Context-dependent effects on attack and defense behaviors in the praying mantis Tenodera sinensis

As predatory animals are often prey themselves, they need to be able to distinguish between cues that represent potential food sources versus threats and react accordingly. Animal behavior can be influenced by many factors from their internal state and surrounding environment. These factors, such as hunger level, reproductive drive, the age of the organism and the presence of predators in the environment, contribute to an animal's behavioral state. Animals may choose a more or less risky food source depending on their level of hunger and those with young may take fewer risks. Indeed, an animal that has located a rich source of food might have a higher threshold for reacting to a potential predator than one that has not. The ability to adjust aggressive or defensive behaviors to specific contexts in the environment is critical for an animal's survival.

There are many examples of species that change their behavior based on their internal state and their perceived predation risk. African wild dogs change their behavior based on whether or not they have pups, with denning packs being more risk averse than non-denning packs (Marneweck et al., 2021). Bat-eared foxes, Otocyon megalotis, stop exploiting food patches earlier during a new moon, when their perceived risk of predation by larger animals is higher (Welch et al., 2017). Pied cormorants change their diving behavior by spending less time on the surface when there is increased predation risk from tiger sharks (Dunphy-Daly et al., 2010). They do this by increasing their dive time, regardless of how successful their feeding is. Intraguild predator avoidance can cause mesopredators to avoid potentially suitable habitats. The little owl, Athene noctua, avoids coming within 150 m of the edge of forests because their predator tawny owl, Strix aluco, may inhabit those habitats (Michel et al., 2016). This behavior was displayed regardless of whether there were actually tawny owls in the forest. Sometimes mesopredators are attracted to dominant predator cues; for instance, the stoat, Mustela erminea, changes its foraging behavior in response to dominant predator odor cues from the ferret, Mustela furo, by being attracted to the odor and using it to identify areas of increased resources. Stoats consumed food earlier in areas marked by apex predators, relative to consumption time in locations with no odor or an unrecognized control (Garvey et al., 2016). When stuffed birds were placed near bromeliads, there was a decrease in damselfly larvae biomass in the plants, presumably as a result of damselflies avoiding the predators and therefore suitable habitats (Breviglieri et al., 2017). Consequently, there was an increase in mesopredator biomass in the bromeliads, dramatically changing the ecology in the bromeliad microenvironment. These various context-dependent behavioral changes show the importance of detecting situations where potential predators might be present and using that to compute a cost–benefit analysis of resource use.

Like the previous examples, praying mantises are both predators and prey and likely contend with similar context-dependent behavioral strategies for survival. Mantises depend on acute vision to detect both their own prey and predators. When attending to a stimulus, they will follow it with their head to keep it in the foveal region of their visual field (Kral and Prete, 2004; Prete et al., 2011; Yamawaki, 2006). Praying mantises are also the only insect confirmed to have stereoscopic vision, where they are able to determine depth by comparing the shifted image on each eye (Nityananda and Read, 2017; Nityananda et al., 2018; Pettigrew, 1986). After observation of a stimulus, the mantis may respond in various ways. When attacking prey that is in range of a strike, typically a maximum of ∼70–80% of the foreleg length away (Maldonado et al., 1967), the mantis will assume a stereotypical ready posture with its raptorial forelegs folded in front of the prothorax before explosively lunging forward to capture the prey. If the prey is too far away for the mantis to successfully lunge, it must either wait for the prey to advance within range (ambush strategy) or advance closer to the prey itself (stalking) (Svenson and Whiting, 2004). With movement comes the risk of discovery by a predator, such that many species of mantises fall on a spectrum of hunting strategies that vary by the amount of risk taken that exposes them to predation, some exhibiting both stalking and ambush depending on a number of factors. Cryptic species such as the floral associated mantises in the genus Creobroter, tend to exhibit an ambush strategy by sitting and waiting for pollinating insects to come close (Svenson et al., 2016). Others that live on the ground, such as members of the family Eremiaphilidae, tend to exhibit an active pursuit strategy to ensure they catch what they need to survive in an open-ground environment (Prete et al., 2012; Svenson and Whiting, 2004). However, generalist species, such as Tenodera sinensis, switch between hunting strategies depending on both external conditions and their internal state of satiety (Bertsch et al., 2019; Prete et al., 2012).

