Evolutionary escalation: the bat–moth arms race

As a well-known predator–prey model system, the interactions of bats and moths have long fascinated scientists (Roeder, 1967), university students (Alcock, 2013) and children alike (Oppel, 1997). Insect-eating bats detect and track moths using echolocation and attempt to capture and consume these insects in flight. In turn, some moths listen for bat echolocation calls to avoid being captured. Although the phrase ‘co-evolutionary arms race’ is often used to describe the moths' anti-bat defences (which include bat-detecting ears, sound-evoked defensive flight and sound production), there is no evidence of reciprocal adaptations between bats and moths at the species level, nor are the causes and effects of this arms race exclusive to the interactions between bats and moths. Most adaptations that make bats better moth hunters also make them more effective hunters of other insects, and such adaptations are therefore not moth specific. Likewise, some moths use their ears to detect not only bats but also insect-eating birds or mates (Conner, 1999; Nakano et al., 2015).

The evolutionary histories of predator and prey also differ. The order Lepidoptera (moths and butterflies) originated ∼150 million years ago (mya; Misof et al., 2014), long before the origin of bats, which first took to the wing sometime between 60 and 95 mya (Bininda-Emonds et al., 2007). Although powered flight is a defining characteristic of bats, laryngeal echolocation may have evolved only ∼50 mya (Simmons et al., 2008; Teeling, 2009; Veselka et al., 2010). The distinctive combination of echolocation and powered flight allowed bats to first exploit, then dominate the foraging niche of night-flying insects. Bat predation drove selection for both bat-detecting ears and non-auditory means of avoiding bats in a variety of insects, including moths and butterflies that never evolved ears and remain deaf to bats (Fullard and Napoleone, 2001; Barber et al., 2015).

The earliest adaptations in the arms race, laryngeal echolocation in bats and tympanal ears in moths, were not necessarily singular events – possibly not in bats, certainly not in moths. For bats, there are two equally parsimonious hypothetical time lines for the origination of laryngeal echolocation; it evolved either once (and was subsequently lost in the Pteropodidae) or twice (once in each of the two suborders Yinpterochiroptera and Yangochiroptera; Fig. 1). Ears have evolved numerous times independently in divergent lepidopteran groups, with bats as the apparent driver in every instance. Today, some lepidopteran lineages remain earless, having evolved other defences against bats. Interestingly, bats, which originated after insect-eating birds, may have driven some previously nocturnal lepidopteran lineages, including some extant butterflies (Yack and Fullard, 2000) and tiger moths (Ratcliffe and Nydam, 2008), into being diurnal.

The interactions of bats and moths do not meet a strict definition of co-evolution in which two species demonstrate reciprocal adaptations (Janzen, 1980; Jones and Rydell, 2003; Jacobs et al., 2008; Ratcliffe, 2009). Co-evolutionary relationships are most common between host and parasite species or mutualistic partners such as plants and pollinators, but are rarely observed between predator and prey. Dawkins and Krebs (1979) suggest that the rarity of co-evolution between predators and prey is due to the ‘life–dinner principle’ – the cost of a mistake being death for the prey versus only a lost meal for the predator – resulting in more specific antipredator defences than predator counter-strategies. Here, we concentrate on the traits of lepidopteran species that reduce their risk of being attacked by bats, emphasizing moths' ears, auditory-evoked evasive flight and defensive sounds. We also review auditory-based defences in other insect groups. Lastly, we discuss the evolutionary refinement of bat echolocation and related adaptations that have allowed bats to better exploit insects at night, including, but not limited to, moths with bat-detecting ears. In summary, we feel the relationship between bats, moths and other night-flying insects continues to represent an ideal system to improve our understanding of the sensory ecology and neuroethology of predator–prey interactions.

The evolution of echolocation in bats set the stage for the vast diversity of ultrasound (>20 kHz) sensitive ears found in modern insects. As an active sensing system using high-energy acoustic signals, echolocation provides potential information to would-be prey. High-frequency-sensitive ears and auditory-evoked avoidance behaviours have evolved in at least six orders of insects (Yack and Dawson, 2008): Orthoptera (crickets, katydids and locusts), Diptera (some tachinid flies), Dictyoptera (mantids), Coleoptera (some tiger and scarab beetles), Neuroptera (green lacewings) and Lepidoptera (moths and butterflies) (Fig. 2A, Table 1). Within the Lepidoptera alone, ears have evolved independently at least six times (Fig. 2A, Table 1).

