Sheep in wolves’ clothing: prey rely on proactive defences when predator and non-predator cues are similar

Abstract

Predation produces intense selection and a diversity of defences. Reactive defences are triggered by predator cues, whereas proactive defences are always in effect. We assess whether prey rely on proactive defences when predator cues do not correlate well with predation risk. Many bats use echolocation to hunt insects, and many insects have evolved to hear bats. However, in species-rich environments like Neotropical forests, bats have extremely diverse foraging strategies, and the presence of echolocation corresponds only weakly to the presence of predators. We assess whether katydids that live in habitats with many non-dangerous bat species stop calling when exposed to echolocation. For 11 species of katydids, we quantified behavioural and neural responses to predator cues, and katydid signalling activity over 24 h periods. Despite having the sensory capacity to detect predators, many Neotropical forest katydids continued calling in the presence of predator cues, displaying proactive defences instead (short, infrequent calls totalling less than 2 cumulative seconds of sound per 24 h). Neotropical katydid signalling illustrates a fascinating case where trophic interactions are probably mediated by a third group: bats with alternative foraging strategies (e.g. frugivory). Although these co-occurring bats are not trophically connected, their mere presence disrupts the correlation between cue and predation risk.

1. Introduction

Predation is a major selective force in nature, and as a result, organisms have evolved an array of anti-predator behaviours. These behavioural defences take many forms [1], including proactive defences that are always present regardless of whether an animal has detected a predator, and reactive defences that are elicited by the detection of a specific threat [2]. Many terms are used to categorize prey defence strategies in the literature, but their meaning can vary depending on whether they refer to morphological, behavioural or physiological defences, whether they are reversible, and whether they are considered from the perspective of predators detecting prey or vice versa (electronic supplementary material). For the purpose of describing behavioural defences from the perspective of the prey animal, we use the terms proactive and reactive defences here. Most animals exhibit both types of defences, switching from proactive to reactive defences at an appropriate stage of a predation sequence [3,4]. Some species, however, clearly invest more than other species in proactive defences such as camouflage and reduced movement [1,5,6]. Although many studies have investigated the cues that elicit reactive anti-predator behaviour [7,8], the question of why some species rely more on proactive behavioural defences whereas others rely more on reactionary defences has received less attention.

Prey species vary in their sensory capacity to detect predators [9,10] and predators vary in their detectability to prey [11–13]. Animals that cannot reliably detect a predator's presence for either reason are predicted to rely more on proactive than reactive defences to avoid predation [14–16]. For example, many nocturnal insects have ears that are sensitive to the ultrasonic echolocation calls of insectivorous bats, allowing them to initiate the reactive defence of evasive flight when they detect an approaching bat [17]. Alternatively, nocturnal moth species that lack ears tend to fly less, closer to vegetation and more erratically than those that have ears [18–20], showing greater reliance on proactive defences than species with the ability to detect the predator cue. The ability of prey to detect predators also varies across predator species and environments [21,22]. For example, dimorphic birds of prey have different levels of foraging success depending on light environment due to their detectability to prey [13,23].

Here, we assess whether predator detection capabilities and the reliability of predator cues relate to the use of proactive or reactive defences in Neotropical katydids (Orthoptera: Tettigoniidae). Male katydids produce acoustic signals (calling song) to attract females for mating. In many habitats, katydids are known for their conspicuous calling song [24,25], reflecting female preference for louder and more repetitive signals [26]. In Neotropical forests, however, it has been noted that forest-dwelling katydid species produce few calls per night, and it was proposed that this low calling rate is a proactive defence against forest-dwelling gleaning bats that eavesdrop on katydid mating calls to locate them on vegetation [27]. Dietary studies have shown that these Neotropical bats are voracious, and in some cases specialist, predators of katydids [27,28]. Due to their reliance on echolocation for orientation in space, bats are continuously providing acoustic cues of their presence as they move through their environment. At least one North American katydid species will stop singing when it hears the echolocation calls of bats [29], and the European katydid Tettigonia viridissima will interrupt calling when hearing pulsed ultrasound [30]. The information content of this predator cue, however, is diluted in the Neotropics by the presence of many other bat species that do not pose a risk to a perched calling katydid, such as aerially hunting and frugivorous bat species. For example, on Barro Colorado Island (BCI), Panama, Symes et al. [31] recorded echolocation sequences approximately once per minute throughout the night, but less than 4% of these echolocation sequences were produced by phyllostomid bats, the family that includes eavesdropping bats. Therefore, the correlation between predator cue and predation risk is very low in this environment.