The defensive behaviors that praying mantises use to avoid predation are qualitatively different and can be easily distinguished from hunting behaviors. Defensive behaviors include evasive flinches, where the animal retreats from the predatory stimulus, and cryptic responses, where the animal crouches against the substrate and extends its forelegs (Watanabe and Yano, 2010; Yamawaki, 2011). A stereotyped deimatic response, a type of startle display, where the animal raises its wings and forelegs to appear as large as possible, can be elicited from live predators (Maldonado, 1970), but is rarely seen in response to artificial looming stimuli, though the other behaviors are readily elicited (Prete et al., 2011; Yamawaki, 2017). A review of startle displays for 58 species found that these displays, as well as morphological traits, were phylogenetically conserved (Vidal-García et al., 2020) and lineages with high diversity were more likely to have evolved these displays. This suggests that displays as defensive responses are both important and aligned with speciose lineages. The deployment of such displays is inherently linked to the context of perceiving an object as a threat or predator.

The flexible hunting strategies and range of defensive responses of generalist praying mantises such as T. sinensis suggest that, rather than being totally stereotyped, both hunting and, to a lesser extent, defensive behaviors are state dependent. Here, we tested this state dependence hypothesis by investigating behavioral responses of T. sinensis to sinusoidal (target) and looming (threat) stimuli in the context of four different states. We found that an animal's level of aggression to a sine stimulus may change its subsequent response to the looming stimulus. An animal's age since eclosion and its sex also strongly affected these behaviors. Finally, rearing location, comparing between lab-raised (L) and wild-caught (W) animals, showed interesting effects. These two groups had very different experiences during their development. The lab-raised insects were kept in separate containers and were always provided with food and water. In contrast, the wild-caught individuals had to actively hunt for prey while dealing with the possibility of being caught by other predators.

Our impetus in examining the last question regarding rearing location stemmed from the realization that environmental influences in pre-adult animals can, in general, lead to changes in adult behaviors. It is well established that stressors during puberty can increase adult aggression in rats, though this is displayed differently by males and females (Cordero et al., 2013), and have a negative effect on their adult problem-solving ability (Denenberg and Morton, 1962; Woods, 1959). In fish, it has been shown that overall warmer temperatures in Lake Tanganyika heighten cichlid aggression (Kua et al., 2020). Furthermore, within-day temperature increases of less than 3°C can increase individual levels of aggression and activity in juvenile damselfish (Biro et al., 2010), which shows that environmental effects can be potent on very different time scales. In insects, there are many examples of early environmental factors impacting adult neurobiology and behavior. In beetles of the Onthophagus family, males that were given more nutrition during development were more likely to be larger and use aggressive mating strategies (Emlen, 1997; Moczek and Emlen, 2000). In three species of mites which share an environment, experience of predation risk in sub-adult stages influenced where adult females would choose to lay their eggs (Walzer and Schausberger, 2011). Environmental influences on young animals can impact their adulthood in complex ways. Here, we add evidence of some of these effects in T. sinensis to this growing body of literature. Taken together with the other variables we examined, our study sheds light on the flexibility of mantis hunting behavior.

Adult Tenodera sinensis (Saussure 1871) were used in all experiments. Animals were either raised in the laboratory or captured from the wild. Lab-raised animals were hatched from ootheca that were produced from individuals that were mated in the lab. These were fed Drosophila in groups until their 3rd instar and then moved to individual 1.8 l plastic containers and kept on a12 h:12 h light:dark cycle at 27°C. They were then fed cockroach nymphs of appropriate size within their containers until adulthood when they were used. Wild-caught animals were captured from three locations: the Bolivar Heights Battlefield in Harpers Ferry, WV, USA; the Forest Hill Park in Cleveland Heights, OH, USA; or the Case Western Reserve University Squire Valleevue Farm in Moreland Hills, OH, USA. Wild animals were caught anywhere between the 5th and 7th instar. All wild animals were caught before the final molt so their ages post-eclosion were accurately measured. All animals were given water and fed 3 times a week. We selected only healthy mantises with all limbs and external sensory structures intact for experiments. All experimental animals were treated appropriately, and we operated in accordance with all ethical animal care guidelines.