In insects with bat-detecting ears, auditory organs apparently evolved from chordotonal organs, which are sensory organs specific to insects and crustaceans that contain neurons sensitive to body movements (Kavlie and Albert, 2013). Chordotonal organs are dispersed throughout the body, explaining why ears are found in many different locations in insects (Fullard and Yack, 1993; Yack and Dawson, 2008; Fig. 2A). Insect ears that evolved from chordotonal organs consist of a tympanum (a membrane that vibrates in response to sound) and auditory receptor cells activated by sound-induced tympanal vibrations (Fig. 2B,C). Given the intense sounds produced by echolocating bats (Waters and Jones, 1995; Holderied and von Helversen, 2003; Holderied et al., 2005; Surlykke and Kalko, 2008), and thus the ability of these sounds to vibrate cuticle, it is easy to imagine an evolutionary progression from a chordotonal organ measuring the movements of particularly thin areas of the exoskeleton to the ears of modern insects (Yack, 2004; Yack and Dawson, 2008). Although some insects have non-tympanal ears, none of these ears are known to function in bat detection because they are insensitive to high-frequency sounds (Yack and Dawson, 2008). The ears of noctuoid moths are the best studied among the numerous families of moths with ears.

Despite the common use of ears for bat detection among the groups listed above, the original selection pressure for hearing varies. Fossil evidence shows that hearing evolved in orthopterans for intraspecific communication before the appearance of bats (Plotnick and Smith, 2012). Tachinid flies that listen for the calling songs of their cricket hosts also had ears before bats evolved (Rosen et al., 2009). For both of these groups, hearing was co-opted for bat detection in addition to still performing its original function. The original selection pressure for ears in some beetles is unclear: the ear is well suited for bat detection and ultrasound triggers evasive flight, but many species also produce sounds themselves, possibly for intraspecific communication (Yager and Spangler, 1995, 1997).

For mantids, green lacewings and moths, it had been assumed that hearing evolved in response to echolocating bats, because bat detection is the primary and often sole function of these ears today (Yack and Dawson, 2008). A recent paper using molecular methods, however, suggests that mantids had ears before bats evolved (Yager and Svenson, 2008). Mantid hearing may have first functioned in detecting other types of danger-indicating acoustic stimuli, and might then have been co-opted to detect bat echolocation calls. Although the evidence that bats drove the evolution of hearing in moths is strong, these surprising results from mantids make it clear that further molecular studies are needed to confirm whether the original function of moth and lacewing ears was indeed bat detection.

What is the evidence that moth ears evolved in response to echolocating bats? First, bat avoidance is the sole function of ears in most moth species (Fullard, 1988), and sound-triggered evasive flight provides a survival advantage for moths under attack by bats (Roeder and Treat, 1962; Acharya and Fenton, 1999; Fullard, 2001). Other uses, such as detecting insect-eating birds or intraspecific communication, are likely to be derived (Conner, 1999; Jacobs et al., 2008; Fournier et al., 2013). That said, ever more moth species are being reported to use sound for communication, and this function might be much more common than previously realized (Nakano et al., 2009). Second, the relative insensitivity of moth ears suggests that they are adapted to detecting high-intensity ultrasound sources, such as bat calls (Waters and Jones, 1995; Holderied and von Helversen, 2003; Holderied et al., 2005; Surlykke and Kalko, 2008), rather than faint ambient sounds like crackling vegetation (Fullard, 1988). Third, moth ears are typically tuned to the frequencies of echolocation calls used by their sympatric bat community and not to other environmental sources of sound (Fullard, 1988, 1998). The auditory receptors also quickly exhibit sensory adaptation to constant sound, but remain sensitive to pulsed sounds like bat calls (Fullard et al., 2008a). Fourth, there is a strong correlation between the sensitivity of moths' ears and their expected exposure to bats (ter Hofstede et al., 2008a).

Moth ears consist of a tympanum (see Glossary) that vibrates in response to sound and 1–4 auditory receptor neurons (depending on the species) that are activated by this movement (Fig. 2B,C; Box 1). Regardless of the number of auditory afferents in different moth taxa, each afferent has the same frequency tuning as the others, meaning that moths are not capable of frequency discrimination (Fullard, 1988, 1998). Each afferent, however, has different sound level thresholds (Fig. 3A; Roeder, 1967). The difference between the most sensitive receptor and the next most sensitive receptor is ∼2–6 dB in pyralids (Skals and Surlykke, 2000), 15 dB in geometrids (Roeder, 1974a; Surlykke and Filskov, 1997) and 15–20 dB in noctuids (ter Hofstede et al., 2013). These differences in threshold between afferents increase the range of amplitudes to which the ear can respond by allowing the less sensitive afferent to continue encoding increasing sound amplitude once the most sensitive afferent has reached its maximum firing capability. This allows for a very large dynamic range for intensity discrimination of 40–60 dB (Roeder, 1967; Coro and Perez, 1993), which should provide valuable information about the relative distance of an approaching bat.

Despite the convergence of form and function in moth ears, there are differences between moth species and taxonomic families that could provide insight into the evolutionary history between bats and moths. For example, the lowest threshold of the most sensitive receptor can range from 20 to 60 dB across different species of noctuids and notodontids (Surlykke, 1984; Surlykke et al., 1999; Fullard et al., 2008b; ter Hofstede et al., 2008a). By comparison, bats have much more sensitive ears, with hearing thresholds similar to those of humans, at 2–20 dB sound pressure level (SPL; Moss and Schnitzler, 1995). The enormous 40 dB range in hearing thresholds across moth species has important behavioural and ecological consequences: sensitivity determines the distance at which moths can detect an oncoming bat, and thus the time available for escape. Below, we consider the selection pressures that could have contributed to, or resulted from, variation in moths' hearing ranges and sensitivity.