To determine how unreliable predator cues might influence anti-predator defences, we conducted acoustic playback experiments, collected electrophysiological recordings of auditory neurons and measured the male signalling rate in 11 species of Neotropical katydids. First, using acoustic playback experiments, we tested whether katydid signalling behaviour is influenced by the presence of bat calls and other acoustic stimuli. Second, if insects disregard predator echolocation calls, this lack of response might arise through two different mechanisms: (i) katydids might not be capable of detecting this predator cue or (ii) katydids might detect the predator cue, but not respond to it. To differentiate between these mechanisms, we measured auditory neural responses of katydids to the same acoustic cues presented in playback experiments. Third, to determine whether katydids adopt a proactive defence strategy that continuously minimizes their conspicuousness to bats, we assessed how often each species signals over 24 h. Previous authors have noted the low signalling rate of some Neotropical forest katydid species [27,32–34]. Here, we make quantitative measurements of signalling activity and daily sound production to assess how often these katydids signal and whether the species that are more responsive to predator cues also have higher signalling rates. Synergistically, these multiple lines of evidence provide a detailed view of the reliance of prey on proactive or reactive defences in response to a dangerous but difficult-to-differentiate predator.

2. Methods

(a). Study animals

Katydids were collected at night from vegetation in the forest and at lights on BCI, Panamá, between 1 January and 25 March of 2016 and 2017. Katydids were identified using published sources [35,36], and maintained on a diet of apple and cat food until testing. We collected data for 11 katydid species from three subfamilies: Arota festae, A. panamae, Chloroscirtus discocercus, Lamprophyllum micans, Microcentrum sp. (morphotype referred to as ‘polka’ in [37]) Orophus conspersus, Phylloptera dimidiata, Steirodon stalii and Viadana brunneri (Phaneropterinae); Copiphora brevirostris (Conocephalinae); Cocconotus wheeleri (Pseudophyllinae). Many of these species are separated by 30–100 Myr of divergence [38], meaning that the sample spans a diverse set of species that now share a similar habitat and suite of predators. Sample sizes varied by experiment and are provided in the figures.

(b). Acoustic playback experiments

We tested whether male katydids produce fewer acoustic signals when exposed to bat echolocation calls compared to silence, or white noise (a control for responses to non-predator acoustic cues). For detailed playback methods, including equipment specifications, recording equipment and stimulus filtering, see electronic supplementary material. In brief, playback experiments were conducted in outdoor mesh greenhouses located less than 5 m from the forest edge, preserving a naturalistic climatic and acoustic environment. During testing, katydids were placed individually in acoustically transparent cylindrical mesh cages and exposed to three playback treatments. The bat playback consisted of calls from Trachops cirrhosus, a gleaning bat species known to hunt by eavesdropping on the calls of katydids [39–41]. Echolocation calls were recorded from three different T. cirrhosus individuals as they flew individually in a large flight enclosure. For playbacks, one representative call was selected from each individual bat and this call was repeated every 30 ms for 1 s, which is typical for T. cirrhosus when flying close to vegetation or hanging from a perch [42]. The 1 s echolocation sequence was broadcast once per minute for 30 min, rotating through the three different T. cirrhosus individuals. The 1 min interval between echolocation sequences is naturalistic for these environments, where it is common for a stationary microphone to record hundreds of echolocation sequences each night [31]. The silence treatment consisted of a 1 min .wav file of silence broadcast once per minute for 30 min. The white noise stimulus was a 50 ms pulse of white noise (with 5 ms linear on and off ramps) from 5 to 100 kHz generated by SASLabPro Acoustic Software (Avisoft Bioacoustics). This acoustic stimulus covers the full spectrum of sound frequencies regularly experienced by katydids in the forest, but it does not have the pulsed temporal structure of most sympatric katydid species. The white noise stimulus was also broadcast once per minute for 30 min, and the order of each 30 min acoustic treatment was randomized for each katydid. Calls of focal katydids were recorded during playback treatments to assess calling activity.

