Main

Each of the five bat species catches flying insects or spiders on threads close to the clutter-producing foliage of vegetation13,14,15,16,17,18, so they are all faced with the problem of distinguishing echoes from prey and from background or clutter2,3,4,6. Two species also take insects from flat water surfaces by trawling, and thus have to cope with clutter echoes from drifting targets or the shore14,17. Using previously published field data on foraging behaviour, diet and microhabitat use13,14,15,16,17,18,19 as well as morphometric data20, we hypothesized that the five species would differ in ability to reject clutter. This, in turn, should be reflected in the minimal distance from background at which the bats could find prey by echolocation. From these data, we predicted the following order of decreasing prey-capture ability: M. nattereri, M. emarginatus, M. mystacinus, M. daubentonii, M. dasycneme. To test this prediction, we encouraged all species to catch prey as close as possible to a standardized clutter-producing background and measured their minimal capture distances (defined as the distance corresponding to 50% capture success).

In a flight tent, we presented mealworms to freshly captured bats at different distances to a vertical ‘clutter screen’ (polypropylene carpet with many latex-clay nubs; see Methods) that produced clutter echoes. In the dark, all bats from all of the five species located and captured silent, motionless prey close to the clutter screen. Bats also attacked rubber prey dummies lacking the odour of prey (total of 359 attacks on dummies: 39 by M. nattereri, 81 by M. emarginatus, 101 by M. mystacinus, 138 by M. daubentonii; dummies not presented to M. dasycneme). We conclude that visual, olfactory and passive acoustic cues were not necessary for prey detection, location and capture, all of which could be accomplished by echolocation alone. This is further corroborated by the fact that the bats of all five species consistently produced approach sequences that terminated with a ‘buzz’ (a series of very short calls with a repetition rate of about 200 Hz) typical of aerial hawking bats3. In our experiments, no bat ever captured a mealworm that was presented directly on the clutter background (Fig. 1). As in previous studies18,21,22, this finding suggests that bats could not use echolocation to distinguish prey directly positioned on clutter-producing background. At distances of 25 and 50 cm almost all mealworms were captured by all bats of all five species. However, capture success differed significantly between species when prey were presented at 5 and 10 cm from clutter (Table 1) and decreased in the order of species we had predicted (Fig. 1). These differences in prey-capture ability suggest that some within-guild structuring of niche space exists.

Figure 1: The capture performance of bats from five sympatric Myotis species searching for prey offered at different distances from a clutter screen that mimicked a vegetation edge.
figure 1

Species' abilities differed at distances of 5 and 10 cm (see Table 1 for statistics). For this graph, individual performances were pooled for each species (for individual performance data see Supplementary Figure). Trials per species per distance of 0, 5, 10, 25 and 50 cm are given below the species names (parentheses enclose trials per individual). We performed a total of 569 trials.

Full size image
Table 1 Capture performance comparison for five sympatric Myotis species
Full size table

Parameters of wing morphology, such as low wing loading or low aspect ratio, are related causally to the ability to sustain slow, manoeuvrable flight5, a prerequisite for consistently capturing prey close to vegetation or a clutter screen. However, although the wing loading is nearly identical in M. emarginatus, M. mystacinus and M. daubentonii5 and they indeed showed similar flight and approach behaviour in our experiments, their capture performance differed considerably. Therefore, we argue that this measured difference in capture success was caused mainly by differences in echolocation call parameters and related differences in echo-processing capabilities. To test this hypothesis we compared the measured minimal capture distances with a number of signal parameters.

The search calls used in our standardized experimental situation were steep broadband frequency-modulated (FM) signals and averaged 1.4–1.8 ms in duration (Fig. 2). The first harmonic was always most prominent. Starting, peak and terminal frequency, bandwidth and pulse interval differed between the species, whereas pulse duration showed no significant differences between the five congeners (Table 2). Capture performance was unrelated to pulse duration, peak frequency or terminal frequency (Table 2). Species with higher starting frequencies and bandwidths and shorter pulse intervals were able to capture prey closer to clutter than those with lower starting frequencies and bandwidths and higher pulse intervals (Table 2, Fig. 3). We qualitatively and quantitatively corroborated the link between bandwidth and performance by building a successful logistic model (see Supplementary Methods) with the individual performance data (see Supplementary Figure). We further examined whether this correlation was a mere product of phylogenetic inertia by performing an independent contrast regression analysis23 using a recent molecular phylogeny of the genus Myotis24. Apparently, the association of broad bandwidth and good capture ability has evolved several times convergently even among the species examined, as there are two clades including a high-bandwidth species with good capture abilities and a low-bandwidth species with lower abilities (an M. nattereri–M. daubentonii clade and an M. emarginatus–M. dasycneme clade). Consequently, the correlation between bandwidth and minimal capture distance remained strong and significant for the phylogenetically corrected data set (R2 = 0.95, P = 0.0047), with a minimal capture distance contrast slope of -0.14 with respect to contrast in bandwidth, strikingly similar to that in the non-corrected data (see Fig. 3).