Determining cause and effect in relationships between moth auditory sensitivity and behavioural ecology is difficult. Here, we present these relationships with no assumptions about which came first, variation in auditory sensitivity or variation in life history. Regardless, we must consider the costs that select against ever more sensitive high-frequency hearing. For a moth, reacting to non-threatening environmental sources of ultrasound, like crackling vegetation (Fullard, 1988) or bats that are too far away to be a viable threat, could result in missed mating or feeding opportunities, or even more dire consequences, such as being exposed to non-aerial predators as a result of evasive diving (Guignion and Fullard, 2004). In addition to such costs, species size, nocturnal activity level, phenology and aspects of the sympatric bat community are all correlated with moth auditory sensitivity – each of these aspects of life history is considered below.

Bats can detect larger moths at greater distances than they can detect smaller moths. Larger moths, however, have more sensitive ears (Surlykke et al., 1999) and detect bats at greater distances than do smaller moths, essentially compensating for their increased conspicuousness. This relationship between size and sensitivity is strongest for frequencies that are used by a large number of sympatric bats for echolocation (ter Hofstede et al., 2013). Forrest et al. (1995) suggest that this pattern of size and sensitivity may apply across all insects with bat-detecting ears. Römer et al. (2008), however, show that there is no relationship between size and high-frequency sensitivity in phaneropterine katydids, which produce high-frequency communication signals.

Moth species can vary in the sensitivity or tuning of their hearing depending on their exposure to bats, either temporally or geographically. Where there is sexual dimorphism in flight ability (flightless females, flight-capable males), males are at greater risk of bat predation, and females have less sensitive ears than males (Cardone and Fullard, 1988; Rydell et al., 1997). Interestingly, this relationship is also found in mantids (Yager, 1990). At a finer-grained level, for moth species in which both sexes fly, males might still be at greater risk of bat predation if they fly more at night in search of perched pheromone-releasing females. Although male-biased moth predation by bats has been observed around streetlights (Acharya, 1995), a male-biased flight-to-light response also exists in some moths (Altermatt et al., 2009). In general, for moth species in which both males and females fly, there is no evidence for differences in sensitivity between male and female moths (Surlykke and Gogala, 1986; Surlykke and Treat, 1995; Skals and Surlykke, 1999, 2000; Fullard, 2006). Sex and size aside, moth species with less sensitive ears fly less at night (ter Hofstede et al., 2008a). Thus, although less sensitive moths experience greater risks in terms of their proximity to bats before initiating evasive flight, they compensate for this cost by reducing their exposure to bats (ter Hofstede et al., 2008a).

Diurnal moth species either do not have ears or have less sensitive ears than nocturnal moths (Fullard et al., 1997, 2000; Surlykke et al., 1998; Muma and Fullard, 2004). Similarly, in butterflies, ultrasound-sensitive ears are only found among nocturnal species (Yack and Fullard, 2000), whereas some diurnal butterflies have ears that are sensitive to low-frequency sounds for communication or hearing avian predators (Yack et al., 2000; Lane et al., 2008; Lucas et al., 2009). Actograms of flight activity have revealed that many moths that were considered diurnal are also active at night, explaining the maintenance of sensitive ears in these species (Fullard and Dawson, 1999; Fullard and Napoleone, 2001). Moths that are active during the day often have greater sensitivity to sonic (<20 kHz) frequencies (Fullard et al., 1997; Surlykke et al., 1998), perhaps as a derived defence to detect the rustling sounds of birds (Jacobs et al., 2008).

Seasonal differences in moth activity also correlate with ultrasound sensitivity. Moths that emerge in early spring before bats are active have higher hearing thresholds (Fullard, 1977; Fullard and Barclay, 1980). Yager et al. (2000) show a similar seasonal pattern for hearing in beetles. Winter moths also have higher thresholds, but only if they are effectively isolated from bats (Surlykke and Treat, 1995; Svensson et al., 1999; Rydell et al., 1997). Defensive sound production in tiger moths (see ‘Sound production in moths: defence, courtship and sexual antagonism’, below) follows the same pattern: species emerging in spring, when bat activity is low, have lost the ability to produce sound, whereas those emerging in summer, when bat activity is high, have not (Fullard, 1977; Fullard and Barclay, 1980; Ratcliffe and Nydam, 2008).