For all experiments, animals were placed on a 14 cm metal perch placed approximately 5–6 cm away from an LCD screen, with the plane of the screen perpendicular to an imaginary line coming at a 45 deg angle from the animal's head. Animals were recorded from the side using a Casio Ex-f1 camera at 120 or 60 frames s⁻¹.

Each animal was presented with a three-stimulus paradigm consisting of an initial sinusoid stimulus presentation, followed by a looming stimulus presentation, followed by another sinusoid stimulus. This block of stimuli was repeated 4 times. Trials started with the ellipse moving sinusoidally from left to right, or top to bottom of the monitor. The ellipse would start at one of 5 randomly selected positions on the track that it followed: the center, one of the ends, or in between the center and endpoints. Both orientations of the sine stimulus took 8.25 s to complete, making approximately 2.5 complete revolutions. When following a sine stimulus, the loom stimulus would appear exactly where the sine stimulus ended. Each looming stimulus took 5.73 s to complete, and the loom expanded non-linearly. Each stimulus began after a 0.01 ms delay after the end of the previous stimulus. There were 20 s of blank white screen in between each block of three stimuli. If an animal jumped off the perch or climbed on the monitor during the trials, it was enticed to climb on a ruler by placing the ruler above it and was then moved back to the perch and allowed to climb back on. Trials where the animal was absent for any part of the three-part stimulus were not included in our analysis.

The loom-only trials were essentially the same as the three-stimulus trials, except only the looming stimulus was shown with 20 s of white screen in between looms. As before, 12 looms were shown per trial presentation. These trials established a baseline response to the looming stimulus which was used to compare responses to the loom after the mantis reacted to a preceding sinusoid stimulus. The baseline for sinusoid-only trials was taken from the first responses of the sine–loom–sine paradigm.

Longitudinal trials followed five adult females from eclosion (designated day 0) until between 41 and 45 days post-eclosion. Each animal was presented with the three-stimulus paradigm every 5 days, coinciding with the 5 day bins. The first experiment could have been presented to the animal any time between day 0 to day 5 post-eclosion, but all subsequent trials for the individuals were presented in strict 5 day increments. These animals were also used for loom-only trials on alternate days from their three-stimulus trials. After 45 days as an adult, death and senescence decrease the reliability of the data to an extent that it was determined no longer useful to test animals of such advanced age.

Animal behaviors in response to the stimuli were subjectively evaluated on a scale ranging from −5 to 5, where more negative numbers represent more defensive behaviors, more positive ones represent more aggressive behaviors, and 0 represents no response (Table 1 and Fig. 1). Thrash responses did not fit into the continuum of behaviors and were coded separately from the number line. Representative video showing examples of several of the behaviors on the scale are presented in Movies 1–3. Videos were scored by multiple personnel independently and differences were resolved offline.

To evaluate differences between the first and second sine responses in Fig. 3, we used paired t-tests. Age effects on behavior magnitude were evaluated with Wilcoxon rank-sum and Kruskal–Wallis (K–W) tests (Fig. 4). t-Tests were used to evaluate differences in rearing environment (Fig. 6), and K–W tests were used to evaluate sex differences (Fig. 7). The number of subjects (N) and trials (n) varied from experiment to experiment and is, therefore, reported on individual figures.