Figure 2: Representative search calls of five sympatric Myotis species in sonagram representation with time signal below and averaged power spectrum on the right.
figure 2

a, Myotis nattereri; b, M. emarginatus; c, M. mystacinus; d, M. daubentonii; and e, M. dasycneme. See Methods for definition of search calls. Notches in signals (a, c) are caused by two-wavefront interference.

Full size image
Table 2 Search call parameters and their correlation with minimal capture distances
Full size table
Figure 3: Correlation of ‘minimal capture distance’ with the bandwidth of search calls for the five sympatric Myotis species.
figure 3

We defined minimal capture distance as the distance of prey to the clutter screen that corresponded to 50% capture success. Minimal capture distance was calculated for every individual by linear interpolation from the two neighbouring distances for which performance was measured (for individual performance data see Supplementary Figure). Data are plotted as species means and error bars give the standard deviation. For number of bats see Methods.

Full size image

When comparing the four species with low to medium capture ability, it became evident that the broadening of the bandwidth was achieved mainly by an increase in the starting frequency. The very high bandwidth of the best-performing species M. nattereri resulted from an additional lowering of the terminal frequency (Table 2). This may indicate that not only absolute frequency, but especially bandwidth is a crucial parameter for the improvement of prey-detection ability.

It could be argued that the bandwidth-dependent increase in range accuracy and range resolution helps bats to find prey25. This explanation might be considered if the search flights were perpendicular to the clutter screen, which would keep the prey echoes in front of the clutter echoes. However, the bats chose an oblique flight path and because of the width of the sonar footprint, the signals hit the screen before they hit the prey and therefore the prey echoes were buried in clutter. Hence, our experimental setting did not constitute a simple ranging task. Rather, we believe that the bats were faced with the classification problem of discriminating clutter echoes without and with prey echo. Our data show that all bats tolerated some overlap of prey echoes and background echoes.

However, until now no convincing hypothesis has been proposed to explain how bats can recognize a desired target amongst clutter, except for flower-visiting bats that use conspicuous echoes provided by the plants they pollinate26. Our data may help to increase the understanding of how bats solve such complex classification problems, because they suggest that high starting frequencies and large bandwidths for good performance in such a task have adaptive value. The width of the sonar footprint on the clutter screen is reduced with increasing frequency, because of the higher directionality of the sonar beam that decreases the number of echoes from clutter at short distances. Therefore, using high-frequency signals could be a good strategy. In addition to an evolutionary increase in starting frequency, bats could achieve a behavioural reduction of the sonar footprint by keeping the search distance short27. To investigate whether bats improve spatial unmasking by optimizing flight behaviour and call directionality during both search and approach, further studies with three-dimensional analysis of flight-path and sonar-beam characteristics will be required.

We believe that an increase in bandwidth is probably advantageous because of the illumination of the sonar scene with a wider range of wavelengths: there is an 8.4-fold decrease from starting to terminal frequency in the case of the best-performing M. nattereri. An M. nattereri call that sweeps from 135 to 16 kHz contains wavelengths from 2.6 to 21.8 mm. Many arthropods (including the mealworms we used) and many reflecting background facets (leaves, grass tips and also the nubs of our clutter screen) fall within this size range. With a broadband call, many reflectors will be illuminated with wavelengths above, at and below their sizes simultaneously. Echo intensity, directionality and other reflection properties depend strongly on whether the wavelength of the incident sound is above, well below or about the same as the reflector size28. Therefore, a given sonar scene may convey clear differences in the information content at frequency channels which are far apart. Bats using broadband calls presumably sample frequency-dependent differences efficiently and develop a detailed characterization of the background contours necessary for a better separation of prey from background.