Geographic variation in moth auditory sensitivity often corresponds with the number of bat species in an area (Fig. 3B). The approach of comparing auditory tuning in moths to the relevant ‘echolocation assemblage’ in which they live has been reviewed previously (Fullard, 1988, 1998). Moth species in areas with bat echolocation assemblages covering a wide frequency range have more sensitive ears that are more broadly tuned than those in areas of low bat diversity (Fullard, 1982a; Fullard et al., 1983). In the Neotropics, where there are many fruit-eating bats, the tuning of moth ears shows better congruence with the insect-eating bat echolocation assemblage than with the bat community as a whole (Fullard, 1988). In addition, moths with extremely sensitive ears tend to be found in areas of high bat predation pressure (Fullard, 1982a; Fullard et al., 2008b). In some cases, particular bat species seem to have an impact on the tuning of local moth ears. For example, the ears of Hawaiian moths have a much narrower tuning curve than those of moths found elsewhere, and they are closely tuned to the echolocation frequency of the one bat species in Hawaii, Lasiurus cinereus semotus (Fullard, 1984a). Horseshoe bats, if present, usually have the highest frequency echolocation calls in a community, and their presence appears to select for improved higher frequency hearing in local moth populations (Jacobs et al., 2008; ter Hofstede et al., 2013) and several Old World mantis species (Triblehorn and Yager, 2001).

Studies have found some evidence of neural regression in moth species that live where bats do not. Moths endemic to bat-free French Polynesia are less sensitive to sounds >25 kHz compared with more recent immigrants (Fullard, 1994; Fullard et al., 2007). This high-frequency deafness might reflect neural regression (Fullard et al., 2007) or thickening of the tympana (Ratcliffe, 2009). The behavioural response to ultrasound is also lost in most of these endemic species (Fullard et al., 2004; Fig. 3B). However, this pattern was not found for bat-free arctic locations, possibly because these moth populations are not isolated from gene flow from southern populations or have not been isolated for long enough for changes to occur (Surlykke, 1986; Rydell et al., 2000).

Given the well-supported hypothesis that the original and primary purpose of moths' ears is to detect predatory bats, we should expect to see adaptive behavioural responses to ultrasonic sounds in moths with ultrasound-sensitive ears. Many studies have shown evasive behaviours by moths in response to bat-like sounds (Table 1). These evasive responses are usually graded, meaning that the responses change in a quantitative or qualitative manner with increasing sound amplitude. Almost all other insects with bat-evasive behaviours also show this pattern of graded response, changing from small alterations in flight direction or speed to drastic, last-ditch behaviours as sound amplitude at their ears increases (green lacewings: Miller and Olesen, 1979; crickets: Nolen and Hoy, 1986; mantids: Yager et al., 1990; beetles: Forrest et al., 1995; some katydids: Schulze and Schul, 2001; tachinid flies: Rosen et al., 2009). This is a highly adaptive pattern for bat avoidance. Many insects can detect bats before the bats can detect the echoes reflected from the insect (Surlykke, 1988; Surlykke et al., 1999; Goerlitz et al., 2010; Fig. 3C), and under these circumstances, directional flight or small changes in flight trajectory should allow the insect to get out of the path of the approaching bat before it is detected (Surlykke, 1988; Surlykke et al., 1999; Goerlitz et al., 2010). Insects with two ears might use differences in the sound arrival time or intensity at each ear as information about the direction of the sound source (Payne et al., 1966). Most of these studies have investigated responses to ultrasound in tethered animals, and it is not entirely clear whether or how some of these graded behaviours translate into differences in behaviour in free flight.

If a bat makes a sudden change in flight direction, a flying insect might find itself confronted by a hunting bat ready to attack. Similarly, some bats will continue to track evasively flying insects, rather than give up. In such situations, the insect may take advantage of its greater manoeuvrability compared with that of the larger bat (Ghose et al., 2009), and behaviours such as loops, dives, zig-zigs and spirals can allow the insect to drop out of the bat's sonar beam. Directional flight away from low-amplitude ultrasonic pulses and last-ditch manoeuvres in response to high-amplitude ultrasonic pulses have been observed in noctuids, geometrids and notodontids (Roeder, 1962, 1964; Agee, 1967, 1969; Surlykke, 1984; Rydell et al., 1997; Svensson et al., 1999). Pyralids (Table 1) probably have a graded response as well; they dive in response to loud ultrasound, but flight responses to low-amplitude ultrasound have been difficult to classify (Skals and Surlykke, 2000; Rodríguez and Greenfield, 2004). Graded responses have also been observed for insects that lack directional hearing (e.g. mantids and sphingids), with low-amplitude ultrasound eliciting turning and higher amplitudes eliciting flight cessation (Yager et al., 1990; Göpfert and Wasserthal, 1999a).

The behavioural responses of moths to bat calls are known to occur at lower sound amplitudes when sound pulses are broadcast at biologically relevant repetition rates, i.e. those used by bats during their search for and approach to prey, rather than isolated sound pulses (noctuids: Agee, 1969; Roeder, 1964; arctiids: Fullard, 1984b; pyralids: Skals and Surlykke, 2000). This pattern also holds for other insect groups with bat-detecting ears (lacewings: Miller and Olesen, 1979; crickets: Nolen and Hoy, 1986; mantids: Triblehorn and Yager, 2005). In addition, moths with ears will fly less at night when subjected to the playbacks of bat calls than they do under silent conditions (Fullard et al., 2003; Ratcliffe et al., 2008). Together, these studies show that insect responses to ultrasound are fine-tuned to the predatory cues of bats.