Our initial analysis established baseline responses of female praying mantises to looming and sinusoidal stimuli. Fig. 2 details the responses of females to an initial sinusoid or looming stimulus. When females encountered a sinusoid stimulus first, their reactions varied widely in degree but were always aggressive or non-responsive. In our dataset, they never displayed defensive responses (Fig. 2A). Conversely, when presented with an initial looming stimulus, the females displayed medium-strength defensive responses, and never showed an aggressive response (Fig. 2B). Animals that did not react to the initial stimuli, but did react to subsequent stimuli, are recorded as 0 (Fig. 2). Thus, the sinusoid stimulus consistently generated an aggressive or 0 response, whereas looming stimuli consistently evoked a defensive or 0 response, showing the behavior segregated in response to our stimulus paradigm. A very small percentage of looming stimuli generate the unique thrash response, which appears to be a somewhat aggressive defensive response. Thrash responses are characterized by a foreleg extension toward the stimulus that could be a defensive grasp or push, but the body or prothorax stays still or withdraws from the stimulus slightly. It is a behavior that does not seem to be focused on prey capture and clearly does not lead to a strike, as strikes were never observed following thrashes and the pre-strike behavior is qualitatively distinct from the thrash. It will be important in later analysis to note that for these initial looming stimuli, the thrash response was exceedingly rare in initial responses, <5% in Fig. 2B.

To examine the interactions between sinusoidal (target) and looming (threat) stimuli, we presented the mantises with a three-part set of stimuli. The first was a sinusoid, followed immediately by a looming stimulus and finally a second sinusoid. This paradigm tests both whether a defensive response to a looming stimulus could make the subject reticent to attack a subsequent sinusoid in the third part of the paradigm and whether the subject would be less likely to perform a totally defensive response to the looming stimulus if it had just performed an aggressive response to the preceding sinusoid.

Fig. 3A displays the behavioral responses of female animals to each of the three stimuli in the sine–loom–sine trial. At the first presentation, sinusoid responses were ignored in 45% of trials. The rest of the trials resulted in aggressive responses. As with single stimuli, no defensive responses were observed. However, unlike the single-stimulus trials, responses to the looming stimuli following the sinusoid showed a new and interesting pattern, with 57% showing defensive reactions similar to those seen in response to an initial looming stimulus (Fig. 2B), but with 23% showing thrashing reactions that were only rarely seen in the response to an initial looming stimulus and could not clearly be categorized as aggressive or defensive (Fig. 3A). The remaining 20% showed no response to looming, which is twice the zero responses to looming as for an initial stimulus.

If the looming stimulus is influenced by previous aggressive responses to the sinusoid stimulus, it would seem logical to expect that this alteration is particularly found when the first sinusoid produced an aggressive response in contrast to a no response trial. When responses to the looming stimulus were divided according to the aggressive or non-aggressive nature of the first sinusoid response, we found that almost all thrash responses to the looming stimuli were preceded by aggressive responses to the sine stimulus (Fig. 3B). In this experiment, almost all the indifferent responses to the looming stimulus (scored 0) were also preceded by aggressive responses to the sinusoid stimulus (Fig. 3B).

The second sinusoid stimulus generated aggressive responses that were not significantly changed from those to the first sinusoid (paired t-test, P<0.05). Thus, the sinusoid does not appear to be similarly affected by being preceded by a defensive response.

To examine whether age had any effect on mantis target versus threat behaviors, we plotted the magnitude of behavioral responses to sine–loom–sine trials throughout the subject's post-metamorphic life. During the first 10 days, animals showed relatively weak aggressive responses to the sine stimulus (Fig. 4A). Between 11 and 25 days after molting, females tended to be more aggressive toward the sinusoid stimuli than their younger counterparts, with a significant difference between the 0–5 and 11–15 days age range (Fig. 4A). After 36 days of age, the animals were significantly less likely to act aggressively towards the first sinusoid stimulus (K–W test, P<0.05).