We used a model comparison approach to assess whether insects produced different numbers of calls across playback treatments. Using the glmer function from the R package lme4, we fit and compared two generalized linear mixed models, both with Poisson link functions [43]. The first model modelled the number of katydid calls using treatment as a fixed effect and individual as a random effect. The second model modelled the number of calls using only individual as a random effect. For one species (Copiphora brevirostris), the models failed to converge, but had a convergence gradient of less than 0.0001 and the parameter estimates were accepted. For each species, we compared the two models using the likelihood ratio test, implemented using the anova function from the base package of the program R [44]. Comparing these two models allowed us to determine whether including the treatment in the model resulted in a significant improvement in fit. If adding treatment to the model improved the fit, we then used the emmeans function from the R package emmeans to conduct a Tukey post hoc test, allowing us to assess differences between the individual treatments [45].

To determine whether katydids habituated to predator playbacks, we also used paired t-tests to determine whether fewer calls were produced after the first predator playback in the series than after the last. Sample sizes are slightly larger in these tests because we were able to include insects that had experienced bat playbacks even if they had not experienced a full set of playback treatments due to power outages or other interruptions.

(c). Electrophysiological recordings of auditory neurons

To determine whether behavioural responses are related to the ability of katydids to detect acoustic stimuli, we recorded neural activity in response to the same acoustic stimuli presented during playback experiments. Specifically, we quantified the activity of TN-1, an ultrasound-sensitive interneuron that is hypothesized to function in bat detection in katydids [46–51]. For an extended description of the stimuli, recording methods and neural preparations, see electronic supplementary material. In brief, the same sound stimuli used in the playback experiments described above (bat calls and white noise) were broadcast at the same intensity (80 dB peSPL) at the katydid's ear. The connective between the prothoracic and suboesophageal ganglia was draped over two extracellular hook electrodes, and neural activity was amplified by a differential amplifier before recording. A microphone (model CM16) was placed behind the katydid to record the sound stimuli on a second channel of the .wav file, allowing us to assign TN-1 action potentials to specific acoustic stimuli.

We compared the number of TN-1 action potentials generated by each acoustic playback. For the bat echolocation call treatment, we calculated the average number of TN-1 action potentials for the three echolocation sequences (1 second sequences of 33 echolocation calls each). For the silence treatment, we counted the number of TN-1 action potentials in 1 s of silence from the start of the neural recording (before acoustic playbacks). For the white noise treatment, we counted the number of TN-1 action potentials in response to the 50 ms stimulus pulse plus 950 ms of silence around the stimulus. Therefore, each measurement was the number of TN-1 action potentials during 1 s.

(d). Measurements of male signalling rate

To determine how often Neotropical katydids produce acoustic signals, we recorded individual insects for 24 h. Individual insects were placed singly into a mesh cage (same type as used for playback experiments) that was placed in an outdoor mesh greenhouse, providing ambient environmental conditions and acoustic exposure from the forest. Each insect was recorded for 24 h, using the internal microphones of a Tascam DR-40 recorder that sampled at 96 kHz. The microphone on this model of recorder has sufficient ultrasonic sensitivity to capture the calls of insects in the high audible and low ultrasonic spectrum. We then processed the recordings using the seewave package in R to detect times when the amplitude exceeded a threshold value and a human reviewer confirmed that only true detections from the focal insect were retained in the dataset. To determine how much sound each katydid produced per day, we multiplied the amount of sound that is produced in the average katydid call for that species by the number of calls detected in 24 h. The amount of sound in the katydid call was calculated as the sum of the durations of the individual sound pulses in the call (figure 1; data from [37]). We quantify the amount of sound per call using this method because some katydid species produce calls with long pulse durations at short intervals over a short period of time, whereas others produce short pulses over long intervals. Therefore, call duration (time from the start of the first pulse to the end of the last pulse) by itself is not an accurate measure of how much sound a katydid produces over time. For one species that was commonly recorded on acoustic monitors in the forest (V. brunneri), calling rates were similar between cages and forest (L.B.S. 2020, unpublished data).