Erratic flight is a pervasive strategy for predator avoidance for many flying insects, providing defence against birds (Jacobs et al., 2008; Fournier et al., 2013) and bats (Roeder, 1967; Acharya and Fenton, 1999) alike. Compared with eared moths, earless moths rely more on passive defences, such as flying less at night (Morrill and Fullard, 1992; Fullard and Napoleone, 2001; Soutar and Fullard, 2004), closer to vegetation (Lewis et al., 1993; Rydell, 1998), more erratically (Lewis et al., 1993; Rydell and Lancaster, 2000) or at times when bats are not active (Yack, 1988; Morrill and Fullard, 1992; Lewis et al., 1993). Hepialis humuli is an earless moth that performs short visual flight displays at dusk, thereby reducing its exposure to birds and bats (Andersson et al., 1998). Although these moths exhibit this and all the other passive defences listed above for earless moths (Rydell, 1998; Rydell and Lancaster, 2000), the bat Eptesicus nilssonii can apparently overcome these defences by emerging early in the evening and using vision to locate these large white moths (Eklöf et al., 2002).

With the appearance of bats, different moth species took parallel paths to bat avoidance: reactionary auditory-evoked evasive flight in eared moths and preventative continuous erratic flight in earless moths. The evolution of ears might have released some moths from the necessity of inefficient erratic flight, allowing for a more efficient straight flight path, only interrupted by erratic flight in the presence of bats.

Until now, we have considered only passive bat avoidance behaviours in earless insects (e.g. meandering flight) and two forms of active bat avoidance in eared moths: auditory-evoked negative phonotaxis (i.e. moving away from a sound source; see Glossary), which occurs in response to distant bats, and erratic flight behaviours, which occur in response to nearby bats. Tiger moths and their allies (family Erebidae, subfamily Arctiinae) are closely related to the noctuids and lymantriids, and, as expected based on their phylogenetic position, possess a pair of ears with two auditory afferents (the A1 and A2 cells found in noctuids; Box 2, Figs 2B, 3A). Chemical defence, by way of compounds produced de novo or sequestered from plants during the caterpillar stage, and high-frequency, high-energy sound production are thought to be ancestral characteristics of this clade (Weller et al., 1999; Nishida, 2002). Tiger moth species that produce sound do so using two tymbals (see Glossary) found on opposite sides of the thorax. These produce a train of clicks as they buckle in and out. Tiger moths produce these sounds in response to tactile stimulation, bat echolocation calls or both. The timing of the clicks varies within and between species, but most sound-producing species appear to time their clicks to the echolocation pulses of nearby bats in the approach phase of attack (Fullard, 1984b; Ratcliffe and Fullard, 2005; Barber and Conner, 2006).

Sound production in tiger moths has long been recognized to have a defensive function against bats, deterring experienced bats from completing their aerial hawking attacks (see Glossary) in both the field and flight room (Acharya and Fenton, 1992; Dunning et al., 1992; Hristov and Conner, 2005; Ratcliffe and Fullard, 2005). Three hypotheses have sought to explain how tiger moth clicks deter bats: (i) by startling naive bats, (ii) by jamming bat echolocation and (iii) by acting as warning signals, reminding experienced bats of the moths' chemical defences (Hristov and Conner, 2005; Ratcliffe and Fullard, 2005; Ratcliffe, 2009; Corcoran et al., 2010). We argue that startle and jamming should be considered together as auditory interference, where startle effects of the tiger moth clicks diminish with bats' experience, and, regardless of experience, tiger moth species able to produce greater numbers of clicks per unit time are better able to interfere with the ranging abilities of attacking bats (in other words, ‘jamming’ bat echolocation).

Moths of the palatable species Bertholdia trigona are at the extreme of this tymbal click rate continuum, producing so many clicks that they inarguably must interfere with an attacking bat's target ranging capabilities. Cycnia tenera and Euchaetes egle, conversely, produce roughly the same number of clicks per unit time and far fewer clicks than B. trigona, and, as follows, these clicks are less likely to fall within the 2 ms window at the bat's ear when it is expecting target echoes (<20% of the time for C. tenera; Ratcliffe and Fullard, 2005). Supporting this contention, behavioural data from naive bats suggest that clicks, in and of themselves, deter bats ∼20% of the time in palatable E. egle and >90% of the time in B. trigona (Hristov and Conner, 2005; Corcoran et al., 2009). Further, when the clicks of chemically defended C. tenera fall within this 2 ms window, they do interfere with echolocation in experienced bats (Ratcliffe and Fullard, 2005). Sound production is not limited to tiger moths, and new evidence suggests that some sphinx moths also deter bats by producing many clicks per unit time with a stridulatory organ (see Glossary; Barber and Kawahara, 2013; Kawahara and Barber, 2015).