Loom responses were more consistent throughout the various age groups. There was no significant change (K–W test, P>0.5) in the mean responses to looming, although there were some decreases in the 11–15 and 16–20 days age groups (Fig. 4A). Moreover, the appearance of thrashing behaviors in response to the looming stimulus increased in the middle age groups (Fig. 4B). This is consistent with the effect on looming of a previous aggressive response to stimuli that is described above (Fig. 3). The increase in aggressive sinusoid stimuli would be expected to be coupled to increased thrashing responses. To further test this notion, we tested a series of loom responses in the absence of preceding sine stimuli (Fig. 4C) throughout adult life and found little or no variation with age (K–W test, P>0.5). Finally, these trials showed little increase in thrashing behaviors (Fig. 4D).

To account for variation among individual subjects within our test group, we repeated the age analysis on a smaller group of 5 animals throughout their entire adult life span. These longitudinal studies each showed similar peaks of aggressive activity in the middle adult life periods, but details varied among individuals (Fig. S1). When all 5 of the longitudinal studies were combined, the pattern was very similar to that seen in Fig. 4A for the entire dataset. However, the smaller number of trials in this study failed to reach significant differences in most cases. As with the larger data set, loom responses in the longitudinal study did not change significantly between any age group (K–W P>0.05).

As the lab-raised subjects had never experienced predators and always had a ready supply of food available, we wondered whether their responses were in any way different from those of the animals caught in the wild. When we separated our data based on the origins of each subject, we did find differences between the wild-caught and lab-raised females in their level of aggression (Fig. 5). In response to the first sinusoid stimulus in the sine–loom–sine paradigm, wild-caught animals had a higher percentage of level 5 (strike) responses (39%) than the lab-raised animals (13%) (Fig. 5A,C). Wild-caught animals also were more likely to ignore the looming stimulus (39%) than lab-raised animals (15%) (Fig. 5A,C). These non-responses to the loom trials were almost exclusively seen following an aggressive response to the preceding sinusoid stimulus (Fig. 5B,D). Thrash response frequency was not affected by the animal's rearing environment.

Comparisons between age groups of lab-raised and wild-caught animals throughout adult life were difficult as there were not enough observations of wild-caught animals in the older periods to show significant differences between age groups (Fig. S2). Thus, only animals between the ages of 6 and 25 were included in the lab-raised versus wild-caught animals analysis in Figs 5 and 6.

To get a more detailed picture of the differences between the two groups as they responded to the first sine stimulus, we plotted each individual response as a data point indicating the behavior level for each age group (Fig. 6A). Behavioral responses to the first sinusoid stimulus can be grouped into less aggressive tracking and stalking behaviors (1–3) versus more aggressive pre-strike and strike behaviors (4–5). A pattern emerges in which the data for lab-raised animals appear to increase in the middle life periods in both stalking (levels 1–3) and strike (4–5) behaviors. In contrast, the wild-caught individuals were always more likely to go completely to striking behaviors rather than just stalking. That is, lab-raised subjects increased both categories of responses, whereas wild-caught individuals always favored the more aggressive strike behaviors.

To confirm this, we compared the incidence of responses in the 1–3 (stalking) range versus the 4–5 (striking) range (Fig. 6B,C) in each 5 day bin for lab-raised and wild-caught animals. These data are presented both as raw numbers (Fig. 6B) and as percentages (Fig. 6C). As expected from our observations of the data in Fig. 6A, the numbers and percentages of stalking and strike responses for the lab-raised subjects were approximately equal for each period (t-test, P=0.45) except the 21–25 day period where stalking behaviors dominated. The increase in stalking and strike behaviors for lab-raised subjects was matched by a decrease in non-responses (0 response) (Fig. 6A). In contrast, wild-caught animals more often generated 4–5 (strike) behaviors rather than 1–3 (stalking) behaviors for all ages up to 25 days (t-test, P<0.05) (Fig. 6B,C), and only showed a decrease in non-responses in the initial (0–5 day) period and the last (21–25 day) period (Fig. 6A).