Figure 1.

Examples of the male call and TN-1 responses to acoustic stimuli for the katydid species O. conspersus. (a) Spectrogram (top trace) and oscillogram (bottom trace) of one male call consisting of four pulses of sound (brackets and labelling indicate how pulse durations were measured from the oscillogram). (b) Electrophysiological recording from the neck connective of a katydid (top trace) showing no TN-1 activity in response to silence playback (bottom trace). (c) Action potentials of TN-1 (top trace) in response to a 1 s sequence of echolocation calls of the bat T. cirrhosus (bottom trace); lower figure shows the same recording but zoomed in to the first 60 ms with TN-1 action potentials marked with red dots. (d) Action potentials of the TN-1 (top trace) in response to a 50 ms pulse of white noise (bottom trace) in a 1 s recording; lower figure shows the same recording but zoomed in to the 60 ms around the white noise pulse with TN-1 action potentials marked with red dots. (Online version in colour.)

3. Results

(a). Acoustic playback experiments

Most katydid species continued to call during predator playbacks, with only one of the 11 katydid species (C. wheeleri) calling significantly less during bat echolocation calls than during silence and the white noise treatments. Calling was not affected by the acoustic treatments in five species (L. micans, Microcentrum sp., O. conspersus, S. stalii, V. brunneri; figure 2; electronic supplementary material, table S1). In the remaining species, four (A. festae, A. panamae, C. brevirostris, P. dimidiata) produced significantly more calls during the white noise treatment than the silence or bat call treatments, with no significant difference between silence or bat call treatments (figure 2; electronic supplementary material, table S1). One species (C. discocercus) produced more calls during silence than white noise, but there was no significant difference in the number of calls between bat calls and either of these treatments (figure 2; electronic supplementary material, table S1). The lack of significant decreases in calling during the bat echolocation call treatment was not related to katydids habituating to the stimulus. There was no difference in the number of calls produced after playback of the first compared the last bat call sequence during the 30 min treatment for any of the katydid species (electronic supplementary material, table S2).

Figure 2.

Boxplots of the number of calls produced by male katydids (11 species) during three acoustic treatments lasting 30 min each: bat echolocation calls, silence and white noise. n, number of individuals tested. Different letters above boxplots indicate significant differences from likelihood ratio tests of model fit and subsequent Tukey post hoc tests.

(b). Neuronal responsiveness

The TN-1 neuron is sensitive to bat calls, with the median number of TN-1 action potentials per species ranging from 26 to 49 spikes s⁻¹ for the bat echolocation treatment, and much lower medians for the silence treatment (0–7 spikes s⁻¹) and the white noise treatment (4–17 spikes s⁻¹) (figure 3). The difference in the number of TN-1 action potentials in response to each treatment was striking and similar across all katydid species tested (figure 3). Due to small sample sizes (2–8 individuals per species; figure 3), we did not apply statistical analyses to data on the number of TN-1 action potentials per acoustic treatment.

Figure 3.

Boxplots of the number of TN-1 action potentials of katydids (11 species) in response to three acoustic stimuli: bat echolocation calls, silence and white noise. n, number of individuals tested. Statistical tests were not applied to these data due to the low sample sizes.