During aerial hawking attacks, tiger moth clicks can act as warning signals for experienced bats. We have suggested previously that this third hypothesized mechanism of bat deterrence (‘acoustic aposematism’) may be related to the effectiveness of the clicks as interference signals (Ratcliffe and Fullard, 2005). Specifically, we suggest that the more high-frequency clicks produced per unit time, the more negative and salient the warning signal. To illustrate, with respect to food flavours, the more negative and salient the signal (e.g. bitter or sour), the more readily the signal is associated with a negative consequence (e.g. gastrointestinal malaise; reviewed in Ratcliffe et al., 2003). If this is also the case in the auditory domain, tiger moths producing more clicks per unit time may be more effective at training individual bats to avoid them in the future (Ratcliffe and Fullard, 2005). At present, the inter- and intra-specific relationships, if any, between palatability and signal complexity are unknown, and these relationships are complicated by mimics – sympatric sound-producing tiger moths that are more palatable than their putative models (Barber and Conner, 2007).

As seen in noctuid moths, the context in which tiger moths detect bats can influence decision making in these species. In the dogbane tiger moth, C. tenera, anti-bat sounds are initially produced in equal numbers to bat-like sounds representing a searching bat (low risk) or an attacking bat (high risk). Over time, however, the moths habituate more (i.e. produce fewer sounds) to the sounds of a low-risk searching bat than to those of an attacking bat, reflecting an adaptive response to different levels of risk. These changes in behaviour are not due to sensory adaptation or muscle fatigue, and thus are the result of changes at the central nervous system level (Ratcliffe et al., 2011a).

In addition to this defensive function, some tiger moths and noctuids use sounds during courtship. Although we will not describe this additional function of sound production in detail, we mention it here because we believe it is an interesting extension of the bat–moth story. For example, males of the pyralid moth Ostrinia furnacalis produce sounds to elicit an anti-bat response (freezing) so they can forcibly mate with perched females, whereas females of the arctiid moth Eilema japonica can distinguish between bat and male calls and only take defensive action to bat calls (Nakano et al., 2013). Producing sounds for mating purposes also comes at the potential cost of attracting eavesdropping bats, as is seen in the pyralid moth Achroia grisella (Alem et al., 2011). Generally considered rare in moths, acoustic communication for mating has been discovered in more and more moth species in recent years, suggesting that it is far more common than previously appreciated, but was previously overlooked in some species because of the low intensity of the signals (Nakano et al., 2008, 2009). For reviews of intraspecific acoustic communication in moths, see Conner (1999) and Nakano et al. (2015).

The case for bats having evolved counter-measures against eared prey is not nearly as strong as the evidence supporting bat predation as the selection pressure for hearing in many insect groups. The adaptive radiation of bats into the ∼20 families and >1300 species seen today occurred ∼50 mya, after laryngeal echolocation had evolved in the ancestral bat (Teeling et al., 2005; Teeling, 2009; Fenton and Ratcliffe, 2010). More than 1000 extant bat species are laryngeal echolocators (Ratcliffe et al., 2013). This diversification was due to the success of bats in exploiting the unoccupied foraging niche of night-flying insects (Griffin, 1958). Although the first bats might not have echolocated (Simmons et al., 2008; Veselka et al., 2010), they most likely ate insects, using passive listening to detect prey on the vegetation and the ground (Simmons and Geisler, 1998). As powered flight further developed, so too, we expect, did echolocation and the ability to track airborne prey (Ratcliffe et al., 2011b, 2013). Emerging hundreds of millions of years after the appearance of insects (Bininda-Emonds et al., 2007; Misof et al., 2014) and ∼100 million years after the origin of the Lepidoptera (Bininda-Emonds et al., 2007; Misof et al., 2014), the first echolocating bats would have encountered many night-flying insects that were defenceless against aerial attacks. Presumably, these insects would, at first, have been easy prey compared with modern night-flying insects.

Aerial hawking was and remains the primary selective pressure for the evolution and maintenance of bat detection and defence in insects (Ratcliffe and Fullard, 2005; Ratcliffe, 2009). Although many bat species take prey from water and terrestrial surfaces, most of these species also capture prey in flight (Ratcliffe et al., 2006; Ratcliffe, 2009). Over time, airborne insects, through the evolution of bat-detecting ears and auditory-evoked defensive flight behaviours and passive anti-bat behaviours (both discussed above), would in general have become more difficult to track and capture, providing selective pressure for bat biosonar systems to better track would-be prey.

Here, we review the evidence for adaptive counter-strategies in bats against eared moths. We propose two criteria for assessing whether a given bat species or clade evolved adaptations specifically to circumvent bat-detecting ears in insects. The first criterion is met if the bat species eats a significantly larger proportion of eared insects than is available in the environment. This is frequently estimated using the proportion of moths in the diet (Box 3). This level of evidence can be considered as a ‘low bar’ that represents only weak support for the existence of counter-adaptation, given that eared moths fly relatively slowly, are soft-bodied and thus should be especially attractive to bats as compared with hard-shelled beetles, fast-flying sphinx moths, measly flies (e.g. mosquitos) and miniscule chironomids. The second ‘high bar’ criterion, which provides a higher level of support for bat counter-adaptation, requires that the putative bat counter-measure against insect hearing has no plausible alternative explanation for its origination and maintenance.