All of the previous analysis was done with female subjects. Male T. sinensis showed slightly higher rates of non-responsiveness to the sinusoid stimulus than females and lower rates of non-responsiveness to the loom (Fig. 7). They tended to have more extreme defensive responses to looming stimuli than did females. However, while they showed very few 2 and 3 responses to the sinusoid stimulus, their overall pattern of behavioral responses to the sine was similar in distribution to the female responses (Fig. 5A). Males showed similar rates of thrash responses to the loom after reacting aggressively to the preceding sinusoid, though they were less likely to ignore the loom in that condition than females (Fig. 5B). The thrash response seemed to be preserved across all of the variables that we studied. Males also had a much higher number of extreme defensive responses to the loom, especially when they did not respond aggressively to the preceding sinusoid. Male animals did not show a significant trend of increased aggression with age, though the trend of being less responsive right after eclosion and with old age appears to be similar. They also showed a much higher level of individual variation than females, with particular individuals being very aggressive and others barely attending to the stimulus at all. Thus, a larger sample size might show a similar trend in aggression to the females.

We found that T. sinensis responds to artificial prey-like sinusoidal stimuli and predator-like looming stimuli in a highly context-dependent manner. When presented with just the sinusoidal stimulus, animals were roughly equally likely to respond aggressively or ignore the stimulus, but never reacted defensively. In contrast, when presented with just the looming stimulus, animals overwhelmingly showed defensive responses and never showed overtly aggressive responses (Fig. 2). When the stimuli were presented in a sequence of sinusoid stimulus–looming stimulus–sinusoid stimulus, responses to both the first and second sinusoid remained consistent with responses to the individual sinusoid stimuli (Fig. 3A). Defensive responses to the looming stimulus, however, were decreased, with more animals either ignoring the stimulus or showing a new type of thrash behavior where the animal possibly tries to push the looming stimulus away (Fig. 3A). These changes were explained by an animal's reaction to the immediately preceding sinusoid stimulus. Animals that did not react to the stimulus showed response patterns similar to those of animals that were presented with only a looming stimulus, while animals that reacted aggressively to the preceding sinusoid stimulus showed the altered response pattern (Fig. 3B). These results highlight an effect on defensive behavior.

We also found effects on the aggressive response to the sinusoid. After eclosion to adulthood, animals, especially females, had a period of low aggression, followed by a period of higher aggression, which then tapered off into a consistent lower level of aggression for the remainder of their lives (Fig. 4). The specific nature of the mid-life spike in aggressive behaviors differed according to whether the animals were laboratory-raised or caught in the wild (Fig. 6). Lab-raised animals continued to show roughly equal amounts of increased attention to the stimulus and very aggressive strike behaviors, whereas wild-caught animals tended to either ignore the stimulus or react very aggressively with strikes (Fig. 6). Therefore, prior aggressive responses, sex, age and rearing location all affected both overall levels of aggression and specific responses to the looming stimulus.

These results are consistent with the notion that animals are not automatons but rather tend to respond to stimuli in a context-dependent manner. That is, under one set of conditions, a stimulus may generate one behavioral response, while in another, the same stimulus may result in a completely different behavior. For example, the wind-sensitive dorsal giant interneurons that are part of the cockroach escape system will evoke running movements when all legs are in contact with the ground. However, if ground contact is removed, the same stimulation of dorsal giants will evoke flying behavior (Ritzmann et al., 1980). In crickets, a single identified interneuron (INT-1) elicits an avoidance maneuver in response to ultrasound, but only if the cricket is flying (Nolen and Hoy, 1984). Moreover, recent studies have shown that the generalist praying mantis, T. senesis, will actively stalk prey when it is hungry, but becomes less active in stalking as it feeds, ultimately switching to an ambush strategy in which the mantis will only strike prey that comes near (Bertsch et al., 2019). In the current study, we take this concept of natural context dependence to a wider range of conditions.

