000, Ocean Optics, Dunedin, FL, USA; see Zeng et al., 2015). A natural spectrum light bulb (R30FF75VLX, Verilux Inc.) was placed on top of the arena and provided ∼600 lx luminance at the arena floor (as measured with a light meter, Extech 401025) (Fig. 4A; Fig. S3).

Five contrast patterns were used to test directional preferences to contrast strength and stripe size. The patterns formed by black and white surfaces were coded as BW11, BW91 and BW19, where the embedded numbers represent the relative proportions of black (B) and white (W) surfaces. In BW91 and BW19, the narrow stripes feature a spatial frequency of ∼3.18 cycles rad⁻¹ (as viewed from the arena's center). Similarly, two patterns formed by gray surfaces paired with black or white surfaces in equal proportions were coded as BG11 and GW11. The effect of contrast strength on the insect's tactic response was tested using BW11, BG11 and GW11, and the effect of stripe size was tested using BW11, BW19 and BW91. Each contrast pattern was tested using 19–38 individuals, with 5 trials per individual.

In each experimental trial, the insect was released through an entrance (diameter 2 cm) at the center of the arena floor, and was observed from above (Movie 1). Three temporal landmarks were recorded: (1) entrance into the arena; (2) moment of arrival at the arena's wall; and (3) initiation of climbing. The insect's directional preference was represented by the angular distance between the point where the insect first reached the arena wall and the nearest contrast edge. Circular directionality of the insects' directional choice was analyzed using ‘CircStats’ package in R software (https://CRAN.R-project.org/package=CircStats; http://www.R-project.org/), which calculates means and confidence intervals of direction θ and concentration κ based on von Mises maximum likelihood estimates. Repeated G-tests were used to test directional preference for the darker and the lighter surfaces in each configuration against null proportions, which were derived based on random association with corresponding areal proportions of different surface patterns. In BW19, for example, the null frequency for an insect walking to the black surface is 0.1. Durations between temporal landmarks, and also the average speed of movement on the arena floor, were compared using repeated-measures ANOVA. When examining the insect's preference for narrow surfaces in BW19 and BW91, we tested whether the frequency of moving toward the narrow surface was associated with contrast edges but not surface size. Thus, the frequency of moving to the narrow surface (either black or white) equals the frequency associated with the corresponding shade in BW11. Based on results with pattern BW11 (Table S1), the null frequency is thus 0.57 for BW19 and 0.43 for BW91.

where U_i is the insect's vertical speed as derived from the camera's view. Given varying and age-specific climbing speeds and intermittent pauses, we used percent ascent time and total ascent height to characterize the insects' endurance capacity.

Simulated predatory attacks and induction of accidental loss of foothold (as a proxy for environmental perturbations) were applied to experimental insects ascending on a vertically oriented cardboard sheet (width 15 cm, height 35 cm) placed beneath a light source. Experimental insects were first released near the bottom of the sheet, and were allowed to ascend under volitional phototaxis and negative gravitaxis. To simulate predatory attacks, a compressed cylinder of paper towel (diameter ∼5 mm) was held against the insect's tibial and tarsal segments for any given leg, pinning it against the substrate for ∼20 ms without causing injury (Fig. 6A; Movie 2). The experiment was conducted on nymphs of three age groups (0, 6 and 12 days old), using 10−15 individuals per group. Each individual was tested in five trials; in each trial, each leg pair received two simulated attacks, with a waiting period of 10 s between consecutive attacks. To induce accidental loss of foothold, a Teflon-coated surface (∼2 cm wide) was introduced onto the cardboard sheet (Fig. 6C). The insects' response to contact with this slippery surface was recorded for nymphs of three age groups (0, 5 and 10 days old), using 7 individuals per group.