(c). Measurements of male signalling rate

Calling rates were highly variable and quite low across all of the focal species, with all but one species producing less than 2 s of sound per night (table 1; electronic supplementary material, figure S1). The timing of peak signalling activity varied across species, but most species produced sound throughout the night. Several of the species, particularly O. conspersus, A. festae and A. panamae, produced numerous calls during the day as well. Sound production was highest in C. wheeleri, the only species that reduced calling in response to echolocation calls, with males producing a mean of 21.9 s of sound per night (table 1).

Table 1.

Total nightly sound production for 11 species of Neotropical forest katydids. The mean duration of sound per call is calculated as the sum of the mean pulse durations for the average number of pulses per species (e.g. figure 1a; data from [37]). The total sound produced per night is the mean duration of sound per call multiplied by the mean number of calls per night for the species.

species	mean calls per night	mean duration of sound per call (s)	total sound produced per night (s)
Arota festae	82	0.008	0.67
Arota panamae	182	0.003	0.56
Chloroscirtus discocercus	27	0.052	1.40
Cocconotus wheeleri	226	0.097	21.90
Copiphora brevirostris	44	0.022	0.98
Lamprophyllum micans	17	0.026	0.43
Microcentrum sp.	86	0.016	1.33
Orophus conspersus	53	0.038	1.99
Phylloptera dimidiata	124	0.006	0.72
Steirodon stalii	16	0.015	0.23
Viadana brunneri	375	0.003	1.09

4. Discussion

Our results suggest that katydids in Neotropical forests rely more on proactive than reactive defences in response to bat echolocation calls. Many species of Neotropical katydids do not alter the number of acoustic signals they produce when exposed to the echolocation sounds of hunting predators, despite having the sensory capability to detect them. Instead, these species demonstrate extremely reduced signalling rates compared to many studied katydid species, making them difficult for eavesdropping gleaning bats to locate [27,29]. The one katydid species that reduced calling during bat echolocation calls (C. wheeleri, family Pseudophyllinae) also produced approximately 20 times more sound per night than any of the other species tested, suggesting a relationship between signalling rate and responsiveness to an ambiguous predator cue. These results support the prediction that, even when they are capable of detecting predator cues, animals will rely on proactive defences when predator cues are not tightly correlated with the risk of attack.

In Neotropical rainforests, where bat predation on insects and katydids is pervasive and intense [28,33,52–54], it is initially surprising that katydids do not stop calling when they hear predators. However, the unresponsiveness of these species makes sense when considered together with information about their acoustic environment. Of the 76 species of echolocating bats on BCI [55], only four have been documented to eavesdrop on the calls of insects [27,39–41], with five additional species suspected to eavesdrop based on their morphology and ecology [55,56]. All of these eavesdropping species belong to the family Phyllostomidae. In an acoustic monitoring survey on BCI, Symes et al. [31] found that less than 4% of the recorded echolocation sequences came from species in the family Phyllostomidae. The vast majority of the echolocation calls recorded by acoustic monitors (and, therefore, available to detection by katydids in the forest) are of aerially hawking bat species, essentially ‘sheep in wolves clothing’ that pose no threat to singing katydids [31,56]. In addition, the family Phyllostomidae, which is endemic to the Neotropics, has diversified to include species with a very broad range of diets (insectivores, carnivores, nectarivores, frugivores and sanguivores: [56–58]). It is unlikely that katydids can distinguish between the calls of frugivorous bats and eavesdropping gleaners, given that the echolocation calls of most of these species cannot be distinguished on the basis of frequency, duration or other parameters [52,56], and given the simplicity of the katydid's auditory system [59]. Combined with the quieter and more directional calls of phyllostomids compared to other bats [42,60], and the tendency for Neotropical eavesdropping gleaning bats to ‘perch hunt’ (hang from a branch and wait for prey cues [61,62]), the number of echolocation calls detected by katydids that actually come from an eavesdropping gleaner becomes vanishingly small. Consequently, while gleaning bats pose a significant threat to singing katydids in the Neotropics [28,33,52,53], echolocation calls are only weakly correlated with the presence of the predator and many species of katydids do not change their calling activity when hearing predator cues. In addition to the echolocation calls of bats, some katydid species themselves produce short ultrasonic calls. Therefore, the calls of other katydid species may further clutter the acoustic space, probably making it more difficult for katydids to distinguish cues of hunting predators [22]. In Neotropical forests, the high probability of false alarms could create substantial selection for enduring the costs and inefficiencies of proactive defences.

