Escape performance decreases during ontogeny in wild crickets

Increasing the ecological relevance of studies on the performance of organisms has been a central issue for functional biologists over the last few years (Schmalhofer and Casey,1999; Irschick,2003; Vanhooydonck and Van Damme, 2003; Irschick et al., 2005). Factors related to the performance of organisms operate within an ecological context and are subject to complex interactions with other intrinsic (e.g. age, experience, learning) and extrinsic variables (e.g. predator type,habitat characteristics). Among them ontogeny is a major factor that can influence the level of ecological performance within natural populations as it affects a broad range of ecological, physiological and behavioural characteristics (see Roth and Johnson,2004). However, studies investigating changes in the performance of animal function with ontogeny are restricted to relatively few groups of organisms, mainly vertebrates, and few traits, such as locomotor and feeding systems. Little consideration is generally given to the ecological context in which these traits take place and are selected(Irschick, 2003; Herrel and Gibb, 2006). Of particular ecological and evolutionary relevance are traits related to escape performance, as they are directly related to the survival and, therefore, the fitness of organisms (O'Steen et al.,2002; Walker et al.,2005). Most studies have focused on the kinematics of various locomotor behaviours displayed during escape, for example sprinting and jumping (Thompson and Kier,2002). However, from an evolutionary perspective, organisms are likely to adapt to predation risks at an earlier detection stage of the predation sequence, as it confers substantial selective advantages for prey(Endler, 1991; Fox et al., 2001).

Arthropods (e.g. insects and spiders) are good models of evolved high performance detection systems; they have hundreds of mechanoreceptive cuticular hairs, which are sensitive to touch (filiform hairs) or wind currents (trichoid hairs) generated by approaching predators(Tautz and Markl, 1978; Gnatzy, 1996; Suter, 2003). The sensing performance of animals greatly influences their ability to detect and escape predators. Prey with high performance levels of detection may achieve greater fitness levels in nature, up to a certain point. Given that juveniles generally suffer higher rates of predation than adults, selection for mechanisms that improve juvenile escape performance is likely to be strong. This is observed in wood crickets (Nemobius sylvestris Bosc), a widespread detritivorous insect, foraging on leaf litter of deciduous European forests. Wood crickets suffer high rates of predation by wolf spiders(Pardosa sp.), especially during juvenile instars(Dangles et al., 2006a). Crickets have evolved one of the highest performance anti-predatory sensory systems (Shimozawa et al.,2003) and because the structure of their mechano-sensory system varies throughout development (Dangles et al., 2006c), we expect there to be strong selective forces acting on their escape performance during ontogeny.

The relationship between performance and body size of organisms remains controversial. Many studies have reported that as animals get larger performance levels increase or remain constant(Irschick, 2000; Trillmich et al., 2003). Recently, there have been reports describing a decrease in performance throughout ontogeny (Gibb et al.,2006). Previous research on performance remains dominated by studies relating to vertebrates and nothing is known about the ontogenic changes relating to invertebrate performance. We measured ontogenic changes and their effects on performance (measured as the escape distance) in wild wood crickets behaving freely under natural conditions. We devised a portable actuator that mimicked the attack dynamics of natural predators on crickets,for example wolf spiders. We quantified the cricket escape performance and behaviour by analysing numeric film sequences obtained with a high-speed video camera.

We examined the escape performance in 30 wood crickets Nemobius sylvestris Bosc in three age classes: two juveniles and one adult. Juveniles (instars II–III, named juveniles I) had tibia II lengths between 1.8 and 2.1 mm, and a mean body length of 0.57±0.02 cm. Intermediate crickets (instar IV–V, named juveniles II) had tibia II lengths between 2.6 and 2.9 mm, and a mean body length of 0.69±0.03 cm. Adults had tibia lengths between 5.4 and 5.6 mm and a mean body length of 0.89±0.03 cm. The morphological characteristics of wood cricket cerci throughout ontogeny have been described(Dangles et al., 2006c). Mean cercus lengths were varied: 1.1 mm in juveniles I, 1.9 mm in Juveniles II, and 4.5 mm in adults. Only young adults displaying a high level of foraging activity were selected to exclude senescent individuals. All crickets were collected from an oak forest in the vicinity of Tours, France(47°17′06″N, 00°47′13″E) the day before the experiment. The crickets were housed in plastic boxes filled with layer of leaves after capture, fed with dry cat food, and watered using wet cotton swabs.

