Adaptive echolocation behavior in bats for the analysis of auditory scenes

Abstract

Echolocating bats emit sonar pulses and listen to returning echoes to probe their surroundings. Bats adapt their echolocation call design to cope with dynamic changes in the acoustic environment, including habitat change or the presence of nearby conspecifics/heterospecifics. Seven pairs of big brown bats, Eptesicus fuscus, were tested in this study to examine how they adjusted their echolocation calls when flying and competing with a conspecific for food. Results showed that differences in five call parameters, start/end frequencies, duration, bandwidth and sweep rate, significantly increased in the two-bat condition compared with the baseline data. In addition, the magnitude of spectral separation of calls was negatively correlated with the baseline call design differences in individual bats. Bats with small baseline call frequency differences showed larger increases in call frequency separation when paired than those with large baseline call frequency differences, suggesting that bats actively change their sonar call structure if pre-existing differences in call design are small. Call design adjustments were also influenced by physical spacing between two bats. Calls of paired bats exhibited the largest design separations when inter-bat distance was shorter than 0.5 m, and the separation decreased as the spacing increased. All individuals modified at least one baseline call parameter in response to the presence of another conspecific. We propose that dissimilarity between the time-frequency features of sonar calls produced by different bats aids each individual in segregating echoes of its own sonar vocalizations from the acoustic signals of neighboring bats.

Fig. 1.

Illustration of assignment of echolocation calls to individual bats. The sound speed is 346.65 m s^–1; T₁ and T₂ is the onset time of recorded calls at microphone 1 and 2, respectively; t₁ and t₂ is the signal travel time from the bat to microphones 1 and 2, respectively, which are estimated from video recordings; d₁ and d₂ is the distance between bat and microphone 1 and 2, respectively. Actual audio delay is calculated from audio recordings and is equal to T₁–T₂. Estimated audio delay is calculated from video recordings and is equal to t₁–t₂. Values of real audio delay and estimated audio delay are the same if one call was correctly assigned to the vocalizing bat.

Fig. 2.

Two examples show the relative position of paired bats and the design of their vocalizations. The 3-D flight paths of each bat in (A) example No. 1 and (B) example No. 2. Arrows in the starting points of each flight curve marked the flight direction of each bat. Flight trajectories of each bat were marked by different colors (blue and red). One bat flew behind the other bat and followed the leading bat's flight trajectory in example No. 1. Two bats flew almost parallel in the beginning of example No. 2. The number beside each flight path is the trial time and matched the x-axis in panel (B) and (D), respectively. Each asterisk and open circle represents one vocalization from bat A (asterisks) and bat B (open circles). The inter-bat distance and call design of bat A and bat B are shown in (B) example No. 1 and (D) example No. 2. The asterisks represent vocalizations from bat A and the open circles represent vocalizations from bat B. From the upper to lower panels are inter-bat distance, start/end frequencies (those two curves with higher values are start frequencies and the other two are end frequencies), duration and sweep rate.

Fig. 3.

Schematic representation of sequential call analysis. Each point represents the start frequency of one vocalization, and different letters mean calls made by different bats. For example, A1 is the first call bat A produced and B3 is the third call bat B generated. The x-axis is the time and y-axis is the start frequency of calls. Curves between two calls represent two consecutive vocalizations produced by different bats and absolute differences between these two sequential calls are used to represent separation in paired bats' call design. Two consecutive calls, which were not connected by curves, were not included in data analysis because they were produced by the same individual.

Fig. 4.

Distribution of call design separation between two sequential calls produced by different bats when they flew together. The thick black line in each histogram indicates the baseline separation, which is the difference in call design between two bats when they each flew alone. The percentages mark the proportion of calls that exceed the baseline separation. Call parameters analyzed here are (A) start frequency, (B) end frequency, (C) bandwidth, (D) duration and (E) sweep rate.

Fig. 5.

The correlation between each pair's baseline separation and the magnitude of adjustment from baseline to two-bat condition in (A) start frequency, (B) end frequency, (C) bandwidth, (D) duration and (E) sweep rate. ^*Means P<0.05 and ^** means P<0.01. Each data point represents one bat pair and the number next to each point refers to different bat pairs. Only spectral parameters, start/end frequencies and bandwidth, show significant negative correlation.

Fig. 6.

The correlation between start/end frequencies separation and proportion of calls with higher start/end frequencies than the preceding call. The closed circles represent start frequency and the open circles represent end frequency in each bat pair. The bat with higher start frequency tended to keep a higher frequency than the other bat in the pair when the start frequency separation is large for this pair. End frequency shows a similar trend to start frequency.

Fig. 7.

The mean comparison of each call design separation between two consecutive calls at different inter-bat distance by one-way analysis of variance (ANOVA). Error bars indicate standard error of mean. Different letters mean that there is a significant difference between these two values. The dotted line in each panel shows the baseline separation. Call designs are measured by five parameters: (A) start frequency, (B) end frequency, (C) bandwidth, (D) duration and (E) sweep rate.

Fig. 8.

The amount of deviation from baseline data in the two-bat condition for each bat in each pair. White and gray bars indicate data from different bats in a pair. Five call parameters were presented here: (A) start frequency, (B) end frequency, (C) bandwidth, (D) duration and (E) sweep rate. All deviated amounts are either significantly larger or smaller than zero, except those marked with n.s. The x-axis shows bat pairs and these bat pair numbers correspond to those shown in Fig. 5. Error bars represent standard error of the mean.

Fig. 9.

The magnitude of call design adjustment from baseline to two-bat condition as a function of pulse intervals. Error bats indicate standard error of mean and different letters mean that there is a significant difference between these two values. The x-axis is the pulse interval of individual bats and the y-axis is the magnitude of adjustment for five call parameters: (A) start frequency, (B) end frequency, (C) bandwidth, (D) duration and (E) sweep rate.