Across subfamilies, signal rates were notably low, probably representing a proactive defence against predation. In most habitats, conocephaline and pseudophylline katydid species are known for their conspicuous and regular calling activity (e.g. [57,63–67]). In Neotropical forests, however, many of these species produce short and infrequent acoustic signals [32,33,37,53], with a forest-dwelling conocephaline (C. brevirostris) and a pseudophylline (C. wheeleri) both producing less than 30 s of sound per night. Previous studies have focused on the short and infrequent calls of Neotropical pseudophylline katydids species [27,32,33,52], whereas our 24 h recordings suggest that low duty cycle calling is typical of Neotropical forest-dwelling phaneropterines as well. In a review of the acoustic characteristics of 330 phaneropterine katydid species from around the world, Heller et al. [68] reported a median call duration of 1 s, whereas ter Hofstede et al. [37] calculated a median call duration of only 70 ms for 31 phaneropterine species on BCI.

Although most katydid species did not decrease calling in response to bat calls, our neurophysiological recordings show that their ears are able to detect these predator cues. The TN-1 auditory interneuron always produced many more action potentials during the bat call playback sequences than silence or white noise. The one katydid species that reduced its calling activity in the presence of bat echolocation calls (C. wheeleri) did not have a more active TN-1 neuron than the other species, suggesting that this difference in behaviour is not related to sensory sensitivity to the predator cue, the mechanism that has been demonstrated in other insects and sensory systems [16,69,70]. For example, many insects initiate escape behaviour in response to looming visual stimuli, typical of an approaching predator, and flies with neural circuits that are more sensitive to these looming stimuli initiate escape at greater distances [16]. While high TN-1 activity does not suppress calling behaviour in these species, it remains possible for future investigations to test whether this neuron is providing information about predatory bats when katydids are in flight, as seen in temperate zone katydids [46,51].

One species in our study displayed more reactive anti-predator defences, reducing calling in response to predator playbacks and also producing comparatively high signalling rates. Cocconotus wheeleri, the only pseudophylline in our study, produced approximately 22 s of sound per 24 h. While still low, this was approximately 20 times higher than the amount of sound produced by the other species we studied. In a 2-year study of the diet of a gleaning bat species in Panama, Belwood [52] found that C. wheeleri formed the largest proportion of katydids in the diet (23%). Pseudophylline katydids are the most common subfamily of katydid found in gleaning bat diets [28,33,52] and often inhabit the forest understorey (less than 10 m [71]), a common hunting location for eavesdropping bats [72,73]. For these reasons, pseudophylline katydid species might be more likely to respond to the calls of bats than phaneropterine katydids, even if the calls do not correlate well with danger. In a prior small-scale study investigating call cessation in five Neotropical katydid species in response to echolocation calls, two of the three pseudophylline species stopped singing in response to bat calls, whereas the conocephaline and phaneropterine species did not [74]. Although we cannot draw conclusions about differences between katydid subfamilies based on one species, the strong difference in response to predator cues, the amount of time producing sound and the predation rate on C. wheeleri suggest areas for future research that could reveal trade-offs between mate finding and predator avoidance behaviour across katydids.