When animals were shown just looming stimuli, 88% of the time, some degree of defensive response was elicited (Fig. 3B). In contrast, responses to a sinusoid alone result in aggressive attack behaviors in 55% of trials. However, when we combined these two stimuli by presenting a sinusoid immediately followed by a looming stimulus and then a second sinusoid, we noted some significant changes in the responses to looming stimuli. They were now bimodal, with some responses resembling those to the single loom while others showed a decrease in defensive responses and the appearance of a novel thrashing behavior. Of the responses to the loom that followed aggressive sinusoid responses, only 27% showed the typical defensive patterns. The rest showed either no response (32%) or the unique thrash response (41%) (Fig. 3B). Conversely, 90% of loom responses that followed non-aggressive responses to the preceding sinusoid were defensive, mirroring the responses to the looming stimulus alone. This seems to indicate that aggression changes the internal state of the animal. However, it should be noted that the second sinusoid stimulus did not evoke any changes in the attack pattern relative to the first sinusoid. Thus, while the aggressive response to the first sinusoid altered the defensive behavior to the loom stimulus, the reverse was not true. That is, a defensive response to the loom stimulus did not reduce the aggressive tendency towards a sinusoid stimulus (Fig. 3).

As the thrash was almost exclusively seen following a previous aggressive stimulus, it seems probable that the behavior itself is being evoked by an internal state change induced by aggression to the sinusoid. Looming responsive neurons have been found in Tenodera aridofoilia (Yamawaki and Toh, 2009) and might be analogous to the DCMD found in locusts which governs defensive behaviors (Santer et al., 2005, 2006, 2008). The aggressive neural circuit governing stalking and ultimately strike behavior may involve the central complex (CX). Navigational movements have been shown to be controlled by neurons in the CX of Drosophila (Ferris et al., 2018; Green et al., 2017; Kim et al., 2015; Seelig and Jayaraman, 2013, 2015; Turner-Evans et al., 2017) and cockroach Blaberus discoidalis (Martin et al., 2015). Extracellular recordings from freely moving cockroaches, a close relative of the praying mantis, revealed neurons in the CX that fired in specific patterns related to turning and forward motion (Martin et al., 2015). Moreover, head direction cells have been found in the CX of both cockroach (Varga and Ritzmann, 2016) and Drosophila (Seelig and Jayaraman, 2015).

Either of these circuits and their resulting behaviors might be modulated by a neuromodulator. Octopamine has been shown to regulate aggression in Drosophila with flies with lower octopamine showing decreased levels of aggression for mates (Hoyer et al., 2008; Jia et al., 2021; Rohrscheib et al., 2015). Moreover, the switch between walking and flight mode in locust neurons is associated with octopamine-induced development of plateau potentials in bimodal neurons (Ramirez and Pearson, 1991). This makes it a possible candidate to regulate aggression in T. sinensis as well. Interestingly, octopmanergic and tyraminergic neurons have been shown to be present in distinct areas of the locust CX (Homberg et al., 2013).

The variance of aggression with age (Fig. 4A) found in this study follows closely the results found in Liske and Davis (1987). They measured the number of crickets eaten per day for 35 days and found that female T. sinensis ate the most crickets between 12 and 25 days post-eclosion (Liske and Davis, 1987). In our study, maximum levels of aggression to the sine stimulus were found between 11 and 25 days. Hunger levels do impact the hunting strategy of T. sinensis, with starved animals more likely to stalk their prey and sated ones more likely to sit and ambush (Bertsch et al., 2019).

As to the older subjects, insects are known to reduce activity dramatically as they approach the end of life (Ridgel and Ritzmann, 2005; Ridgel et al., 2003). While our subjects were not as immobile as the cockroaches described by Ridgel et al. (2003), they did seem to be less active. An extensive analysis of their activity patterns was beyond the scope of this paper. However, a parallel project that observed mantises of various ages for 24 h in a large arena with and without prey present indicated that the older mantises were far less mobile and would only strike prey that came near to them (J.W.B., B.M.B.-S. and R.E.R., in preparation).