We evaluated jumping behavior in nymphs by allowing them to ascend to an elevated platform surrounded by distant visual targets such that jumping would be required to cross the gap to any given target. The platform (2 cm×5 cm) was mounted atop a vertically oriented rod and ∼50 cm above the experimental chamber's floor and >40 cm away from its walls. The floor and walls were covered with white felt, and environmental luminance was ∼600 lx. Two landing targets were used: (1) a horizontally oriented rod (diameter of 5 mm) wrapped with black felt and positioned 20 cm below the platform, with a white background ∼30 cm below the platform (Fig. 7A); and (2) a vertically oriented rod, also wrapped with black felt, positioned ∼8 cm away from the platform (Fig. 7A). For each of these two configurations, jumping behavior was tested for nymphs of three age groups (0, 6 and 12 days old; horizontal target, 6–9 individuals per age group and 5 trials per individual; vertical target, 5–9 individuals per age group and 2–4 trials per individual). In each trial, the experimenter released the insect onto the middle of the vertical rod and let it ascend to the platform. If the insect stopped moving, the experimenter gently tapped the vertical rod to stimulate the insect upward. The experimenter then observed the insect's response for 4 min after it ascended to the platform. If the insect climbed down the vertical rod, the experimenter would restart the trial by letting the insect walk onto a small rod, and then reversed the rod to allow the insect to walk back to the vertical rod in a head-up orientation. Once on the platform, the insects could descend, rest or jump, with transitions between cursorial and exploratory movements (Fig. 7B; see Results). The ensuing behavior sequences were coded accordingly, whereby repeated descent was treated as a single action). A minimum of 2 min resting period was provided between trials. Control experiments were conducted in five 0 day old individuals following the same protocols, but without any visual landing targets.

After being released into a vertically oriented T-maze within a visually homogeneous environment (Fig. 2A), nymphs moved to the intersection and ascended along the vertical rod under various environmental luminances (0–3000 lx), exhibiting frequencies of gravitactic response (f_g ranges from 0.81 to 0.97) significantly different from the null (f_g=0.5) (Fig. 2B). When there is no light gradient in the vertical direction, gravity alone can thus be used by nymphal E. tiaratum as a directional cue. Furthermore, ascent frequency was inversely correlated with environmental luminance in newly hatched (0 day old) nymphs, with the greatest frequency (f_g=0.97±0.02; mean±s.e.m.) occurring in total darkness and the lowest (f_g=0.79±0.08) occurring in 600 lx (Fig. 2C).

After being released into a T-maze of horizontal dark tunnels, nymphs moved into a controlled luminance gradient. Three luminance contrasts were used, with the mean value ranging from ∼2 to 1.5×10⁴ lx (Fig. 3A). Newly hatched nymphs (0 days old) exhibited strong phototactic displacement under all luminance conditions, with the frequency of phototactic response (f_p)>0.8 under all light conditions. In older nymphs (i.e. 5 and 10 days old), f_p declined with both age and increasing mean luminance (Fig. 3B). For example, 10 day old nymphs preferred the brighter direction under a mean luminance of 2 lx (f_p=0.90±0.09; mean±s.e.m.) and tended to move toward the darker direction when mean luminance exceeded 1.5×10⁴ lx (f_p=0.27±0.07; mean±s.e.m.). In control experiments with no luminance gradient, insects showed no directional bias (repeated G-test: G=0.722, d.f.=1, P=0.396), supporting the hypothesis that nymphal E. tiaratum can use luminance gradients for directional reference.

Experimental insects were released into a visual arena surrounded by vertically oriented contrast lines (Fig. 4A). When contrast patterns were formed by equally sized surfaces (black–white, black–gray and gray–white), insects rapidly moved toward contrast lines (Movie 1), showing the strongest directional preference for a black–gray contrast. Specifically, movements toward a black–gray contrast exhibited the smallest angular deviation from the contrast line (<1 deg), and accordingly the highest concentration (κ>700) (Fig. 4B,C; Tables S2 and S3). When exposed to black–gray contrasts, they also exhibited the fastest movement speeds (3.51±0.25 cm s⁻¹; mean±s.e.m.) on the arena floor, and initiated climbing immediately after reaching the wall (<2 s) (Fig. 4D,E). Black–white and gray–white contrasts were less attractive; the insects showed the slowest movements when exposed to the gray–white contrast (2.18±0.3 cm s⁻¹; mean±s.e.m.). The insects showed a preference for the darker surface paired with a white surface (frequency of moving to the darker surface was 0.57 in BW11 and 0.76 in GW11), but preferred the gray surface when paired with a black surface (frequency of moving to gray surface of 0.62) (Table S1).

Furthermore, when exposed to contrast patterns formed between surfaces of different sizes, insects showed a general preference for the narrower surface independent of its brightness. For such contrast patterns formed between black and white surfaces, the frequency of moving toward the narrower surface (0.76 for BW19 and 0.82 for BW91) was significantly greater than the null expectation (Fig. 4F,G). The tactic response to vertically oriented contrast lines and preference for narrower surfaces may help in localizing vegetational structures with which to initiate ascent (see Discussion).