Most studies of sexually selected signals focus on species with conspicuous and repetitive sexual signals (e.g. [26,75,76]). Conspicuous signals are often used to advertise male quality, indicating that a given male is capable of surviving and evading predators despite the conspicuous trait [26,77]. However, when males have little option to detect or evade predators, conspicuous signals just make them a conspicuous meal with little opportunity for males to differentiate themselves by superior evasive abilities. In systems characterized by predators that are indistinguishable from non-threatening stimuli, animals may evolve proactive defences and alternative signalling and mating strategies. The mating and communication systems that arise when predator cues are ambiguous may be vastly different from the intensively studied systems where prey species conspicuously advertise their ability to evade predators [78,79]. The katydid species in our study are from three subfamilies (Conocephalinae, Pseudophyllinae and Phaneropterinae), and characteristics of the mating and signalling systems of these subfamilies hint at the types of mating and communication systems that could facilitate reduced calling rates and other proactive defences against predation. Neotropical forest katydid species rely on a diversity of additional modalities and strategies for attracting mates and differentiating themselves from competitors. For example, in Phaneropterinae, it is common for males and females to engage in acoustic duets, a behaviour that may allow them to form pairs while minimizing conspicuousness [68,80]. In a study of 12 congeneric phaneropterine species in Greece, Heller & von Helversen [81] found that in duetting species, males produced fewer calls than non-duetting species. Some conocephaline and pseudophylline species, including focal species C. brevirostris and C. wheeleri [52], produce species-specific spontaneous vibrational signals, potentially circumventing the trade-offs between bat and mate attraction [27,33,82–85]. Using a private channel for communication by producing signals outside the sensory range of a predator has been demonstrated in fish and primates [86–89]. Vibrational signalling is not observed in pseudophylline species from tropical forests lacking eavesdropping gleaning bats [57] and may offer Neotropical katydid species an alternative modality for mate attraction that is private from acoustically orienting eavesdropping bats [27,82] (but see [90]). Males of some species are more likely to call when hearing heterospecifics [91] or white noise, suggesting that they may target their calls to times when there is acoustic cover and their calls are less isolated and conspicuous. Finally, it is common, at least in Phaneropterinae and Pseudophyllinae, for males to present females with a large spermatophore. In the face of intense predation on signalling males, these large spermatophores may represent a reallocation from investment in signalling to investment in direct benefits [92].

Katydids and bats provide an excellent opportunity to assess how predation risk and predation avoidance differ across habitats. To date, few studies have investigated the responses of singing katydids to bat echolocation calls [30,47,74,93,94], making it difficult to compare the results seen here to other species and locations. Understanding the full diversity of anti-predator strategies and trade-offs will require in-depth comparisons within and between katydid subfamilies and locations, along with the integration of emerging phylogenetic information.

The interaction among katydids, eavesdropping bats and other bat guilds highlights how dyadic predator–prey interactions can be influenced by trophically distant members of the community. If the only bats in a habitat were eavesdropping gleaning bats, then echolocation would be a highly informative cue, making it possible for katydids to detect and respond to predator presence with reactive defences. However, when predator cues are diluted by the cues from non-predators, it becomes difficult for prey to anticipate and respond to predators, probably favouring reliance on proactive defences. Most organisms exist in complex communities and it is likely that the ability to detect predators reliably is often compromised by the presence of third-party species. For example, in environments with a loud cicada chorus, it may be much harder for prey to detect the sounds of approaching predators [95]. Similarly, eastern fence lizards flee from invasive predatory fire ants, but generalize this response to native non-predatory ants as well [96]. This type of ecological interaction is currently under-modelled in theoretical community ecology, and understanding the ecology of such systems will require dedicated and consistent efforts to disentangle the complicated interactions that can occur among species. Our study highlights the potential for trophically distant community members to influence the information content of predator cues and the expression of proactive or reactive defences in prey.

Ethics

Research permissions were obtained through the Smithsonian Tropical Research Institute.

Data accessibility

Data are contained in the electronic supplementary Material.

Authors' contributions

L.B.S., S.J.M. and H.M.t.H. designed the study, collected the data and drafted and revised the manuscript. C.E.K. contributed to the framing of the manuscript and revised multiple drafts. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by Dartmouth College, Smithsonian Tropical Research Institute and the Neukom Institute and an Artificial Intelligence for Earth Innovation grant from Microsoft/National Geographic (NG5-57246T-18).