A considerable amount of anecdotal evidence suggests differences between insects that are raised totally under lab conditions versus those caught in the wild. Here, we provided data to support this notion. Differences between lab-raised and wild-caught animals show that while in many ways their behavior is similar, wild-caught animals are more likely to strike than their lab-raised cohorts (Fig. 5A,C). Wild-caught animals were also more likely to ignore the looming stimulus than lab-raised animals (Fig. 5), consistent with our observations on the effect of aggressive responses to the first sine upon the subsequent loom stimulus.

These distinctions were made clearer when the categories were split into the 5 day age groups (Fig. 6). While lab-raised animals showed increased levels of aggressive responses between 6–10 and 11–20 days post-eclosion, the wild-caught animals largely maintained similar levels of aggression from 0–20 days post-eclosion and then showed increased aggression in the final 21–25 day age range. In the younger age groups, lab-raised animal responses were equally divided between movements associated with detecting the stimulus (head and body orientation toward the stimulus; our categories 1–3) and actual striking behaviors (categories 4–5), whereas wild-caught animals either ignored the stimulus or responded very aggressively (categories 4–5). Upon entering the more aggressive ages, the lab-raised animals showed equal increases in low-level (1–3) and high-level (4–5) aggressive responses. This suggests that as they entered the more aggressive periods of their lives, they became more sensitive to the sinusoid stimuli. They detected the stimulus more often, but the resulting behaviors were still evenly divided between just orienting toward the stimulus source and actual striking behaviors. In contrast, wild-caught animals entered an even more aggressive period than lab-raised animals, but did so later, showing increased overall responsiveness and strike frequency in the 21–25 day range. As a result, when the wild-caught animals did respond, they were more likely to fully commit to an aggressive response that culminated in a strike.

Thus, as the lab-raised subjects transitioned from early adults into mid-adults around 15 days after eclosion, they appeared to become more aggressive because they were better at detecting potential prey, but the chances of these subjects either striking at the target or just turning toward it remained about equal. In the same period of adulthood, wild-caught subjects usually committed to striking at the target whenever they responded to it. Intriguingly, while both lab-raised and wild-caught animals showed a period of increased aggression, the period when they showed maximum levels of aggression did not overlap. Lab-raised animals showed a peak between the ages of 11 and 20 days, with no strikes (behavior 5) recorded from the lab-raised animals in the 21–25 day range (Fig. 6A). Wild-caught animals seem to be most aggressive in exactly this 21–25 day group, showing a large increase in strike frequency and a decreased tendency to ignore the stimulus. This means that the period of aggression from day 11 to day 25 post-eclosion is due to lab-raised animals at 11–20 days and wild-caught animals at 21–25 days and not necessarily driven by the same groups of insects (Fig. 4A). Had our dataset come from only one or other cohort, the graph in Fig. 4A would have been different from that presented here.

It is beyond the scope of this study to determine why the wild-caught animals are more likely to fully commit to a strike than the lab-raised animals. Permanent changes during embryonic development have been demonstrated in cockroaches. A study on the effect of temperature during development showed that individual cockroaches raised in warm conditions were significantly better at flying (Diekman and Ritzmann, 1987). Interestingly, groups that experienced warmer temperatures while still in their egg cases and were then switched to cooler temperatures after hatching were still better fliers than those that were placed at cooler temperatures during egg case development (Diekman and Ritzmann, 1987). Thus, a critical period appears to exist during development that occurs in the egg case, which permanently alters the cockroach's ability to initiate flight months after adult metamorphosis. A similar change could account for the differences we saw between lab-raised and wild-caught specimens.

Our study contributes to the extensive literature on context-dependent behavior. While praying mantises are known to be aggressive predators, their response to even a standardized stimulus is greatly affected by the conditions that they are in when the stimulus is presented as well as at other stages of their life. As praying mantises are complex individuals, analysis of the behaviors they show in response to a visual stimulus must take into account the age and sex of the individual. It also must consider whether the animal is in an aggressive state as a result of recent aggressive responses. Finally, the impacts of raising praying mantises in captivity should be further explored.

The authors would like to thank Dr Matthew Clemens for assistance with creating the figures in this paper as well as productive discussions.