Measuring the escape performance of freely behaving crickets in the field requires portable equipment, including an attack simulator and a video recording system (Fig. 1). To simulate a spider attack, a `spider operator' used an original device based on the controlled displacement of an actuator-triggered `piston'. The piston measured 0.8 cm in diameter (Fig. 1A; A) (LAL35 actuator/LAC1 controller, Cedrat Technologies, Meylan, France), and has been previously tested and used(Dangles et al., 2006b). The acceleration and velocity of the piston displacement was driven easily and with high precision (±4%) using specific macro commands viahyperterminal software and the rs232 port of a laptop(Fig. 1A; B). This allowed us to control and set it to mimic a spider attack. We had previously validated this device using particle image velocimetry (PIV) analyses of the air-flow in front of the moving piston (see Dangles et al., 2006b). The piston displacement produces a flow pattern very similar to that produced in front of a running spider. Initially, the piston was fixed to an actuator, which could not be easily moved under field conditions to follow wandering crickets. Therefore, we added a 1.5 m wire between the actuator and the tip of the piston to facilitate its use(Fig. 1A; A,C).

Cricket escape behaviour was recorded using a lightning RDT1 high-speed(HS) digital video camera (Fig. 1A; D) (DRS Data & Imaging Systems, Inc., Oakland, NJ, USA)with a frame rate of 1000 frames s^–1 and resolution of 1280×512 pixels. The escape response for freely behaving animals was recorded by following a cricket with the camera after its initial release until it stopped moving. The camera was mounted on a ramp(Fig. 1A; E) fixed on the rotating head of a stable tripod 0.5 m above the ground(Fig. 1A; F). This system allowed us to follow cricket displacements within a 2.5 m diameter. The size of the image on the camera was only about twice the size of the cricket,providing sufficient detail for measuring escape distances, but making it difficult to keep the moving cricket centred on the camera screen. We overcame this problem by fixing a red laser pointer(Fig. 1A; G) to the camera support in such a manner that pointing the laser on the cricket body ensured that the cricket was well-centred on the camera. The ground was illuminated by a halogen lamp (Fig. 1A; H)located 1 m above the ground surface. The video acquisition was controlled using a laptop (Fig. 1A; I)equipped with a DA170M data acquisition board (4 Gb of memory) and MIDAS 2.0 software (Xcitex Inc, Cambridge, USA).

The escape response of crickets to predatory signals produced by the piston was recorded on the ground litter of a beechwood tree at the University of Tours, where wood crickets and wolf spiders can be found. The experiment duration was a single summer day (ground temperature illuminated by the lamp:30°C). Crickets were individually placed on the ground and followed by our tripod-ramp-HS camera system as they started to walk freely. We waited for the crickets to stop moving before launching the piston because cricket behaviour(such as walking, running) at the time of simulation influences their behavioural latency [as reported for cockroaches(Camhi et al., 1978)]. Once the cricket was motionless, the piston was placed rear-on 4.5 cm from the cricket:this is the median distance at which spiders attack wood crickets(Dangles et al., 2006b). The piston was positioned at the height of a spider, 0.2 cm above ground. The piston was launched at an attack angle of ±20° from the rear of the cricket, to make sure that crickets relied only on their cercal system for detection. The piston travelled far enough to touch the cricket. The cricket escape behaviour was recorded by the HS camera(Fig. 2).