The majority of eared insects are most sensitive to sounds within the range of most bat echolocation signals, between 20 and 50 kHz (Fullard, 1998; Ratcliffe, 2009). Therefore, one potential counter-strategy for bats against insect hearing, referred to as the allotonic frequency hypothesis, would be to produce either very high- or very low-frequency echolocation calls outside the best hearing range of most insects. A handful of species, including several molossids and at least one vespertilionid – the spotted bat, Euderma maculatum – are known to produce exceptionally low-frequency echolocation calls (Fullard and Dawson, 1997). Euderma maculatum is a small- to medium-sized bat. Based on the relationship between size and echolocation call frequency that is well established among vespertilionid species (Jones, 1999), E. maculatum would be expected to use echolocation calls with maximum energy around or above 30 kHz. Instead, they produce narrowband calls with peak frequencies of ∼10 kHz, outside the sensitivity range of sympatric eared moth species (Fullard and Dawson, 1997). Genetic diet analyses have not been conducted, but faecal analysis suggests that these bats consume many more moths than would be predicted from moth abundance alone (Painter et al., 2009).

Although E. maculatum and some of the molossids with low-frequency calls (e.g. Eumops spp.) satisfy our criterion for the low bar, they do not unequivocally cross our high bar for evidence of counter-adaptation. This is because the use of very low-frequency echolocation calls has other advantages for the species that produce them. All of these species are fast-flying, open-space aerial hawkers. Low frequencies attenuate less rapidly in air (Griffin, 1971; Lawrence and Simmons, 1982) and, all else being equal, these bats should enjoy the advantage of greater detection ranges (reviewed in Ratcliffe, 2009).

At the other extreme of the echolocation call frequency spectrum is the little-studied African hipposiderid, Cleotis percivalis, which produces echolocation calls with maximum energy at a frequency higher than any animal ever recorded (∼212 kHz: Fenton and Bell, 1981); these signals are almost certainly inaudible to even the most sensitive moth species throughout the entire pursuit (Fullard, 1998; Ratcliffe, 2009). Owing to the extreme atmospheric attenuation of such frequencies, however, the detection range of C. pervalis will be very short, imposing a cost for orientation and prey detection, and suggesting that this extreme frequency might have evolved to circumvent auditory detection by insects. However, using high-frequency calls can also provide the benefit of more-detailed information about one's immediate surroundings, and therefore only fulfils the low bar criterion as a counter-adaptation to insect hearing.

Among aerially hawking laryngeal-echolocating bats, there are those that produce short-duration, often frequency-modulated calls and separate call and echo in time to estimate the position of prey and other objects, and those that produce longer-duration calls of constant frequency and separate call from echo using frequency differences generated by Doppler shifts (Neuweiler, 1990; Fenton et al., 2012). Bats from the first group, called frequency-modulating or low duty cycle (LDC) bats, are found in both suborders and have a worldwide distribution. Bats in the second group, called constant-frequency or high duty cycle (HDC) bats, are, with the exception of a few mormoopid species, found only in one suborder and confined to the Old World. Relative to LDC bats of the same size, HDC bats tend to use higher peak frequencies, and, as a result, have been proposed to circumvent the auditory defences of moths. Among communities of HDC bats, those that use the highest frequencies consume the most moths (Schoeman and Jacobs, 2003). However, owing to the increased sensitivity of moth ears to longer-duration sounds (Tougaard, 1996, 1998; Ratcliffe et al., 2009), any advantage conferred to HDC bats by using higher frequencies appears to be nullified by their very long duration calls (Jacobs et al., 2008). HDC bats are able to detect fluttering insect wings (through acoustic glints) and distinguish them from background echoes (e.g. those from vegetation); LDC bats cannot. For this reason, insects flying close to clutter, whether eared or earless, may be more susceptible to capture by HDC bats.

During an aerial attack, bats produce calls more rapidly as they approach prey. A feeding buzz (i.e. echolocation call rates of >100 calls s⁻¹, and ≥200 calls s⁻¹ in some species; see Glossary) has been described for almost all bats studied when taking airborne prey, a rare exception being the primarily substrate gleaning bat (see Glossary) Megaderma lyra from the ancient, species-poor family Megadermatidae (Schmidt et al., 2011). During the terminal sub-phase of a feeding buzz (often referred to as buzz II), species in the sister LDC bat families Vespertilionidae and Molossidae drop the peak frequency of their echolocation calls by an octave and, consequently, roughly double the breadth of their sonar beam (Jakobsen et al., 2013; Ratcliffe et al., 2013). HDC rhinolophids, through an unknown mechanism, also broaden their beam during the buzz (Matsuta et al., 2013). For all hawking species, we suggest that the buzz evolved to allow better tracking of ever more evasive prey (Elemans et al., 2011; Ratcliffe et al., 2013; Hulgard and Ratcliffe, 2016). For vespertilionids and molossids, the octave drop of buzz II might represent a further counter-adaptation to track evasively flying insects, including, but not limited to, eared moths and other insects with bat-detecting ears (Jakobsen and Surlykke, 2010; Ratcliffe et al., 2013).