We evaluated ascent behavior of nymphal E. tiaratum for six age groups ranging from 2 to 240 h old. Experimental insects were placed on a vertically oriented and speed-controlled treadmill, and were filmed for 60 min (Fig. 5A). Ascents on the treadmill were then digitized to generate speed profiles, consisting of intermittently ascending and pausing intervals (Fig. 5B). Ascent speed (range 2.3–5.3 cm s⁻¹) generally tended to decline with increasing age. The 4 and 48 h old nymphs exhibited the highest speed (5.1–5.4 cm s⁻¹), peaking within the first 3 days of hatching (Fig. 5C). Among insects younger than 3 days, a low ascent speed for 2 h old nymphs derived from frequent pauses (0.42±0.07 times min⁻¹), which is ∼2 times greater than that of older ones (Fig. 5D). The 2 h old nymphs exhibited the greatest ascent capacity, including the greatest ascent time (95±5% active time; mean±s.d.) and the greatest total ascent height (125 m, or ∼7350 body lengths); older (>5 days old) nymphs were active for <1 min (<5% active time) and otherwise remained immobile throughout the trial (Fig. 5D,E).

To explore mechanisms of the aerial phase initiation, experimental insects were tested with two tactile perturbations, namely simulated predatory attacks and accidental loss of foothold (see Materials and Methods), of which the latter was a proxy for environmental perturbation and initiated a perturbed descent as in previous gliding experiments (see Zeng et al., 2015). For newly hatched nymphs, simulated attacks led to self-dropping (frequency 0.15–0.19; Fig. 6A,B). By contrast, older nymphs generally oriented away from the direction of attack (frequency of dropping 0–0.07). High-speed filming revealed that self-dropping was initiated by voluntary withdrawal of the tarsus from the substrate within ∼50 ms following termination of the simulated attack, followed by mid-air tucking (i.e. flexion of the tibia–femur joint and elevation of the femur) of all leg pairs (N=30 trials; Fig. 6A; Movie 2). Preliminary high-speed recording showed that such leg tucking allows the insect to reduce contact with surrounding structures during descent, and to roll on a 70 deg incline at an average speed of 40.5±2.0 cm s⁻¹ (mean±s.d.; N=2 trials; ∼24 body lengths per second; Fig. 6C; Movie 3). This strategy may allow rapid descent through the clustered space of foliage under natural conditions (see Discussion). Slippery surfaces were used to induce loss of foothold during ascent, such that experimental insects were subjected to an unexpected sudden imbalance and loss of foothold (Fig. 6D). Newly hatched nymphs (0 days old) then exhibited self-dropping behavior at a frequency of 0.56±0.07 (mean±s.e.m.; Fig. 6E). High-speed videos revealed that, following loss of foothold by one leg, other leg pairs showed similar flexion and tucking to that described above, leading to complete removal of tarsi from the substrate and self-dropping (Movie 2). Self-dropping after touching the slippery surface was not observed in older nymphs.

To test whether nymphal E. tiaratum jump to cross gaps, we placed individual nymphs at the base of an elevated platform with a distant visual target (either a horizontal rod or a vertical rod) which contrasted with a white background (Fig. 7A). After ascending to the platform, the newly hatched insects (0 days old) typically explored the platform, ultimately oriented toward the visual target, and tried to reach out using their forelegs; having failed to cross the gap using the forelegs, these insects would then jump toward the visual target. Sometimes, insects also descended via the vertical rod, in which case the experimenter would re-introduce them to the vertical rod in an upward orientation. In other cases, the insects rested by the edge of or beneath the platform before either jumping or resting until the end of trial (Fig. 7B). Whereas newly hatched insects exhibited active exploration and jumping on the platform, older insects (i.e. 6 and 12 days old) showed no jumping, and either climbed down the vertical rod of the platform or rested in a still posture after trying to reach out with their forelegs (Fig. 7C).

In newly hatched insects, jumping was observed for 19 out of 24 tested individuals (Movie 2). When exposed to the horizontal target, insects spent 1.8±0.2 min (mean±s.e.m., N=28 trials from N=9 individuals) before jumping and successfully landing on the target in 17 out of 28 trials (success rate ∼60%); for trials with successful landing, they traveled a mean horizontal distance of 8.2±1.6 cm with a 20 cm descent (and thus with a ratio of horizontal to vertical distance of ∼0.41) (Fig. 7D). When exposed to the vertical target, the insects spent 2.8±0.3 min (N=21 trials from 9 individuals) before jumping, and successfully landed in 13 out of 23 trials (success rate ∼56%); for trials with successful landings, they descended 14.4±6.1 cm with an 8 cm horizontal translation and a horizontal to vertical distance ratio of ∼0.55. No jumping behavior was elicited in control experiments with no visual target (N=5 individuals).