The kinematics of the spider run is composed of a short initial acceleration over 1 cm followed by a phase of relatively constant velocity up to prey (Dangles et al.,2006b). Piston kinematic was set up to mimic these properties. Moreover, we simulated various piston velocities because wolf spiders modulate their attack velocity when preying on crickets (see Dangles et al., 2006b), thus increasing the ecological relevance of our biological test. We chose three attack velocities within the range of those recorded for spider attacks (see Dangles et al., 2006b): slow(40 mm s^–1), intermediate (150 mm s^–1) and fast (300 mm s^–1) attack. In total, 90 individuals (10 replicates × 3 instars × 3 velocities) were tested in a randomised set-up to avoid any of them adapting to the piston velocity.

Crickets regularly respond to a wind stimulus by first turning their bodies consistently away from the danger source (by e.g. 60–90°) with a short latency and then by walking/running or jumping(Stabel et al., 1985; Tauber and Camhi, 1995). Pivoting is time-consuming, but this strategy appears to be adaptive in escaping from natural predators [as suggested for cockroaches(Camhi et al., 1978)]. In the present study, we determined two variables for each HS video recording: (1)the type of escape (walking or jumping) and (2) the escape angle, i.e. the body angle of the cricket at the moment of escape relative to the direction of the piston.

The escape performance by evading crickets was expressed as a function of their escape distance, i.e. the distance between the cricket and the piston at the time the cricket initiates its escape. The escape performance was divided into two categories, depending on whether the cricket escaped before or after being touched by the piston. (1) The escape distance before contact between the cricket and the piston; this was the straight-line distance between the tip of the cercus and the tip of the piston in the frame prior to cricket-leg flexion, either to turn (most cases) or directly jump/walk (few cases). This measured the air-evoked escape performance of the cricket. (2) The escape distance after contact between the piston and the cricket; this was the distance between the location of the tip of the piston in the frame of contact and its location in the frame prior to cricket-leg flexion. This is an inverse measurement for touch-evoked escape performances in crickets. Therefore, an escape distance of zero corresponded to the tip of the cerci being touched by the piston. Positive and negative values for escape distances corresponded to air- and touch-evoked responses, respectively. We normalised escape distances to the total body length of the cricket to provide a more ecologically relevant measure of cricket escape performance. This measurement indicates how many cricket body lengths remain between the cricket and the mimicked spider before the cricket escapes (Fig. 3). All image analyses were performed using R (R development core team 2004).

Log-transformed proportions of various types of escape behaviours(Table 1) were analysed using a Tukey-type multiple comparison test (Zar,1998). Log-transformed escape distance data(Table 2) were analysed using two-way repeated-measures ANOVA, in which `instar' and `attack velocity' were the first and second factors. In cases of statistically significant changes,data were further analysed using a post hoc Fisher's least significant difference (LSD) test. Data were considered significant for P values less than 0.05. All statistical and image analyses were performed using R.

In 90 trials, only five crickets (two juveniles and three adults) did not respond to attacks simulated by the piston, i.e. they remained motionless. We observed a significantly higher proportion of air-evoked escapes in juvenile I and II (32.1% and 33.3%, respectively) than in adults (18.5%)(Table 1, Tukey test: P=0.030 and P=0.027, respectively); this was not the case between juveniles I and II (Tukey test, P=0.56). We also observed ontogenic differences in cricket escape behaviour: juveniles almost always(96.5%) responded to predatory stimuli by jumping, whereas a substantial proportion of adults (25.9%) escaped by walking. Most (64.3–66.6%)instar escape turn angles were between 0° and 30°. Only six crickets in total escaped with a turn angle greater than 60°. No difference was observed for the frequency or magnitude of turns between walking and jumping responses.

We detected significant differences in normalised cricket escape distance among instars for all three piston velocities in both air- and touch-evoked escapes (Fig. 3, Table 2). Post-hoccomparisons revealed that adults had lower air- and touch-evoked escape distances than juveniles (I and II), for all attack velocities. Interestingly,we observed an increase in interindividual variability of escape distance with cricket development: few adult crickets can react as quickly as juveniles after being touched by the piston and most adults have a two- to fivefold increase in normalised escape distances. This ontogenic difference in escape distance was more pronounced for air-evoked escapes: the greatest escape distances to wind signal (up to 2.3 times the cricket body length) were always observed in juveniles for all three piston velocities(Fig. 3). However, this result was a consequence of the normalisation of escape distances to cricket body length. When applied to absolute escape distance, the wind-evoked response was the same among age classes (ANOVA, F=1.9361, P=0.1766, data not shown). No effects of piston velocity on the escape distance of crickets for any type of response were detected(Table 2).