When flying in open space, bats are far louder than previously recognized, producing signals between 120 and 140 dB SPL when measured at 10 cm from the bat's mouth (Surlykke and Kalko, 2008). Bats therefore have much longer detection ranges than previously appreciated but are detected by eared prey at even greater distances. This occurs because, although bats' ears are more sensitive than those of insects, prey echoes returning to the bat have travelled – and attenuated – over the distance both to and from the insect, and consequently contain much less energy than the signal reaching the insect (Fig. 3C). Through substrate gleaning of prey from terrestrial surfaces, some bats might circumvent eared insects' defences. Gleaning is likely to be the ancestral condition of predatory bats, and has independently evolved many times since (Simmons and Geisler, 1998). Extant gleaning bats – and those which both glean and hawk prey – use echolocation calls of much lower intensity than hawking bats, and are relatively inaudible to eared moths (Faure et al., 1990, 1993). The presence of gleaning bats might not select for more sensitive ears in moths if the risk of being eaten by a gleaning bat is less than the risk of never mating owing to unnecessary evasion of aerially hawking bats that are too far away to pose a threat.

There is scant evidence that gleaning bats take more eared prey than would be expected by chance, or that low-intensity echolocation calls evolved in the context of gleaning to circumvent the ears of moths or other insects. Bats use lower-intensity echolocation as they approach objects – whether to land, fly by or take prey – to prevent self-deafening from intense echoes and reduce the costs of producing high-intensity sounds when they are unnecessary to produce audible echoes (Kick and Simmons, 1984; Hartley, 1992).

The only bat thus far that meets both of our criteria for a counter-strategy to insect hearing is the hawking vespertilionid, Barbastella barbastellus (Goerlitz et al., 2010). When searching for prey, B. barbastellus produces echolocation calls of much lower intensity (∼94 dB SPL) than those of the sympatric vespertilionid Nyctalus leisleri (∼127 dB SPL). Both bat species produce calls with peak frequencies within the best hearing range of sympatric moths (B. barbastellus: 33 kHz; N. leisleri: 28 kHz). Owing to its low-intensity calls, however, B. barbastellus is not detected by noctuid moths until the bat is ∼4 m away. Conversely, N. leisleri can be detected at >30 m. A stealth approach has benefits and costs; B. barbastellus is the only bat species known that should detect eared moths before the moths detect them (Fig. 3C). However, the prey detection range of B. barbastellus is only ∼5 m, whereas that of N. leisleri is ∼20 m. Molecular diet analyses of these two species reveal that ∼85% of the diet of B. barbastellus consists of eared moths, whereas for N. leisleri the range is 0–56%. There is no plausible functional explanation for using echolocation calls of such low intensity in open space other than to circumvent the bat-detecting ears of insects, which is likely to explain the fact that the diet of B. barbastellus consists almost entirely of eared moths. This suggests that circumventing moth hearing is the selective pressure that maintains, and perhaps originally selected for, the use of low-intensity calls by B. barbastellus. Further phylogenetic analyses are needed to determine the ancestral state of calling amplitude in the bat lineages closely related to B. barbastellus to provide stronger evidence for this hypothesis.

Bats and their insect prey provide a fascinating study system to investigate questions related to the evolution of sensory systems and behaviour. The evolution of echolocation in bats for orientation and finding food was a major change for the entire nocturnal animal community at the time, and the number of times that bat-detecting ears have independently evolved in the insects is a testament to the intense selection pressure that bats exert on these species. Designating these relationships as co-evolutionary processes is problematic because of the very diffuse interactions between these groups, but research into potential counter-adaptations by bats to the anti-bat adaptations in insects provides exciting examples of how bat echolocation can be influenced by their prey. Several aspects of bat–insect interactions make them excellent candidates for future work on predator–prey dynamics. First, whereas most predator–prey interactions may be multimodal, and thus require many combinations of sensory stimuli for adequate behavioural tests, bats and moths present a simple, almost exclusively acoustic-based system. Second, sound is relatively easily recorded, analysed and reproduced for experimentation. Third, recent advances in 3D microphone arrays and 3D infrared videography mean that researchers have begun to investigate these interactions in the wild (Corcoran and Conner, 2012). Fourth, using combinations of real-world interactions, naturalistic acoustic stimuli and neurophysiological recordings, researchers will soon be able to elucidate the neural underpinnings of bat–moth interactions, thus providing us with further insight into this fascinating system.

We dedicate this review to James Fullard and Annemarie Surlykke. We thank Jeneni Thiagavel, Signe Brinkløv, Jesse Barber, Charlotte Rutledge and one anonymous reviewer for comments on the manuscript.