The tactic uses of gravity and environmental light gradients by newly hatched E. tiaratum support our initial hypotheses on the roles of these two cues for legged movement in canopy space, as found in other ecosystems (e.g. other larval insects, and mites; Perkins et al., 2008; Zhang, 1992). Nymphal E. tiaratum lack ocelli and thus depend on compound eyes for visual perception. Previous work on a different phasmid species suggests that eyes are less nocturnally adapted in nymphs than in adults (Meyer-Rochow and Keskinen, 2003), which probably derives from dispersal behavior and a broader temporal niche. In newly hatched E. tiaratum, the positive phototaxis persists across a broad luminance range (∼1 to ∼2×10⁴ lx; Fig. 3B), and thus facilitates localization of open canopy space under a variety of light conditions. This behavior contrasts with negative phototaxis in most phasmid species, except for some winged species attracted by light at night (see Brock and Hasenpusch, 2009). Natural spectrum light was used in our experiments, and E. tiaratum may possess wavelength-dependent sensitivity typical of other phototactic insects (Chen et al., 2012, 2013), and probably with the highest sensitivity to blue and green light as in orthopteran insects more generally (Bailey and Harris, 1991). The newly hatched nymphs showed an increased tendency to descend under low light (Fig. 2C), which may represent a detouring strategy for alternative paths or for seeking shelters.

Tactic movement toward vertically oriented contrast lines suggests searching behavior for plant structures such as stems and tree trunks that can be used in ascent; similar behavior has been reported for nymphal crickets (Meille et al., 1994). Black–gray contrasts were more attractive than black–white and gray–white contrasts (Fig. 4A–E), and this preference for dark or shaded surfaces may reduce exposure to visual predators during dispersal. A similar preference was shown in gliding nymphal stick insects (Zeng et al., 2015). Nymphal E. tiaratum also showed a preference for narrow strips regardless of the contrast mode (Fig. 4F–H), suggesting a certain level of pattern recognition using spatial frequency characteristics of potential targets (e.g. a stem in the foreground contrasting with a differently shaded background). Overall, these visually mediated behaviors facilitate diurnal dispersal and may serve as a precursor to the evolution of gliding from plants (see below). More systematic investigation is needed to understand the influence of environmental luminance on the insect's visual acuity and how different visual signals are used to enable navigation in a complex canopy space.

Newly hatched nymphs (e.g. 2 h old) showed the longest ascent time, suggesting that they have the lowest fatigue rate; they also exhibited the most frequent pauses, which may be associated with an increased tendency for spatial exploration. A high climbing speed was shown by 4–48 h old nymphs (5.1–5.4 cm s⁻¹, or about 3–3.2 body lengths per second), which may be the highest fastest known relative speed for walking phasmids (e.g. mean walking speed in Carausius morosus is ∼2 cm s⁻¹ or ∼1.5 body lengths per second in the first instar, and ∼5 cm s⁻¹ or ∼0.7 body lengths per second in adults; Graham, 1972). This rapid walking, as fueled by embryonic reserves, likely reflects a strong selective pressure for rapid dispersal in a rainforest ecosystem (see below). Future work may examine the ontogenetic variation of climbing energetics (e.g. aerobic metabolism) to further understand the physiological specializations underlying such rapid movements.

Our experimental setups were repeatedly used with different subjects, and any potential influence of pheromones was assumed to be negligible. Unlike social insects (e.g. ants), phasmids are not known to deposit pheromone trails, which nevertheless does not exclude the possibility that E. tiaratum nymphs may use chemical cues for host search during dispersal. The integrative use of various environmental cues for spatial navigation remains to be studied in arboreal arthropods. To further understand cognitive abilities underlying dispersal in nymphal E. tiaratum, studies of their movements within structurally and visually more complex environments would be informative. The potential use of other directional cues (e.g. polarized light, chromatic signals, chemical cues, etc.) in the canopy space can be tested using binary choice experiments. Recording movement on natural vegetation (e.g. via a Lagrangian approach; Baguette et al., 2014) can provide direct evidence for integrative use of various short-range and long-range cues.