The use of air cues by crickets and cockroaches to escape their predators is well documented (Camhi,1984; Gnatzy,1996). However, our field study further describes various types of predator signals. We observed that most crickets escaped running predator signals only after physical contact and not by using their air-detection system, as it is generally assumed. This result was observed for all tested piston velocities and suggests a crucial, but previously disregarded, role for cercal touch perception in cricket escape performance. Crickets can rely on powerful legs to escape predators and this may explain why they wait until the last moment to initiate a jump. This late reaction may explain why most crickets do not consistently achieve escape angles between 60 and 90° from the source of danger (e.g. 60–90°) (see Tauber and Camhi, 1995) before jumping. The prominent role of touch-sensing for escaping from predators may have been reinforced by the field conditions of the experiment. Air signals from a predator may be barely detectable and poorly reliable in the presence of background noise created by wind turbulence, a feature that has been observed in foraging bats (Jones and Rydell, 1994). Moreover, touch-sensing may be more efficient than air-sensing in detecting predators (such as spiders) at close range in leaf litter.

Most studies on the development of performance of traits related to escape behaviour have focused on locomotion, generally reporting that adults have higher or similar absolute capacities than juveniles [e.g. lizard speed(Irschick, 2000), mammal speed(Trillmich et al., 2003)]. However, Gibb et al. (Gibb et al.,2006) proposed that the development of the escape response in Teleost fish would decrease as juveniles become larger, as a result of biomechanical constraints in body shape. Measurements of the escape distance in crickets lead us to similar conclusions: escape performances in juveniles were higher than those in older instars. This increase in the escape distance for juveniles was in most cases related to a lower reaction distance after being touched by the piston: they respond faster than adults to an immediate danger. The decline in the escape performance for all tested speeds and for both escape types, despite strong differences in stimulus transmission to the sensory structures, is striking and calls for further interpretation in terms of neural control processes related to the age of the animal.

The negative relationship between body size and escape performance may have important implications in an evolutionary and ecological context. For instance, optimality theory suggests that the behavioural response of an individual in response to a predator is influenced by the risk of predation(Ydenberg and Dill, 1986). Although juvenile crickets forage in the same environment as adults and are faced with the same predators, they suffer higher rates of mortality, partly because their smaller size makes them more vulnerable to predators(Dangles et al., 2006a). An increase in absolute size allows crickets to reach a size-refuge, the size at which they become too big to be caught by most wandering predators. This hypothesis is supported by the greater proportion of walking responses in adults than juveniles; it is also supported by a previous experiment reporting more frequent hiding behaviour in juvenile crickets than in adults exposed to predators (Dangles et al.,2006a). A broad range of morphological, physiological, behavioural and ecological changes through ontogeny affect performance(Irschick, 2003); however, the specific concept of size-refuge has seldom been considered by functional ecologists [for an exception, see Wainwright(Wainwright, 1996)]. We expect that the concept of size-refuge would improve our understanding in the development of performance, given its importance in community and population ecology theories [see Aljetlawi et al.(Aljetlawi et al., 2004) and references therein]. This is another reason to move swiftly towards more field experiments carried out in ecologically relevant settings.

The authors are grateful to two anonymous referees for their helpful comments on a previous version of the manuscript, especially on the scaling issue of escape performance. Part of this study was conducted within the Cricket Inspired perCeption and Autonomous Decision Automata (CICADA) project(IST-2001-34718) and within the Customized Intelligent Life Inspired Arrays(CILIA) project (FP6-IST-016039). These projects are both funded by the European Community under the `Information Society Technologies-IST' Program,Future and emergent Technologies (FET), Lifelike Perception Systems action.