Gleaning bats from the families Megadermatidae and Phyllostomidae produce broad bandwidths by using multiharmonic signals and often, but not always, rely on passive acoustic prey detection2,3,27. The Myotis in our study as well as the palaeotropical Kerivoulinae and Murinae29 (all family Vespertilionidae) achieve large call bandwidths with the first harmonic only, which may possibly prove to be a key innovation for prey detection by echolocation close to clutter.

Our data suggest that the differences in prey-detection abilities of different bats even within groups of morphologically and ecologically similar species are linked to differences in their echolocation signals and associated sensory abilities. Our data further indicate that differences in sensory abilities generate differences in prey availability for different bat species, even when foraging in exactly the same place. Therefore, differences in echolocation signals and associated sensory abilities contribute to within-guild resource partitioning. From our findings in bats, we suggest that the sensory ecologies of potentially competing species might play an important role in the structuring of animal communities.

Methods

Animals and flight tent

Bats were captured in Germany under licence from the responsible authorities (BR Hannover licence no. 503.41-22201/3; RP Karlsruhe 73c1-8852.15; RP Tübingen 73-8/8852.21; RP Stuttgart 73-8850.68-14/ Uni Tübingen; RP Freiburg Az: 73/8852.46-2) and released to the wild at the site of capture following completion of experiments. The freshly captured, experimentally naive bats were released into a mobile flight tent with a ground area of 3.5 by 7 m and about 2.5 m height, erected close to the site where the bats were captured. It had a natural light regime. We present data from three M. dasycneme (male), two M. daubentonii (male), eight M. emarginatus (female), two M. mystacinus (one male, one female) and four M. nattereri (one male, three female).

Behavioural experiments

Experiments were run during the natural activity period of the bats, either in complete darkness (moonless night) or with dim light (artificial or moonlight) from outside the flight tent. To assess the capture ability of the different species close to clutter-producing background, one mealworm (Tenebrio molitor Coleoptera) at a time was suspended on a thread (of diameter 0.1 mm) 0, 5, 10, 25 and 50 cm in front of a vertical clutter screen. A trial was scored as successful if the bat found and captured the mealworm within a 1-min time window of continuous search flight. We continuously modified the position of the mealworms in front of the clutter screen (left–right, up–down) to avoid conditioning the bats to a certain location of the screen and pseudo-randomly alternated between distances within any one session. It is important to note that we examined the bats in rather an unnatural situation, and so we did not necessarily measure the minimal detection distances bats will show in the wild, when searching for insects in front of vegetation. In natural situations, the classification problem of discriminating clutter echoes without and with prey echo will be far more complex because vegetation contours will be more irregular than our clutter screen and will deliver clutter echoes not only from the outer contour but also from the reflecting facets below30.

We performed the experiments separately with only one bat flying at a time, with the exception of two M. nattereri, and three and four M. emarginatus that had to be flown together for motivational reasons. Their performances and echolocation calls were combined to form one ‘individual unit’ (IU) each. Thus we have presented data for 13 individuals (bats or IUs, respectively).

We used a vertical clutter screen to mimic vegetation edges in a standardized way. It consisted of a polypropylene carpet (208 cm wide, 170 cm high) with latex-clay half-spheres (nubs) of 5-mm diameter in a regular pattern with 12-mm spacing between them. When ensonified at an acute angle with a bat-like FM pulse, it returned many overlapping copies of this signal22, which is an echo complex somewhat similar to the echoes of natural vegetation30.

To investigate the importance of arthropod-specific cues for prey detection, rubber dummies (electrical shrinkwrap tubing ranging from 1.6 to 12.7 mm in diameter and 1 to 47 mm in length) were offered to the bats by suspending them on nylon threads in the flight tent.

Search call recordings and sound analysis

Sequences of echolocation calls were recorded from each bat when flying in front of the clutter screen during trials in which they did not find the mealworm. We assumed that the bats were searching for (and not approaching) prey and we therefore considered these echolocation signals to be search calls. Sound recording and analysis were performed with custom-built hard- and software as described in detail elsewhere18 (256 FFT, 93.75% overlap). Sound duration and pulse interval were measured from the time signal. Starting frequency and terminal frequency were determined from a sonagram representation at 25 dB below peak frequency (that is, frequency with maximum amplitude) on the first harmonic of each signal. The -25 dB bandwidth (that is, sweep range) of each signal was computed as starting frequency minus terminal frequency.