A multimodal locomotor strategy, consisting of both aerial and legged phases, can assist small insects moving in the canopy space. In E. tiaratum, mid-air tucking of legs after defensive self-dropping can reduce the body moment of inertia and overall projected size, which also permits rolling along inclines (Fig. 6C). In aggregate, these behaviors may allow for rapid descent through clusters of leaves and branches. Their widespread occurrence in phasmids (e.g. Phaenopharus khaoyaiensis; Y.Z., unpublished observation) suggests a general adaptation to arboreal life. Jumping is effective for gap crossing but can be risky given the high chances of a missed landing (Fig. 7D). Our observations suggest that nymphal E. tiaratum rely on a variety of mechanisms to assess opportunities for jumping, including leg-search and probing behavior used in gap crossing (Dürr, 2001; Blaesing and Cruse, 2004), and active visual scanning for depth perception (e.g. sideways movement of the head or with the legs; Kral, 2009). We predict that, under natural conditions, dispersing E. tiaratum may frequently alternate movements on vegetational structures and in air, thus forming a behavioral loop (Fig. 8A,B). An aerial phase may be initiated by missed landings, environmental perturbations and predatory attacks. After dropping, energy required for subsequent ascent can be indirectly reduced by aerial righting and gliding, rather than falling to the ground. The self-dropping reflex that triggers the withdrawal of all tarsi may present a reflex arc similar to avoidance reflexes at the single-leg level (Kittmann et al., 1996). Foothold withdrawal may underlie a variety of voluntary self-dropping mechanisms in phasmids (Bedford, 1978) and some other insects (e.g. ants and aphids; see Haemig, 1997; Humphreys and Ruxton, 2019).

The diurnal niche plays a key role in the gliding behavior of nymphal E. tiaratum. Control of mid-air maneuvering and landing depends on the strength and continuity of vertically oriented contrast edges (e.g. tree trunks; see Zeng et al., 2015), the qualities of which decrease with and would be less salient at lower light levels. Compound eyes of arthropods generally possess much lower acuity compared with vertebrate eyes (Land, 1997), and will be more dependent in flight on visual contrast associated with tree trunks. Therefore, a diurnal niche may be essential to the evolution of gliding in arthropods, and perhaps less so in vertebrates. This conclusion is indirectly supported by the observation of widespread diurnal gliding in various arthropods (see Dudley and Yanoviak, 2011; Yanoviak et al., 2015), in contrast to crepuscular and nocturnal gliding in various vertebrate gliders (e.g. colugos; Byrnes et al., 2011a,b). Spatiotemporal variation of visual quality of landing targets within vegetational canopies has not yet been examined relative to arthropod gliding, but clearly is relevant to the evolution of this behavior. Similarly, few data exist pertaining to the occurrence and behavioral contexts of aerial descent (e.g. as elicited as a defensive strategy or in response to environmental perturbation) in a range of arboreal invertebrates (see Yanoviak et al., 2011, 2015).

Phasmids are generally nocturnal (Bedford, 1978), whereas diurnal hatching in E. tiaratum likely evolved under selection for rapid dispersal within a rainforest ecosystem. The ascent of ground-hatched nymphs is a transient behavior that utilizes environmental cues for navigation and movement. As environmental brightness is strongest around midday (Shuttleworth, 1985), hatching around this time maximizes the use of this cue (see Fig. 9A). Even for cloudy days, the luminance-insensitive phototactic response can allow newly hatched E. tiaratum to disperse upward toward canopies. During the rainy season of north Queensland, for example, >50% of consecutive wet day intervals last less than 2 days (Cook and Heerdegen, 2001), and the sky is on average only partially cloudy (i.e. 3–5 oktas) with >8 h of daily sunshine (Australian Government Bureau of Meteorology Weather Data Services; www.bom.gov.au/catalogue/data-feeds.shtml; accessed 8 Jun 2020). Furthermore, ant mimicry by newly hatched E. tiaratum may help to deter visual predators during movement on vegetational structures. Similar ant-mimicking phenotypes have been described from newly hatched nymphs in several unrelated phasmid species (Fig. 9B; see also Hanibeltz et al., 1995). Nymphs of all such taxa hatch on the ground and are diurnally active. Convergent evolution of this strategy implies a common demand for dispersal efficiency, particularly given the high abundance of nocturnally active predators on vegetation (see Berger and Wirth, 2004). Hatching location, diurnality and vegetational features thus underlie the evolution of diurnal ascent in nymphal phasmids, although a broad phylogenetic survey is now warranted to evaluate possible correlates of other dispersal strategies.

We thank Douglas Ewing for inspiration, and David Wake and members of Berkeley Animal Flight Laboratory for comments. We also thank Andrea Lucky for communication about Leptomyrmex ants, and Paul Brock and Jack Hasenpusch for advice on E. tiaratum natural history. We further acknowledge research facilities provided by the Center for Interdisciplinary Biological Inspiration in Education and Research (CiBER), UC-Berkeley.